update libvpx to hash 133f83e2c tag v1.7.0 from https://chromium.googlesource.com/webm/libvpx

This commit is contained in:
Anthony Minessale 2019-01-17 20:16:34 +00:00 committed by Andrey Volk
parent 9bfd5255b2
commit 731ee28fe7
599 changed files with 103914 additions and 65606 deletions

View File

@ -3,11 +3,13 @@
Aaron Watry <awatry@gmail.com>
Abo Talib Mahfoodh <ab.mahfoodh@gmail.com>
Adam Xu <adam@xuyaowu.com>
Adrian Grange <agrange@google.com>
Aex Converse <aconverse@google.com>
Ahmad Sharif <asharif@google.com>
Aleksey Vasenev <margtu-fivt@ya.ru>
Alexander Potapenko <glider@google.com>
Alexander Voronov <avoronov@graphics.cs.msu.ru>
Alexandra Hájková <alexandra.khirnova@gmail.com>
Alexis Ballier <aballier@gentoo.org>
Alok Ahuja <waveletcoeff@gmail.com>
Alpha Lam <hclam@google.com>
@ -15,6 +17,7 @@ A.Mahfoodh <ab.mahfoodh@gmail.com>
Ami Fischman <fischman@chromium.org>
Andoni Morales Alastruey <ylatuya@gmail.com>
Andres Mejia <mcitadel@gmail.com>
Andrew Lewis <andrewlewis@google.com>
Andrew Russell <anrussell@google.com>
Angie Chiang <angiebird@google.com>
Aron Rosenberg <arosenberg@logitech.com>
@ -22,11 +25,14 @@ Attila Nagy <attilanagy@google.com>
Brion Vibber <bvibber@wikimedia.org>
changjun.yang <changjun.yang@intel.com>
Charles 'Buck' Krasic <ckrasic@google.com>
Cheng Chen <chengchen@google.com>
chm <chm@rock-chips.com>
Chris Cunningham <chcunningham@chromium.org>
Christian Duvivier <cduvivier@google.com>
Daniele Castagna <dcastagna@chromium.org>
Daniel Kang <ddkang@google.com>
Deb Mukherjee <debargha@google.com>
Deepa K G <deepa.kg@ittiam.com>
Dim Temp <dimtemp0@gmail.com>
Dmitry Kovalev <dkovalev@google.com>
Dragan Mrdjan <dmrdjan@mips.com>
@ -37,17 +43,21 @@ Fabio Pedretti <fabio.ped@libero.it>
Frank Galligan <fgalligan@google.com>
Fredrik Söderquist <fs@opera.com>
Fritz Koenig <frkoenig@google.com>
Gabriel Marin <gmx@chromium.org>
Gaute Strokkenes <gaute.strokkenes@broadcom.com>
Geza Lore <gezalore@gmail.com>
Ghislain MARY <ghislainmary2@gmail.com>
Giuseppe Scrivano <gscrivano@gnu.org>
Gordana Cmiljanovic <gordana.cmiljanovic@imgtec.com>
Gregor Jasny <gjasny@gmail.com>
Guillaume Martres <gmartres@google.com>
Guillermo Ballester Valor <gbvalor@gmail.com>
Hangyu Kuang <hkuang@google.com>
Hanno Böck <hanno@hboeck.de>
Han Shen <shenhan@google.com>
Henrik Lundin <hlundin@google.com>
Hui Su <huisu@google.com>
Ivan Krasin <krasin@chromium.org>
Ivan Maltz <ivanmaltz@google.com>
Jacek Caban <cjacek@gmail.com>
Jacky Chen <jackychen@google.com>
@ -61,6 +71,7 @@ Jean-Yves Avenard <jyavenard@mozilla.com>
Jeff Faust <jfaust@google.com>
Jeff Muizelaar <jmuizelaar@mozilla.com>
Jeff Petkau <jpet@chromium.org>
Jerome Jiang <jianj@google.com>
Jia Jia <jia.jia@linaro.org>
Jian Zhou <zhoujian@google.com>
Jim Bankoski <jimbankoski@google.com>
@ -75,7 +86,9 @@ Joshua Litt <joshualitt@google.com>
Julia Robson <juliamrobson@gmail.com>
Justin Clift <justin@salasaga.org>
Justin Lebar <justin.lebar@gmail.com>
Kaustubh Raste <kaustubh.raste@imgtec.com>
KO Myung-Hun <komh@chollian.net>
Kyle Siefring <kylesiefring@gmail.com>
Lawrence Velázquez <larryv@macports.org>
Linfeng Zhang <linfengz@google.com>
Lou Quillio <louquillio@google.com>
@ -91,8 +104,12 @@ Michael Kohler <michaelkohler@live.com>
Mike Frysinger <vapier@chromium.org>
Mike Hommey <mhommey@mozilla.com>
Mikhal Shemer <mikhal@google.com>
Min Chen <chenm003@gmail.com>
Minghai Shang <minghai@google.com>
Min Ye <yeemmi@google.com>
Moriyoshi Koizumi <mozo@mozo.jp>
Morton Jonuschat <yabawock@gmail.com>
Nathan E. Egge <negge@mozilla.com>
Nico Weber <thakis@chromium.org>
Parag Salasakar <img.mips1@gmail.com>
Pascal Massimino <pascal.massimino@gmail.com>
@ -101,16 +118,22 @@ Paul Wilkins <paulwilkins@google.com>
Pavol Rusnak <stick@gk2.sk>
Paweł Hajdan <phajdan@google.com>
Pengchong Jin <pengchong@google.com>
Peter Boström <pbos@chromium.org>
Peter Collingbourne <pcc@chromium.org>
Peter de Rivaz <peter.derivaz@gmail.com>
Philip Jägenstedt <philipj@opera.com>
Priit Laes <plaes@plaes.org>
Rafael Ávila de Espíndola <rafael.espindola@gmail.com>
Rafaël Carré <funman@videolan.org>
Rafael de Lucena Valle <rafaeldelucena@gmail.com>
Rahul Chaudhry <rahulchaudhry@google.com>
Ralph Giles <giles@xiph.org>
Ranjit Kumar Tulabandu <ranjit.tulabandu@ittiam.com>
Rob Bradford <rob@linux.intel.com>
Ronald S. Bultje <rsbultje@gmail.com>
Rui Ueyama <ruiu@google.com>
Sami Pietilä <samipietila@google.com>
Sarah Parker <sarahparker@google.com>
Sasi Inguva <isasi@google.com>
Scott Graham <scottmg@chromium.org>
Scott LaVarnway <slavarnway@google.com>
@ -118,9 +141,11 @@ Sean McGovern <gseanmcg@gmail.com>
Sergey Kolomenkin <kolomenkin@gmail.com>
Sergey Ulanov <sergeyu@chromium.org>
Shimon Doodkin <helpmepro1@gmail.com>
Shiyou Yin <yinshiyou-hf@loongson.cn>
Shunyao Li <shunyaoli@google.com>
Stefan Holmer <holmer@google.com>
Suman Sunkara <sunkaras@google.com>
Sylvestre Ledru <sylvestre@mozilla.com>
Taekhyun Kim <takim@nvidia.com>
Takanori MATSUURA <t.matsuu@gmail.com>
Tamar Levy <tamar.levy@intel.com>
@ -130,7 +155,10 @@ Thijs Vermeir <thijsvermeir@gmail.com>
Tim Kopp <tkopp@google.com>
Timothy B. Terriberry <tterribe@xiph.org>
Tom Finegan <tomfinegan@google.com>
Tristan Matthews <le.businessman@gmail.com>
Urvang Joshi <urvang@google.com>
Vignesh Venkatasubramanian <vigneshv@google.com>
Vlad Tsyrklevich <vtsyrklevich@chromium.org>
Yaowu Xu <yaowu@google.com>
Yi Luo <luoyi@google.com>
Yongzhe Wang <yongzhe@google.com>

View File

@ -1,3 +1,44 @@
2017-01-04 v1.7.0 "Mandarin Duck"
This release focused on high bit depth performance (10/12 bit) and vp9
encoding improvements.
- Upgrading:
This release is ABI incompatible due to new vp9 encoder features.
Frame parallel decoding for vp9 has been removed.
- Enhancements:
vp9 encoding supports additional threads with --row-mt. This can be greater
than the number of tiles.
Two new vp9 encoder options have been added:
--corpus-complexity
--tune-content=film
Additional tooling for respecting the vp9 "level" profiles has been added.
- Bug fixes:
A variety of fuzzing issues.
vp8 threading fix for ARM.
Codec control VP9_SET_SKIP_LOOP_FILTER fixed.
Reject invalid multi resolution configurations.
2017-01-09 v1.6.1 "Long Tailed Duck"
This release improves upon the VP9 encoder and speeds up the encoding and
decoding processes.
- Upgrading:
This release is ABI compatible with 1.6.0.
- Enhancements:
Faster VP9 encoding and decoding.
High bit depth builds now provide similar speed for 8 bit encode and decode
for x86 targets. Other platforms and higher bit depth improvements are in
progress.
- Bug Fixes:
A variety of fuzzing issues.
2016-07-20 v1.6.0 "Khaki Campbell Duck"
This release improves upon the VP9 encoder and speeds up the encoding and
decoding processes.

View File

@ -1,4 +1,4 @@
README - 20 July 2016
README - 24 January 2018
Welcome to the WebM VP8/VP9 Codec SDK!
@ -47,6 +47,7 @@ COMPILING THE APPLICATIONS/LIBRARIES:
--help output of the configure script. As of this writing, the list of
available targets is:
arm64-android-gcc
arm64-darwin-gcc
arm64-linux-gcc
armv7-android-gcc
@ -57,10 +58,13 @@ COMPILING THE APPLICATIONS/LIBRARIES:
armv7-win32-vs11
armv7-win32-vs12
armv7-win32-vs14
armv7-win32-vs15
armv7s-darwin-gcc
armv8-linux-gcc
mips32-linux-gcc
mips64-linux-gcc
ppc64-linux-gcc
ppc64le-linux-gcc
sparc-solaris-gcc
x86-android-gcc
x86-darwin8-gcc
@ -73,6 +77,7 @@ COMPILING THE APPLICATIONS/LIBRARIES:
x86-darwin13-gcc
x86-darwin14-gcc
x86-darwin15-gcc
x86-darwin16-gcc
x86-iphonesimulator-gcc
x86-linux-gcc
x86-linux-icc
@ -83,6 +88,7 @@ COMPILING THE APPLICATIONS/LIBRARIES:
x86-win32-vs11
x86-win32-vs12
x86-win32-vs14
x86-win32-vs15
x86_64-android-gcc
x86_64-darwin9-gcc
x86_64-darwin10-gcc
@ -91,6 +97,7 @@ COMPILING THE APPLICATIONS/LIBRARIES:
x86_64-darwin13-gcc
x86_64-darwin14-gcc
x86_64-darwin15-gcc
x86_64-darwin16-gcc
x86_64-iphonesimulator-gcc
x86_64-linux-gcc
x86_64-linux-icc
@ -100,6 +107,7 @@ COMPILING THE APPLICATIONS/LIBRARIES:
x86_64-win64-vs11
x86_64-win64-vs12
x86_64-win64-vs14
x86_64-win64-vs15
generic-gnu
The generic-gnu target, in conjunction with the CROSS environment variable,

View File

@ -13,6 +13,7 @@
#include <limits.h>
#include "args.h"
#include "vpx/vpx_integer.h"
#include "vpx_ports/msvc.h"
#if defined(__GNUC__) && __GNUC__
@ -118,13 +119,13 @@ void arg_show_usage(FILE *fp, const struct arg_def *const *defs) {
}
unsigned int arg_parse_uint(const struct arg *arg) {
long int rawval;
uint32_t rawval;
char *endptr;
rawval = strtol(arg->val, &endptr, 10);
rawval = (uint32_t)strtoul(arg->val, &endptr, 10);
if (arg->val[0] != '\0' && endptr[0] == '\0') {
if (rawval >= 0 && rawval <= UINT_MAX) return (unsigned int)rawval;
if (rawval <= UINT_MAX) return rawval;
die("Option %s: Value %ld out of range for unsigned int\n", arg->name,
rawval);
@ -135,10 +136,10 @@ unsigned int arg_parse_uint(const struct arg *arg) {
}
int arg_parse_int(const struct arg *arg) {
long int rawval;
int32_t rawval;
char *endptr;
rawval = strtol(arg->val, &endptr, 10);
rawval = (int32_t)strtol(arg->val, &endptr, 10);
if (arg->val[0] != '\0' && endptr[0] == '\0') {
if (rawval >= INT_MIN && rawval <= INT_MAX) return (int)rawval;

View File

@ -64,6 +64,9 @@ CONFIG_DIR := $(LOCAL_PATH)/
LIBVPX_PATH := $(LOCAL_PATH)/libvpx
ASM_CNV_PATH_LOCAL := $(TARGET_ARCH_ABI)/ads2gas
ASM_CNV_PATH := $(LOCAL_PATH)/$(ASM_CNV_PATH_LOCAL)
ifneq ($(V),1)
qexec := @
endif
# Use the makefiles generated by upstream configure to determine which files to
# build. Also set any architecture-specific flags.
@ -71,7 +74,7 @@ ifeq ($(TARGET_ARCH_ABI),armeabi-v7a)
include $(CONFIG_DIR)libs-armv7-android-gcc.mk
LOCAL_ARM_MODE := arm
else ifeq ($(TARGET_ARCH_ABI),arm64-v8a)
include $(CONFIG_DIR)libs-armv8-android-gcc.mk
include $(CONFIG_DIR)libs-arm64-android-gcc.mk
LOCAL_ARM_MODE := arm
else ifeq ($(TARGET_ARCH_ABI),x86)
include $(CONFIG_DIR)libs-x86-android-gcc.mk
@ -101,10 +104,10 @@ LOCAL_CFLAGS := -O3
# like x86inc.asm and x86_abi_support.asm
LOCAL_ASMFLAGS := -I$(LIBVPX_PATH)
.PRECIOUS: %.asm.s
$(ASM_CNV_PATH)/libvpx/%.asm.s: $(LIBVPX_PATH)/%.asm
@mkdir -p $(dir $@)
@$(CONFIG_DIR)$(ASM_CONVERSION) <$< > $@
.PRECIOUS: %.asm.S
$(ASM_CNV_PATH)/libvpx/%.asm.S: $(LIBVPX_PATH)/%.asm
$(qexec)mkdir -p $(dir $@)
$(qexec)$(CONFIG_DIR)$(ASM_CONVERSION) <$< > $@
# For building *_rtcd.h, which have rules in libs.mk
TGT_ISA:=$(word 1, $(subst -, ,$(TOOLCHAIN)))
@ -132,7 +135,7 @@ endif
# Pull out assembly files, splitting NEON from the rest. This is
# done to specify that the NEON assembly files use NEON assembler flags.
# x86 assembly matches %.asm, arm matches %.asm.s
# x86 assembly matches %.asm, arm matches %.asm.S
# x86:
@ -140,31 +143,44 @@ CODEC_SRCS_ASM_X86 = $(filter %.asm, $(CODEC_SRCS_UNIQUE))
LOCAL_SRC_FILES += $(foreach file, $(CODEC_SRCS_ASM_X86), libvpx/$(file))
# arm:
CODEC_SRCS_ASM_ARM_ALL = $(filter %.asm.s, $(CODEC_SRCS_UNIQUE))
CODEC_SRCS_ASM_ARM_ALL = $(filter %.asm.S, $(CODEC_SRCS_UNIQUE))
CODEC_SRCS_ASM_ARM = $(foreach v, \
$(CODEC_SRCS_ASM_ARM_ALL), \
$(if $(findstring neon,$(v)),,$(v)))
CODEC_SRCS_ASM_ADS2GAS = $(patsubst %.s, \
$(ASM_CNV_PATH_LOCAL)/libvpx/%.s, \
CODEC_SRCS_ASM_ADS2GAS = $(patsubst %.S, \
$(ASM_CNV_PATH_LOCAL)/libvpx/%.S, \
$(CODEC_SRCS_ASM_ARM))
LOCAL_SRC_FILES += $(CODEC_SRCS_ASM_ADS2GAS)
ifeq ($(TARGET_ARCH_ABI),armeabi-v7a)
ASM_INCLUDES := vpx_dsp/arm/idct_neon.asm.S
CODEC_SRCS_ASM_NEON = $(foreach v, \
$(CODEC_SRCS_ASM_ARM_ALL),\
$(if $(findstring neon,$(v)),$(v),))
CODEC_SRCS_ASM_NEON_ADS2GAS = $(patsubst %.s, \
$(ASM_CNV_PATH_LOCAL)/libvpx/%.s, \
CODEC_SRCS_ASM_NEON := $(filter-out $(addprefix %, $(ASM_INCLUDES)), \
$(CODEC_SRCS_ASM_NEON))
CODEC_SRCS_ASM_NEON_ADS2GAS = $(patsubst %.S, \
$(ASM_CNV_PATH_LOCAL)/libvpx/%.S, \
$(CODEC_SRCS_ASM_NEON))
LOCAL_SRC_FILES += $(patsubst %.s, \
%.s.neon, \
LOCAL_SRC_FILES += $(patsubst %.S, \
%.S.neon, \
$(CODEC_SRCS_ASM_NEON_ADS2GAS))
NEON_ASM_TARGETS = $(patsubst %.S, \
$(ASM_CNV_PATH)/libvpx/%.S, \
$(CODEC_SRCS_ASM_NEON))
# add a dependency to the full path to the ads2gas output to ensure the
# includes are converted first.
ifneq ($(strip $(NEON_ASM_TARGETS)),)
$(NEON_ASM_TARGETS): $(addprefix $(ASM_CNV_PATH)/libvpx/, $(ASM_INCLUDES))
endif
endif
LOCAL_CFLAGS += \
-DHAVE_CONFIG_H=vpx_config.h \
-I$(LIBVPX_PATH) \
-I$(ASM_CNV_PATH)
-I$(ASM_CNV_PATH) \
-I$(ASM_CNV_PATH)/libvpx
LOCAL_MODULE := libvpx
@ -185,7 +201,8 @@ endif
$$(rtcd_dep_template_SRCS): vpx_scale_rtcd.h
$$(rtcd_dep_template_SRCS): vpx_dsp_rtcd.h
ifneq ($(findstring $(TARGET_ARCH_ABI),x86 x86_64),)
rtcd_dep_template_CONFIG_ASM_ABIS := x86 x86_64 armeabi-v7a
ifneq ($$(findstring $(TARGET_ARCH_ABI),$$(rtcd_dep_template_CONFIG_ASM_ABIS)),)
$$(rtcd_dep_template_SRCS): vpx_config.asm
endif
endef
@ -195,16 +212,17 @@ $(eval $(call rtcd_dep_template))
.PHONY: clean
clean:
@echo "Clean: ads2gas files [$(TARGET_ARCH_ABI)]"
@$(RM) $(CODEC_SRCS_ASM_ADS2GAS) $(CODEC_SRCS_ASM_NEON_ADS2GAS)
@$(RM) -r $(ASM_CNV_PATH)
@$(RM) $(CLEAN-OBJS)
$(qexec)$(RM) $(CODEC_SRCS_ASM_ADS2GAS) $(CODEC_SRCS_ASM_NEON_ADS2GAS)
$(qexec)$(RM) -r $(ASM_CNV_PATH)
$(qexec)$(RM) $(CLEAN-OBJS)
ifeq ($(ENABLE_SHARED),1)
LOCAL_CFLAGS += -fPIC
include $(BUILD_SHARED_LIBRARY)
else
include $(BUILD_STATIC_LIBRARY)
endif
ifeq ($(CONFIG_RUNTIME_CPU_DETECT),yes)
$(call import-module,cpufeatures)
$(call import-module,android/cpufeatures)
endif

View File

@ -90,7 +90,7 @@ all:
.PHONY: clean
clean::
rm -f $(OBJS-yes) $(OBJS-yes:.o=.d) $(OBJS-yes:.asm.s.o=.asm.s)
rm -f $(OBJS-yes) $(OBJS-yes:.o=.d) $(OBJS-yes:.asm.S.o=.asm.S)
rm -f $(CLEAN-OBJS)
.PHONY: clean
@ -124,6 +124,7 @@ ifeq ($(TOOLCHAIN), x86-os2-gcc)
CFLAGS += -mstackrealign
endif
# x86[_64]
$(BUILD_PFX)%_mmx.c.d: CFLAGS += -mmmx
$(BUILD_PFX)%_mmx.c.o: CFLAGS += -mmmx
$(BUILD_PFX)%_sse2.c.d: CFLAGS += -msse2
@ -138,6 +139,12 @@ $(BUILD_PFX)%_avx.c.d: CFLAGS += -mavx
$(BUILD_PFX)%_avx.c.o: CFLAGS += -mavx
$(BUILD_PFX)%_avx2.c.d: CFLAGS += -mavx2
$(BUILD_PFX)%_avx2.c.o: CFLAGS += -mavx2
$(BUILD_PFX)%_avx512.c.d: CFLAGS += -mavx512f -mavx512cd -mavx512bw -mavx512dq -mavx512vl
$(BUILD_PFX)%_avx512.c.o: CFLAGS += -mavx512f -mavx512cd -mavx512bw -mavx512dq -mavx512vl
# POWER
$(BUILD_PFX)%_vsx.c.d: CFLAGS += -maltivec -mvsx
$(BUILD_PFX)%_vsx.c.o: CFLAGS += -maltivec -mvsx
$(BUILD_PFX)%.c.d: %.c
$(if $(quiet),@echo " [DEP] $@")
@ -180,13 +187,13 @@ $(BUILD_PFX)%.asm.o: %.asm
$(qexec)$(if $(CONFIG_DEPENDENCY_TRACKING),,mkdir -p $(dir $@))
$(qexec)$(AS) $(ASFLAGS) -o $@ $<
$(BUILD_PFX)%.s.d: %.s
$(BUILD_PFX)%.S.d: %.S
$(if $(quiet),@echo " [DEP] $@")
$(qexec)mkdir -p $(dir $@)
$(qexec)$(SRC_PATH_BARE)/build/make/gen_asm_deps.sh \
--build-pfx=$(BUILD_PFX) --depfile=$@ $(ASFLAGS) $< > $@
$(BUILD_PFX)%.s.o: %.s
$(BUILD_PFX)%.S.o: %.S
$(if $(quiet),@echo " [AS] $@")
$(qexec)$(if $(CONFIG_DEPENDENCY_TRACKING),,mkdir -p $(dir $@))
$(qexec)$(AS) $(ASFLAGS) -o $@ $<
@ -198,8 +205,8 @@ $(BUILD_PFX)%.c.S: %.c
$(qexec)$(if $(CONFIG_DEPENDENCY_TRACKING),,mkdir -p $(dir $@))
$(qexec)$(CC) -S $(CFLAGS) -o $@ $<
.PRECIOUS: %.asm.s
$(BUILD_PFX)%.asm.s: %.asm
.PRECIOUS: %.asm.S
$(BUILD_PFX)%.asm.S: %.asm
$(if $(quiet),@echo " [ASM CONVERSION] $@")
$(qexec)mkdir -p $(dir $@)
$(qexec)$(ASM_CONVERSION) <$< >$@

View File

@ -138,14 +138,6 @@ while (<STDIN>)
s/DCD(.*)/.long $1/;
s/DCB(.*)/.byte $1/;
# RN to .req
if (s/RN\s+([Rr]\d+|lr)/.req $1/)
{
print;
print "$comment_sub$comment\n" if defined $comment;
next;
}
# Make function visible to linker, and make additional symbol with
# prepended underscore
s/EXPORT\s+\|([\$\w]*)\|/.global $1 \n\t.type $1, function/;

View File

@ -120,18 +120,6 @@ while (<STDIN>)
s/DCD(.*)/.long $1/;
s/DCB(.*)/.byte $1/;
# Build a hash of all the register - alias pairs.
if (s/(.*)RN(.*)/$1 .req $2/g)
{
$register_aliases{trim($1)} = trim($2);
next;
}
while (($key, $value) = each(%register_aliases))
{
s/\b$key\b/$value/g;
}
# Make function visible to linker, and make additional symbol with
# prepended underscore
s/EXPORT\s+\|([\$\w]*)\|/.globl _$1\n\t.globl $1/;

View File

@ -332,20 +332,10 @@ EOF
}
check_cflags() {
log check_cflags "$@"
case "$CC" in
*gcc*|*clang)
check_cc -Werror "$@" <<EOF
log check_cflags "$@"
check_cc -Werror "$@" <<EOF
int x;
EOF
;;
*)
check_cc "$@" <<EOF
int x;
EOF
;;
esac
}
check_cxxflags() {
@ -413,6 +403,23 @@ check_gcc_machine_option() {
fi
}
# tests for -m$2, -m$3, -m$4... toggling the feature given in $1.
check_gcc_machine_options() {
feature="$1"
shift
flags="-m$1"
shift
for opt in $*; do
flags="$flags -m$opt"
done
if enabled gcc && ! disabled "$feature" && ! check_cflags $flags; then
RTCD_OPTIONS="${RTCD_OPTIONS}--disable-$feature "
else
soft_enable "$feature"
fi
}
write_common_config_banner() {
print_webm_license config.mk "##" ""
echo '# This file automatically generated by configure. Do not edit!' >> config.mk
@ -645,7 +652,7 @@ setup_gnu_toolchain() {
AS=${AS:-${CROSS}as}
STRIP=${STRIP:-${CROSS}strip}
NM=${NM:-${CROSS}nm}
AS_SFX=.s
AS_SFX=.S
EXE_SFX=
}
@ -684,7 +691,6 @@ check_xcode_minimum_version() {
process_common_toolchain() {
if [ -z "$toolchain" ]; then
gcctarget="${CHOST:-$(gcc -dumpmachine 2> /dev/null)}"
# detect tgt_isa
case "$gcctarget" in
aarch64*)
@ -707,6 +713,18 @@ process_common_toolchain() {
*sparc*)
tgt_isa=sparc
;;
power*64*-*)
tgt_isa=ppc64
;;
power*)
tgt_isa=ppc
;;
*mips64el*)
tgt_isa=mips64
;;
*mips32el*)
tgt_isa=mips32
;;
esac
# detect tgt_os
@ -735,9 +753,16 @@ process_common_toolchain() {
tgt_isa=x86_64
tgt_os=darwin15
;;
*darwin16*)
tgt_isa=x86_64
tgt_os=darwin16
;;
x86_64*mingw32*)
tgt_os=win64
;;
x86_64*cygwin*)
tgt_os=win64
;;
*mingw32*|*cygwin*)
[ -z "$tgt_isa" ] && tgt_isa=x86
tgt_os=win32
@ -785,6 +810,9 @@ process_common_toolchain() {
mips*)
enable_feature mips
;;
ppc*)
enable_feature ppc
;;
esac
# PIC is probably what we want when building shared libs
@ -853,6 +881,10 @@ process_common_toolchain() {
add_cflags "-mmacosx-version-min=10.11"
add_ldflags "-mmacosx-version-min=10.11"
;;
*-darwin16-*)
add_cflags "-mmacosx-version-min=10.12"
add_ldflags "-mmacosx-version-min=10.12"
;;
*-iphonesimulator-*)
add_cflags "-miphoneos-version-min=${IOS_VERSION_MIN}"
add_ldflags "-miphoneos-version-min=${IOS_VERSION_MIN}"
@ -936,7 +968,7 @@ EOF
;;
vs*)
asm_conversion_cmd="${source_path}/build/make/ads2armasm_ms.pl"
AS_SFX=.s
AS_SFX=.S
msvs_arch_dir=arm-msvs
disable_feature multithread
disable_feature unit_tests
@ -946,6 +978,7 @@ EOF
# only "AppContainerApplication" which requires an AppxManifest.
# Therefore disable the examples, just build the library.
disable_feature examples
disable_feature tools
fi
;;
rvct)
@ -1044,7 +1077,7 @@ EOF
STRIP="$(${XCRUN_FIND} strip)"
NM="$(${XCRUN_FIND} nm)"
RANLIB="$(${XCRUN_FIND} ranlib)"
AS_SFX=.s
AS_SFX=.S
LD="${CXX:-$(${XCRUN_FIND} ld)}"
# ASFLAGS is written here instead of using check_add_asflags
@ -1153,10 +1186,20 @@ EOF
fi
fi
if enabled mmi; then
tgt_isa=loongson3a
check_add_ldflags -march=loongson3a
fi
check_add_cflags -march=${tgt_isa}
check_add_asflags -march=${tgt_isa}
check_add_asflags -KPIC
;;
ppc*)
link_with_cc=gcc
setup_gnu_toolchain
check_gcc_machine_option "vsx"
;;
x86*)
case ${tgt_os} in
win*)
@ -1211,6 +1254,13 @@ EOF
AS=msvs
msvs_arch_dir=x86-msvs
vc_version=${tgt_cc##vs}
case $vc_version in
7|8|9|10|11|12|13|14)
echo "${tgt_cc} does not support avx512, disabling....."
RTCD_OPTIONS="${RTCD_OPTIONS}--disable-avx512 "
soft_disable avx512
;;
esac
case $vc_version in
7|8|9|10)
echo "${tgt_cc} does not support avx/avx2, disabling....."
@ -1255,9 +1305,18 @@ EOF
elif disabled $ext; then
disable_exts="yes"
else
# use the shortened version for the flag: sse4_1 -> sse4
check_gcc_machine_option ${ext%_*} $ext
if [ "$ext" = "avx512" ]; then
check_gcc_machine_options $ext avx512f avx512cd avx512bw avx512dq avx512vl
else
# use the shortened version for the flag: sse4_1 -> sse4
check_gcc_machine_option ${ext%_*} $ext
fi
fi
# https://bugs.chromium.org/p/webm/issues/detail?id=1464
# The assembly optimizations for vpx_sub_pixel_variance do not link with
# gcc 6.
enabled sse2 && soft_enable pic
done
if enabled external_build; then
@ -1282,7 +1341,6 @@ EOF
esac
log_echo " using $AS"
fi
[ "${AS##*/}" = nasm ] && add_asflags -Ox
AS_SFX=.asm
case ${tgt_os} in
win32)
@ -1291,7 +1349,7 @@ EOF
EXE_SFX=.exe
;;
win64)
add_asflags -f x64
add_asflags -f win64
enabled debug && add_asflags -g cv8
EXE_SFX=.exe
;;
@ -1425,6 +1483,10 @@ EOF
echo "msa optimizations are available only for little endian platforms"
disable_feature msa
fi
if enabled mmi; then
echo "mmi optimizations are available only for little endian platforms"
disable_feature mmi
fi
fi
;;
esac

View File

@ -25,7 +25,7 @@ files.
Options:
--help Print this message
--out=outfile Redirect output to a file
--ver=version Version (7,8,9,10,11,12,14) of visual studio to generate for
--ver=version Version (7,8,9,10,11,12,14,15) of visual studio to generate for
--target=isa-os-cc Target specifier
EOF
exit 1
@ -215,7 +215,7 @@ for opt in "$@"; do
;;
--ver=*) vs_ver="$optval"
case $optval in
10|11|12|14)
10|11|12|14|15)
;;
*) die Unrecognized Visual Studio Version in $opt
;;
@ -240,9 +240,12 @@ case "${vs_ver:-10}" in
12) sln_vers="12.00"
sln_vers_str="Visual Studio 2013"
;;
14) sln_vers="14.00"
14) sln_vers="12.00"
sln_vers_str="Visual Studio 2015"
;;
15) sln_vers="12.00"
sln_vers_str="Visual Studio 2017"
;;
esac
sfx=vcxproj

View File

@ -34,7 +34,7 @@ Options:
--name=project_name Name of the project (required)
--proj-guid=GUID GUID to use for the project
--module-def=filename File containing export definitions (for DLLs)
--ver=version Version (10,11,12,14) of visual studio to generate for
--ver=version Version (10,11,12,14,15) of visual studio to generate for
--src-path-bare=dir Path to root of source tree
-Ipath/to/include Additional include directories
-DFLAG[=value] Preprocessor macros to define
@ -82,7 +82,7 @@ generate_filter() {
| sed -e "s,$src_path_bare,," \
-e 's/^[\./]\+//g' -e 's,[:/ ],_,g')
if ([ "$pat" == "asm" ] || [ "$pat" == "s" ]) && $asm_use_custom_step; then
if ([ "$pat" == "asm" ] || [ "$pat" == "s" ] || [ "$pat" == "S" ]) && $asm_use_custom_step; then
# Avoid object file name collisions, i.e. vpx_config.c and
# vpx_config.asm produce the same object file without
# this additional suffix.
@ -168,7 +168,7 @@ for opt in "$@"; do
--ver=*)
vs_ver="$optval"
case "$optval" in
10|11|12|14)
10|11|12|14|15)
;;
*) die Unrecognized Visual Studio Version in $opt
;;
@ -203,7 +203,7 @@ for opt in "$@"; do
# The paths in file_list are fixed outside of the loop.
file_list[${#file_list[@]}]="$opt"
case "$opt" in
*.asm|*.s) uses_asm=true
*.asm|*.[Ss]) uses_asm=true
;;
esac
;;
@ -218,7 +218,7 @@ guid=${guid:-`generate_uuid`}
asm_use_custom_step=false
uses_asm=${uses_asm:-false}
case "${vs_ver:-11}" in
10|11|12|14)
10|11|12|14|15)
asm_use_custom_step=$uses_asm
;;
esac
@ -347,6 +347,9 @@ generate_vcxproj() {
if [ "$vs_ver" = "14" ]; then
tag_content PlatformToolset v140
fi
if [ "$vs_ver" = "15" ]; then
tag_content PlatformToolset v141
fi
tag_content CharacterSet Unicode
if [ "$config" = "Release" ]; then
tag_content WholeProgramOptimization true
@ -452,7 +455,7 @@ generate_vcxproj() {
done
open_tag ItemGroup
generate_filter "Source Files" "c;cc;cpp;def;odl;idl;hpj;bat;asm;asmx;s"
generate_filter "Source Files" "c;cc;cpp;def;odl;idl;hpj;bat;asm;asmx;s;S"
close_tag ItemGroup
open_tag ItemGroup
generate_filter "Header Files" "h;hm;inl;inc;xsd"

View File

@ -35,8 +35,8 @@ ARM_TARGETS="arm64-darwin-gcc
armv7s-darwin-gcc"
SIM_TARGETS="x86-iphonesimulator-gcc
x86_64-iphonesimulator-gcc"
OSX_TARGETS="x86-darwin15-gcc
x86_64-darwin15-gcc"
OSX_TARGETS="x86-darwin16-gcc
x86_64-darwin16-gcc"
TARGETS="${ARM_TARGETS} ${SIM_TARGETS}"
# Configures for the target specified by $1, and invokes make with the dist
@ -271,7 +271,7 @@ cat << EOF
--help: Display this message and exit.
--enable-shared: Build a dynamic framework for use on iOS 8 or later.
--extra-configure-args <args>: Extra args to pass when configuring libvpx.
--macosx: Uses darwin15 targets instead of iphonesimulator targets for x86
--macosx: Uses darwin16 targets instead of iphonesimulator targets for x86
and x86_64. Allows linking to framework when builds target MacOSX
instead of iOS.
--preserve-build-output: Do not delete the build directory.

View File

@ -1,4 +1,13 @@
#!/usr/bin/env perl
##
## Copyright (c) 2017 The WebM project authors. All Rights Reserved.
##
## Use of this source code is governed by a BSD-style license
## that can be found in the LICENSE file in the root of the source
## tree. An additional intellectual property rights grant can be found
## in the file PATENTS. All contributing project authors may
## be found in the AUTHORS file in the root of the source tree.
##
no strict 'refs';
use warnings;
@ -200,6 +209,7 @@ sub filter {
sub common_top() {
my $include_guard = uc($opts{sym})."_H_";
print <<EOF;
// This file is generated. Do not edit.
#ifndef ${include_guard}
#define ${include_guard}
@ -335,6 +345,36 @@ EOF
common_bottom;
}
sub ppc() {
determine_indirection("c", @ALL_ARCHS);
# Assign the helper variable for each enabled extension
foreach my $opt (@ALL_ARCHS) {
my $opt_uc = uc $opt;
eval "\$have_${opt}=\"flags & HAS_${opt_uc}\"";
}
common_top;
print <<EOF;
#include "vpx_config.h"
#ifdef RTCD_C
#include "vpx_ports/ppc.h"
static void setup_rtcd_internal(void)
{
int flags = ppc_simd_caps();
(void)flags;
EOF
set_function_pointers("c", @ALL_ARCHS);
print <<EOF;
}
#endif
EOF
common_bottom;
}
sub unoptimized() {
determine_indirection "c";
common_top;
@ -361,10 +401,10 @@ EOF
&require("c");
if ($opts{arch} eq 'x86') {
@ALL_ARCHS = filter(qw/mmx sse sse2 sse3 ssse3 sse4_1 avx avx2/);
@ALL_ARCHS = filter(qw/mmx sse sse2 sse3 ssse3 sse4_1 avx avx2 avx512/);
x86;
} elsif ($opts{arch} eq 'x86_64') {
@ALL_ARCHS = filter(qw/mmx sse sse2 sse3 ssse3 sse4_1 avx avx2/);
@ALL_ARCHS = filter(qw/mmx sse sse2 sse3 ssse3 sse4_1 avx avx2 avx512/);
@REQUIRES = filter(keys %required ? keys %required : qw/mmx sse sse2/);
&require(@REQUIRES);
x86;
@ -381,6 +421,10 @@ if ($opts{arch} eq 'x86') {
@ALL_ARCHS = filter("$opts{arch}", qw/msa/);
last;
}
if (/HAVE_MMI=yes/) {
@ALL_ARCHS = filter("$opts{arch}", qw/mmi/);
last;
}
}
close CONFIG_FILE;
mips;
@ -390,6 +434,9 @@ if ($opts{arch} eq 'x86') {
} elsif ($opts{arch} eq 'armv8' || $opts{arch} eq 'arm64' ) {
@ALL_ARCHS = filter(qw/neon/);
arm;
} elsif ($opts{arch} =~ /^ppc/ ) {
@ALL_ARCHS = filter(qw/vsx/);
ppc;
} else {
unoptimized;
}

View File

@ -60,6 +60,7 @@ if [ ${bare} ]; then
echo "${changelog_version}${git_version_id}" > $$.tmp
else
cat<<EOF>$$.tmp
// This file is generated. Do not edit.
#define VERSION_MAJOR $major_version
#define VERSION_MINOR $minor_version
#define VERSION_PATCH $patch_version

43
libs/libvpx/configure vendored
View File

@ -22,6 +22,7 @@ show_help(){
Advanced options:
${toggle_libs} libraries
${toggle_examples} examples
${toggle_tools} tools
${toggle_docs} documentation
${toggle_unit_tests} unit tests
${toggle_decode_perf_tests} build decoder perf tests with unit tests
@ -108,10 +109,13 @@ all_platforms="${all_platforms} armv7-none-rvct" #neon Cortex-A8
all_platforms="${all_platforms} armv7-win32-vs11"
all_platforms="${all_platforms} armv7-win32-vs12"
all_platforms="${all_platforms} armv7-win32-vs14"
all_platforms="${all_platforms} armv7-win32-vs15"
all_platforms="${all_platforms} armv7s-darwin-gcc"
all_platforms="${all_platforms} armv8-linux-gcc"
all_platforms="${all_platforms} mips32-linux-gcc"
all_platforms="${all_platforms} mips64-linux-gcc"
all_platforms="${all_platforms} ppc64-linux-gcc"
all_platforms="${all_platforms} ppc64le-linux-gcc"
all_platforms="${all_platforms} sparc-solaris-gcc"
all_platforms="${all_platforms} x86-android-gcc"
all_platforms="${all_platforms} x86-darwin8-gcc"
@ -124,6 +128,7 @@ all_platforms="${all_platforms} x86-darwin12-gcc"
all_platforms="${all_platforms} x86-darwin13-gcc"
all_platforms="${all_platforms} x86-darwin14-gcc"
all_platforms="${all_platforms} x86-darwin15-gcc"
all_platforms="${all_platforms} x86-darwin16-gcc"
all_platforms="${all_platforms} x86-iphonesimulator-gcc"
all_platforms="${all_platforms} x86-linux-gcc"
all_platforms="${all_platforms} x86-linux-icc"
@ -134,6 +139,7 @@ all_platforms="${all_platforms} x86-win32-vs10"
all_platforms="${all_platforms} x86-win32-vs11"
all_platforms="${all_platforms} x86-win32-vs12"
all_platforms="${all_platforms} x86-win32-vs14"
all_platforms="${all_platforms} x86-win32-vs15"
all_platforms="${all_platforms} x86_64-android-gcc"
all_platforms="${all_platforms} x86_64-darwin9-gcc"
all_platforms="${all_platforms} x86_64-darwin10-gcc"
@ -142,6 +148,7 @@ all_platforms="${all_platforms} x86_64-darwin12-gcc"
all_platforms="${all_platforms} x86_64-darwin13-gcc"
all_platforms="${all_platforms} x86_64-darwin14-gcc"
all_platforms="${all_platforms} x86_64-darwin15-gcc"
all_platforms="${all_platforms} x86_64-darwin16-gcc"
all_platforms="${all_platforms} x86_64-iphonesimulator-gcc"
all_platforms="${all_platforms} x86_64-linux-gcc"
all_platforms="${all_platforms} x86_64-linux-icc"
@ -151,22 +158,26 @@ all_platforms="${all_platforms} x86_64-win64-vs10"
all_platforms="${all_platforms} x86_64-win64-vs11"
all_platforms="${all_platforms} x86_64-win64-vs12"
all_platforms="${all_platforms} x86_64-win64-vs14"
all_platforms="${all_platforms} x86_64-win64-vs15"
all_platforms="${all_platforms} generic-gnu"
# all_targets is a list of all targets that can be configured
# note that these should be in dependency order for now.
all_targets="libs examples docs"
all_targets="libs examples tools docs"
# all targets available are enabled, by default.
for t in ${all_targets}; do
[ -f "${source_path}/${t}.mk" ] && enable_feature ${t}
done
if ! diff --version >/dev/null; then
die "diff missing: Try installing diffutils via your package manager."
fi
if ! perl --version >/dev/null; then
die "Perl is required to build"
fi
if [ "`cd \"${source_path}\" && pwd`" != "`pwd`" ]; then
# test to see if source_path already configured
if [ -f "${source_path}/vpx_config.h" ]; then
@ -222,6 +233,7 @@ ARCH_LIST="
mips
x86
x86_64
ppc
"
ARCH_EXT_LIST_X86="
mmx
@ -232,7 +244,13 @@ ARCH_EXT_LIST_X86="
sse4_1
avx
avx2
avx512
"
ARCH_EXT_LIST_LOONGSON="
mmi
"
ARCH_EXT_LIST="
neon
neon_asm
@ -243,6 +261,10 @@ ARCH_EXT_LIST="
mips64
${ARCH_EXT_LIST_X86}
vsx
${ARCH_EXT_LIST_LOONGSON}
"
HAVE_LIST="
${ARCH_EXT_LIST}
@ -254,7 +276,6 @@ EXPERIMENT_LIST="
spatial_svc
fp_mb_stats
emulate_hardware
misc_fixes
"
CONFIG_LIST="
dependency_tracking
@ -309,6 +330,7 @@ CONFIG_LIST="
better_hw_compatibility
experimental
size_limit
always_adjust_bpm
${EXPERIMENT_LIST}
"
CMDLINE_SELECT="
@ -331,6 +353,7 @@ CMDLINE_SELECT="
libs
examples
tools
docs
libc
as
@ -367,6 +390,7 @@ CMDLINE_SELECT="
better_hw_compatibility
vp9_highbitdepth
experimental
always_adjust_bpm
"
process_cmdline() {
@ -476,7 +500,7 @@ EOF
#
# Write makefiles for all enabled targets
#
for tgt in libs examples docs solution; do
for tgt in libs examples tools docs solution; do
tgt_fn="$tgt-$toolchain.mk"
if enabled $tgt; then
@ -568,6 +592,7 @@ process_toolchain() {
check_add_cflags -Wdeclaration-after-statement
check_add_cflags -Wdisabled-optimization
check_add_cflags -Wfloat-conversion
check_add_cflags -Wparentheses-equality
check_add_cflags -Wpointer-arith
check_add_cflags -Wtype-limits
check_add_cflags -Wcast-qual
@ -575,12 +600,20 @@ process_toolchain() {
check_add_cflags -Wimplicit-function-declaration
check_add_cflags -Wuninitialized
check_add_cflags -Wunused
# -Wextra has some tricky cases. Rather than fix them all now, get the
# flag for as many files as possible and fix the remaining issues
# piecemeal.
# https://bugs.chromium.org/p/webm/issues/detail?id=1069
check_add_cflags -Wextra
# check_add_cflags also adds to cxxflags. gtest does not do well with
# -Wundef so add it explicitly to CFLAGS only.
check_cflags -Wundef && add_cflags_only -Wundef
if enabled mips || [ -z "${INLINE}" ]; then
enabled extra_warnings || check_add_cflags -Wno-unused-function
fi
# Avoid this warning for third_party C++ sources. Some reorganization
# would be needed to apply this only to test/*.cc.
check_cflags -Wshorten-64-to-32 && add_cflags_only -Wshorten-64-to-32
fi
if enabled icc; then
@ -632,7 +665,7 @@ process_toolchain() {
gen_vcproj_cmd=${source_path}/build/make/gen_msvs_vcxproj.sh
enabled werror && gen_vcproj_cmd="${gen_vcproj_cmd} --enable-werror"
all_targets="${all_targets} solution"
INLINE="__forceinline"
INLINE="__inline"
;;
esac

View File

@ -76,6 +76,7 @@ vpxdec.SRCS += tools_common.c tools_common.h
vpxdec.SRCS += y4menc.c y4menc.h
ifeq ($(CONFIG_LIBYUV),yes)
vpxdec.SRCS += $(LIBYUV_SRCS)
$(BUILD_PFX)third_party/libyuv/%.cc.o: CXXFLAGS += -Wno-unused-parameter
endif
ifeq ($(CONFIG_WEBM_IO),yes)
vpxdec.SRCS += $(LIBWEBM_COMMON_SRCS)

View File

@ -151,7 +151,7 @@ static void write_ivf_frame_header(FILE *outfile,
if (pkt->kind != VPX_CODEC_CX_FRAME_PKT) return;
pts = pkt->data.frame.pts;
mem_put_le32(header, pkt->data.frame.sz);
mem_put_le32(header, (int)pkt->data.frame.sz);
mem_put_le32(header + 4, pts & 0xFFFFFFFF);
mem_put_le32(header + 8, pts >> 32);
@ -190,7 +190,7 @@ static void set_temporal_layer_pattern(int num_temporal_layers,
cfg->ts_layer_id[0] = 0;
cfg->ts_layer_id[1] = 1;
// Use 60/40 bit allocation as example.
cfg->ts_target_bitrate[0] = 0.6f * bitrate;
cfg->ts_target_bitrate[0] = (int)(0.6f * bitrate);
cfg->ts_target_bitrate[1] = bitrate;
/* 0=L, 1=GF */
@ -240,9 +240,9 @@ static void set_temporal_layer_pattern(int num_temporal_layers,
cfg->ts_layer_id[1] = 2;
cfg->ts_layer_id[2] = 1;
cfg->ts_layer_id[3] = 2;
// Use 40/20/40 bit allocation as example.
cfg->ts_target_bitrate[0] = 0.4f * bitrate;
cfg->ts_target_bitrate[1] = 0.6f * bitrate;
// Use 45/20/35 bit allocation as example.
cfg->ts_target_bitrate[0] = (int)(0.45f * bitrate);
cfg->ts_target_bitrate[1] = (int)(0.65f * bitrate);
cfg->ts_target_bitrate[2] = bitrate;
/* 0=L, 1=GF, 2=ARF */
@ -294,8 +294,8 @@ int main(int argc, char **argv) {
vpx_codec_err_t res[NUM_ENCODERS];
int i;
long width;
long height;
int width;
int height;
int length_frame;
int frame_avail;
int got_data;
@ -347,9 +347,9 @@ int main(int argc, char **argv) {
printf("Using %s\n", vpx_codec_iface_name(interface));
width = strtol(argv[1], NULL, 0);
height = strtol(argv[2], NULL, 0);
framerate = strtol(argv[3], NULL, 0);
width = (int)strtol(argv[1], NULL, 0);
height = (int)strtol(argv[2], NULL, 0);
framerate = (int)strtol(argv[3], NULL, 0);
if (width < 16 || width % 2 || height < 16 || height % 2)
die("Invalid resolution: %ldx%ld", width, height);
@ -371,12 +371,13 @@ int main(int argc, char **argv) {
// Bitrates per spatial layer: overwrite default rates above.
for (i = 0; i < NUM_ENCODERS; i++) {
target_bitrate[i] = strtol(argv[NUM_ENCODERS + 5 + i], NULL, 0);
target_bitrate[i] = (int)strtol(argv[NUM_ENCODERS + 5 + i], NULL, 0);
}
// Temporal layers per spatial layers: overwrite default settings above.
for (i = 0; i < NUM_ENCODERS; i++) {
num_temporal_layers[i] = strtol(argv[2 * NUM_ENCODERS + 5 + i], NULL, 0);
num_temporal_layers[i] =
(int)strtol(argv[2 * NUM_ENCODERS + 5 + i], NULL, 0);
if (num_temporal_layers[i] < 1 || num_temporal_layers[i] > 3)
die("Invalid temporal layers: %d, Must be 1, 2, or 3. \n",
num_temporal_layers);
@ -391,9 +392,9 @@ int main(int argc, char **argv) {
downsampled_input[i] = fopen(filename, "wb");
}
key_frame_insert = strtol(argv[3 * NUM_ENCODERS + 5], NULL, 0);
key_frame_insert = (int)strtol(argv[3 * NUM_ENCODERS + 5], NULL, 0);
show_psnr = strtol(argv[3 * NUM_ENCODERS + 6], NULL, 0);
show_psnr = (int)strtol(argv[3 * NUM_ENCODERS + 6], NULL, 0);
/* Populate default encoder configuration */
for (i = 0; i < NUM_ENCODERS; i++) {
@ -460,7 +461,7 @@ int main(int argc, char **argv) {
// Set the number of threads per encode/spatial layer.
// (1, 1, 1) means no encoder threading.
cfg[0].g_threads = 2;
cfg[0].g_threads = 1;
cfg[1].g_threads = 1;
cfg[2].g_threads = 1;
@ -469,7 +470,7 @@ int main(int argc, char **argv) {
if (!vpx_img_alloc(&raw[i], VPX_IMG_FMT_I420, cfg[i].g_w, cfg[i].g_h, 32))
die("Failed to allocate image", cfg[i].g_w, cfg[i].g_h);
if (raw[0].stride[VPX_PLANE_Y] == raw[0].d_w)
if (raw[0].stride[VPX_PLANE_Y] == (int)raw[0].d_w)
read_frame_p = read_frame;
else
read_frame_p = read_frame_by_row;
@ -507,9 +508,11 @@ int main(int argc, char **argv) {
/* Set NOISE_SENSITIVITY to do TEMPORAL_DENOISING */
/* Enable denoising for the highest-resolution encoder. */
if (vpx_codec_control(&codec[0], VP8E_SET_NOISE_SENSITIVITY, 4))
if (vpx_codec_control(&codec[0], VP8E_SET_NOISE_SENSITIVITY, 1))
die_codec(&codec[0], "Failed to set noise_sensitivity");
for (i = 1; i < NUM_ENCODERS; i++) {
if (vpx_codec_control(&codec[1], VP8E_SET_NOISE_SENSITIVITY, 1))
die_codec(&codec[1], "Failed to set noise_sensitivity");
for (i = 2; i < NUM_ENCODERS; i++) {
if (vpx_codec_control(&codec[i], VP8E_SET_NOISE_SENSITIVITY, 0))
die_codec(&codec[i], "Failed to set noise_sensitivity");
}
@ -556,7 +559,8 @@ int main(int argc, char **argv) {
/* Write out down-sampled input. */
length_frame = cfg[i].g_w * cfg[i].g_h * 3 / 2;
if (fwrite(raw[i].planes[0], 1, length_frame,
downsampled_input[NUM_ENCODERS - i - 1]) != length_frame) {
downsampled_input[NUM_ENCODERS - i - 1]) !=
(unsigned int)length_frame) {
return EXIT_FAILURE;
}
}
@ -617,10 +621,6 @@ int main(int argc, char **argv) {
break;
default: break;
}
printf(pkt[i]->kind == VPX_CODEC_CX_FRAME_PKT &&
(pkt[i]->data.frame.flags & VPX_FRAME_IS_KEY)
? "K"
: "");
fflush(stdout);
}
}
@ -661,7 +661,6 @@ int main(int argc, char **argv) {
write_ivf_file_header(outfile[i], &cfg[i], frame_cnt - 1);
fclose(outfile[i]);
}
printf("\n");
return EXIT_SUCCESS;
}

View File

@ -84,6 +84,8 @@ static const arg_def_t speed_arg =
ARG_DEF("sp", "speed", 1, "speed configuration");
static const arg_def_t aqmode_arg =
ARG_DEF("aq", "aqmode", 1, "aq-mode off/on");
static const arg_def_t bitrates_arg =
ARG_DEF("bl", "bitrates", 1, "bitrates[sl * num_tl + tl]");
#if CONFIG_VP9_HIGHBITDEPTH
static const struct arg_enum_list bitdepth_enum[] = {
@ -124,6 +126,7 @@ static const arg_def_t *svc_args[] = { &frames_arg,
#endif
&speed_arg,
&rc_end_usage_arg,
&bitrates_arg,
NULL };
static const uint32_t default_frames_to_skip = 0;
@ -165,7 +168,7 @@ void usage_exit(void) {
static void parse_command_line(int argc, const char **argv_,
AppInput *app_input, SvcContext *svc_ctx,
vpx_codec_enc_cfg_t *enc_cfg) {
struct arg arg = { 0 };
struct arg arg;
char **argv = NULL;
char **argi = NULL;
char **argj = NULL;
@ -250,6 +253,9 @@ static void parse_command_line(int argc, const char **argv_,
} else if (arg_match(&arg, &scale_factors_arg, argi)) {
snprintf(string_options, sizeof(string_options), "%s scale-factors=%s",
string_options, arg.val);
} else if (arg_match(&arg, &bitrates_arg, argi)) {
snprintf(string_options, sizeof(string_options), "%s bitrates=%s",
string_options, arg.val);
} else if (arg_match(&arg, &passes_arg, argi)) {
passes = arg_parse_uint(&arg);
if (passes < 1 || passes > 2) {
@ -417,7 +423,6 @@ static void set_rate_control_stats(struct RateControlStats *rc,
for (sl = 0; sl < cfg->ss_number_layers; ++sl) {
for (tl = 0; tl < cfg->ts_number_layers; ++tl) {
const int layer = sl * cfg->ts_number_layers + tl;
const int tlayer0 = sl * cfg->ts_number_layers;
if (cfg->ts_number_layers == 1)
rc->layer_framerate[layer] = framerate;
else
@ -428,8 +433,8 @@ static void set_rate_control_stats(struct RateControlStats *rc,
cfg->layer_target_bitrate[layer - 1]) /
(rc->layer_framerate[layer] - rc->layer_framerate[layer - 1]);
} else {
rc->layer_pfb[tlayer0] = 1000.0 * cfg->layer_target_bitrate[tlayer0] /
rc->layer_framerate[tlayer0];
rc->layer_pfb[layer] = 1000.0 * cfg->layer_target_bitrate[layer] /
rc->layer_framerate[layer];
}
rc->layer_input_frames[layer] = 0;
rc->layer_enc_frames[layer] = 0;
@ -449,12 +454,13 @@ static void printout_rate_control_summary(struct RateControlStats *rc,
vpx_codec_enc_cfg_t *cfg,
int frame_cnt) {
unsigned int sl, tl;
int tot_num_frames = 0;
double perc_fluctuation = 0.0;
int tot_num_frames = 0;
printf("Total number of processed frames: %d\n\n", frame_cnt - 1);
printf("Rate control layer stats for sl%d tl%d layer(s):\n\n",
cfg->ss_number_layers, cfg->ts_number_layers);
for (sl = 0; sl < cfg->ss_number_layers; ++sl) {
tot_num_frames = 0;
for (tl = 0; tl < cfg->ts_number_layers; ++tl) {
const int layer = sl * cfg->ts_number_layers + tl;
const int num_dropped =
@ -462,7 +468,7 @@ static void printout_rate_control_summary(struct RateControlStats *rc,
? (rc->layer_input_frames[layer] - rc->layer_enc_frames[layer])
: (rc->layer_input_frames[layer] - rc->layer_enc_frames[layer] -
1);
if (!sl) tot_num_frames += rc->layer_input_frames[layer];
tot_num_frames += rc->layer_input_frames[layer];
rc->layer_encoding_bitrate[layer] = 0.001 * rc->layer_framerate[layer] *
rc->layer_encoding_bitrate[layer] /
tot_num_frames;
@ -503,7 +509,7 @@ static void printout_rate_control_summary(struct RateControlStats *rc,
}
vpx_codec_err_t parse_superframe_index(const uint8_t *data, size_t data_sz,
uint32_t sizes[8], int *count) {
uint64_t sizes[8], int *count) {
// A chunk ending with a byte matching 0xc0 is an invalid chunk unless
// it is a super frame index. If the last byte of real video compression
// data is 0xc0 the encoder must add a 0 byte. If we have the marker but
@ -600,9 +606,9 @@ void set_frame_flags_bypass_mode(int sl, int tl, int num_spatial_layers,
}
int main(int argc, const char **argv) {
AppInput app_input = { 0 };
AppInput app_input;
VpxVideoWriter *writer = NULL;
VpxVideoInfo info = { 0 };
VpxVideoInfo info;
vpx_codec_ctx_t codec;
vpx_codec_enc_cfg_t enc_cfg;
SvcContext svc_ctx;
@ -620,7 +626,7 @@ int main(int argc, const char **argv) {
struct RateControlStats rc;
vpx_svc_layer_id_t layer_id;
vpx_svc_ref_frame_config_t ref_frame_config;
int sl, tl;
unsigned int sl, tl;
double sum_bitrate = 0.0;
double sum_bitrate2 = 0.0;
double framerate = 30.0;
@ -634,8 +640,9 @@ int main(int argc, const char **argv) {
// Allocate image buffer
#if CONFIG_VP9_HIGHBITDEPTH
if (!vpx_img_alloc(&raw, enc_cfg.g_input_bit_depth == 8 ? VPX_IMG_FMT_I420
: VPX_IMG_FMT_I42016,
if (!vpx_img_alloc(&raw,
enc_cfg.g_input_bit_depth == 8 ? VPX_IMG_FMT_I420
: VPX_IMG_FMT_I42016,
enc_cfg.g_w, enc_cfg.g_h, 32)) {
die("Failed to allocate image %dx%d\n", enc_cfg.g_w, enc_cfg.g_h);
}
@ -673,7 +680,7 @@ int main(int argc, const char **argv) {
}
#if OUTPUT_RC_STATS
// For now, just write temporal layer streams.
// TODO(wonkap): do spatial by re-writing superframe.
// TODO(marpan): do spatial by re-writing superframe.
if (svc_ctx.output_rc_stat) {
for (tl = 0; tl < enc_cfg.ts_number_layers; ++tl) {
char file_name[PATH_MAX];
@ -691,10 +698,18 @@ int main(int argc, const char **argv) {
if (svc_ctx.speed != -1)
vpx_codec_control(&codec, VP8E_SET_CPUUSED, svc_ctx.speed);
if (svc_ctx.threads)
if (svc_ctx.threads) {
vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, (svc_ctx.threads >> 1));
if (svc_ctx.threads > 1)
vpx_codec_control(&codec, VP9E_SET_ROW_MT, 1);
else
vpx_codec_control(&codec, VP9E_SET_ROW_MT, 0);
}
if (svc_ctx.speed >= 5 && svc_ctx.aqmode == 1)
vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
if (svc_ctx.speed >= 5)
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
vpx_codec_control(&codec, VP8E_SET_MAX_INTRA_BITRATE_PCT, 900);
// Encode frames
while (!end_of_stream) {
@ -730,7 +745,7 @@ int main(int argc, const char **argv) {
&ref_frame_config);
// Keep track of input frames, to account for frame drops in rate control
// stats/metrics.
for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
for (sl = 0; sl < (unsigned int)enc_cfg.ss_number_layers; ++sl) {
++rc.layer_input_frames[sl * enc_cfg.ts_number_layers +
layer_id.temporal_layer_id];
}
@ -755,18 +770,20 @@ int main(int argc, const char **argv) {
SvcInternal_t *const si = (SvcInternal_t *)svc_ctx.internal;
if (cx_pkt->data.frame.sz > 0) {
#if OUTPUT_RC_STATS
uint32_t sizes[8];
uint64_t sizes[8];
int count = 0;
#endif
vpx_video_writer_write_frame(writer, cx_pkt->data.frame.buf,
cx_pkt->data.frame.sz,
cx_pkt->data.frame.pts);
#if OUTPUT_RC_STATS
// TODO(marpan/wonkap): Put this (to line728) in separate function.
// TODO(marpan): Put this (to line728) in separate function.
if (svc_ctx.output_rc_stat) {
vpx_codec_control(&codec, VP9E_GET_SVC_LAYER_ID, &layer_id);
parse_superframe_index(cx_pkt->data.frame.buf,
cx_pkt->data.frame.sz, sizes, &count);
if (enc_cfg.ss_number_layers == 1)
sizes[0] = cx_pkt->data.frame.sz;
// Note computing input_layer_frames here won't account for frame
// drops in rate control stats.
// TODO(marpan): Fix this for non-bypass mode so we can get stats
@ -793,7 +810,7 @@ int main(int argc, const char **argv) {
rc.layer_encoding_bitrate[layer] += 8.0 * sizes[sl];
// Keep count of rate control stats per layer, for non-key
// frames.
if (tl == layer_id.temporal_layer_id &&
if (tl == (unsigned int)layer_id.temporal_layer_id &&
!(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY)) {
rc.layer_avg_frame_size[layer] += 8.0 * sizes[sl];
rc.layer_avg_rate_mismatch[layer] +=
@ -807,7 +824,7 @@ int main(int argc, const char **argv) {
// Update for short-time encoding bitrate states, for moving
// window of size rc->window, shifted by rc->window / 2.
// Ignore first window segment, due to key frame.
if (frame_cnt > rc.window_size) {
if (frame_cnt > (unsigned int)rc.window_size) {
tl = layer_id.temporal_layer_id;
for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
sum_bitrate += 0.001 * 8.0 * sizes[sl] * framerate;
@ -823,13 +840,14 @@ int main(int argc, const char **argv) {
}
// Second shifted window.
if (frame_cnt > rc.window_size + rc.window_size / 2) {
if (frame_cnt >
(unsigned int)(rc.window_size + rc.window_size / 2)) {
tl = layer_id.temporal_layer_id;
for (sl = 0; sl < enc_cfg.ss_number_layers; ++sl) {
sum_bitrate2 += 0.001 * 8.0 * sizes[sl] * framerate;
}
if (frame_cnt > 2 * rc.window_size &&
if (frame_cnt > (unsigned int)(2 * rc.window_size) &&
frame_cnt % rc.window_size == 0) {
rc.window_count += 1;
rc.avg_st_encoding_bitrate += sum_bitrate2 / rc.window_size;
@ -842,10 +860,11 @@ int main(int argc, const char **argv) {
}
#endif
}
/*
printf("SVC frame: %d, kf: %d, size: %d, pts: %d\n", frames_received,
!!(cx_pkt->data.frame.flags & VPX_FRAME_IS_KEY),
(int)cx_pkt->data.frame.sz, (int)cx_pkt->data.frame.pts);
*/
if (enc_cfg.ss_number_layers == 1 && enc_cfg.ts_number_layers == 1)
si->bytes_sum[0] += (int)cx_pkt->data.frame.sz;
++frames_received;

View File

@ -304,6 +304,7 @@ int main(int argc, char **argv) {
const char *height_arg = NULL;
const char *infile_arg = NULL;
const char *outfile_arg = NULL;
const char *update_frame_num_arg = NULL;
unsigned int limit = 0;
vp9_zero(ecodec);
@ -318,18 +319,21 @@ int main(int argc, char **argv) {
height_arg = argv[2];
infile_arg = argv[3];
outfile_arg = argv[4];
update_frame_num_arg = argv[5];
encoder = get_vpx_encoder_by_name("vp9");
if (!encoder) die("Unsupported codec.");
update_frame_num = atoi(argv[5]);
update_frame_num = (unsigned int)strtoul(update_frame_num_arg, NULL, 0);
// In VP9, the reference buffers (cm->buffer_pool->frame_bufs[i].buf) are
// allocated while calling vpx_codec_encode(), thus, setting reference for
// 1st frame isn't supported.
if (update_frame_num <= 1) die("Couldn't parse frame number '%s'\n", argv[5]);
if (update_frame_num <= 1) {
die("Couldn't parse frame number '%s'\n", update_frame_num_arg);
}
if (argc > 6) {
limit = atoi(argv[6]);
limit = (unsigned int)strtoul(argv[6], NULL, 0);
if (update_frame_num > limit)
die("Update frame number couldn't larger than limit\n");
}

View File

@ -26,17 +26,27 @@
#include "../tools_common.h"
#include "../video_writer.h"
#define VP8_ROI_MAP 0
static const char *exec_name;
void usage_exit(void) { exit(EXIT_FAILURE); }
// Denoiser states, for temporal denoising.
enum denoiserState {
kDenoiserOff,
kDenoiserOnYOnly,
kDenoiserOnYUV,
kDenoiserOnYUVAggressive,
kDenoiserOnAdaptive
// Denoiser states for vp8, for temporal denoising.
enum denoiserStateVp8 {
kVp8DenoiserOff,
kVp8DenoiserOnYOnly,
kVp8DenoiserOnYUV,
kVp8DenoiserOnYUVAggressive,
kVp8DenoiserOnAdaptive
};
// Denoiser states for vp9, for temporal denoising.
enum denoiserStateVp9 {
kVp9DenoiserOff,
kVp9DenoiserOnYOnly,
// For SVC: denoise the top two spatial layers.
kVp9DenoiserOnYTwoSpatialLayers
};
static int mode_to_num_layers[13] = { 1, 2, 2, 3, 3, 3, 3, 5, 2, 3, 3, 3, 3 };
@ -154,6 +164,53 @@ static void printout_rate_control_summary(struct RateControlMetrics *rc,
die("Error: Number of input frames not equal to output! \n");
}
#if VP8_ROI_MAP
static void vp8_set_roi_map(vpx_codec_enc_cfg_t *cfg, vpx_roi_map_t *roi) {
unsigned int i, j;
memset(roi, 0, sizeof(*roi));
// ROI is based on the segments (4 for vp8, 8 for vp9), smallest unit for
// segment is 16x16 for vp8, 8x8 for vp9.
roi->rows = (cfg->g_h + 15) / 16;
roi->cols = (cfg->g_w + 15) / 16;
// Applies delta QP on the segment blocks, varies from -63 to 63.
// Setting to negative means lower QP (better quality).
// Below we set delta_q to the extreme (-63) to show strong effect.
roi->delta_q[0] = 0;
roi->delta_q[1] = -63;
roi->delta_q[2] = 0;
roi->delta_q[3] = 0;
// Applies delta loopfilter strength on the segment blocks, varies from -63 to
// 63. Setting to positive means stronger loopfilter.
roi->delta_lf[0] = 0;
roi->delta_lf[1] = 0;
roi->delta_lf[2] = 0;
roi->delta_lf[3] = 0;
// Applies skip encoding threshold on the segment blocks, varies from 0 to
// UINT_MAX. Larger value means more skipping of encoding is possible.
// This skip threshold only applies on delta frames.
roi->static_threshold[0] = 0;
roi->static_threshold[1] = 0;
roi->static_threshold[2] = 0;
roi->static_threshold[3] = 0;
// Use 2 states: 1 is center square, 0 is the rest.
roi->roi_map =
(uint8_t *)calloc(roi->rows * roi->cols, sizeof(*roi->roi_map));
for (i = 0; i < roi->rows; ++i) {
for (j = 0; j < roi->cols; ++j) {
if (i > (roi->rows >> 2) && i < ((roi->rows * 3) >> 2) &&
j > (roi->cols >> 2) && j < ((roi->cols * 3) >> 2)) {
roi->roi_map[i * roi->cols + j] = 1;
}
}
}
}
#endif
// Temporal scaling parameters:
// NOTE: The 3 prediction frames cannot be used interchangeably due to
// differences in the way they are handled throughout the code. The
@ -495,6 +552,7 @@ int main(int argc, char **argv) {
vpx_codec_err_t res;
unsigned int width;
unsigned int height;
uint32_t error_resilient = 0;
int speed;
int frame_avail;
int got_data;
@ -505,16 +563,15 @@ int main(int argc, char **argv) {
int layering_mode = 0;
int layer_flags[VPX_TS_MAX_PERIODICITY] = { 0 };
int flag_periodicity = 1;
#if VPX_ENCODER_ABI_VERSION > (4 + VPX_CODEC_ABI_VERSION)
vpx_svc_layer_id_t layer_id = { 0, 0 };
#else
vpx_svc_layer_id_t layer_id = { 0 };
#if VP8_ROI_MAP
vpx_roi_map_t roi;
#endif
vpx_svc_layer_id_t layer_id = { 0, 0 };
const VpxInterface *encoder = NULL;
FILE *infile = NULL;
struct RateControlMetrics rc;
int64_t cx_time = 0;
const int min_args_base = 11;
const int min_args_base = 13;
#if CONFIG_VP9_HIGHBITDEPTH
vpx_bit_depth_t bit_depth = VPX_BITS_8;
int input_bit_depth = 8;
@ -531,12 +588,14 @@ int main(int argc, char **argv) {
if (argc < min_args) {
#if CONFIG_VP9_HIGHBITDEPTH
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> "
"<error_resilient> <threads> <mode> "
"<Rate_0> ... <Rate_nlayers-1> <bit-depth> \n",
argv[0]);
#else
die("Usage: %s <infile> <outfile> <codec_type(vp8/vp9)> <width> <height> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> <mode> "
"<rate_num> <rate_den> <speed> <frame_drop_threshold> "
"<error_resilient> <threads> <mode> "
"<Rate_0> ... <Rate_nlayers-1> \n",
argv[0]);
#endif // CONFIG_VP9_HIGHBITDEPTH
@ -553,9 +612,9 @@ int main(int argc, char **argv) {
die("Invalid resolution: %d x %d", width, height);
}
layering_mode = (int)strtol(argv[10], NULL, 0);
layering_mode = (int)strtol(argv[12], NULL, 0);
if (layering_mode < 0 || layering_mode > 13) {
die("Invalid layering mode (0..12) %s", argv[10]);
die("Invalid layering mode (0..12) %s", argv[12]);
}
if (argc != min_args + mode_to_num_layers[layering_mode]) {
@ -619,11 +678,11 @@ int main(int argc, char **argv) {
for (i = min_args_base;
(int)i < min_args_base + mode_to_num_layers[layering_mode]; ++i) {
rc.layer_target_bitrate[i - 11] = (int)strtol(argv[i], NULL, 0);
rc.layer_target_bitrate[i - 13] = (int)strtol(argv[i], NULL, 0);
if (strncmp(encoder->name, "vp8", 3) == 0)
cfg.ts_target_bitrate[i - 11] = rc.layer_target_bitrate[i - 11];
cfg.ts_target_bitrate[i - 13] = rc.layer_target_bitrate[i - 13];
else if (strncmp(encoder->name, "vp9", 3) == 0)
cfg.layer_target_bitrate[i - 11] = rc.layer_target_bitrate[i - 11];
cfg.layer_target_bitrate[i - 13] = rc.layer_target_bitrate[i - 13];
}
// Real time parameters.
@ -634,7 +693,7 @@ int main(int argc, char **argv) {
if (strncmp(encoder->name, "vp9", 3) == 0) cfg.rc_max_quantizer = 52;
cfg.rc_undershoot_pct = 50;
cfg.rc_overshoot_pct = 50;
cfg.rc_buf_initial_sz = 500;
cfg.rc_buf_initial_sz = 600;
cfg.rc_buf_optimal_sz = 600;
cfg.rc_buf_sz = 1000;
@ -642,10 +701,14 @@ int main(int argc, char **argv) {
cfg.rc_resize_allowed = 0;
// Use 1 thread as default.
cfg.g_threads = 1;
cfg.g_threads = (unsigned int)strtoul(argv[11], NULL, 0);
error_resilient = (uint32_t)strtoul(argv[10], NULL, 0);
if (error_resilient != 0 && error_resilient != 1) {
die("Invalid value for error resilient (0, 1): %d.", error_resilient);
}
// Enable error resilient mode.
cfg.g_error_resilient = 1;
cfg.g_error_resilient = error_resilient;
cfg.g_lag_in_frames = 0;
cfg.kf_mode = VPX_KF_AUTO;
@ -700,18 +763,33 @@ int main(int argc, char **argv) {
if (strncmp(encoder->name, "vp8", 3) == 0) {
vpx_codec_control(&codec, VP8E_SET_CPUUSED, -speed);
vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kDenoiserOff);
vpx_codec_control(&codec, VP8E_SET_NOISE_SENSITIVITY, kVp8DenoiserOff);
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
vpx_codec_control(&codec, VP8E_SET_GF_CBR_BOOST_PCT, 0);
#if VP8_ROI_MAP
vp8_set_roi_map(&cfg, &roi);
if (vpx_codec_control(&codec, VP8E_SET_ROI_MAP, &roi))
die_codec(&codec, "Failed to set ROI map");
#endif
} else if (strncmp(encoder->name, "vp9", 3) == 0) {
vpx_svc_extra_cfg_t svc_params;
memset(&svc_params, 0, sizeof(svc_params));
vpx_codec_control(&codec, VP8E_SET_CPUUSED, speed);
vpx_codec_control(&codec, VP9E_SET_AQ_MODE, 3);
vpx_codec_control(&codec, VP9E_SET_GF_CBR_BOOST_PCT, 0);
vpx_codec_control(&codec, VP9E_SET_FRAME_PARALLEL_DECODING, 0);
vpx_codec_control(&codec, VP9E_SET_FRAME_PERIODIC_BOOST, 0);
vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, kDenoiserOff);
vpx_codec_control(&codec, VP9E_SET_NOISE_SENSITIVITY, kVp9DenoiserOff);
vpx_codec_control(&codec, VP8E_SET_STATIC_THRESHOLD, 1);
vpx_codec_control(&codec, VP9E_SET_TUNE_CONTENT, 0);
vpx_codec_control(&codec, VP9E_SET_TILE_COLUMNS, (cfg.g_threads >> 1));
// TODO(marpan/jianj): There is an issue with row-mt for low resolutons at
// high speed settings, disable its use for those cases for now.
if (cfg.g_threads > 1 && ((cfg.g_w > 320 && cfg.g_h > 240) || speed < 7))
vpx_codec_control(&codec, VP9E_SET_ROW_MT, 1);
else
vpx_codec_control(&codec, VP9E_SET_ROW_MT, 0);
if (vpx_codec_control(&codec, VP9E_SET_SVC, layering_mode > 0 ? 1 : 0))
die_codec(&codec, "Failed to set SVC");
for (i = 0; i < cfg.ts_number_layers; ++i) {
@ -730,7 +808,7 @@ int main(int argc, char **argv) {
// For generating smaller key frames, use a smaller max_intra_size_pct
// value, like 100 or 200.
{
const int max_intra_size_pct = 900;
const int max_intra_size_pct = 1000;
vpx_codec_control(&codec, VP8E_SET_MAX_INTRA_BITRATE_PCT,
max_intra_size_pct);
}
@ -740,10 +818,8 @@ int main(int argc, char **argv) {
struct vpx_usec_timer timer;
vpx_codec_iter_t iter = NULL;
const vpx_codec_cx_pkt_t *pkt;
#if VPX_ENCODER_ABI_VERSION > (4 + VPX_CODEC_ABI_VERSION)
// Update the temporal layer_id. No spatial layers in this test.
layer_id.spatial_layer_id = 0;
#endif
layer_id.temporal_layer_id =
cfg.ts_layer_id[frame_cnt % cfg.ts_periodicity];
if (strncmp(encoder->name, "vp9", 3) == 0) {

View File

@ -12,7 +12,7 @@
# ARM assembly files are written in RVCT-style. We use some make magic to
# filter those files to allow GCC compilation
ifeq ($(ARCH_ARM),yes)
ASM:=$(if $(filter yes,$(CONFIG_GCC)$(CONFIG_MSVS)),.asm.s,.asm)
ASM:=$(if $(filter yes,$(CONFIG_GCC)$(CONFIG_MSVS)),.asm.S,.asm)
else
ASM:=.asm
endif
@ -106,9 +106,6 @@ ifeq ($(CONFIG_VP9_DECODER),yes)
CODEC_DOC_SECTIONS += vp9 vp9_decoder
endif
VP9_PREFIX=vp9/
$(BUILD_PFX)$(VP9_PREFIX)%.c.o:
ifeq ($(CONFIG_ENCODERS),yes)
CODEC_DOC_SECTIONS += encoder
endif
@ -116,6 +113,12 @@ ifeq ($(CONFIG_DECODERS),yes)
CODEC_DOC_SECTIONS += decoder
endif
# Suppress -Wextra warnings in third party code.
$(BUILD_PFX)third_party/googletest/%.cc.o: CXXFLAGS += -Wno-missing-field-initializers
# Suppress -Wextra warnings in first party code pending investigation.
# https://bugs.chromium.org/p/webm/issues/detail?id=1069
$(BUILD_PFX)vp8/encoder/onyx_if.c.o: CFLAGS += -Wno-unknown-warning-option -Wno-clobbered
$(BUILD_PFX)vp8/decoder/onyxd_if.c.o: CFLAGS += -Wno-unknown-warning-option -Wno-clobbered
ifeq ($(CONFIG_MSVS),yes)
CODEC_LIB=$(if $(CONFIG_STATIC_MSVCRT),vpxmt,vpxmd)
@ -146,6 +149,7 @@ CODEC_SRCS-yes += $(BUILD_PFX)vpx_config.c
INSTALL-SRCS-no += $(BUILD_PFX)vpx_config.c
ifeq ($(ARCH_X86)$(ARCH_X86_64),yes)
INSTALL-SRCS-$(CONFIG_CODEC_SRCS) += third_party/x86inc/x86inc.asm
INSTALL-SRCS-$(CONFIG_CODEC_SRCS) += vpx_dsp/x86/bitdepth_conversion_sse2.asm
endif
CODEC_EXPORTS-yes += vpx/exports_com
CODEC_EXPORTS-$(CONFIG_ENCODERS) += vpx/exports_enc
@ -184,6 +188,13 @@ libvpx_srcs.txt:
@echo $(CODEC_SRCS) | xargs -n1 echo | LC_ALL=C sort -u > $@
CLEAN-OBJS += libvpx_srcs.txt
# Assembly files that are included, but don't define symbols themselves.
# Filtered out to avoid Windows build warnings.
ASM_INCLUDES := \
third_party/x86inc/x86inc.asm \
vpx_config.asm \
vpx_ports/x86_abi_support.asm \
vpx_dsp/x86/bitdepth_conversion_sse2.asm \
ifeq ($(CONFIG_EXTERNAL_BUILD),yes)
ifeq ($(CONFIG_MSVS),yes)
@ -195,13 +206,6 @@ vpx.def: $(call enabled,CODEC_EXPORTS)
--out=$@ $^
CLEAN-OBJS += vpx.def
# Assembly files that are included, but don't define symbols themselves.
# Filtered out to avoid Visual Studio build warnings.
ASM_INCLUDES := \
third_party/x86inc/x86inc.asm \
vpx_config.asm \
vpx_ports/x86_abi_support.asm \
vpx.$(VCPROJ_SFX): $(CODEC_SRCS) vpx.def
@echo " [CREATE] $@"
$(qexec)$(GEN_VCPROJ) \
@ -224,12 +228,12 @@ vpx.$(VCPROJ_SFX): $(RTCD)
endif
else
LIBVPX_OBJS=$(call objs,$(CODEC_SRCS))
LIBVPX_OBJS=$(call objs, $(filter-out $(ASM_INCLUDES), $(CODEC_SRCS)))
OBJS-yes += $(LIBVPX_OBJS)
LIBS-$(if yes,$(CONFIG_STATIC)) += $(BUILD_PFX)libvpx.a $(BUILD_PFX)libvpx_g.a
$(BUILD_PFX)libvpx_g.a: $(LIBVPX_OBJS)
SO_VERSION_MAJOR := 4
SO_VERSION_MAJOR := 5
SO_VERSION_MINOR := 0
SO_VERSION_PATCH := 0
ifeq ($(filter darwin%,$(TGT_OS)),$(TGT_OS))
@ -363,7 +367,7 @@ endif
#
# Add assembler dependencies for configuration.
#
$(filter %.s.o,$(OBJS-yes)): $(BUILD_PFX)vpx_config.asm
$(filter %.S.o,$(OBJS-yes)): $(BUILD_PFX)vpx_config.asm
$(filter %$(ASM).o,$(OBJS-yes)): $(BUILD_PFX)vpx_config.asm
@ -388,7 +392,7 @@ LIBVPX_TEST_SRCS=$(addprefix test/,$(call enabled,LIBVPX_TEST_SRCS))
LIBVPX_TEST_BIN=./test_libvpx$(EXE_SFX)
LIBVPX_TEST_DATA=$(addprefix $(LIBVPX_TEST_DATA_PATH)/,\
$(call enabled,LIBVPX_TEST_DATA))
libvpx_test_data_url=http://downloads.webmproject.org/test_data/libvpx/$(1)
libvpx_test_data_url=https://storage.googleapis.com/downloads.webmproject.org/test_data/libvpx/$(1)
TEST_INTRA_PRED_SPEED_BIN=./test_intra_pred_speed$(EXE_SFX)
TEST_INTRA_PRED_SPEED_SRCS=$(addprefix test/,$(call enabled,TEST_INTRA_PRED_SPEED_SRCS))
@ -401,8 +405,16 @@ CLEAN-OBJS += libvpx_test_srcs.txt
$(LIBVPX_TEST_DATA): $(SRC_PATH_BARE)/test/test-data.sha1
@echo " [DOWNLOAD] $@"
$(qexec)trap 'rm -f $@' INT TERM &&\
curl -L -o $@ $(call libvpx_test_data_url,$(@F))
# Attempt to download the file using curl, retrying once if it fails for a
# partial file (18).
$(qexec)( \
trap 'rm -f $@' INT TERM; \
curl="curl --retry 1 -L -o $@ $(call libvpx_test_data_url,$(@F))"; \
$$curl; \
case "$$?" in \
18) $$curl -C -;; \
esac \
)
testdata:: $(LIBVPX_TEST_DATA)
$(qexec)[ -x "$$(which sha1sum)" ] && sha1sum=sha1sum;\

View File

@ -37,7 +37,13 @@ struct rate_hist {
struct rate_hist *init_rate_histogram(const vpx_codec_enc_cfg_t *cfg,
const vpx_rational_t *fps) {
int i;
struct rate_hist *hist = malloc(sizeof(*hist));
struct rate_hist *hist = calloc(1, sizeof(*hist));
if (hist == NULL || cfg == NULL || fps == NULL || fps->num == 0 ||
fps->den == 0) {
destroy_rate_histogram(hist);
return NULL;
}
// Determine the number of samples in the buffer. Use the file's framerate
// to determine the number of frames in rc_buf_sz milliseconds, with an
@ -80,7 +86,11 @@ void update_rate_histogram(struct rate_hist *hist,
(uint64_t)cfg->g_timebase.num /
(uint64_t)cfg->g_timebase.den;
int idx = hist->frames++ % hist->samples;
int idx;
if (hist == NULL || cfg == NULL || pkt == NULL) return;
idx = hist->frames++ % hist->samples;
hist->pts[idx] = now;
hist->sz[idx] = (int)pkt->data.frame.sz;
@ -116,9 +126,14 @@ void update_rate_histogram(struct rate_hist *hist,
static int merge_hist_buckets(struct hist_bucket *bucket, int max_buckets,
int *num_buckets) {
int small_bucket = 0, merge_bucket = INT_MAX, big_bucket = 0;
int buckets = *num_buckets;
int buckets;
int i;
assert(bucket != NULL);
assert(num_buckets != NULL);
buckets = *num_buckets;
/* Find the extrema for this list of buckets */
big_bucket = small_bucket = 0;
for (i = 0; i < buckets; i++) {
@ -181,6 +196,8 @@ static void show_histogram(const struct hist_bucket *bucket, int buckets,
const char *pat1, *pat2;
int i;
assert(bucket != NULL);
switch ((int)(log(bucket[buckets - 1].high) / log(10)) + 1) {
case 1:
case 2:
@ -259,6 +276,8 @@ void show_rate_histogram(struct rate_hist *hist, const vpx_codec_enc_cfg_t *cfg,
int i, scale;
int buckets = 0;
if (hist == NULL || cfg == NULL) return;
for (i = 0; i < RATE_BINS; i++) {
if (hist->bucket[i].low == INT_MAX) continue;
hist->bucket[buckets++] = hist->bucket[i];

View File

@ -11,6 +11,10 @@
#ifndef TEST_ACM_RANDOM_H_
#define TEST_ACM_RANDOM_H_
#include <assert.h>
#include <limits>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "vpx/vpx_integer.h"
@ -50,6 +54,13 @@ class ACMRandom {
return r < 128 ? r << 4 : r >> 4;
}
uint32_t RandRange(const uint32_t range) {
// testing::internal::Random::Generate provides values in the range
// testing::internal::Random::kMaxRange.
assert(range <= testing::internal::Random::kMaxRange);
return random_.Generate(range);
}
int PseudoUniform(int range) { return random_.Generate(range); }
int operator()(int n) { return PseudoUniform(n); }

View File

@ -32,6 +32,7 @@ LOCAL_CPP_EXTENSION := .cc
LOCAL_MODULE := gtest
LOCAL_C_INCLUDES := $(LOCAL_PATH)/third_party/googletest/src/
LOCAL_C_INCLUDES += $(LOCAL_PATH)/third_party/googletest/src/include/
LOCAL_EXPORT_C_INCLUDES := $(LOCAL_PATH)/third_party/googletest/src/include/
LOCAL_SRC_FILES := ./third_party/googletest/src/src/gtest-all.cc
include $(BUILD_STATIC_LIBRARY)

View File

@ -14,6 +14,7 @@
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vp9_rtcd.h"
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"
@ -22,6 +23,7 @@
#include "test/register_state_check.h"
#include "test/util.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/vpx_timer.h"
using libvpx_test::ACMRandom;
@ -53,7 +55,7 @@ class AverageTestBase : public ::testing::Test {
}
// Sum Pixels
unsigned int ReferenceAverage8x8(const uint8_t *source, int pitch) {
static unsigned int ReferenceAverage8x8(const uint8_t *source, int pitch) {
unsigned int average = 0;
for (int h = 0; h < 8; ++h) {
for (int w = 0; w < 8; ++w) average += source[h * pitch + w];
@ -61,7 +63,7 @@ class AverageTestBase : public ::testing::Test {
return ((average + 32) >> 6);
}
unsigned int ReferenceAverage4x4(const uint8_t *source, int pitch) {
static unsigned int ReferenceAverage4x4(const uint8_t *source, int pitch) {
unsigned int average = 0;
for (int h = 0; h < 4; ++h) {
for (int w = 0; w < 4; ++w) average += source[h * pitch + w];
@ -98,11 +100,12 @@ class AverageTest : public AverageTestBase,
protected:
void CheckAverages() {
const int block_size = GET_PARAM(3);
unsigned int expected = 0;
if (GET_PARAM(3) == 8) {
if (block_size == 8) {
expected =
ReferenceAverage8x8(source_data_ + GET_PARAM(2), source_stride_);
} else if (GET_PARAM(3) == 4) {
} else if (block_size == 4) {
expected =
ReferenceAverage4x4(source_data_ + GET_PARAM(2), source_stride_);
}
@ -185,7 +188,7 @@ class IntProColTest : public AverageTestBase,
int16_t sum_c_;
};
typedef int (*SatdFunc)(const int16_t *coeffs, int length);
typedef int (*SatdFunc)(const tran_low_t *coeffs, int length);
typedef std::tr1::tuple<int, SatdFunc> SatdTestParam;
class SatdTest : public ::testing::Test,
@ -195,7 +198,7 @@ class SatdTest : public ::testing::Test,
satd_size_ = GET_PARAM(0);
satd_func_ = GET_PARAM(1);
rnd_.Reset(ACMRandom::DeterministicSeed());
src_ = reinterpret_cast<int16_t *>(
src_ = reinterpret_cast<tran_low_t *>(
vpx_memalign(16, sizeof(*src_) * satd_size_));
ASSERT_TRUE(src_ != NULL);
}
@ -205,12 +208,15 @@ class SatdTest : public ::testing::Test,
vpx_free(src_);
}
void FillConstant(const int16_t val) {
void FillConstant(const tran_low_t val) {
for (int i = 0; i < satd_size_; ++i) src_[i] = val;
}
void FillRandom() {
for (int i = 0; i < satd_size_; ++i) src_[i] = rnd_.Rand16();
for (int i = 0; i < satd_size_; ++i) {
const int16_t tmp = rnd_.Rand16();
src_[i] = (tran_low_t)tmp;
}
}
void Check(const int expected) {
@ -222,11 +228,66 @@ class SatdTest : public ::testing::Test,
int satd_size_;
private:
int16_t *src_;
tran_low_t *src_;
SatdFunc satd_func_;
ACMRandom rnd_;
};
typedef int64_t (*BlockErrorFunc)(const tran_low_t *coeff,
const tran_low_t *dqcoeff, int block_size);
typedef std::tr1::tuple<int, BlockErrorFunc> BlockErrorTestFPParam;
class BlockErrorTestFP
: public ::testing::Test,
public ::testing::WithParamInterface<BlockErrorTestFPParam> {
protected:
virtual void SetUp() {
txfm_size_ = GET_PARAM(0);
block_error_func_ = GET_PARAM(1);
rnd_.Reset(ACMRandom::DeterministicSeed());
coeff_ = reinterpret_cast<tran_low_t *>(
vpx_memalign(16, sizeof(*coeff_) * txfm_size_));
dqcoeff_ = reinterpret_cast<tran_low_t *>(
vpx_memalign(16, sizeof(*dqcoeff_) * txfm_size_));
ASSERT_TRUE(coeff_ != NULL);
ASSERT_TRUE(dqcoeff_ != NULL);
}
virtual void TearDown() {
libvpx_test::ClearSystemState();
vpx_free(coeff_);
vpx_free(dqcoeff_);
}
void FillConstant(const tran_low_t coeff_val, const tran_low_t dqcoeff_val) {
for (int i = 0; i < txfm_size_; ++i) coeff_[i] = coeff_val;
for (int i = 0; i < txfm_size_; ++i) dqcoeff_[i] = dqcoeff_val;
}
void FillRandom() {
// Just two fixed seeds
rnd_.Reset(0xb0b9);
for (int i = 0; i < txfm_size_; ++i) coeff_[i] = rnd_.Rand16() >> 1;
rnd_.Reset(0xb0c8);
for (int i = 0; i < txfm_size_; ++i) dqcoeff_[i] = rnd_.Rand16() >> 1;
}
void Check(const int64_t expected) {
int64_t total;
ASM_REGISTER_STATE_CHECK(
total = block_error_func_(coeff_, dqcoeff_, txfm_size_));
EXPECT_EQ(expected, total);
}
int txfm_size_;
private:
tran_low_t *coeff_;
tran_low_t *dqcoeff_;
BlockErrorFunc block_error_func_;
ACMRandom rnd_;
};
uint8_t *AverageTestBase::source_data_ = NULL;
TEST_P(AverageTest, MinValue) {
@ -307,6 +368,66 @@ TEST_P(SatdTest, Random) {
Check(expected);
}
TEST_P(SatdTest, DISABLED_Speed) {
const int kCountSpeedTestBlock = 20000;
vpx_usec_timer timer;
DECLARE_ALIGNED(16, tran_low_t, coeff[1024]);
const int blocksize = GET_PARAM(0);
vpx_usec_timer_start(&timer);
for (int i = 0; i < kCountSpeedTestBlock; ++i) {
GET_PARAM(1)(coeff, blocksize);
}
vpx_usec_timer_mark(&timer);
const int elapsed_time = static_cast<int>(vpx_usec_timer_elapsed(&timer));
printf("blocksize: %4d time: %4d us\n", blocksize, elapsed_time);
}
TEST_P(BlockErrorTestFP, MinValue) {
const int64_t kMin = -32640;
const int64_t expected = kMin * kMin * txfm_size_;
FillConstant(kMin, 0);
Check(expected);
}
TEST_P(BlockErrorTestFP, MaxValue) {
const int64_t kMax = 32640;
const int64_t expected = kMax * kMax * txfm_size_;
FillConstant(kMax, 0);
Check(expected);
}
TEST_P(BlockErrorTestFP, Random) {
int64_t expected;
switch (txfm_size_) {
case 16: expected = 2051681432; break;
case 64: expected = 11075114379; break;
case 256: expected = 44386271116; break;
case 1024: expected = 184774996089; break;
default:
FAIL() << "Invalid satd size (" << txfm_size_
<< ") valid: 16/64/256/1024";
}
FillRandom();
Check(expected);
}
TEST_P(BlockErrorTestFP, DISABLED_Speed) {
const int kCountSpeedTestBlock = 20000;
vpx_usec_timer timer;
DECLARE_ALIGNED(16, tran_low_t, coeff[1024]);
DECLARE_ALIGNED(16, tran_low_t, dqcoeff[1024]);
const int blocksize = GET_PARAM(0);
vpx_usec_timer_start(&timer);
for (int i = 0; i < kCountSpeedTestBlock; ++i) {
GET_PARAM(1)(coeff, dqcoeff, blocksize);
}
vpx_usec_timer_mark(&timer);
const int elapsed_time = static_cast<int>(vpx_usec_timer_elapsed(&timer));
printf("blocksize: %4d time: %4d us\n", blocksize, elapsed_time);
}
using std::tr1::make_tuple;
INSTANTIATE_TEST_CASE_P(
@ -320,6 +441,13 @@ INSTANTIATE_TEST_CASE_P(C, SatdTest,
make_tuple(256, &vpx_satd_c),
make_tuple(1024, &vpx_satd_c)));
INSTANTIATE_TEST_CASE_P(
C, BlockErrorTestFP,
::testing::Values(make_tuple(16, &vp9_block_error_fp_c),
make_tuple(64, &vp9_block_error_fp_c),
make_tuple(256, &vp9_block_error_fp_c),
make_tuple(1024, &vp9_block_error_fp_c)));
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(
SSE2, AverageTest,
@ -349,6 +477,28 @@ INSTANTIATE_TEST_CASE_P(SSE2, SatdTest,
make_tuple(64, &vpx_satd_sse2),
make_tuple(256, &vpx_satd_sse2),
make_tuple(1024, &vpx_satd_sse2)));
INSTANTIATE_TEST_CASE_P(
SSE2, BlockErrorTestFP,
::testing::Values(make_tuple(16, &vp9_block_error_fp_sse2),
make_tuple(64, &vp9_block_error_fp_sse2),
make_tuple(256, &vp9_block_error_fp_sse2),
make_tuple(1024, &vp9_block_error_fp_sse2)));
#endif // HAVE_SSE2
#if HAVE_AVX2
INSTANTIATE_TEST_CASE_P(AVX2, SatdTest,
::testing::Values(make_tuple(16, &vpx_satd_avx2),
make_tuple(64, &vpx_satd_avx2),
make_tuple(256, &vpx_satd_avx2),
make_tuple(1024, &vpx_satd_avx2)));
INSTANTIATE_TEST_CASE_P(
AVX2, BlockErrorTestFP,
::testing::Values(make_tuple(16, &vp9_block_error_fp_avx2),
make_tuple(64, &vp9_block_error_fp_avx2),
make_tuple(256, &vp9_block_error_fp_avx2),
make_tuple(1024, &vp9_block_error_fp_avx2)));
#endif
#if HAVE_NEON
@ -380,7 +530,18 @@ INSTANTIATE_TEST_CASE_P(NEON, SatdTest,
make_tuple(64, &vpx_satd_neon),
make_tuple(256, &vpx_satd_neon),
make_tuple(1024, &vpx_satd_neon)));
#endif
// TODO(jianj): Remove the highbitdepth flag once the SIMD functions are
// in place.
#if !CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
NEON, BlockErrorTestFP,
::testing::Values(make_tuple(16, &vp9_block_error_fp_neon),
make_tuple(64, &vp9_block_error_fp_neon),
make_tuple(256, &vp9_block_error_fp_neon),
make_tuple(1024, &vp9_block_error_fp_neon)));
#endif // !CONFIG_VP9_HIGHBITDEPTH
#endif // HAVE_NEON
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(
@ -391,6 +552,30 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(16, 16, 0, 4, &vpx_avg_4x4_msa),
make_tuple(16, 16, 5, 4, &vpx_avg_4x4_msa),
make_tuple(32, 32, 15, 4, &vpx_avg_4x4_msa)));
#endif
INSTANTIATE_TEST_CASE_P(
MSA, IntProRowTest,
::testing::Values(make_tuple(16, &vpx_int_pro_row_msa, &vpx_int_pro_row_c),
make_tuple(32, &vpx_int_pro_row_msa, &vpx_int_pro_row_c),
make_tuple(64, &vpx_int_pro_row_msa,
&vpx_int_pro_row_c)));
INSTANTIATE_TEST_CASE_P(
MSA, IntProColTest,
::testing::Values(make_tuple(16, &vpx_int_pro_col_msa, &vpx_int_pro_col_c),
make_tuple(32, &vpx_int_pro_col_msa, &vpx_int_pro_col_c),
make_tuple(64, &vpx_int_pro_col_msa,
&vpx_int_pro_col_c)));
// TODO(jingning): Remove the highbitdepth flag once the SIMD functions are
// in place.
#if !CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(MSA, SatdTest,
::testing::Values(make_tuple(16, &vpx_satd_msa),
make_tuple(64, &vpx_satd_msa),
make_tuple(256, &vpx_satd_msa),
make_tuple(1024, &vpx_satd_msa)));
#endif // !CONFIG_VP9_HIGHBITDEPTH
#endif // HAVE_MSA
} // namespace

382
libs/libvpx/test/buffer.h Normal file
View File

@ -0,0 +1,382 @@
/*
* Copyright (c) 2016 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef TEST_BUFFER_H_
#define TEST_BUFFER_H_
#include <stdio.h>
#include <limits>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/acm_random.h"
#include "vpx/vpx_integer.h"
#include "vpx_mem/vpx_mem.h"
namespace libvpx_test {
template <typename T>
class Buffer {
public:
Buffer(int width, int height, int top_padding, int left_padding,
int right_padding, int bottom_padding)
: width_(width), height_(height), top_padding_(top_padding),
left_padding_(left_padding), right_padding_(right_padding),
bottom_padding_(bottom_padding), alignment_(0), padding_value_(0),
stride_(0), raw_size_(0), num_elements_(0), raw_buffer_(NULL) {}
Buffer(int width, int height, int top_padding, int left_padding,
int right_padding, int bottom_padding, unsigned int alignment)
: width_(width), height_(height), top_padding_(top_padding),
left_padding_(left_padding), right_padding_(right_padding),
bottom_padding_(bottom_padding), alignment_(alignment),
padding_value_(0), stride_(0), raw_size_(0), num_elements_(0),
raw_buffer_(NULL) {}
Buffer(int width, int height, int padding)
: width_(width), height_(height), top_padding_(padding),
left_padding_(padding), right_padding_(padding),
bottom_padding_(padding), alignment_(0), padding_value_(0), stride_(0),
raw_size_(0), num_elements_(0), raw_buffer_(NULL) {}
Buffer(int width, int height, int padding, unsigned int alignment)
: width_(width), height_(height), top_padding_(padding),
left_padding_(padding), right_padding_(padding),
bottom_padding_(padding), alignment_(alignment), padding_value_(0),
stride_(0), raw_size_(0), num_elements_(0), raw_buffer_(NULL) {}
~Buffer() {
if (alignment_) {
vpx_free(raw_buffer_);
} else {
delete[] raw_buffer_;
}
}
T *TopLeftPixel() const;
int stride() const { return stride_; }
// Set the buffer (excluding padding) to 'value'.
void Set(const T value);
// Set the buffer (excluding padding) to the output of ACMRandom function
// 'rand_func'.
void Set(ACMRandom *rand_class, T (ACMRandom::*rand_func)());
// Set the buffer (excluding padding) to the output of ACMRandom function
// 'RandRange' with range 'low' to 'high' which typically must be within
// testing::internal::Random::kMaxRange (1u << 31). However, because we want
// to allow negative low (and high) values, it is restricted to INT32_MAX
// here.
void Set(ACMRandom *rand_class, const T low, const T high);
// Copy the contents of Buffer 'a' (excluding padding).
void CopyFrom(const Buffer<T> &a);
void DumpBuffer() const;
// Highlight the differences between two buffers if they are the same size.
void PrintDifference(const Buffer<T> &a) const;
bool HasPadding() const;
// Sets all the values in the buffer to 'padding_value'.
void SetPadding(const T padding_value);
// Checks if all the values (excluding padding) are equal to 'value' if the
// Buffers are the same size.
bool CheckValues(const T value) const;
// Check that padding matches the expected value or there is no padding.
bool CheckPadding() const;
// Compare the non-padding portion of two buffers if they are the same size.
bool CheckValues(const Buffer<T> &a) const;
bool Init() {
if (raw_buffer_ != NULL) return false;
EXPECT_GT(width_, 0);
EXPECT_GT(height_, 0);
EXPECT_GE(top_padding_, 0);
EXPECT_GE(left_padding_, 0);
EXPECT_GE(right_padding_, 0);
EXPECT_GE(bottom_padding_, 0);
stride_ = left_padding_ + width_ + right_padding_;
num_elements_ = stride_ * (top_padding_ + height_ + bottom_padding_);
raw_size_ = num_elements_ * sizeof(T);
if (alignment_) {
EXPECT_GE(alignment_, sizeof(T));
// Ensure alignment of the first value will be preserved.
EXPECT_EQ((left_padding_ * sizeof(T)) % alignment_, 0u);
// Ensure alignment of the subsequent rows will be preserved when there is
// a stride.
if (stride_ != width_) {
EXPECT_EQ((stride_ * sizeof(T)) % alignment_, 0u);
}
raw_buffer_ = reinterpret_cast<T *>(vpx_memalign(alignment_, raw_size_));
} else {
raw_buffer_ = new (std::nothrow) T[num_elements_];
}
EXPECT_TRUE(raw_buffer_ != NULL);
SetPadding(std::numeric_limits<T>::max());
return !::testing::Test::HasFailure();
}
private:
bool BufferSizesMatch(const Buffer<T> &a) const;
const int width_;
const int height_;
const int top_padding_;
const int left_padding_;
const int right_padding_;
const int bottom_padding_;
const unsigned int alignment_;
T padding_value_;
int stride_;
int raw_size_;
int num_elements_;
T *raw_buffer_;
};
template <typename T>
T *Buffer<T>::TopLeftPixel() const {
if (!raw_buffer_) return NULL;
return raw_buffer_ + (top_padding_ * stride_) + left_padding_;
}
template <typename T>
void Buffer<T>::Set(const T value) {
if (!raw_buffer_) return;
T *src = TopLeftPixel();
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < width_; ++width) {
src[width] = value;
}
src += stride_;
}
}
template <typename T>
void Buffer<T>::Set(ACMRandom *rand_class, T (ACMRandom::*rand_func)()) {
if (!raw_buffer_) return;
T *src = TopLeftPixel();
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < width_; ++width) {
src[width] = (*rand_class.*rand_func)();
}
src += stride_;
}
}
template <typename T>
void Buffer<T>::Set(ACMRandom *rand_class, const T low, const T high) {
if (!raw_buffer_) return;
EXPECT_LE(low, high);
EXPECT_LE(static_cast<int64_t>(high) - low,
std::numeric_limits<int32_t>::max());
T *src = TopLeftPixel();
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < width_; ++width) {
// 'low' will be promoted to unsigned given the return type of RandRange.
// Store the value as an int to avoid unsigned overflow warnings when
// 'low' is negative.
const int32_t value =
static_cast<int32_t>((*rand_class).RandRange(high - low));
src[width] = static_cast<T>(value + low);
}
src += stride_;
}
}
template <typename T>
void Buffer<T>::CopyFrom(const Buffer<T> &a) {
if (!raw_buffer_) return;
if (!BufferSizesMatch(a)) return;
T *a_src = a.TopLeftPixel();
T *b_src = this->TopLeftPixel();
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < width_; ++width) {
b_src[width] = a_src[width];
}
a_src += a.stride();
b_src += this->stride();
}
}
template <typename T>
void Buffer<T>::DumpBuffer() const {
if (!raw_buffer_) return;
for (int height = 0; height < height_ + top_padding_ + bottom_padding_;
++height) {
for (int width = 0; width < stride_; ++width) {
printf("%4d", raw_buffer_[height + width * stride_]);
}
printf("\n");
}
}
template <typename T>
bool Buffer<T>::HasPadding() const {
if (!raw_buffer_) return false;
return top_padding_ || left_padding_ || right_padding_ || bottom_padding_;
}
template <typename T>
void Buffer<T>::PrintDifference(const Buffer<T> &a) const {
if (!raw_buffer_) return;
if (!BufferSizesMatch(a)) return;
T *a_src = a.TopLeftPixel();
T *b_src = TopLeftPixel();
printf("This buffer:\n");
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < width_; ++width) {
if (a_src[width] != b_src[width]) {
printf("*%3d", b_src[width]);
} else {
printf("%4d", b_src[width]);
}
}
printf("\n");
a_src += a.stride();
b_src += this->stride();
}
a_src = a.TopLeftPixel();
b_src = TopLeftPixel();
printf("Reference buffer:\n");
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < width_; ++width) {
if (a_src[width] != b_src[width]) {
printf("*%3d", a_src[width]);
} else {
printf("%4d", a_src[width]);
}
}
printf("\n");
a_src += a.stride();
b_src += this->stride();
}
}
template <typename T>
void Buffer<T>::SetPadding(const T padding_value) {
if (!raw_buffer_) return;
padding_value_ = padding_value;
T *src = raw_buffer_;
for (int i = 0; i < num_elements_; ++i) {
src[i] = padding_value;
}
}
template <typename T>
bool Buffer<T>::CheckValues(const T value) const {
if (!raw_buffer_) return false;
T *src = TopLeftPixel();
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < width_; ++width) {
if (value != src[width]) {
return false;
}
}
src += stride_;
}
return true;
}
template <typename T>
bool Buffer<T>::CheckPadding() const {
if (!raw_buffer_) return false;
if (!HasPadding()) return true;
// Top padding.
T const *top = raw_buffer_;
for (int i = 0; i < stride_ * top_padding_; ++i) {
if (padding_value_ != top[i]) {
return false;
}
}
// Left padding.
T const *left = TopLeftPixel() - left_padding_;
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < left_padding_; ++width) {
if (padding_value_ != left[width]) {
return false;
}
}
left += stride_;
}
// Right padding.
T const *right = TopLeftPixel() + width_;
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < right_padding_; ++width) {
if (padding_value_ != right[width]) {
return false;
}
}
right += stride_;
}
// Bottom padding
T const *bottom = raw_buffer_ + (top_padding_ + height_) * stride_;
for (int i = 0; i < stride_ * bottom_padding_; ++i) {
if (padding_value_ != bottom[i]) {
return false;
}
}
return true;
}
template <typename T>
bool Buffer<T>::CheckValues(const Buffer<T> &a) const {
if (!raw_buffer_) return false;
if (!BufferSizesMatch(a)) return false;
T *a_src = a.TopLeftPixel();
T *b_src = this->TopLeftPixel();
for (int height = 0; height < height_; ++height) {
for (int width = 0; width < width_; ++width) {
if (a_src[width] != b_src[width]) {
return false;
}
}
a_src += a.stride();
b_src += this->stride();
}
return true;
}
template <typename T>
bool Buffer<T>::BufferSizesMatch(const Buffer<T> &a) const {
if (!raw_buffer_) return false;
if (a.width_ != this->width_ || a.height_ != this->height_) {
printf(
"Reference buffer of size %dx%d does not match this buffer which is "
"size %dx%d\n",
a.width_, a.height_, this->width_, this->height_);
return false;
}
return true;
}
} // namespace libvpx_test
#endif // TEST_BUFFER_H_

View File

@ -128,8 +128,8 @@ class ByteAlignmentTest
// TODO(fgalligan): Move the MD5 testing code into another class.
void OpenMd5File(const std::string &md5_file_name_) {
md5_file_ = libvpx_test::OpenTestDataFile(md5_file_name_);
ASSERT_TRUE(md5_file_ != NULL) << "MD5 file open failed. Filename: "
<< md5_file_name_;
ASSERT_TRUE(md5_file_ != NULL)
<< "MD5 file open failed. Filename: " << md5_file_name_;
}
void CheckMd5(const vpx_image_t &img) {

View File

@ -65,6 +65,12 @@ class CodecTestWith3Params
: public ::testing::TestWithParam<
std::tr1::tuple<const libvpx_test::CodecFactory *, T1, T2, T3> > {};
template <class T1, class T2, class T3, class T4>
class CodecTestWith4Params
: public ::testing::TestWithParam<
std::tr1::tuple<const libvpx_test::CodecFactory *, T1, T2, T3, T4> > {
};
/*
* VP8 Codec Definitions
*/
@ -115,6 +121,8 @@ class VP8CodecFactory : public CodecFactory {
#if CONFIG_VP8_DECODER
return new VP8Decoder(cfg, flags);
#else
(void)cfg;
(void)flags;
return NULL;
#endif
}
@ -126,6 +134,10 @@ class VP8CodecFactory : public CodecFactory {
#if CONFIG_VP8_ENCODER
return new VP8Encoder(cfg, deadline, init_flags, stats);
#else
(void)cfg;
(void)deadline;
(void)init_flags;
(void)stats;
return NULL;
#endif
}
@ -135,6 +147,8 @@ class VP8CodecFactory : public CodecFactory {
#if CONFIG_VP8_ENCODER
return vpx_codec_enc_config_default(&vpx_codec_vp8_cx_algo, cfg, usage);
#else
(void)cfg;
(void)usage;
return VPX_CODEC_INCAPABLE;
#endif
}
@ -203,6 +217,8 @@ class VP9CodecFactory : public CodecFactory {
#if CONFIG_VP9_DECODER
return new VP9Decoder(cfg, flags);
#else
(void)cfg;
(void)flags;
return NULL;
#endif
}
@ -214,6 +230,10 @@ class VP9CodecFactory : public CodecFactory {
#if CONFIG_VP9_ENCODER
return new VP9Encoder(cfg, deadline, init_flags, stats);
#else
(void)cfg;
(void)deadline;
(void)init_flags;
(void)stats;
return NULL;
#endif
}
@ -223,6 +243,8 @@ class VP9CodecFactory : public CodecFactory {
#if CONFIG_VP9_ENCODER
return vpx_codec_enc_config_default(&vpx_codec_vp9_cx_algo, cfg, usage);
#else
(void)cfg;
(void)usage;
return VPX_CODEC_INCAPABLE;
#endif
}

View File

@ -0,0 +1,182 @@
/*
* Copyright (c) 2017 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vpx_dsp_rtcd.h"
#include "test/acm_random.h"
#include "test/buffer.h"
#include "test/register_state_check.h"
#include "vpx_ports/vpx_timer.h"
namespace {
using ::libvpx_test::ACMRandom;
using ::libvpx_test::Buffer;
typedef void (*AvgPredFunc)(uint8_t *a, const uint8_t *b, int w, int h,
const uint8_t *c, int c_stride);
uint8_t avg_with_rounding(uint8_t a, uint8_t b) { return (a + b + 1) >> 1; }
void reference_pred(const Buffer<uint8_t> &pred, const Buffer<uint8_t> &ref,
int width, int height, Buffer<uint8_t> *avg) {
for (int y = 0; y < height; ++y) {
for (int x = 0; x < width; ++x) {
avg->TopLeftPixel()[y * avg->stride() + x] =
avg_with_rounding(pred.TopLeftPixel()[y * pred.stride() + x],
ref.TopLeftPixel()[y * ref.stride() + x]);
}
}
}
class AvgPredTest : public ::testing::TestWithParam<AvgPredFunc> {
public:
virtual void SetUp() {
avg_pred_func_ = GetParam();
rnd_.Reset(ACMRandom::DeterministicSeed());
}
protected:
AvgPredFunc avg_pred_func_;
ACMRandom rnd_;
};
TEST_P(AvgPredTest, SizeCombinations) {
// This is called as part of the sub pixel variance. As such it must be one of
// the variance block sizes.
for (int width_pow = 2; width_pow <= 6; ++width_pow) {
for (int height_pow = width_pow - 1; height_pow <= width_pow + 1;
++height_pow) {
// Don't test 4x2 or 64x128
if (height_pow == 1 || height_pow == 7) continue;
// The sse2 special-cases when ref width == stride, so make sure to test
// it.
for (int ref_padding = 0; ref_padding < 2; ref_padding++) {
const int width = 1 << width_pow;
const int height = 1 << height_pow;
// Only the reference buffer may have a stride not equal to width.
Buffer<uint8_t> ref =
Buffer<uint8_t>(width, height, ref_padding ? 8 : 0);
ASSERT_TRUE(ref.Init());
Buffer<uint8_t> pred = Buffer<uint8_t>(width, height, 0, 16);
ASSERT_TRUE(pred.Init());
Buffer<uint8_t> avg_ref = Buffer<uint8_t>(width, height, 0, 16);
ASSERT_TRUE(avg_ref.Init());
Buffer<uint8_t> avg_chk = Buffer<uint8_t>(width, height, 0, 16);
ASSERT_TRUE(avg_chk.Init());
ref.Set(&rnd_, &ACMRandom::Rand8);
pred.Set(&rnd_, &ACMRandom::Rand8);
reference_pred(pred, ref, width, height, &avg_ref);
ASM_REGISTER_STATE_CHECK(
avg_pred_func_(avg_chk.TopLeftPixel(), pred.TopLeftPixel(), width,
height, ref.TopLeftPixel(), ref.stride()));
EXPECT_TRUE(avg_chk.CheckValues(avg_ref));
if (HasFailure()) {
printf("Width: %d Height: %d\n", width, height);
avg_chk.PrintDifference(avg_ref);
return;
}
}
}
}
}
TEST_P(AvgPredTest, CompareReferenceRandom) {
const int width = 64;
const int height = 32;
Buffer<uint8_t> ref = Buffer<uint8_t>(width, height, 8);
ASSERT_TRUE(ref.Init());
Buffer<uint8_t> pred = Buffer<uint8_t>(width, height, 0, 16);
ASSERT_TRUE(pred.Init());
Buffer<uint8_t> avg_ref = Buffer<uint8_t>(width, height, 0, 16);
ASSERT_TRUE(avg_ref.Init());
Buffer<uint8_t> avg_chk = Buffer<uint8_t>(width, height, 0, 16);
ASSERT_TRUE(avg_chk.Init());
for (int i = 0; i < 500; ++i) {
ref.Set(&rnd_, &ACMRandom::Rand8);
pred.Set(&rnd_, &ACMRandom::Rand8);
reference_pred(pred, ref, width, height, &avg_ref);
ASM_REGISTER_STATE_CHECK(avg_pred_func_(avg_chk.TopLeftPixel(),
pred.TopLeftPixel(), width, height,
ref.TopLeftPixel(), ref.stride()));
EXPECT_TRUE(avg_chk.CheckValues(avg_ref));
if (HasFailure()) {
printf("Width: %d Height: %d\n", width, height);
avg_chk.PrintDifference(avg_ref);
return;
}
}
}
TEST_P(AvgPredTest, DISABLED_Speed) {
for (int width_pow = 2; width_pow <= 6; ++width_pow) {
for (int height_pow = width_pow - 1; height_pow <= width_pow + 1;
++height_pow) {
// Don't test 4x2 or 64x128
if (height_pow == 1 || height_pow == 7) continue;
for (int ref_padding = 0; ref_padding < 2; ref_padding++) {
const int width = 1 << width_pow;
const int height = 1 << height_pow;
Buffer<uint8_t> ref =
Buffer<uint8_t>(width, height, ref_padding ? 8 : 0);
ASSERT_TRUE(ref.Init());
Buffer<uint8_t> pred = Buffer<uint8_t>(width, height, 0, 16);
ASSERT_TRUE(pred.Init());
Buffer<uint8_t> avg = Buffer<uint8_t>(width, height, 0, 16);
ASSERT_TRUE(avg.Init());
ref.Set(&rnd_, &ACMRandom::Rand8);
pred.Set(&rnd_, &ACMRandom::Rand8);
vpx_usec_timer timer;
vpx_usec_timer_start(&timer);
for (int i = 0; i < 10000000 / (width * height); ++i) {
avg_pred_func_(avg.TopLeftPixel(), pred.TopLeftPixel(), width, height,
ref.TopLeftPixel(), ref.stride());
}
vpx_usec_timer_mark(&timer);
const int elapsed_time =
static_cast<int>(vpx_usec_timer_elapsed(&timer));
printf("Average Test (ref_padding: %d) %dx%d time: %5d us\n",
ref_padding, width, height, elapsed_time);
}
}
}
}
INSTANTIATE_TEST_CASE_P(C, AvgPredTest,
::testing::Values(&vpx_comp_avg_pred_c));
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(SSE2, AvgPredTest,
::testing::Values(&vpx_comp_avg_pred_sse2));
#endif // HAVE_SSE2
#if HAVE_NEON
INSTANTIATE_TEST_CASE_P(NEON, AvgPredTest,
::testing::Values(&vpx_comp_avg_pred_neon));
#endif // HAVE_NEON
#if HAVE_VSX
INSTANTIATE_TEST_CASE_P(VSX, AvgPredTest,
::testing::Values(&vpx_comp_avg_pred_vsx));
#endif // HAVE_VSX
} // namespace

File diff suppressed because it is too large Load Diff

File diff suppressed because it is too large Load Diff

View File

@ -255,11 +255,11 @@ void iht16x16_ref(const tran_low_t *in, uint8_t *dest, int stride,
#if CONFIG_VP9_HIGHBITDEPTH
void idct16x16_10(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct16x16_256_add_c(in, out, stride, 10);
vpx_highbd_idct16x16_256_add_c(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct16x16_12(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct16x16_256_add_c(in, out, stride, 12);
vpx_highbd_idct16x16_256_add_c(in, CAST_TO_SHORTPTR(out), stride, 12);
}
void idct16x16_10_ref(const tran_low_t *in, uint8_t *out, int stride,
@ -273,36 +273,36 @@ void idct16x16_12_ref(const tran_low_t *in, uint8_t *out, int stride,
}
void iht16x16_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
vp9_highbd_iht16x16_256_add_c(in, out, stride, tx_type, 10);
vp9_highbd_iht16x16_256_add_c(in, CAST_TO_SHORTPTR(out), stride, tx_type, 10);
}
void iht16x16_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
vp9_highbd_iht16x16_256_add_c(in, out, stride, tx_type, 12);
vp9_highbd_iht16x16_256_add_c(in, CAST_TO_SHORTPTR(out), stride, tx_type, 12);
}
#if HAVE_SSE2
void idct16x16_10_add_10_c(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct16x16_10_add_c(in, out, stride, 10);
vpx_highbd_idct16x16_10_add_c(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct16x16_10_add_12_c(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct16x16_10_add_c(in, out, stride, 12);
vpx_highbd_idct16x16_10_add_c(in, CAST_TO_SHORTPTR(out), stride, 12);
}
void idct16x16_256_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct16x16_256_add_sse2(in, out, stride, 10);
vpx_highbd_idct16x16_256_add_sse2(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct16x16_256_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct16x16_256_add_sse2(in, out, stride, 12);
vpx_highbd_idct16x16_256_add_sse2(in, CAST_TO_SHORTPTR(out), stride, 12);
}
void idct16x16_10_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct16x16_10_add_sse2(in, out, stride, 10);
vpx_highbd_idct16x16_10_add_sse2(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct16x16_10_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct16x16_10_add_sse2(in, out, stride, 12);
vpx_highbd_idct16x16_10_add_sse2(in, CAST_TO_SHORTPTR(out), stride, 12);
}
#endif // HAVE_SSE2
#endif // CONFIG_VP9_HIGHBITDEPTH
@ -353,7 +353,7 @@ class Trans16x16TestBase {
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(test_temp_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
RunInvTxfm(test_temp_block, CAST_TO_BYTEPTR(dst16), pitch_));
#endif
}
@ -475,10 +475,10 @@ class Trans16x16TestBase {
ASM_REGISTER_STATE_CHECK(RunInvTxfm(output_ref_block, dst, pitch_));
#if CONFIG_VP9_HIGHBITDEPTH
} else {
inv_txfm_ref(output_ref_block, CONVERT_TO_BYTEPTR(ref16), pitch_,
inv_txfm_ref(output_ref_block, CAST_TO_BYTEPTR(ref16), pitch_,
tx_type_);
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(output_ref_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
RunInvTxfm(output_ref_block, CAST_TO_BYTEPTR(dst16), pitch_));
#endif
}
if (bit_depth_ == VPX_BITS_8) {
@ -530,8 +530,7 @@ class Trans16x16TestBase {
ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, 16));
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), 16));
ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, CAST_TO_BYTEPTR(dst16), 16));
#endif // CONFIG_VP9_HIGHBITDEPTH
}
@ -543,8 +542,8 @@ class Trans16x16TestBase {
const uint32_t diff = dst[j] - src[j];
#endif // CONFIG_VP9_HIGHBITDEPTH
const uint32_t error = diff * diff;
EXPECT_GE(1u, error) << "Error: 16x16 IDCT has error " << error
<< " at index " << j;
EXPECT_GE(1u, error)
<< "Error: 16x16 IDCT has error " << error << " at index " << j;
}
}
}
@ -585,9 +584,9 @@ class Trans16x16TestBase {
ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, pitch_));
} else {
#if CONFIG_VP9_HIGHBITDEPTH
ref_txfm(coeff, CONVERT_TO_BYTEPTR(ref16), pitch_);
ref_txfm(coeff, CAST_TO_BYTEPTR(ref16), pitch_);
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), pitch_));
RunInvTxfm(coeff, CAST_TO_BYTEPTR(dst16), pitch_));
#endif // CONFIG_VP9_HIGHBITDEPTH
}
@ -745,66 +744,6 @@ TEST_P(InvTrans16x16DCT, CompareReference) {
CompareInvReference(ref_txfm_, thresh_);
}
class PartialTrans16x16Test : public ::testing::TestWithParam<
std::tr1::tuple<FdctFunc, vpx_bit_depth_t> > {
public:
virtual ~PartialTrans16x16Test() {}
virtual void SetUp() {
fwd_txfm_ = GET_PARAM(0);
bit_depth_ = GET_PARAM(1);
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
vpx_bit_depth_t bit_depth_;
FdctFunc fwd_txfm_;
};
TEST_P(PartialTrans16x16Test, Extremes) {
#if CONFIG_VP9_HIGHBITDEPTH
const int16_t maxval =
static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
#else
const int16_t maxval = 255;
#endif
const int minval = -maxval;
DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
for (int i = 0; i < kNumCoeffs; ++i) input[i] = maxval;
output[0] = 0;
ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
EXPECT_EQ((maxval * kNumCoeffs) >> 1, output[0]);
for (int i = 0; i < kNumCoeffs; ++i) input[i] = minval;
output[0] = 0;
ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
EXPECT_EQ((minval * kNumCoeffs) >> 1, output[0]);
}
TEST_P(PartialTrans16x16Test, Random) {
#if CONFIG_VP9_HIGHBITDEPTH
const int16_t maxval =
static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
#else
const int16_t maxval = 255;
#endif
DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
ACMRandom rnd(ACMRandom::DeterministicSeed());
int sum = 0;
for (int i = 0; i < kNumCoeffs; ++i) {
const int val = (i & 1) ? -rnd(maxval + 1) : rnd(maxval + 1);
input[i] = val;
sum += val;
}
output[0] = 0;
ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 16));
EXPECT_EQ(sum >> 1, output[0]);
}
using std::tr1::make_tuple;
#if CONFIG_VP9_HIGHBITDEPTH
@ -837,11 +776,6 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 1, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 2, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 3, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
C, PartialTrans16x16Test,
::testing::Values(make_tuple(&vpx_highbd_fdct16x16_1_c, VPX_BITS_8),
make_tuple(&vpx_highbd_fdct16x16_1_c, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct16x16_1_c, VPX_BITS_12)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans16x16HT,
@ -850,17 +784,14 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 1, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 2, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 3, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(C, PartialTrans16x16Test,
::testing::Values(make_tuple(&vpx_fdct16x16_1_c,
VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_NEON && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
NEON, Trans16x16DCT,
::testing::Values(make_tuple(&vpx_fdct16x16_c, &vpx_idct16x16_256_add_neon,
0, VPX_BITS_8)));
#endif
::testing::Values(make_tuple(&vpx_fdct16x16_neon,
&vpx_idct16x16_256_add_neon, 0, VPX_BITS_8)));
#endif // HAVE_NEON && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
@ -877,9 +808,6 @@ INSTANTIATE_TEST_CASE_P(
2, VPX_BITS_8),
make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2,
3, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(SSE2, PartialTrans16x16Test,
::testing::Values(make_tuple(&vpx_fdct16x16_1_sse2,
VPX_BITS_8)));
#endif // HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -914,9 +842,6 @@ INSTANTIATE_TEST_CASE_P(
&idct16x16_10_add_12_sse2, 3167, VPX_BITS_12),
make_tuple(&idct16x16_12, &idct16x16_256_add_12_sse2,
3167, VPX_BITS_12)));
INSTANTIATE_TEST_CASE_P(SSE2, PartialTrans16x16Test,
::testing::Values(make_tuple(&vpx_fdct16x16_1_sse2,
VPX_BITS_8)));
#endif // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -932,8 +857,12 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp9_fht16x16_msa, &vp9_iht16x16_256_add_msa, 2, VPX_BITS_8),
make_tuple(&vp9_fht16x16_msa, &vp9_iht16x16_256_add_msa, 3,
VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(MSA, PartialTrans16x16Test,
::testing::Values(make_tuple(&vpx_fdct16x16_1_msa,
VPX_BITS_8)));
#endif // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(VSX, Trans16x16DCT,
::testing::Values(make_tuple(&vpx_fdct16x16_c,
&vpx_idct16x16_256_add_vsx,
0, VPX_BITS_8)));
#endif // HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
} // namespace

View File

@ -71,11 +71,11 @@ typedef std::tr1::tuple<FwdTxfmFunc, InvTxfmFunc, int, vpx_bit_depth_t>
#if CONFIG_VP9_HIGHBITDEPTH
void idct32x32_10(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct32x32_1024_add_c(in, out, stride, 10);
vpx_highbd_idct32x32_1024_add_c(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct32x32_12(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct32x32_1024_add_c(in, out, stride, 12);
vpx_highbd_idct32x32_1024_add_c(in, CAST_TO_SHORTPTR(out), stride, 12);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
@ -137,7 +137,7 @@ TEST_P(Trans32x32Test, AccuracyCheck) {
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
inv_txfm_(test_temp_block, CONVERT_TO_BYTEPTR(dst16), 32));
inv_txfm_(test_temp_block, CAST_TO_BYTEPTR(dst16), 32));
#endif
}
@ -275,7 +275,7 @@ TEST_P(Trans32x32Test, InverseAccuracy) {
ASM_REGISTER_STATE_CHECK(inv_txfm_(coeff, dst, 32));
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(inv_txfm_(coeff, CONVERT_TO_BYTEPTR(dst16), 32));
ASM_REGISTER_STATE_CHECK(inv_txfm_(coeff, CAST_TO_BYTEPTR(dst16), 32));
#endif
}
for (int j = 0; j < kNumCoeffs; ++j) {
@ -292,67 +292,6 @@ TEST_P(Trans32x32Test, InverseAccuracy) {
}
}
class PartialTrans32x32Test
: public ::testing::TestWithParam<
std::tr1::tuple<FwdTxfmFunc, vpx_bit_depth_t> > {
public:
virtual ~PartialTrans32x32Test() {}
virtual void SetUp() {
fwd_txfm_ = GET_PARAM(0);
bit_depth_ = GET_PARAM(1);
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
vpx_bit_depth_t bit_depth_;
FwdTxfmFunc fwd_txfm_;
};
TEST_P(PartialTrans32x32Test, Extremes) {
#if CONFIG_VP9_HIGHBITDEPTH
const int16_t maxval =
static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
#else
const int16_t maxval = 255;
#endif
const int minval = -maxval;
DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
for (int i = 0; i < kNumCoeffs; ++i) input[i] = maxval;
output[0] = 0;
ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
EXPECT_EQ((maxval * kNumCoeffs) >> 3, output[0]);
for (int i = 0; i < kNumCoeffs; ++i) input[i] = minval;
output[0] = 0;
ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
EXPECT_EQ((minval * kNumCoeffs) >> 3, output[0]);
}
TEST_P(PartialTrans32x32Test, Random) {
#if CONFIG_VP9_HIGHBITDEPTH
const int16_t maxval =
static_cast<int16_t>(clip_pixel_highbd(1 << 30, bit_depth_));
#else
const int16_t maxval = 255;
#endif
DECLARE_ALIGNED(16, int16_t, input[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, output[kNumCoeffs]);
ACMRandom rnd(ACMRandom::DeterministicSeed());
int sum = 0;
for (int i = 0; i < kNumCoeffs; ++i) {
const int val = (i & 1) ? -rnd(maxval + 1) : rnd(maxval + 1);
input[i] = val;
sum += val;
}
output[0] = 0;
ASM_REGISTER_STATE_CHECK(fwd_txfm_(input, output, 32));
EXPECT_EQ(sum >> 3, output[0]);
}
using std::tr1::make_tuple;
#if CONFIG_VP9_HIGHBITDEPTH
@ -366,11 +305,6 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vpx_fdct32x32_c, &vpx_idct32x32_1024_add_c, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_c, &vpx_idct32x32_1024_add_c, 1,
VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
C, PartialTrans32x32Test,
::testing::Values(make_tuple(&vpx_highbd_fdct32x32_1_c, VPX_BITS_8),
make_tuple(&vpx_highbd_fdct32x32_1_c, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct32x32_1_c, VPX_BITS_12)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans32x32Test,
@ -378,19 +312,16 @@ INSTANTIATE_TEST_CASE_P(
VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_c, &vpx_idct32x32_1024_add_c,
1, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(C, PartialTrans32x32Test,
::testing::Values(make_tuple(&vpx_fdct32x32_1_c,
VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_NEON && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
NEON, Trans32x32Test,
::testing::Values(make_tuple(&vpx_fdct32x32_c, &vpx_idct32x32_1024_add_neon,
0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_c,
::testing::Values(make_tuple(&vpx_fdct32x32_neon,
&vpx_idct32x32_1024_add_neon, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_neon,
&vpx_idct32x32_1024_add_neon, 1, VPX_BITS_8)));
#endif // HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#endif // HAVE_NEON && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
@ -399,9 +330,6 @@ INSTANTIATE_TEST_CASE_P(
&vpx_idct32x32_1024_add_sse2, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_sse2,
&vpx_idct32x32_1024_add_sse2, 1, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(SSE2, PartialTrans32x32Test,
::testing::Values(make_tuple(&vpx_fdct32x32_1_sse2,
VPX_BITS_8)));
#endif // HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -418,9 +346,6 @@ INSTANTIATE_TEST_CASE_P(
VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_sse2, &vpx_idct32x32_1024_add_c, 1,
VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(SSE2, PartialTrans32x32Test,
::testing::Values(make_tuple(&vpx_fdct32x32_1_sse2,
VPX_BITS_8)));
#endif // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_AVX2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -439,8 +364,14 @@ INSTANTIATE_TEST_CASE_P(
&vpx_idct32x32_1024_add_msa, 0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_msa,
&vpx_idct32x32_1024_add_msa, 1, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(MSA, PartialTrans32x32Test,
::testing::Values(make_tuple(&vpx_fdct32x32_1_msa,
VPX_BITS_8)));
#endif // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
VSX, Trans32x32Test,
::testing::Values(make_tuple(&vpx_fdct32x32_c, &vpx_idct32x32_1024_add_vsx,
0, VPX_BITS_8),
make_tuple(&vpx_fdct32x32_rd_c,
&vpx_idct32x32_1024_add_vsx, 1, VPX_BITS_8)));
#endif // HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
} // namespace

View File

@ -0,0 +1,169 @@
/*
* Copyright (c) 2017 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <math.h>
#include <stdlib.h>
#include <string.h>
#include <limits>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vpx_dsp_rtcd.h"
#include "test/acm_random.h"
#include "test/buffer.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "test/util.h"
#include "vpx/vpx_codec.h"
#include "vpx/vpx_integer.h"
#include "vpx_dsp/vpx_dsp_common.h"
using libvpx_test::ACMRandom;
using libvpx_test::Buffer;
using std::tr1::tuple;
using std::tr1::make_tuple;
namespace {
typedef void (*PartialFdctFunc)(const int16_t *in, tran_low_t *out, int stride);
typedef tuple<PartialFdctFunc, int /* size */, vpx_bit_depth_t>
PartialFdctParam;
tran_low_t partial_fdct_ref(const Buffer<int16_t> &in, int size) {
int64_t sum = 0;
for (int y = 0; y < size; ++y) {
for (int x = 0; x < size; ++x) {
sum += in.TopLeftPixel()[y * in.stride() + x];
}
}
switch (size) {
case 4: sum *= 2; break;
case 8: /*sum = sum;*/ break;
case 16: sum >>= 1; break;
case 32: sum >>= 3; break;
}
return static_cast<tran_low_t>(sum);
}
class PartialFdctTest : public ::testing::TestWithParam<PartialFdctParam> {
public:
PartialFdctTest() {
fwd_txfm_ = GET_PARAM(0);
size_ = GET_PARAM(1);
bit_depth_ = GET_PARAM(2);
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
void RunTest() {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int16_t maxvalue =
clip_pixel_highbd(std::numeric_limits<int16_t>::max(), bit_depth_);
const int16_t minvalue = -maxvalue;
Buffer<int16_t> input_block =
Buffer<int16_t>(size_, size_, 8, size_ == 4 ? 0 : 16);
ASSERT_TRUE(input_block.Init());
Buffer<tran_low_t> output_block = Buffer<tran_low_t>(size_, size_, 0, 16);
ASSERT_TRUE(output_block.Init());
for (int i = 0; i < 100; ++i) {
if (i == 0) {
input_block.Set(maxvalue);
} else if (i == 1) {
input_block.Set(minvalue);
} else {
input_block.Set(&rnd, minvalue, maxvalue);
}
ASM_REGISTER_STATE_CHECK(fwd_txfm_(input_block.TopLeftPixel(),
output_block.TopLeftPixel(),
input_block.stride()));
EXPECT_EQ(partial_fdct_ref(input_block, size_),
output_block.TopLeftPixel()[0]);
}
}
PartialFdctFunc fwd_txfm_;
vpx_bit_depth_t bit_depth_;
int size_;
};
TEST_P(PartialFdctTest, PartialFdctTest) { RunTest(); }
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
C, PartialFdctTest,
::testing::Values(make_tuple(&vpx_highbd_fdct32x32_1_c, 32, VPX_BITS_12),
make_tuple(&vpx_highbd_fdct32x32_1_c, 32, VPX_BITS_10),
make_tuple(&vpx_fdct32x32_1_c, 32, VPX_BITS_8),
make_tuple(&vpx_highbd_fdct16x16_1_c, 16, VPX_BITS_12),
make_tuple(&vpx_highbd_fdct16x16_1_c, 16, VPX_BITS_10),
make_tuple(&vpx_fdct16x16_1_c, 16, VPX_BITS_8),
make_tuple(&vpx_highbd_fdct8x8_1_c, 8, VPX_BITS_12),
make_tuple(&vpx_highbd_fdct8x8_1_c, 8, VPX_BITS_10),
make_tuple(&vpx_fdct8x8_1_c, 8, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_1_c, 4, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, PartialFdctTest,
::testing::Values(make_tuple(&vpx_fdct32x32_1_c, 32, VPX_BITS_8),
make_tuple(&vpx_fdct16x16_1_c, 16, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_1_c, 8, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_1_c, 4, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(
SSE2, PartialFdctTest,
::testing::Values(make_tuple(&vpx_fdct32x32_1_sse2, 32, VPX_BITS_8),
make_tuple(&vpx_fdct16x16_1_sse2, 16, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_1_sse2, 8, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_1_sse2, 4, VPX_BITS_8)));
#endif // HAVE_SSE2
#if HAVE_NEON
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
NEON, PartialFdctTest,
::testing::Values(make_tuple(&vpx_fdct32x32_1_neon, 32, VPX_BITS_8),
make_tuple(&vpx_fdct16x16_1_neon, 16, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_1_neon, 8, VPX_BITS_12),
make_tuple(&vpx_fdct8x8_1_neon, 8, VPX_BITS_10),
make_tuple(&vpx_fdct8x8_1_neon, 8, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_1_neon, 4, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
NEON, PartialFdctTest,
::testing::Values(make_tuple(&vpx_fdct32x32_1_neon, 32, VPX_BITS_8),
make_tuple(&vpx_fdct16x16_1_neon, 16, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_1_neon, 8, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_1_neon, 4, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#endif // HAVE_NEON
#if HAVE_MSA
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(MSA, PartialFdctTest,
::testing::Values(make_tuple(&vpx_fdct8x8_1_msa, 8,
VPX_BITS_8)));
#else // !CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
MSA, PartialFdctTest,
::testing::Values(make_tuple(&vpx_fdct32x32_1_msa, 32, VPX_BITS_8),
make_tuple(&vpx_fdct16x16_1_msa, 16, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_1_msa, 8, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#endif // HAVE_MSA
} // namespace

View File

@ -0,0 +1,737 @@
/*
* Copyright (c) 2017 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <math.h>
#include <stdlib.h>
#include <string.h>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vp9_rtcd.h"
#include "./vpx_dsp_rtcd.h"
#include "test/acm_random.h"
#include "test/buffer.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "test/util.h"
#include "vp9/common/vp9_entropy.h"
#include "vpx/vpx_codec.h"
#include "vpx/vpx_integer.h"
#include "vpx_ports/mem.h"
using libvpx_test::ACMRandom;
using libvpx_test::Buffer;
using std::tr1::tuple;
using std::tr1::make_tuple;
namespace {
typedef void (*FdctFunc)(const int16_t *in, tran_low_t *out, int stride);
typedef void (*IdctFunc)(const tran_low_t *in, uint8_t *out, int stride);
typedef void (*FhtFunc)(const int16_t *in, tran_low_t *out, int stride,
int tx_type);
typedef void (*FhtFuncRef)(const Buffer<int16_t> &in, Buffer<tran_low_t> *out,
int size, int tx_type);
typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
int tx_type);
/* forward transform, inverse transform, size, transform type, bit depth */
typedef tuple<FdctFunc, IdctFunc, int, int, vpx_bit_depth_t> DctParam;
typedef tuple<FhtFunc, IhtFunc, int, int, vpx_bit_depth_t> HtParam;
void fdct_ref(const Buffer<int16_t> &in, Buffer<tran_low_t> *out, int size,
int /*tx_type*/) {
const int16_t *i = in.TopLeftPixel();
const int i_stride = in.stride();
tran_low_t *o = out->TopLeftPixel();
if (size == 4) {
vpx_fdct4x4_c(i, o, i_stride);
} else if (size == 8) {
vpx_fdct8x8_c(i, o, i_stride);
} else if (size == 16) {
vpx_fdct16x16_c(i, o, i_stride);
} else if (size == 32) {
vpx_fdct32x32_c(i, o, i_stride);
}
}
void fht_ref(const Buffer<int16_t> &in, Buffer<tran_low_t> *out, int size,
int tx_type) {
const int16_t *i = in.TopLeftPixel();
const int i_stride = in.stride();
tran_low_t *o = out->TopLeftPixel();
if (size == 4) {
vp9_fht4x4_c(i, o, i_stride, tx_type);
} else if (size == 8) {
vp9_fht8x8_c(i, o, i_stride, tx_type);
} else if (size == 16) {
vp9_fht16x16_c(i, o, i_stride, tx_type);
}
}
void fwht_ref(const Buffer<int16_t> &in, Buffer<tran_low_t> *out, int size,
int /*tx_type*/) {
ASSERT_EQ(size, 4);
vp9_fwht4x4_c(in.TopLeftPixel(), out->TopLeftPixel(), in.stride());
}
#if CONFIG_VP9_HIGHBITDEPTH
#define idctNxN(n, coeffs, bitdepth) \
void idct##n##x##n##_##bitdepth(const tran_low_t *in, uint8_t *out, \
int stride) { \
vpx_highbd_idct##n##x##n##_##coeffs##_add_c(in, CAST_TO_SHORTPTR(out), \
stride, bitdepth); \
}
idctNxN(4, 16, 10);
idctNxN(4, 16, 12);
idctNxN(8, 64, 10);
idctNxN(8, 64, 12);
idctNxN(16, 256, 10);
idctNxN(16, 256, 12);
idctNxN(32, 1024, 10);
idctNxN(32, 1024, 12);
#define ihtNxN(n, coeffs, bitdepth) \
void iht##n##x##n##_##bitdepth(const tran_low_t *in, uint8_t *out, \
int stride, int tx_type) { \
vp9_highbd_iht##n##x##n##_##coeffs##_add_c(in, CAST_TO_SHORTPTR(out), \
stride, tx_type, bitdepth); \
}
ihtNxN(4, 16, 10);
ihtNxN(4, 16, 12);
ihtNxN(8, 64, 10);
ihtNxN(8, 64, 12);
ihtNxN(16, 256, 10);
// ihtNxN(16, 256, 12);
void iwht4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_iwht4x4_16_add_c(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void iwht4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_iwht4x4_16_add_c(in, CAST_TO_SHORTPTR(out), stride, 12);
}
#endif // CONFIG_VP9_HIGHBITDEPTH
class TransTestBase {
public:
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
virtual void RunFwdTxfm(const Buffer<int16_t> &in,
Buffer<tran_low_t> *out) = 0;
virtual void RunInvTxfm(const Buffer<tran_low_t> &in, uint8_t *out) = 0;
void RunAccuracyCheck(int limit) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
Buffer<int16_t> test_input_block =
Buffer<int16_t>(size_, size_, 8, size_ == 4 ? 0 : 16);
ASSERT_TRUE(test_input_block.Init());
Buffer<tran_low_t> test_temp_block =
Buffer<tran_low_t>(size_, size_, 0, 16);
ASSERT_TRUE(test_temp_block.Init());
Buffer<uint8_t> dst = Buffer<uint8_t>(size_, size_, 0, 16);
ASSERT_TRUE(dst.Init());
Buffer<uint8_t> src = Buffer<uint8_t>(size_, size_, 0, 16);
ASSERT_TRUE(src.Init());
#if CONFIG_VP9_HIGHBITDEPTH
Buffer<uint16_t> dst16 = Buffer<uint16_t>(size_, size_, 0, 16);
ASSERT_TRUE(dst16.Init());
Buffer<uint16_t> src16 = Buffer<uint16_t>(size_, size_, 0, 16);
ASSERT_TRUE(src16.Init());
#endif // CONFIG_VP9_HIGHBITDEPTH
uint32_t max_error = 0;
int64_t total_error = 0;
const int count_test_block = 10000;
for (int i = 0; i < count_test_block; ++i) {
if (bit_depth_ == 8) {
src.Set(&rnd, &ACMRandom::Rand8);
dst.Set(&rnd, &ACMRandom::Rand8);
// Initialize a test block with input range [-255, 255].
for (int h = 0; h < size_; ++h) {
for (int w = 0; w < size_; ++w) {
test_input_block.TopLeftPixel()[h * test_input_block.stride() + w] =
src.TopLeftPixel()[h * src.stride() + w] -
dst.TopLeftPixel()[h * dst.stride() + w];
}
}
#if CONFIG_VP9_HIGHBITDEPTH
} else {
src16.Set(&rnd, 0, max_pixel_value_);
dst16.Set(&rnd, 0, max_pixel_value_);
for (int h = 0; h < size_; ++h) {
for (int w = 0; w < size_; ++w) {
test_input_block.TopLeftPixel()[h * test_input_block.stride() + w] =
src16.TopLeftPixel()[h * src16.stride() + w] -
dst16.TopLeftPixel()[h * dst16.stride() + w];
}
}
#endif // CONFIG_VP9_HIGHBITDEPTH
}
ASM_REGISTER_STATE_CHECK(RunFwdTxfm(test_input_block, &test_temp_block));
if (bit_depth_ == VPX_BITS_8) {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(test_temp_block, dst.TopLeftPixel()));
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(test_temp_block, CAST_TO_BYTEPTR(dst16.TopLeftPixel())));
#endif // CONFIG_VP9_HIGHBITDEPTH
}
for (int h = 0; h < size_; ++h) {
for (int w = 0; w < size_; ++w) {
int diff;
#if CONFIG_VP9_HIGHBITDEPTH
if (bit_depth_ != 8) {
diff = dst16.TopLeftPixel()[h * dst16.stride() + w] -
src16.TopLeftPixel()[h * src16.stride() + w];
} else {
#endif // CONFIG_VP9_HIGHBITDEPTH
diff = dst.TopLeftPixel()[h * dst.stride() + w] -
src.TopLeftPixel()[h * src.stride() + w];
#if CONFIG_VP9_HIGHBITDEPTH
}
#endif // CONFIG_VP9_HIGHBITDEPTH
const uint32_t error = diff * diff;
if (max_error < error) max_error = error;
total_error += error;
}
}
}
EXPECT_GE(static_cast<uint32_t>(limit), max_error)
<< "Error: 4x4 FHT/IHT has an individual round trip error > " << limit;
EXPECT_GE(count_test_block * limit, total_error)
<< "Error: 4x4 FHT/IHT has average round trip error > " << limit
<< " per block";
}
void RunCoeffCheck() {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 5000;
Buffer<int16_t> input_block =
Buffer<int16_t>(size_, size_, 8, size_ == 4 ? 0 : 16);
ASSERT_TRUE(input_block.Init());
Buffer<tran_low_t> output_ref_block = Buffer<tran_low_t>(size_, size_, 0);
ASSERT_TRUE(output_ref_block.Init());
Buffer<tran_low_t> output_block = Buffer<tran_low_t>(size_, size_, 0, 16);
ASSERT_TRUE(output_block.Init());
for (int i = 0; i < count_test_block; ++i) {
// Initialize a test block with input range [-max_pixel_value_,
// max_pixel_value_].
input_block.Set(&rnd, -max_pixel_value_, max_pixel_value_);
fwd_txfm_ref(input_block, &output_ref_block, size_, tx_type_);
ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_block, &output_block));
// The minimum quant value is 4.
EXPECT_TRUE(output_block.CheckValues(output_ref_block));
if (::testing::Test::HasFailure()) {
printf("Size: %d Transform type: %d\n", size_, tx_type_);
output_block.PrintDifference(output_ref_block);
return;
}
}
}
void RunMemCheck() {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 5000;
Buffer<int16_t> input_extreme_block =
Buffer<int16_t>(size_, size_, 8, size_ == 4 ? 0 : 16);
ASSERT_TRUE(input_extreme_block.Init());
Buffer<tran_low_t> output_ref_block = Buffer<tran_low_t>(size_, size_, 0);
ASSERT_TRUE(output_ref_block.Init());
Buffer<tran_low_t> output_block = Buffer<tran_low_t>(size_, size_, 0, 16);
ASSERT_TRUE(output_block.Init());
for (int i = 0; i < count_test_block; ++i) {
// Initialize a test block with -max_pixel_value_ or max_pixel_value_.
if (i == 0) {
input_extreme_block.Set(max_pixel_value_);
} else if (i == 1) {
input_extreme_block.Set(-max_pixel_value_);
} else {
for (int h = 0; h < size_; ++h) {
for (int w = 0; w < size_; ++w) {
input_extreme_block
.TopLeftPixel()[h * input_extreme_block.stride() + w] =
rnd.Rand8() % 2 ? max_pixel_value_ : -max_pixel_value_;
}
}
}
fwd_txfm_ref(input_extreme_block, &output_ref_block, size_, tx_type_);
ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_extreme_block, &output_block));
// The minimum quant value is 4.
EXPECT_TRUE(output_block.CheckValues(output_ref_block));
for (int h = 0; h < size_; ++h) {
for (int w = 0; w < size_; ++w) {
EXPECT_GE(
4 * DCT_MAX_VALUE << (bit_depth_ - 8),
abs(output_block.TopLeftPixel()[h * output_block.stride() + w]))
<< "Error: 4x4 FDCT has coefficient larger than "
"4*DCT_MAX_VALUE"
<< " at " << w << "," << h;
if (::testing::Test::HasFailure()) {
printf("Size: %d Transform type: %d\n", size_, tx_type_);
output_block.DumpBuffer();
return;
}
}
}
}
}
void RunInvAccuracyCheck(int limit) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 1000;
Buffer<int16_t> in = Buffer<int16_t>(size_, size_, 4);
ASSERT_TRUE(in.Init());
Buffer<tran_low_t> coeff = Buffer<tran_low_t>(size_, size_, 0, 16);
ASSERT_TRUE(coeff.Init());
Buffer<uint8_t> dst = Buffer<uint8_t>(size_, size_, 0, 16);
ASSERT_TRUE(dst.Init());
Buffer<uint8_t> src = Buffer<uint8_t>(size_, size_, 0);
ASSERT_TRUE(src.Init());
Buffer<uint16_t> dst16 = Buffer<uint16_t>(size_, size_, 0, 16);
ASSERT_TRUE(dst16.Init());
Buffer<uint16_t> src16 = Buffer<uint16_t>(size_, size_, 0);
ASSERT_TRUE(src16.Init());
for (int i = 0; i < count_test_block; ++i) {
// Initialize a test block with input range [-max_pixel_value_,
// max_pixel_value_].
if (bit_depth_ == VPX_BITS_8) {
src.Set(&rnd, &ACMRandom::Rand8);
dst.Set(&rnd, &ACMRandom::Rand8);
for (int h = 0; h < size_; ++h) {
for (int w = 0; w < size_; ++w) {
in.TopLeftPixel()[h * in.stride() + w] =
src.TopLeftPixel()[h * src.stride() + w] -
dst.TopLeftPixel()[h * dst.stride() + w];
}
}
#if CONFIG_VP9_HIGHBITDEPTH
} else {
src16.Set(&rnd, 0, max_pixel_value_);
dst16.Set(&rnd, 0, max_pixel_value_);
for (int h = 0; h < size_; ++h) {
for (int w = 0; w < size_; ++w) {
in.TopLeftPixel()[h * in.stride() + w] =
src16.TopLeftPixel()[h * src16.stride() + w] -
dst16.TopLeftPixel()[h * dst16.stride() + w];
}
}
#endif // CONFIG_VP9_HIGHBITDEPTH
}
fwd_txfm_ref(in, &coeff, size_, tx_type_);
if (bit_depth_ == VPX_BITS_8) {
ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst.TopLeftPixel()));
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(coeff, CAST_TO_BYTEPTR(dst16.TopLeftPixel())));
#endif // CONFIG_VP9_HIGHBITDEPTH
}
for (int h = 0; h < size_; ++h) {
for (int w = 0; w < size_; ++w) {
int diff;
#if CONFIG_VP9_HIGHBITDEPTH
if (bit_depth_ != 8) {
diff = dst16.TopLeftPixel()[h * dst16.stride() + w] -
src16.TopLeftPixel()[h * src16.stride() + w];
} else {
#endif // CONFIG_VP9_HIGHBITDEPTH
diff = dst.TopLeftPixel()[h * dst.stride() + w] -
src.TopLeftPixel()[h * src.stride() + w];
#if CONFIG_VP9_HIGHBITDEPTH
}
#endif // CONFIG_VP9_HIGHBITDEPTH
const uint32_t error = diff * diff;
EXPECT_GE(static_cast<uint32_t>(limit), error)
<< "Error: " << size_ << "x" << size_ << " IDCT has error "
<< error << " at " << w << "," << h;
}
}
}
}
FhtFuncRef fwd_txfm_ref;
vpx_bit_depth_t bit_depth_;
int tx_type_;
int max_pixel_value_;
int size_;
};
class TransDCT : public TransTestBase,
public ::testing::TestWithParam<DctParam> {
public:
TransDCT() {
fwd_txfm_ref = fdct_ref;
fwd_txfm_ = GET_PARAM(0);
inv_txfm_ = GET_PARAM(1);
size_ = GET_PARAM(2);
tx_type_ = GET_PARAM(3);
bit_depth_ = GET_PARAM(4);
max_pixel_value_ = (1 << bit_depth_) - 1;
}
protected:
void RunFwdTxfm(const Buffer<int16_t> &in, Buffer<tran_low_t> *out) {
fwd_txfm_(in.TopLeftPixel(), out->TopLeftPixel(), in.stride());
}
void RunInvTxfm(const Buffer<tran_low_t> &in, uint8_t *out) {
inv_txfm_(in.TopLeftPixel(), out, in.stride());
}
FdctFunc fwd_txfm_;
IdctFunc inv_txfm_;
};
TEST_P(TransDCT, AccuracyCheck) { RunAccuracyCheck(1); }
TEST_P(TransDCT, CoeffCheck) { RunCoeffCheck(); }
TEST_P(TransDCT, MemCheck) { RunMemCheck(); }
TEST_P(TransDCT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
C, TransDCT,
::testing::Values(
make_tuple(&vpx_highbd_fdct32x32_c, &idct32x32_10, 32, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct32x32_c, &idct32x32_12, 32, 0, VPX_BITS_10),
make_tuple(&vpx_fdct32x32_c, &vpx_idct32x32_1024_add_c, 32, 0,
VPX_BITS_8),
make_tuple(&vpx_highbd_fdct16x16_c, &idct16x16_10, 16, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct16x16_c, &idct16x16_12, 16, 0, VPX_BITS_10),
make_tuple(&vpx_fdct16x16_c, &vpx_idct16x16_256_add_c, 16, 0,
VPX_BITS_8),
make_tuple(&vpx_highbd_fdct8x8_c, &idct8x8_10, 8, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct8x8_c, &idct8x8_12, 8, 0, VPX_BITS_10),
make_tuple(&vpx_fdct8x8_c, &vpx_idct8x8_64_add_c, 8, 0, VPX_BITS_8),
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_10, 4, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12, 4, 0, VPX_BITS_12),
make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 4, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, TransDCT,
::testing::Values(
make_tuple(&vpx_fdct32x32_c, &vpx_idct32x32_1024_add_c, 32, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct16x16_c, &vpx_idct16x16_256_add_c, 16, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct8x8_c, &vpx_idct8x8_64_add_c, 8, 0, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 4, 0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_SSE2
#if !CONFIG_EMULATE_HARDWARE
#if CONFIG_VP9_HIGHBITDEPTH
/* TODO:(johannkoenig) Determine why these fail AccuracyCheck
make_tuple(&vpx_highbd_fdct32x32_sse2, &idct32x32_12, 32, 0, VPX_BITS_12),
make_tuple(&vpx_highbd_fdct16x16_sse2, &idct16x16_12, 16, 0, VPX_BITS_12),
*/
INSTANTIATE_TEST_CASE_P(
SSE2, TransDCT,
::testing::Values(
make_tuple(&vpx_highbd_fdct32x32_sse2, &idct32x32_10, 32, 0,
VPX_BITS_10),
make_tuple(&vpx_fdct32x32_sse2, &vpx_idct32x32_1024_add_sse2, 32, 0,
VPX_BITS_8),
make_tuple(&vpx_highbd_fdct16x16_sse2, &idct16x16_10, 16, 0,
VPX_BITS_10),
make_tuple(&vpx_fdct16x16_sse2, &vpx_idct16x16_256_add_sse2, 16, 0,
VPX_BITS_8),
make_tuple(&vpx_highbd_fdct8x8_sse2, &idct8x8_10, 8, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct8x8_sse2, &idct8x8_12, 8, 0, VPX_BITS_12),
make_tuple(&vpx_fdct8x8_sse2, &vpx_idct8x8_64_add_sse2, 8, 0,
VPX_BITS_8),
make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_10, 4, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_12, 4, 0, VPX_BITS_12),
make_tuple(&vpx_fdct4x4_sse2, &vpx_idct4x4_16_add_sse2, 4, 0,
VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
SSE2, TransDCT,
::testing::Values(make_tuple(&vpx_fdct32x32_sse2,
&vpx_idct32x32_1024_add_sse2, 32, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct16x16_sse2,
&vpx_idct16x16_256_add_sse2, 16, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct8x8_sse2, &vpx_idct8x8_64_add_sse2, 8,
0, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_sse2, &vpx_idct4x4_16_add_sse2, 4,
0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#endif // !CONFIG_EMULATE_HARDWARE
#endif // HAVE_SSE2
#if !CONFIG_VP9_HIGHBITDEPTH
#if HAVE_SSSE3 && !CONFIG_EMULATE_HARDWARE
#if !ARCH_X86_64
// TODO(johannkoenig): high bit depth fdct8x8.
INSTANTIATE_TEST_CASE_P(
SSSE3, TransDCT,
::testing::Values(make_tuple(&vpx_fdct32x32_c, &vpx_idct32x32_1024_add_sse2,
32, 0, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_c, &vpx_idct8x8_64_add_sse2, 8, 0,
VPX_BITS_8)));
#else
// vpx_fdct8x8_ssse3 is only available in 64 bit builds.
INSTANTIATE_TEST_CASE_P(
SSSE3, TransDCT,
::testing::Values(make_tuple(&vpx_fdct32x32_c, &vpx_idct32x32_1024_add_sse2,
32, 0, VPX_BITS_8),
make_tuple(&vpx_fdct8x8_ssse3, &vpx_idct8x8_64_add_sse2,
8, 0, VPX_BITS_8)));
#endif // !ARCH_X86_64
#endif // HAVE_SSSE3 && !CONFIG_EMULATE_HARDWARE
#endif // !CONFIG_VP9_HIGHBITDEPTH
#if !CONFIG_VP9_HIGHBITDEPTH && HAVE_AVX2 && !CONFIG_EMULATE_HARDWARE
// TODO(johannkoenig): high bit depth fdct32x32.
INSTANTIATE_TEST_CASE_P(
AVX2, TransDCT, ::testing::Values(make_tuple(&vpx_fdct32x32_avx2,
&vpx_idct32x32_1024_add_sse2,
32, 0, VPX_BITS_8)));
#endif // !CONFIG_VP9_HIGHBITDEPTH && HAVE_AVX2 && !CONFIG_EMULATE_HARDWARE
#if HAVE_NEON
#if !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
NEON, TransDCT,
::testing::Values(make_tuple(&vpx_fdct32x32_neon,
&vpx_idct32x32_1024_add_neon, 32, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct16x16_neon,
&vpx_idct16x16_256_add_neon, 16, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct8x8_neon, &vpx_idct8x8_64_add_neon, 8,
0, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_neon, &vpx_idct4x4_16_add_neon, 4,
0, VPX_BITS_8)));
#endif // !CONFIG_EMULATE_HARDWARE
#endif // HAVE_NEON
#if HAVE_MSA
#if !CONFIG_VP9_HIGHBITDEPTH
#if !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
MSA, TransDCT,
::testing::Values(
make_tuple(&vpx_fdct32x32_msa, &vpx_idct32x32_1024_add_msa, 32, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct16x16_msa, &vpx_idct16x16_256_add_msa, 16, 0,
VPX_BITS_8),
make_tuple(&vpx_fdct8x8_msa, &vpx_idct8x8_64_add_msa, 8, 0, VPX_BITS_8),
make_tuple(&vpx_fdct4x4_msa, &vpx_idct4x4_16_add_msa, 4, 0,
VPX_BITS_8)));
#endif // !CONFIG_EMULATE_HARDWARE
#endif // !CONFIG_VP9_HIGHBITDEPTH
#endif // HAVE_MSA
#if HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(VSX, TransDCT,
::testing::Values(make_tuple(&vpx_fdct4x4_c,
&vpx_idct4x4_16_add_vsx, 4,
0, VPX_BITS_8)));
#endif // HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
class TransHT : public TransTestBase, public ::testing::TestWithParam<HtParam> {
public:
TransHT() {
fwd_txfm_ref = fht_ref;
fwd_txfm_ = GET_PARAM(0);
inv_txfm_ = GET_PARAM(1);
size_ = GET_PARAM(2);
tx_type_ = GET_PARAM(3);
bit_depth_ = GET_PARAM(4);
max_pixel_value_ = (1 << bit_depth_) - 1;
}
protected:
void RunFwdTxfm(const Buffer<int16_t> &in, Buffer<tran_low_t> *out) {
fwd_txfm_(in.TopLeftPixel(), out->TopLeftPixel(), in.stride(), tx_type_);
}
void RunInvTxfm(const Buffer<tran_low_t> &in, uint8_t *out) {
inv_txfm_(in.TopLeftPixel(), out, in.stride(), tx_type_);
}
FhtFunc fwd_txfm_;
IhtFunc inv_txfm_;
};
TEST_P(TransHT, AccuracyCheck) { RunAccuracyCheck(1); }
TEST_P(TransHT, CoeffCheck) { RunCoeffCheck(); }
TEST_P(TransHT, MemCheck) { RunMemCheck(); }
TEST_P(TransHT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
/* TODO:(johannkoenig) Determine why these fail AccuracyCheck
make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_12, 16, 0, VPX_BITS_12),
make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_12, 16, 1, VPX_BITS_12),
make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_12, 16, 2, VPX_BITS_12),
make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_12, 16, 3, VPX_BITS_12),
*/
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
C, TransHT,
::testing::Values(
make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_10, 16, 0, VPX_BITS_10),
make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_10, 16, 1, VPX_BITS_10),
make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_10, 16, 2, VPX_BITS_10),
make_tuple(&vp9_highbd_fht16x16_c, &iht16x16_10, 16, 3, VPX_BITS_10),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 16, 0, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 16, 1, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 16, 2, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 16, 3, VPX_BITS_8),
make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_10, 8, 0, VPX_BITS_10),
make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_10, 8, 1, VPX_BITS_10),
make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_10, 8, 2, VPX_BITS_10),
make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_10, 8, 3, VPX_BITS_10),
make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_12, 8, 0, VPX_BITS_12),
make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_12, 8, 1, VPX_BITS_12),
make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_12, 8, 2, VPX_BITS_12),
make_tuple(&vp9_highbd_fht8x8_c, &iht8x8_12, 8, 3, VPX_BITS_12),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 8, 0, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 8, 1, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 8, 2, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 8, 3, VPX_BITS_8),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 4, 0, VPX_BITS_10),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 4, 1, VPX_BITS_10),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 4, 2, VPX_BITS_10),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 4, 3, VPX_BITS_10),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 4, 0, VPX_BITS_12),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 4, 1, VPX_BITS_12),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 4, 2, VPX_BITS_12),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 4, 3, VPX_BITS_12),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 4, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 4, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 4, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 4, 3, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, TransHT,
::testing::Values(
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 16, 0, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 16, 1, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 16, 2, VPX_BITS_8),
make_tuple(&vp9_fht16x16_c, &vp9_iht16x16_256_add_c, 16, 3, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 8, 0, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 8, 1, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 8, 2, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 8, 3, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 4, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 4, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 4, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 4, 3, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(
SSE2, TransHT,
::testing::Values(
make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 16, 0,
VPX_BITS_8),
make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 16, 1,
VPX_BITS_8),
make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 16, 2,
VPX_BITS_8),
make_tuple(&vp9_fht16x16_sse2, &vp9_iht16x16_256_add_sse2, 16, 3,
VPX_BITS_8),
make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 8, 0, VPX_BITS_8),
make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 8, 1, VPX_BITS_8),
make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 8, 2, VPX_BITS_8),
make_tuple(&vp9_fht8x8_sse2, &vp9_iht8x8_64_add_sse2, 8, 3, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 4, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 4, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 4, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 4, 3,
VPX_BITS_8)));
#endif // HAVE_SSE2
class TransWHT : public TransTestBase,
public ::testing::TestWithParam<DctParam> {
public:
TransWHT() {
fwd_txfm_ref = fwht_ref;
fwd_txfm_ = GET_PARAM(0);
inv_txfm_ = GET_PARAM(1);
size_ = GET_PARAM(2);
tx_type_ = GET_PARAM(3);
bit_depth_ = GET_PARAM(4);
max_pixel_value_ = (1 << bit_depth_) - 1;
}
protected:
void RunFwdTxfm(const Buffer<int16_t> &in, Buffer<tran_low_t> *out) {
fwd_txfm_(in.TopLeftPixel(), out->TopLeftPixel(), in.stride());
}
void RunInvTxfm(const Buffer<tran_low_t> &in, uint8_t *out) {
inv_txfm_(in.TopLeftPixel(), out, in.stride());
}
FdctFunc fwd_txfm_;
IdctFunc inv_txfm_;
};
TEST_P(TransWHT, AccuracyCheck) { RunAccuracyCheck(0); }
TEST_P(TransWHT, CoeffCheck) { RunCoeffCheck(); }
TEST_P(TransWHT, MemCheck) { RunMemCheck(); }
TEST_P(TransWHT, InvAccuracyCheck) { RunInvAccuracyCheck(0); }
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
C, TransWHT,
::testing::Values(
make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_10, 4, 0, VPX_BITS_10),
make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_12, 4, 0, VPX_BITS_12),
make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_c, 4, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(C, TransWHT,
::testing::Values(make_tuple(&vp9_fwht4x4_c,
&vpx_iwht4x4_16_add_c, 4,
0, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(SSE2, TransWHT,
::testing::Values(make_tuple(&vp9_fwht4x4_sse2,
&vpx_iwht4x4_16_add_sse2,
4, 0, VPX_BITS_8)));
#endif // HAVE_SSE2
} // namespace

View File

@ -172,4 +172,21 @@ TEST(DecodeAPI, Vp9PeekSI) {
}
#endif // CONFIG_VP9_DECODER
TEST(DecodeAPI, HighBitDepthCapability) {
// VP8 should not claim VP9 HBD as a capability.
#if CONFIG_VP8_DECODER
const vpx_codec_caps_t vp8_caps = vpx_codec_get_caps(&vpx_codec_vp8_dx_algo);
EXPECT_EQ(vp8_caps & VPX_CODEC_CAP_HIGHBITDEPTH, 0);
#endif
#if CONFIG_VP9_DECODER
const vpx_codec_caps_t vp9_caps = vpx_codec_get_caps(&vpx_codec_vp9_dx_algo);
#if CONFIG_VP9_HIGHBITDEPTH
EXPECT_EQ(vp9_caps & VPX_CODEC_CAP_HIGHBITDEPTH, VPX_CODEC_CAP_HIGHBITDEPTH);
#else
EXPECT_EQ(vp9_caps & VPX_CODEC_CAP_HIGHBITDEPTH, 0);
#endif
#endif
}
} // namespace

View File

@ -0,0 +1,124 @@
/*
* Copyright (c) 2016 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <string>
#include "test/codec_factory.h"
#include "test/decode_test_driver.h"
#include "test/ivf_video_source.h"
#include "test/test_vectors.h"
#include "test/util.h"
namespace {
const unsigned int kNumFrames = 19;
class DecodeSvcTest : public ::libvpx_test::DecoderTest,
public ::libvpx_test::CodecTestWithParam<const char *> {
protected:
DecodeSvcTest() : DecoderTest(GET_PARAM(::libvpx_test::kCodecFactoryParam)) {}
virtual ~DecodeSvcTest() {}
virtual void PreDecodeFrameHook(
const libvpx_test::CompressedVideoSource &video,
libvpx_test::Decoder *decoder) {
if (video.frame_number() == 0)
decoder->Control(VP9_DECODE_SVC_SPATIAL_LAYER, spatial_layer_);
}
virtual void DecompressedFrameHook(const vpx_image_t &img,
const unsigned int frame_number) {
ASSERT_EQ(img.d_w, width_);
ASSERT_EQ(img.d_h, height_);
total_frames_ = frame_number;
}
int spatial_layer_;
unsigned int width_;
unsigned int height_;
unsigned int total_frames_;
};
// SVC test vector is 1280x720, with 3 spatial layers, and 20 frames.
// Decode the SVC test vector, which has 3 spatial layers, and decode up to
// spatial layer 0. Verify the resolution of each decoded frame and the total
// number of frames decoded. This results in 1/4x1/4 resolution (320x180).
TEST_P(DecodeSvcTest, DecodeSvcTestUpToSpatialLayer0) {
const std::string filename = GET_PARAM(1);
testing::internal::scoped_ptr<libvpx_test::CompressedVideoSource> video;
video.reset(new libvpx_test::IVFVideoSource(filename));
ASSERT_TRUE(video.get() != NULL);
video->Init();
total_frames_ = 0;
spatial_layer_ = 0;
width_ = 320;
height_ = 180;
ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
ASSERT_EQ(total_frames_, kNumFrames);
}
// Decode the SVC test vector, which has 3 spatial layers, and decode up to
// spatial layer 1. Verify the resolution of each decoded frame and the total
// number of frames decoded. This results in 1/2x1/2 resolution (640x360).
TEST_P(DecodeSvcTest, DecodeSvcTestUpToSpatialLayer1) {
const std::string filename = GET_PARAM(1);
testing::internal::scoped_ptr<libvpx_test::CompressedVideoSource> video;
video.reset(new libvpx_test::IVFVideoSource(filename));
ASSERT_TRUE(video.get() != NULL);
video->Init();
total_frames_ = 0;
spatial_layer_ = 1;
width_ = 640;
height_ = 360;
ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
ASSERT_EQ(total_frames_, kNumFrames);
}
// Decode the SVC test vector, which has 3 spatial layers, and decode up to
// spatial layer 2. Verify the resolution of each decoded frame and the total
// number of frames decoded. This results in the full resolution (1280x720).
TEST_P(DecodeSvcTest, DecodeSvcTestUpToSpatialLayer2) {
const std::string filename = GET_PARAM(1);
testing::internal::scoped_ptr<libvpx_test::CompressedVideoSource> video;
video.reset(new libvpx_test::IVFVideoSource(filename));
ASSERT_TRUE(video.get() != NULL);
video->Init();
total_frames_ = 0;
spatial_layer_ = 2;
width_ = 1280;
height_ = 720;
ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
ASSERT_EQ(total_frames_, kNumFrames);
}
// Decode the SVC test vector, which has 3 spatial layers, and decode up to
// spatial layer 10. Verify the resolution of each decoded frame and the total
// number of frames decoded. This is beyond the number of spatial layers, so
// the decoding should result in the full resolution (1280x720).
TEST_P(DecodeSvcTest, DecodeSvcTestUpToSpatialLayer10) {
const std::string filename = GET_PARAM(1);
testing::internal::scoped_ptr<libvpx_test::CompressedVideoSource> video;
video.reset(new libvpx_test::IVFVideoSource(filename));
ASSERT_TRUE(video.get() != NULL);
video->Init();
total_frames_ = 0;
spatial_layer_ = 10;
width_ = 1280;
height_ = 720;
ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
ASSERT_EQ(total_frames_, kNumFrames);
}
VP9_INSTANTIATE_TEST_CASE(
DecodeSvcTest, ::testing::ValuesIn(libvpx_test::kVP9TestVectorsSvc,
libvpx_test::kVP9TestVectorsSvc +
libvpx_test::kNumVP9TestVectorsSvc));
} // namespace

View File

@ -53,13 +53,13 @@ void DecoderTest::HandlePeekResult(Decoder *const decoder,
* pass it is not a keyframe, so we only expect VPX_CODEC_OK on the first
* frame, which must be a keyframe. */
if (video->frame_number() == 0)
ASSERT_EQ(VPX_CODEC_OK, res_peek) << "Peek return failed: "
<< vpx_codec_err_to_string(res_peek);
ASSERT_EQ(VPX_CODEC_OK, res_peek)
<< "Peek return failed: " << vpx_codec_err_to_string(res_peek);
} else {
/* The Vp9 implementation of PeekStream returns an error only if the
* data passed to it isn't a valid Vp9 chunk. */
ASSERT_EQ(VPX_CODEC_OK, res_peek) << "Peek return failed: "
<< vpx_codec_err_to_string(res_peek);
ASSERT_EQ(VPX_CODEC_OK, res_peek)
<< "Peek return failed: " << vpx_codec_err_to_string(res_peek);
}
}

View File

@ -62,4 +62,134 @@ TEST(EncodeAPI, InvalidParams) {
}
}
TEST(EncodeAPI, HighBitDepthCapability) {
// VP8 should not claim VP9 HBD as a capability.
#if CONFIG_VP8_ENCODER
const vpx_codec_caps_t vp8_caps = vpx_codec_get_caps(&vpx_codec_vp8_cx_algo);
EXPECT_EQ(vp8_caps & VPX_CODEC_CAP_HIGHBITDEPTH, 0);
#endif
#if CONFIG_VP9_ENCODER
const vpx_codec_caps_t vp9_caps = vpx_codec_get_caps(&vpx_codec_vp9_cx_algo);
#if CONFIG_VP9_HIGHBITDEPTH
EXPECT_EQ(vp9_caps & VPX_CODEC_CAP_HIGHBITDEPTH, VPX_CODEC_CAP_HIGHBITDEPTH);
#else
EXPECT_EQ(vp9_caps & VPX_CODEC_CAP_HIGHBITDEPTH, 0);
#endif
#endif
}
#if CONFIG_VP8_ENCODER
TEST(EncodeAPI, ImageSizeSetting) {
const int width = 711;
const int height = 360;
const int bps = 12;
vpx_image_t img;
vpx_codec_ctx_t enc;
vpx_codec_enc_cfg_t cfg;
uint8_t *img_buf = reinterpret_cast<uint8_t *>(
calloc(width * height * bps / 8, sizeof(*img_buf)));
vpx_codec_enc_config_default(vpx_codec_vp8_cx(), &cfg, 0);
cfg.g_w = width;
cfg.g_h = height;
vpx_img_wrap(&img, VPX_IMG_FMT_I420, width, height, 1, img_buf);
vpx_codec_enc_init(&enc, vpx_codec_vp8_cx(), &cfg, 0);
EXPECT_EQ(VPX_CODEC_OK, vpx_codec_encode(&enc, &img, 0, 1, 0, 0));
free(img_buf);
vpx_codec_destroy(&enc);
}
#endif
// Set up 2 spatial streams with 2 temporal layers per stream, and generate
// invalid configuration by setting the temporal layer rate allocation
// (ts_target_bitrate[]) to 0 for both layers. This should fail independent of
// CONFIG_MULTI_RES_ENCODING.
TEST(EncodeAPI, MultiResEncode) {
static const vpx_codec_iface_t *kCodecs[] = {
#if CONFIG_VP8_ENCODER
&vpx_codec_vp8_cx_algo,
#endif
#if CONFIG_VP9_ENCODER
&vpx_codec_vp9_cx_algo,
#endif
};
const int width = 1280;
const int height = 720;
const int width_down = width / 2;
const int height_down = height / 2;
const int target_bitrate = 1000;
const int framerate = 30;
for (int c = 0; c < NELEMENTS(kCodecs); ++c) {
const vpx_codec_iface_t *const iface = kCodecs[c];
vpx_codec_ctx_t enc[2];
vpx_codec_enc_cfg_t cfg[2];
vpx_rational_t dsf[2] = { { 2, 1 }, { 2, 1 } };
memset(enc, 0, sizeof(enc));
for (int i = 0; i < 2; i++) {
vpx_codec_enc_config_default(iface, &cfg[i], 0);
}
/* Highest-resolution encoder settings */
cfg[0].g_w = width;
cfg[0].g_h = height;
cfg[0].rc_dropframe_thresh = 0;
cfg[0].rc_end_usage = VPX_CBR;
cfg[0].rc_resize_allowed = 0;
cfg[0].rc_min_quantizer = 2;
cfg[0].rc_max_quantizer = 56;
cfg[0].rc_undershoot_pct = 100;
cfg[0].rc_overshoot_pct = 15;
cfg[0].rc_buf_initial_sz = 500;
cfg[0].rc_buf_optimal_sz = 600;
cfg[0].rc_buf_sz = 1000;
cfg[0].g_error_resilient = 1; /* Enable error resilient mode */
cfg[0].g_lag_in_frames = 0;
cfg[0].kf_mode = VPX_KF_AUTO;
cfg[0].kf_min_dist = 3000;
cfg[0].kf_max_dist = 3000;
cfg[0].rc_target_bitrate = target_bitrate; /* Set target bitrate */
cfg[0].g_timebase.num = 1; /* Set fps */
cfg[0].g_timebase.den = framerate;
memcpy(&cfg[1], &cfg[0], sizeof(cfg[0]));
cfg[1].rc_target_bitrate = 500;
cfg[1].g_w = width_down;
cfg[1].g_h = height_down;
for (int i = 0; i < 2; i++) {
cfg[i].ts_number_layers = 2;
cfg[i].ts_periodicity = 2;
cfg[i].ts_rate_decimator[0] = 2;
cfg[i].ts_rate_decimator[1] = 1;
cfg[i].ts_layer_id[0] = 0;
cfg[i].ts_layer_id[1] = 1;
// Invalid parameters.
cfg[i].ts_target_bitrate[0] = 0;
cfg[i].ts_target_bitrate[1] = 0;
}
// VP9 should report incapable, VP8 invalid for all configurations.
const char kVP9Name[] = "WebM Project VP9";
const bool is_vp9 = strncmp(kVP9Name, vpx_codec_iface_name(iface),
sizeof(kVP9Name) - 1) == 0;
EXPECT_EQ(is_vp9 ? VPX_CODEC_INCAPABLE : VPX_CODEC_INVALID_PARAM,
vpx_codec_enc_init_multi(&enc[0], iface, &cfg[0], 2, 0, &dsf[0]));
for (int i = 0; i < 2; i++) {
vpx_codec_destroy(&enc[i]);
}
}
}
} // namespace

View File

@ -201,6 +201,8 @@ void EncoderTest::RunLoop(VideoSource *video) {
PreEncodeFrameHook(video, encoder.get());
encoder->EncodeFrame(video, frame_flags_);
PostEncodeFrameHook(encoder.get());
CxDataIterator iter = encoder->GetCxData();
bool has_cxdata = false;
@ -226,6 +228,8 @@ void EncoderTest::RunLoop(VideoSource *video) {
case VPX_CODEC_PSNR_PKT: PSNRPktHook(pkt); break;
case VPX_CODEC_STATS_PKT: StatsPktHook(pkt); break;
default: break;
}
}

View File

@ -139,6 +139,13 @@ class Encoder {
}
#endif
#if CONFIG_VP8_ENCODER
void Control(int ctrl_id, vpx_roi_map_t *arg) {
const vpx_codec_err_t res = vpx_codec_control_(&encoder_, ctrl_id, arg);
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
}
#endif
void Config(const vpx_codec_enc_cfg_t *cfg) {
const vpx_codec_err_t res = vpx_codec_enc_config_set(&encoder_, cfg);
ASSERT_EQ(VPX_CODEC_OK, res) << EncoderError();
@ -212,12 +219,17 @@ class EncoderTest {
virtual void PreEncodeFrameHook(VideoSource * /*video*/,
Encoder * /*encoder*/) {}
virtual void PostEncodeFrameHook(Encoder * /*encoder*/) {}
// Hook to be called on every compressed data packet.
virtual void FramePktHook(const vpx_codec_cx_pkt_t * /*pkt*/) {}
// Hook to be called on every PSNR packet.
virtual void PSNRPktHook(const vpx_codec_cx_pkt_t * /*pkt*/) {}
// Hook to be called on every first pass stats packet.
virtual void StatsPktHook(const vpx_codec_cx_pkt_t * /*pkt*/) {}
// Hook to determine whether the encode loop should continue.
virtual bool Continue() const {
return !(::testing::Test::HasFatalFailure() || abort_);

View File

@ -90,8 +90,7 @@ class ErrorResilienceTestLarge
return frame_flags;
}
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video,
::libvpx_test::Encoder * /*encoder*/) {
virtual void PreEncodeFrameHook(libvpx_test::VideoSource *video) {
frame_flags_ &=
~(VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF);
// For temporal layer case.

View File

@ -15,7 +15,7 @@
example_tests=$(ls $(dirname $0)/*.sh)
# List of script names to exclude.
exclude_list="examples tools_common"
exclude_list="examples stress tools_common"
# Filter out the scripts in $exclude_list.
for word in ${exclude_list}; do

View File

@ -34,7 +34,8 @@ struct ExternalFrameBuffer {
// Class to manipulate a list of external frame buffers.
class ExternalFrameBufferList {
public:
ExternalFrameBufferList() : num_buffers_(0), ext_fb_list_(NULL) {}
ExternalFrameBufferList()
: num_buffers_(0), num_used_buffers_(0), ext_fb_list_(NULL) {}
virtual ~ExternalFrameBufferList() {
for (int i = 0; i < num_buffers_; ++i) {
@ -71,6 +72,8 @@ class ExternalFrameBufferList {
}
SetFrameBuffer(idx, fb);
num_used_buffers_++;
return 0;
}
@ -106,6 +109,7 @@ class ExternalFrameBufferList {
}
EXPECT_EQ(1, ext_fb->in_use);
ext_fb->in_use = 0;
num_used_buffers_--;
return 0;
}
@ -121,6 +125,8 @@ class ExternalFrameBufferList {
}
}
int num_used_buffers() const { return num_used_buffers_; }
private:
// Returns the index of the first free frame buffer. Returns |num_buffers_|
// if there are no free frame buffers.
@ -145,6 +151,7 @@ class ExternalFrameBufferList {
}
int num_buffers_;
int num_used_buffers_;
ExternalFrameBuffer *ext_fb_list_;
};
@ -220,8 +227,8 @@ class ExternalFrameBufferMD5Test
void OpenMD5File(const std::string &md5_file_name_) {
md5_file_ = libvpx_test::OpenTestDataFile(md5_file_name_);
ASSERT_TRUE(md5_file_ != NULL) << "Md5 file open failed. Filename: "
<< md5_file_name_;
ASSERT_TRUE(md5_file_ != NULL)
<< "Md5 file open failed. Filename: " << md5_file_name_;
}
virtual void DecompressedFrameHook(const vpx_image_t &img,
@ -273,6 +280,7 @@ class ExternalFrameBufferMD5Test
#if CONFIG_WEBM_IO
const char kVP9TestFile[] = "vp90-2-02-size-lf-1920x1080.webm";
const char kVP9NonRefTestFile[] = "vp90-2-22-svc_1280x720_1.webm";
// Class for testing passing in external frame buffers to libvpx.
class ExternalFrameBufferTest : public ::testing::Test {
@ -292,7 +300,9 @@ class ExternalFrameBufferTest : public ::testing::Test {
virtual void TearDown() {
delete decoder_;
decoder_ = NULL;
delete video_;
video_ = NULL;
}
// Passes the external frame buffer information to libvpx.
@ -325,7 +335,7 @@ class ExternalFrameBufferTest : public ::testing::Test {
return VPX_CODEC_OK;
}
private:
protected:
void CheckDecodedFrames() {
libvpx_test::DxDataIterator dec_iter = decoder_->GetDxData();
const vpx_image_t *img = NULL;
@ -341,6 +351,25 @@ class ExternalFrameBufferTest : public ::testing::Test {
int num_buffers_;
ExternalFrameBufferList fb_list_;
};
class ExternalFrameBufferNonRefTest : public ExternalFrameBufferTest {
protected:
virtual void SetUp() {
video_ = new libvpx_test::WebMVideoSource(kVP9NonRefTestFile);
ASSERT_TRUE(video_ != NULL);
video_->Init();
video_->Begin();
vpx_codec_dec_cfg_t cfg = vpx_codec_dec_cfg_t();
decoder_ = new libvpx_test::VP9Decoder(cfg, 0);
ASSERT_TRUE(decoder_ != NULL);
}
virtual void CheckFrameBufferRelease() {
TearDown();
ASSERT_EQ(0, fb_list_.num_used_buffers());
}
};
#endif // CONFIG_WEBM_IO
// This test runs through the set of test vectors, and decodes them.
@ -419,6 +448,8 @@ TEST_F(ExternalFrameBufferTest, NotEnoughBuffers) {
SetFrameBufferFunctions(num_buffers, get_vp9_frame_buffer,
release_vp9_frame_buffer));
ASSERT_EQ(VPX_CODEC_OK, DecodeOneFrame());
// Only run this on long clips. Decoding a very short clip will return
// VPX_CODEC_OK even with only 2 buffers.
ASSERT_EQ(VPX_CODEC_MEM_ERROR, DecodeRemainingFrames());
}
@ -467,6 +498,15 @@ TEST_F(ExternalFrameBufferTest, SetAfterDecode) {
SetFrameBufferFunctions(num_buffers, get_vp9_frame_buffer,
release_vp9_frame_buffer));
}
TEST_F(ExternalFrameBufferNonRefTest, ReleaseNonRefFrameBuffer) {
const int num_buffers = VP9_MAXIMUM_REF_BUFFERS + VPX_MAXIMUM_WORK_BUFFERS;
ASSERT_EQ(VPX_CODEC_OK,
SetFrameBufferFunctions(num_buffers, get_vp9_frame_buffer,
release_vp9_frame_buffer));
ASSERT_EQ(VPX_CODEC_OK, DecodeRemainingFrames());
CheckFrameBufferRelease();
}
#endif // CONFIG_WEBM_IO
VP9_INSTANTIATE_TEST_CASE(

View File

@ -1,512 +0,0 @@
/*
* Copyright (c) 2012 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <math.h>
#include <stdlib.h>
#include <string.h>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vp9_rtcd.h"
#include "./vpx_dsp_rtcd.h"
#include "test/acm_random.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "test/util.h"
#include "vp9/common/vp9_entropy.h"
#include "vpx/vpx_codec.h"
#include "vpx/vpx_integer.h"
#include "vpx_ports/mem.h"
using libvpx_test::ACMRandom;
namespace {
const int kNumCoeffs = 16;
typedef void (*FdctFunc)(const int16_t *in, tran_low_t *out, int stride);
typedef void (*IdctFunc)(const tran_low_t *in, uint8_t *out, int stride);
typedef void (*FhtFunc)(const int16_t *in, tran_low_t *out, int stride,
int tx_type);
typedef void (*IhtFunc)(const tran_low_t *in, uint8_t *out, int stride,
int tx_type);
typedef std::tr1::tuple<FdctFunc, IdctFunc, int, vpx_bit_depth_t> Dct4x4Param;
typedef std::tr1::tuple<FhtFunc, IhtFunc, int, vpx_bit_depth_t> Ht4x4Param;
void fdct4x4_ref(const int16_t *in, tran_low_t *out, int stride,
int /*tx_type*/) {
vpx_fdct4x4_c(in, out, stride);
}
void fht4x4_ref(const int16_t *in, tran_low_t *out, int stride, int tx_type) {
vp9_fht4x4_c(in, out, stride, tx_type);
}
void fwht4x4_ref(const int16_t *in, tran_low_t *out, int stride,
int /*tx_type*/) {
vp9_fwht4x4_c(in, out, stride);
}
#if CONFIG_VP9_HIGHBITDEPTH
void idct4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct4x4_16_add_c(in, out, stride, 10);
}
void idct4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct4x4_16_add_c(in, out, stride, 12);
}
void iht4x4_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
vp9_highbd_iht4x4_16_add_c(in, out, stride, tx_type, 10);
}
void iht4x4_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
vp9_highbd_iht4x4_16_add_c(in, out, stride, tx_type, 12);
}
void iwht4x4_10(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_iwht4x4_16_add_c(in, out, stride, 10);
}
void iwht4x4_12(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_iwht4x4_16_add_c(in, out, stride, 12);
}
#if HAVE_SSE2
void idct4x4_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct4x4_16_add_sse2(in, out, stride, 10);
}
void idct4x4_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct4x4_16_add_sse2(in, out, stride, 12);
}
#endif // HAVE_SSE2
#endif // CONFIG_VP9_HIGHBITDEPTH
class Trans4x4TestBase {
public:
virtual ~Trans4x4TestBase() {}
protected:
virtual void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) = 0;
virtual void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) = 0;
void RunAccuracyCheck(int limit) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
uint32_t max_error = 0;
int64_t total_error = 0;
const int count_test_block = 10000;
for (int i = 0; i < count_test_block; ++i) {
DECLARE_ALIGNED(16, int16_t, test_input_block[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, test_temp_block[kNumCoeffs]);
DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
#if CONFIG_VP9_HIGHBITDEPTH
DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
#endif
// Initialize a test block with input range [-255, 255].
for (int j = 0; j < kNumCoeffs; ++j) {
if (bit_depth_ == VPX_BITS_8) {
src[j] = rnd.Rand8();
dst[j] = rnd.Rand8();
test_input_block[j] = src[j] - dst[j];
#if CONFIG_VP9_HIGHBITDEPTH
} else {
src16[j] = rnd.Rand16() & mask_;
dst16[j] = rnd.Rand16() & mask_;
test_input_block[j] = src16[j] - dst16[j];
#endif
}
}
ASM_REGISTER_STATE_CHECK(
RunFwdTxfm(test_input_block, test_temp_block, pitch_));
if (bit_depth_ == VPX_BITS_8) {
ASM_REGISTER_STATE_CHECK(RunInvTxfm(test_temp_block, dst, pitch_));
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(test_temp_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
#endif
}
for (int j = 0; j < kNumCoeffs; ++j) {
#if CONFIG_VP9_HIGHBITDEPTH
const int diff =
bit_depth_ == VPX_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
#else
ASSERT_EQ(VPX_BITS_8, bit_depth_);
const int diff = dst[j] - src[j];
#endif
const uint32_t error = diff * diff;
if (max_error < error) max_error = error;
total_error += error;
}
}
EXPECT_GE(static_cast<uint32_t>(limit), max_error)
<< "Error: 4x4 FHT/IHT has an individual round trip error > " << limit;
EXPECT_GE(count_test_block * limit, total_error)
<< "Error: 4x4 FHT/IHT has average round trip error > " << limit
<< " per block";
}
void RunCoeffCheck() {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 5000;
DECLARE_ALIGNED(16, int16_t, input_block[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, output_block[kNumCoeffs]);
for (int i = 0; i < count_test_block; ++i) {
// Initialize a test block with input range [-mask_, mask_].
for (int j = 0; j < kNumCoeffs; ++j) {
input_block[j] = (rnd.Rand16() & mask_) - (rnd.Rand16() & mask_);
}
fwd_txfm_ref(input_block, output_ref_block, pitch_, tx_type_);
ASM_REGISTER_STATE_CHECK(RunFwdTxfm(input_block, output_block, pitch_));
// The minimum quant value is 4.
for (int j = 0; j < kNumCoeffs; ++j)
EXPECT_EQ(output_block[j], output_ref_block[j]);
}
}
void RunMemCheck() {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 5000;
DECLARE_ALIGNED(16, int16_t, input_extreme_block[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, output_ref_block[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, output_block[kNumCoeffs]);
for (int i = 0; i < count_test_block; ++i) {
// Initialize a test block with input range [-mask_, mask_].
for (int j = 0; j < kNumCoeffs; ++j) {
input_extreme_block[j] = rnd.Rand8() % 2 ? mask_ : -mask_;
}
if (i == 0) {
for (int j = 0; j < kNumCoeffs; ++j) input_extreme_block[j] = mask_;
} else if (i == 1) {
for (int j = 0; j < kNumCoeffs; ++j) input_extreme_block[j] = -mask_;
}
fwd_txfm_ref(input_extreme_block, output_ref_block, pitch_, tx_type_);
ASM_REGISTER_STATE_CHECK(
RunFwdTxfm(input_extreme_block, output_block, pitch_));
// The minimum quant value is 4.
for (int j = 0; j < kNumCoeffs; ++j) {
EXPECT_EQ(output_block[j], output_ref_block[j]);
EXPECT_GE(4 * DCT_MAX_VALUE << (bit_depth_ - 8), abs(output_block[j]))
<< "Error: 4x4 FDCT has coefficient larger than 4*DCT_MAX_VALUE";
}
}
}
void RunInvAccuracyCheck(int limit) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
const int count_test_block = 1000;
DECLARE_ALIGNED(16, int16_t, in[kNumCoeffs]);
DECLARE_ALIGNED(16, tran_low_t, coeff[kNumCoeffs]);
DECLARE_ALIGNED(16, uint8_t, dst[kNumCoeffs]);
DECLARE_ALIGNED(16, uint8_t, src[kNumCoeffs]);
#if CONFIG_VP9_HIGHBITDEPTH
DECLARE_ALIGNED(16, uint16_t, dst16[kNumCoeffs]);
DECLARE_ALIGNED(16, uint16_t, src16[kNumCoeffs]);
#endif
for (int i = 0; i < count_test_block; ++i) {
// Initialize a test block with input range [-mask_, mask_].
for (int j = 0; j < kNumCoeffs; ++j) {
if (bit_depth_ == VPX_BITS_8) {
src[j] = rnd.Rand8();
dst[j] = rnd.Rand8();
in[j] = src[j] - dst[j];
#if CONFIG_VP9_HIGHBITDEPTH
} else {
src16[j] = rnd.Rand16() & mask_;
dst16[j] = rnd.Rand16() & mask_;
in[j] = src16[j] - dst16[j];
#endif
}
}
fwd_txfm_ref(in, coeff, pitch_, tx_type_);
if (bit_depth_ == VPX_BITS_8) {
ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, pitch_));
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), pitch_));
#endif
}
for (int j = 0; j < kNumCoeffs; ++j) {
#if CONFIG_VP9_HIGHBITDEPTH
const int diff =
bit_depth_ == VPX_BITS_8 ? dst[j] - src[j] : dst16[j] - src16[j];
#else
const int diff = dst[j] - src[j];
#endif
const uint32_t error = diff * diff;
EXPECT_GE(static_cast<uint32_t>(limit), error)
<< "Error: 4x4 IDCT has error " << error << " at index " << j;
}
}
}
int pitch_;
int tx_type_;
FhtFunc fwd_txfm_ref;
vpx_bit_depth_t bit_depth_;
int mask_;
};
class Trans4x4DCT : public Trans4x4TestBase,
public ::testing::TestWithParam<Dct4x4Param> {
public:
virtual ~Trans4x4DCT() {}
virtual void SetUp() {
fwd_txfm_ = GET_PARAM(0);
inv_txfm_ = GET_PARAM(1);
tx_type_ = GET_PARAM(2);
pitch_ = 4;
fwd_txfm_ref = fdct4x4_ref;
bit_depth_ = GET_PARAM(3);
mask_ = (1 << bit_depth_) - 1;
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
fwd_txfm_(in, out, stride);
}
void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
inv_txfm_(out, dst, stride);
}
FdctFunc fwd_txfm_;
IdctFunc inv_txfm_;
};
TEST_P(Trans4x4DCT, AccuracyCheck) { RunAccuracyCheck(1); }
TEST_P(Trans4x4DCT, CoeffCheck) { RunCoeffCheck(); }
TEST_P(Trans4x4DCT, MemCheck) { RunMemCheck(); }
TEST_P(Trans4x4DCT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
class Trans4x4HT : public Trans4x4TestBase,
public ::testing::TestWithParam<Ht4x4Param> {
public:
virtual ~Trans4x4HT() {}
virtual void SetUp() {
fwd_txfm_ = GET_PARAM(0);
inv_txfm_ = GET_PARAM(1);
tx_type_ = GET_PARAM(2);
pitch_ = 4;
fwd_txfm_ref = fht4x4_ref;
bit_depth_ = GET_PARAM(3);
mask_ = (1 << bit_depth_) - 1;
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
fwd_txfm_(in, out, stride, tx_type_);
}
void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
inv_txfm_(out, dst, stride, tx_type_);
}
FhtFunc fwd_txfm_;
IhtFunc inv_txfm_;
};
TEST_P(Trans4x4HT, AccuracyCheck) { RunAccuracyCheck(1); }
TEST_P(Trans4x4HT, CoeffCheck) { RunCoeffCheck(); }
TEST_P(Trans4x4HT, MemCheck) { RunMemCheck(); }
TEST_P(Trans4x4HT, InvAccuracyCheck) { RunInvAccuracyCheck(1); }
class Trans4x4WHT : public Trans4x4TestBase,
public ::testing::TestWithParam<Dct4x4Param> {
public:
virtual ~Trans4x4WHT() {}
virtual void SetUp() {
fwd_txfm_ = GET_PARAM(0);
inv_txfm_ = GET_PARAM(1);
tx_type_ = GET_PARAM(2);
pitch_ = 4;
fwd_txfm_ref = fwht4x4_ref;
bit_depth_ = GET_PARAM(3);
mask_ = (1 << bit_depth_) - 1;
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
void RunFwdTxfm(const int16_t *in, tran_low_t *out, int stride) {
fwd_txfm_(in, out, stride);
}
void RunInvTxfm(const tran_low_t *out, uint8_t *dst, int stride) {
inv_txfm_(out, dst, stride);
}
FdctFunc fwd_txfm_;
IdctFunc inv_txfm_;
};
TEST_P(Trans4x4WHT, AccuracyCheck) { RunAccuracyCheck(0); }
TEST_P(Trans4x4WHT, CoeffCheck) { RunCoeffCheck(); }
TEST_P(Trans4x4WHT, MemCheck) { RunMemCheck(); }
TEST_P(Trans4x4WHT, InvAccuracyCheck) { RunInvAccuracyCheck(0); }
using std::tr1::make_tuple;
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
C, Trans4x4DCT,
::testing::Values(
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_10, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12, 0, VPX_BITS_12),
make_tuple(&vpx_fdct4x4_c, &vpx_idct4x4_16_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(C, Trans4x4DCT,
::testing::Values(make_tuple(&vpx_fdct4x4_c,
&vpx_idct4x4_16_add_c, 0,
VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
C, Trans4x4HT,
::testing::Values(
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 0, VPX_BITS_10),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 1, VPX_BITS_10),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 2, VPX_BITS_10),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_10, 3, VPX_BITS_10),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 0, VPX_BITS_12),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 1, VPX_BITS_12),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 2, VPX_BITS_12),
make_tuple(&vp9_highbd_fht4x4_c, &iht4x4_12, 3, VPX_BITS_12),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(
C, Trans4x4HT,
::testing::Values(
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
C, Trans4x4WHT,
::testing::Values(
make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_10, 0, VPX_BITS_10),
make_tuple(&vp9_highbd_fwht4x4_c, &iwht4x4_12, 0, VPX_BITS_12),
make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8)));
#else
INSTANTIATE_TEST_CASE_P(C, Trans4x4WHT,
::testing::Values(make_tuple(&vp9_fwht4x4_c,
&vpx_iwht4x4_16_add_c, 0,
VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(NEON, Trans4x4DCT,
::testing::Values(make_tuple(&vpx_fdct4x4_c,
&vpx_idct4x4_16_add_neon,
0, VPX_BITS_8)));
#endif // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
NEON, Trans4x4HT,
::testing::Values(
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_c, &vp9_iht4x4_16_add_neon, 3, VPX_BITS_8)));
#endif // HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4WHT,
::testing::Values(
make_tuple(&vp9_fwht4x4_sse2, &vpx_iwht4x4_16_add_c, 0, VPX_BITS_8),
make_tuple(&vp9_fwht4x4_c, &vpx_iwht4x4_16_add_sse2, 0, VPX_BITS_8)));
#endif
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(SSE2, Trans4x4DCT,
::testing::Values(make_tuple(&vpx_fdct4x4_sse2,
&vpx_idct4x4_16_add_sse2,
0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4HT,
::testing::Values(
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_sse2, 3, VPX_BITS_8)));
#endif // HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4DCT,
::testing::Values(
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_10_sse2, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_10_sse2, 0, VPX_BITS_10),
make_tuple(&vpx_highbd_fdct4x4_c, &idct4x4_12_sse2, 0, VPX_BITS_12),
make_tuple(&vpx_highbd_fdct4x4_sse2, &idct4x4_12_sse2, 0, VPX_BITS_12),
make_tuple(&vpx_fdct4x4_sse2, &vpx_idct4x4_16_add_c, 0, VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
SSE2, Trans4x4HT,
::testing::Values(
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_sse2, &vp9_iht4x4_16_add_c, 3, VPX_BITS_8)));
#endif // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(MSA, Trans4x4DCT,
::testing::Values(make_tuple(&vpx_fdct4x4_msa,
&vpx_idct4x4_16_add_msa, 0,
VPX_BITS_8)));
INSTANTIATE_TEST_CASE_P(
MSA, Trans4x4HT,
::testing::Values(
make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 0, VPX_BITS_8),
make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 1, VPX_BITS_8),
make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 2, VPX_BITS_8),
make_tuple(&vp9_fht4x4_msa, &vp9_iht4x4_16_add_msa, 3, VPX_BITS_8)));
#endif // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
} // namespace

View File

@ -88,45 +88,45 @@ void fht8x8_ref(const int16_t *in, tran_low_t *out, int stride, int tx_type) {
#if CONFIG_VP9_HIGHBITDEPTH
void idct8x8_10(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_64_add_c(in, out, stride, 10);
vpx_highbd_idct8x8_64_add_c(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct8x8_12(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_64_add_c(in, out, stride, 12);
vpx_highbd_idct8x8_64_add_c(in, CAST_TO_SHORTPTR(out), stride, 12);
}
void iht8x8_10(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
vp9_highbd_iht8x8_64_add_c(in, out, stride, tx_type, 10);
vp9_highbd_iht8x8_64_add_c(in, CAST_TO_SHORTPTR(out), stride, tx_type, 10);
}
void iht8x8_12(const tran_low_t *in, uint8_t *out, int stride, int tx_type) {
vp9_highbd_iht8x8_64_add_c(in, out, stride, tx_type, 12);
vp9_highbd_iht8x8_64_add_c(in, CAST_TO_SHORTPTR(out), stride, tx_type, 12);
}
#if HAVE_SSE2
void idct8x8_10_add_10_c(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_10_add_c(in, out, stride, 10);
void idct8x8_12_add_10_c(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_12_add_c(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct8x8_10_add_12_c(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_10_add_c(in, out, stride, 12);
void idct8x8_12_add_12_c(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_12_add_c(in, CAST_TO_SHORTPTR(out), stride, 12);
}
void idct8x8_10_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_10_add_sse2(in, out, stride, 10);
void idct8x8_12_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_12_add_sse2(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct8x8_10_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_10_add_sse2(in, out, stride, 12);
void idct8x8_12_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_12_add_sse2(in, CAST_TO_SHORTPTR(out), stride, 12);
}
void idct8x8_64_add_10_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_64_add_sse2(in, out, stride, 10);
vpx_highbd_idct8x8_64_add_sse2(in, CAST_TO_SHORTPTR(out), stride, 10);
}
void idct8x8_64_add_12_sse2(const tran_low_t *in, uint8_t *out, int stride) {
vpx_highbd_idct8x8_64_add_sse2(in, out, stride, 12);
vpx_highbd_idct8x8_64_add_sse2(in, CAST_TO_SHORTPTR(out), stride, 12);
}
#endif // HAVE_SSE2
#endif // CONFIG_VP9_HIGHBITDEPTH
@ -257,7 +257,7 @@ class FwdTrans8x8TestBase {
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(test_temp_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
RunInvTxfm(test_temp_block, CAST_TO_BYTEPTR(dst16), pitch_));
#endif
}
@ -340,7 +340,7 @@ class FwdTrans8x8TestBase {
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(test_temp_block, CONVERT_TO_BYTEPTR(dst16), pitch_));
RunInvTxfm(test_temp_block, CAST_TO_BYTEPTR(dst16), pitch_));
#endif
}
@ -413,7 +413,7 @@ class FwdTrans8x8TestBase {
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), pitch_));
RunInvTxfm(coeff, CAST_TO_BYTEPTR(dst16), pitch_));
#endif
}
@ -497,9 +497,9 @@ class FwdTrans8x8TestBase {
ASM_REGISTER_STATE_CHECK(RunInvTxfm(coeff, dst, pitch_));
#if CONFIG_VP9_HIGHBITDEPTH
} else {
ref_txfm(coeff, CONVERT_TO_BYTEPTR(ref16), pitch_);
ref_txfm(coeff, CAST_TO_BYTEPTR(ref16), pitch_);
ASM_REGISTER_STATE_CHECK(
RunInvTxfm(coeff, CONVERT_TO_BYTEPTR(dst16), pitch_));
RunInvTxfm(coeff, CAST_TO_BYTEPTR(dst16), pitch_));
#endif
}
@ -511,8 +511,8 @@ class FwdTrans8x8TestBase {
const int diff = dst[j] - ref[j];
#endif
const uint32_t error = diff * diff;
EXPECT_EQ(0u, error) << "Error: 8x8 IDCT has error " << error
<< " at index " << j;
EXPECT_EQ(0u, error)
<< "Error: 8x8 IDCT has error " << error << " at index " << j;
}
}
}
@ -670,14 +670,12 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_c, 3, VPX_BITS_8)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#if HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_NEON && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(NEON, FwdTrans8x8DCT,
::testing::Values(make_tuple(&vpx_fdct8x8_neon,
&vpx_idct8x8_64_add_neon,
0, VPX_BITS_8)));
#endif // HAVE_NEON_ASM && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if !CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
NEON, FwdTrans8x8HT,
::testing::Values(
@ -685,7 +683,8 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_neon, 1, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_neon, 2, VPX_BITS_8),
make_tuple(&vp9_fht8x8_c, &vp9_iht8x8_64_add_neon, 3, VPX_BITS_8)));
#endif // HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#endif // !CONFIG_VP9_HIGHBITDEPTH
#endif // HAVE_NEON && !CONFIG_EMULATE_HARDWARE
#if HAVE_SSE2 && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(SSE2, FwdTrans8x8DCT,
@ -728,10 +727,10 @@ INSTANTIATE_TEST_CASE_P(
INSTANTIATE_TEST_CASE_P(
SSE2, InvTrans8x8DCT,
::testing::Values(
make_tuple(&idct8x8_10_add_10_c, &idct8x8_10_add_10_sse2, 6225,
make_tuple(&idct8x8_12_add_10_c, &idct8x8_12_add_10_sse2, 6225,
VPX_BITS_10),
make_tuple(&idct8x8_10, &idct8x8_64_add_10_sse2, 6225, VPX_BITS_10),
make_tuple(&idct8x8_10_add_12_c, &idct8x8_10_add_12_sse2, 6225,
make_tuple(&idct8x8_12_add_12_c, &idct8x8_12_add_12_sse2, 6225,
VPX_BITS_12),
make_tuple(&idct8x8_12, &idct8x8_64_add_12_sse2, 6225, VPX_BITS_12)));
#endif // HAVE_SSE2 && CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
@ -740,7 +739,7 @@ INSTANTIATE_TEST_CASE_P(
!CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(SSSE3, FwdTrans8x8DCT,
::testing::Values(make_tuple(&vpx_fdct8x8_ssse3,
&vpx_idct8x8_64_add_ssse3,
&vpx_idct8x8_64_add_sse2,
0, VPX_BITS_8)));
#endif
@ -757,4 +756,11 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp9_fht8x8_msa, &vp9_iht8x8_64_add_msa, 2, VPX_BITS_8),
make_tuple(&vp9_fht8x8_msa, &vp9_iht8x8_64_add_msa, 3, VPX_BITS_8)));
#endif // HAVE_MSA && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
#if HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
INSTANTIATE_TEST_CASE_P(VSX, FwdTrans8x8DCT,
::testing::Values(make_tuple(&vpx_fdct8x8_c,
&vpx_idct8x8_64_add_vsx, 0,
VPX_BITS_8)));
#endif // HAVE_VSX && !CONFIG_VP9_HIGHBITDEPTH && !CONFIG_EMULATE_HARDWARE
} // namespace

View File

@ -13,6 +13,7 @@
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vpx_dsp_rtcd.h"
#include "vpx_ports/vpx_timer.h"
#include "test/acm_random.h"
#include "test/register_state_check.h"
@ -21,7 +22,8 @@ namespace {
using ::libvpx_test::ACMRandom;
typedef void (*HadamardFunc)(const int16_t *a, int a_stride, int16_t *b);
typedef void (*HadamardFunc)(const int16_t *a, ptrdiff_t a_stride,
tran_low_t *b);
void hadamard_loop(const int16_t *a, int a_stride, int16_t *out) {
int16_t b[8];
@ -46,18 +48,16 @@ void hadamard_loop(const int16_t *a, int a_stride, int16_t *out) {
out[5] = c[3] - c[7];
}
void reference_hadamard8x8(const int16_t *a, int a_stride, int16_t *b) {
void reference_hadamard8x8(const int16_t *a, int a_stride, tran_low_t *b) {
int16_t buf[64];
for (int i = 0; i < 8; ++i) {
hadamard_loop(a + i, a_stride, buf + i * 8);
}
int16_t buf2[64];
for (int i = 0; i < 8; ++i) hadamard_loop(a + i, a_stride, buf + i * 8);
for (int i = 0; i < 8; ++i) hadamard_loop(buf + i, 8, buf2 + i * 8);
for (int i = 0; i < 8; ++i) {
hadamard_loop(buf + i, 8, b + i * 8);
}
for (int i = 0; i < 64; ++i) b[i] = (tran_low_t)buf2[i];
}
void reference_hadamard16x16(const int16_t *a, int a_stride, int16_t *b) {
void reference_hadamard16x16(const int16_t *a, int a_stride, tran_low_t *b) {
/* The source is a 16x16 block. The destination is rearranged to 8x32.
* Input is 9 bit. */
reference_hadamard8x8(a + 0 + 0 * a_stride, a_stride, b + 0);
@ -68,16 +68,16 @@ void reference_hadamard16x16(const int16_t *a, int a_stride, int16_t *b) {
/* Overlay the 8x8 blocks and combine. */
for (int i = 0; i < 64; ++i) {
/* 8x8 steps the range up to 15 bits. */
const int16_t a0 = b[0];
const int16_t a1 = b[64];
const int16_t a2 = b[128];
const int16_t a3 = b[192];
const tran_low_t a0 = b[0];
const tran_low_t a1 = b[64];
const tran_low_t a2 = b[128];
const tran_low_t a3 = b[192];
/* Prevent the result from escaping int16_t. */
const int16_t b0 = (a0 + a1) >> 1;
const int16_t b1 = (a0 - a1) >> 1;
const int16_t b2 = (a2 + a3) >> 1;
const int16_t b3 = (a2 - a3) >> 1;
const tran_low_t b0 = (a0 + a1) >> 1;
const tran_low_t b1 = (a0 - a1) >> 1;
const tran_low_t b2 = (a2 + a3) >> 1;
const tran_low_t b3 = (a2 - a3) >> 1;
/* Store a 16 bit value. */
b[0] = b0 + b2;
@ -101,12 +101,35 @@ class HadamardTestBase : public ::testing::TestWithParam<HadamardFunc> {
ACMRandom rnd_;
};
void HadamardSpeedTest(const char *name, HadamardFunc const func,
const int16_t *input, int stride, tran_low_t *output,
int times) {
int i;
vpx_usec_timer timer;
vpx_usec_timer_start(&timer);
for (i = 0; i < times; ++i) {
func(input, stride, output);
}
vpx_usec_timer_mark(&timer);
const int elapsed_time = static_cast<int>(vpx_usec_timer_elapsed(&timer));
printf("%s[%12d runs]: %d us\n", name, times, elapsed_time);
}
class Hadamard8x8Test : public HadamardTestBase {};
void HadamardSpeedTest8x8(HadamardFunc const func, int times) {
DECLARE_ALIGNED(16, int16_t, input[64]);
DECLARE_ALIGNED(16, tran_low_t, output[64]);
memset(input, 1, sizeof(input));
HadamardSpeedTest("Hadamard8x8", func, input, 8, output, times);
}
TEST_P(Hadamard8x8Test, CompareReferenceRandom) {
DECLARE_ALIGNED(16, int16_t, a[64]);
DECLARE_ALIGNED(16, int16_t, b[64]);
int16_t b_ref[64];
DECLARE_ALIGNED(16, tran_low_t, b[64]);
tran_low_t b_ref[64];
for (int i = 0; i < 64; ++i) {
a[i] = rnd_.Rand9Signed();
}
@ -124,8 +147,8 @@ TEST_P(Hadamard8x8Test, CompareReferenceRandom) {
TEST_P(Hadamard8x8Test, VaryStride) {
DECLARE_ALIGNED(16, int16_t, a[64 * 8]);
DECLARE_ALIGNED(16, int16_t, b[64]);
int16_t b_ref[64];
DECLARE_ALIGNED(16, tran_low_t, b[64]);
tran_low_t b_ref[64];
for (int i = 0; i < 64 * 8; ++i) {
a[i] = rnd_.Rand9Signed();
}
@ -144,6 +167,12 @@ TEST_P(Hadamard8x8Test, VaryStride) {
}
}
TEST_P(Hadamard8x8Test, DISABLED_Speed) {
HadamardSpeedTest8x8(h_func_, 10);
HadamardSpeedTest8x8(h_func_, 10000);
HadamardSpeedTest8x8(h_func_, 10000000);
}
INSTANTIATE_TEST_CASE_P(C, Hadamard8x8Test,
::testing::Values(&vpx_hadamard_8x8_c));
@ -162,12 +191,33 @@ INSTANTIATE_TEST_CASE_P(NEON, Hadamard8x8Test,
::testing::Values(&vpx_hadamard_8x8_neon));
#endif // HAVE_NEON
// TODO(jingning): Remove highbitdepth flag when the SIMD functions are
// in place and turn on the unit test.
#if !CONFIG_VP9_HIGHBITDEPTH
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(MSA, Hadamard8x8Test,
::testing::Values(&vpx_hadamard_8x8_msa));
#endif // HAVE_MSA
#endif // !CONFIG_VP9_HIGHBITDEPTH
#if HAVE_VSX
INSTANTIATE_TEST_CASE_P(VSX, Hadamard8x8Test,
::testing::Values(&vpx_hadamard_8x8_vsx));
#endif // HAVE_VSX
class Hadamard16x16Test : public HadamardTestBase {};
void HadamardSpeedTest16x16(HadamardFunc const func, int times) {
DECLARE_ALIGNED(16, int16_t, input[256]);
DECLARE_ALIGNED(16, tran_low_t, output[256]);
memset(input, 1, sizeof(input));
HadamardSpeedTest("Hadamard16x16", func, input, 16, output, times);
}
TEST_P(Hadamard16x16Test, CompareReferenceRandom) {
DECLARE_ALIGNED(16, int16_t, a[16 * 16]);
DECLARE_ALIGNED(16, int16_t, b[16 * 16]);
int16_t b_ref[16 * 16];
DECLARE_ALIGNED(16, tran_low_t, b[16 * 16]);
tran_low_t b_ref[16 * 16];
for (int i = 0; i < 16 * 16; ++i) {
a[i] = rnd_.Rand9Signed();
}
@ -185,8 +235,8 @@ TEST_P(Hadamard16x16Test, CompareReferenceRandom) {
TEST_P(Hadamard16x16Test, VaryStride) {
DECLARE_ALIGNED(16, int16_t, a[16 * 16 * 8]);
DECLARE_ALIGNED(16, int16_t, b[16 * 16]);
int16_t b_ref[16 * 16];
DECLARE_ALIGNED(16, tran_low_t, b[16 * 16]);
tran_low_t b_ref[16 * 16];
for (int i = 0; i < 16 * 16 * 8; ++i) {
a[i] = rnd_.Rand9Signed();
}
@ -205,6 +255,12 @@ TEST_P(Hadamard16x16Test, VaryStride) {
}
}
TEST_P(Hadamard16x16Test, DISABLED_Speed) {
HadamardSpeedTest16x16(h_func_, 10);
HadamardSpeedTest16x16(h_func_, 10000);
HadamardSpeedTest16x16(h_func_, 10000000);
}
INSTANTIATE_TEST_CASE_P(C, Hadamard16x16Test,
::testing::Values(&vpx_hadamard_16x16_c));
@ -213,8 +269,25 @@ INSTANTIATE_TEST_CASE_P(SSE2, Hadamard16x16Test,
::testing::Values(&vpx_hadamard_16x16_sse2));
#endif // HAVE_SSE2
#if HAVE_AVX2
INSTANTIATE_TEST_CASE_P(AVX2, Hadamard16x16Test,
::testing::Values(&vpx_hadamard_16x16_avx2));
#endif // HAVE_AVX2
#if HAVE_VSX
INSTANTIATE_TEST_CASE_P(VSX, Hadamard16x16Test,
::testing::Values(&vpx_hadamard_16x16_vsx));
#endif // HAVE_VSX
#if HAVE_NEON
INSTANTIATE_TEST_CASE_P(NEON, Hadamard16x16Test,
::testing::Values(&vpx_hadamard_16x16_neon));
#endif // HAVE_NEON
#if !CONFIG_VP9_HIGHBITDEPTH
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(MSA, Hadamard16x16Test,
::testing::Values(&vpx_hadamard_16x16_msa));
#endif // HAVE_MSA
#endif // !CONFIG_VP9_HIGHBITDEPTH
} // namespace

View File

@ -13,6 +13,7 @@
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/buffer.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "vpx/vpx_integer.h"
@ -21,106 +22,156 @@ typedef void (*IdctFunc)(int16_t *input, unsigned char *pred_ptr,
int pred_stride, unsigned char *dst_ptr,
int dst_stride);
namespace {
using libvpx_test::Buffer;
class IDCTTest : public ::testing::TestWithParam<IdctFunc> {
protected:
virtual void SetUp() {
int i;
UUT = GetParam();
memset(input, 0, sizeof(input));
/* Set up guard blocks */
for (i = 0; i < 256; i++) output[i] = ((i & 0xF) < 4 && (i < 64)) ? 0 : -1;
input = new Buffer<int16_t>(4, 4, 0);
ASSERT_TRUE(input != NULL);
ASSERT_TRUE(input->Init());
predict = new Buffer<uint8_t>(4, 4, 3);
ASSERT_TRUE(predict != NULL);
ASSERT_TRUE(predict->Init());
output = new Buffer<uint8_t>(4, 4, 3);
ASSERT_TRUE(output != NULL);
ASSERT_TRUE(output->Init());
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
virtual void TearDown() {
delete input;
delete predict;
delete output;
libvpx_test::ClearSystemState();
}
IdctFunc UUT;
int16_t input[16];
unsigned char output[256];
unsigned char predict[256];
Buffer<int16_t> *input;
Buffer<uint8_t> *predict;
Buffer<uint8_t> *output;
};
TEST_P(IDCTTest, TestGuardBlocks) {
int i;
for (i = 0; i < 256; i++) {
if ((i & 0xF) < 4 && i < 64)
EXPECT_EQ(0, output[i]) << i;
else
EXPECT_EQ(255, output[i]);
}
}
TEST_P(IDCTTest, TestAllZeros) {
int i;
// When the input is '0' the output will be '0'.
input->Set(0);
predict->Set(0);
output->Set(0);
ASM_REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
ASM_REGISTER_STATE_CHECK(UUT(input->TopLeftPixel(), predict->TopLeftPixel(),
predict->stride(), output->TopLeftPixel(),
output->stride()));
for (i = 0; i < 256; i++) {
if ((i & 0xF) < 4 && i < 64)
EXPECT_EQ(0, output[i]) << "i==" << i;
else
EXPECT_EQ(255, output[i]) << "i==" << i;
}
ASSERT_TRUE(input->CheckValues(0));
ASSERT_TRUE(input->CheckPadding());
ASSERT_TRUE(output->CheckValues(0));
ASSERT_TRUE(output->CheckPadding());
}
TEST_P(IDCTTest, TestAllOnes) {
int i;
input->Set(0);
// When the first element is '4' it will fill the output buffer with '1'.
input->TopLeftPixel()[0] = 4;
predict->Set(0);
output->Set(0);
input[0] = 4;
ASM_REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
ASM_REGISTER_STATE_CHECK(UUT(input->TopLeftPixel(), predict->TopLeftPixel(),
predict->stride(), output->TopLeftPixel(),
output->stride()));
for (i = 0; i < 256; i++) {
if ((i & 0xF) < 4 && i < 64)
EXPECT_EQ(1, output[i]) << "i==" << i;
else
EXPECT_EQ(255, output[i]) << "i==" << i;
}
ASSERT_TRUE(output->CheckValues(1));
ASSERT_TRUE(output->CheckPadding());
}
TEST_P(IDCTTest, TestAddOne) {
int i;
// Set the transform output to '1' and make sure it gets added to the
// prediction buffer.
input->Set(0);
input->TopLeftPixel()[0] = 4;
output->Set(0);
for (i = 0; i < 256; i++) predict[i] = i;
input[0] = 4;
ASM_REGISTER_STATE_CHECK(UUT(input, predict, 16, output, 16));
for (i = 0; i < 256; i++) {
if ((i & 0xF) < 4 && i < 64)
EXPECT_EQ(i + 1, output[i]) << "i==" << i;
else
EXPECT_EQ(255, output[i]) << "i==" << i;
uint8_t *pred = predict->TopLeftPixel();
for (int y = 0; y < 4; ++y) {
for (int x = 0; x < 4; ++x) {
pred[y * predict->stride() + x] = y * 4 + x;
}
}
ASM_REGISTER_STATE_CHECK(UUT(input->TopLeftPixel(), predict->TopLeftPixel(),
predict->stride(), output->TopLeftPixel(),
output->stride()));
uint8_t const *out = output->TopLeftPixel();
for (int y = 0; y < 4; ++y) {
for (int x = 0; x < 4; ++x) {
EXPECT_EQ(1 + y * 4 + x, out[y * output->stride() + x]);
}
}
if (HasFailure()) {
output->DumpBuffer();
}
ASSERT_TRUE(output->CheckPadding());
}
TEST_P(IDCTTest, TestWithData) {
int i;
// Test a single known input.
predict->Set(0);
for (i = 0; i < 16; i++) input[i] = i;
ASM_REGISTER_STATE_CHECK(UUT(input, output, 16, output, 16));
for (i = 0; i < 256; i++) {
if ((i & 0xF) > 3 || i > 63)
EXPECT_EQ(255, output[i]) << "i==" << i;
else if (i == 0)
EXPECT_EQ(11, output[i]) << "i==" << i;
else if (i == 34)
EXPECT_EQ(1, output[i]) << "i==" << i;
else if (i == 2 || i == 17 || i == 32)
EXPECT_EQ(3, output[i]) << "i==" << i;
else
EXPECT_EQ(0, output[i]) << "i==" << i;
int16_t *in = input->TopLeftPixel();
for (int y = 0; y < 4; ++y) {
for (int x = 0; x < 4; ++x) {
in[y * input->stride() + x] = y * 4 + x;
}
}
ASM_REGISTER_STATE_CHECK(UUT(input->TopLeftPixel(), predict->TopLeftPixel(),
predict->stride(), output->TopLeftPixel(),
output->stride()));
uint8_t *out = output->TopLeftPixel();
for (int y = 0; y < 4; ++y) {
for (int x = 0; x < 4; ++x) {
switch (y * 4 + x) {
case 0: EXPECT_EQ(11, out[y * output->stride() + x]); break;
case 2:
case 5:
case 8: EXPECT_EQ(3, out[y * output->stride() + x]); break;
case 10: EXPECT_EQ(1, out[y * output->stride() + x]); break;
default: EXPECT_EQ(0, out[y * output->stride() + x]);
}
}
}
if (HasFailure()) {
output->DumpBuffer();
}
ASSERT_TRUE(output->CheckPadding());
}
INSTANTIATE_TEST_CASE_P(C, IDCTTest, ::testing::Values(vp8_short_idct4x4llm_c));
#if HAVE_NEON
INSTANTIATE_TEST_CASE_P(NEON, IDCTTest,
::testing::Values(vp8_short_idct4x4llm_neon));
#endif // HAVE_NEON
#if HAVE_MMX
INSTANTIATE_TEST_CASE_P(MMX, IDCTTest,
::testing::Values(vp8_short_idct4x4llm_mmx));
#endif
#endif // HAVE_MMX
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(MSA, IDCTTest,
::testing::Values(vp8_short_idct4x4llm_msa));
#endif
#endif // HAVE_MSA
#if HAVE_MMI
INSTANTIATE_TEST_CASE_P(MMI, IDCTTest,
::testing::Values(vp8_short_idct4x4llm_mmi));
#endif // HAVE_MMI
}

View File

@ -45,8 +45,8 @@ class InvalidFileTest : public ::libvpx_test::DecoderTest,
void OpenResFile(const std::string &res_file_name_) {
res_file_ = libvpx_test::OpenTestDataFile(res_file_name_);
ASSERT_TRUE(res_file_ != NULL) << "Result file open failed. Filename: "
<< res_file_name_;
ASSERT_TRUE(res_file_ != NULL)
<< "Result file open failed. Filename: " << res_file_name_;
}
virtual bool HandleDecodeResult(
@ -120,10 +120,23 @@ class InvalidFileTest : public ::libvpx_test::DecoderTest,
TEST_P(InvalidFileTest, ReturnCode) { RunTest(); }
#if CONFIG_VP8_DECODER
const DecodeParam kVP8InvalidFileTests[] = {
{ 1, "invalid-bug-1443.ivf" },
};
VP8_INSTANTIATE_TEST_CASE(InvalidFileTest,
::testing::ValuesIn(kVP8InvalidFileTests));
#endif // CONFIG_VP8_DECODER
#if CONFIG_VP9_DECODER
const DecodeParam kVP9InvalidFileTests[] = {
{ 1, "invalid-vp90-02-v2.webm" },
#if CONFIG_VP9_HIGHBITDEPTH
{ 1, "invalid-vp90-2-00-quantizer-00.webm.ivf.s5861_r01-05_b6-.v2.ivf" },
{ 1,
"invalid-vp90-2-21-resize_inter_320x180_5_3-4.webm.ivf.s45551_r01-05_b6-."
"ivf" },
#endif
{ 1, "invalid-vp90-03-v3.webm" },
{ 1, "invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-.ivf" },
@ -141,10 +154,14 @@ const DecodeParam kVP9InvalidFileTests[] = {
{ 1, "invalid-vp90-2-12-droppable_1.ivf.s73804_r01-05_b6-.ivf" },
{ 1, "invalid-vp90-2-03-size-224x196.webm.ivf.s44156_r01-05_b6-.ivf" },
{ 1, "invalid-vp90-2-03-size-202x210.webm.ivf.s113306_r01-05_b6-.ivf" },
{ 1,
"invalid-vp90-2-10-show-existing-frame.webm.ivf.s180315_r01-05_b6-.ivf" },
{ 1, "invalid-crbug-667044.webm" },
};
VP9_INSTANTIATE_TEST_CASE(InvalidFileTest,
::testing::ValuesIn(kVP9InvalidFileTests));
#endif // CONFIG_VP9_DECODER
// This class will include test vectors that are expected to fail
// peek. However they are still expected to have no fatal failures.
@ -159,12 +176,12 @@ class InvalidFileInvalidPeekTest : public InvalidFileTest {
TEST_P(InvalidFileInvalidPeekTest, ReturnCode) { RunTest(); }
#if CONFIG_VP8_DECODER
const DecodeParam kVP8InvalidFileTests[] = {
const DecodeParam kVP8InvalidPeekTests[] = {
{ 1, "invalid-vp80-00-comprehensive-018.ivf.2kf_0x6.ivf" },
};
VP8_INSTANTIATE_TEST_CASE(InvalidFileInvalidPeekTest,
::testing::ValuesIn(kVP8InvalidFileTests));
::testing::ValuesIn(kVP8InvalidPeekTests));
#endif // CONFIG_VP8_DECODER
#if CONFIG_VP9_DECODER
@ -184,6 +201,7 @@ const DecodeParam kMultiThreadedVP9InvalidFileTests[] = {
"invalid-vp90-2-08-tile_1x8_frame_parallel.webm.ivf.s288_r01-05_b6-.ivf" },
{ 2, "invalid-vp90-2-09-aq2.webm.ivf.s3984_r01-05_b6-.v2.ivf" },
{ 4, "invalid-vp90-2-09-subpixel-00.ivf.s19552_r01-05_b6-.v2.ivf" },
{ 2, "invalid-crbug-629481.webm" },
};
INSTANTIATE_TEST_CASE_P(

View File

@ -47,8 +47,8 @@ class IVFVideoSource : public CompressedVideoSource {
virtual void Begin() {
input_file_ = OpenTestDataFile(file_name_);
ASSERT_TRUE(input_file_ != NULL) << "Input file open failed. Filename: "
<< file_name_;
ASSERT_TRUE(input_file_ != NULL)
<< "Input file open failed. Filename: " << file_name_;
// Read file header
uint8_t file_hdr[kIvfFileHdrSize];

View File

@ -135,8 +135,8 @@ TEST_P(KeyframeTest, TestAutoKeyframe) {
for (std::vector<vpx_codec_pts_t>::const_iterator iter = kf_pts_list_.begin();
iter != kf_pts_list_.end(); ++iter) {
if (deadline_ == VPX_DL_REALTIME && *iter > 0)
EXPECT_EQ(0, (*iter - 1) % 30) << "Unexpected keyframe at frame "
<< *iter;
EXPECT_EQ(0, (*iter - 1) % 30)
<< "Unexpected keyframe at frame " << *iter;
else
EXPECT_EQ(0, *iter % 30) << "Unexpected keyframe at frame " << *iter;
}

View File

@ -66,6 +66,36 @@ class LevelTest
int level_;
};
TEST_P(LevelTest, TestTargetLevel11Large) {
ASSERT_NE(encoding_mode_, ::libvpx_test::kRealTime);
::libvpx_test::I420VideoSource video("hantro_odd.yuv", 208, 144, 30, 1, 0,
60);
target_level_ = 11;
cfg_.rc_target_bitrate = 150;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(target_level_, level_);
}
TEST_P(LevelTest, TestTargetLevel20Large) {
ASSERT_NE(encoding_mode_, ::libvpx_test::kRealTime);
::libvpx_test::I420VideoSource video("hantro_collage_w352h288.yuv", 352, 288,
30, 1, 0, 60);
target_level_ = 20;
cfg_.rc_target_bitrate = 1200;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(target_level_, level_);
}
TEST_P(LevelTest, TestTargetLevel31Large) {
ASSERT_NE(encoding_mode_, ::libvpx_test::kRealTime);
::libvpx_test::I420VideoSource video("niklas_1280_720_30.y4m", 1280, 720, 30,
1, 0, 60);
target_level_ = 31;
cfg_.rc_target_bitrate = 8000;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_GE(target_level_, level_);
}
// Test for keeping level stats only
TEST_P(LevelTest, TestTargetLevel0) {
::libvpx_test::I420VideoSource video("hantro_odd.yuv", 208, 144, 30, 1, 0,
@ -73,11 +103,11 @@ TEST_P(LevelTest, TestTargetLevel0) {
target_level_ = 0;
min_gf_internal_ = 4;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_EQ(11, level_);
ASSERT_GE(11, level_);
cfg_.rc_target_bitrate = 1600;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
ASSERT_EQ(20, level_);
ASSERT_GE(20, level_);
}
// Test for level control being turned off
@ -94,12 +124,13 @@ TEST_P(LevelTest, TestTargetLevelApi) {
vpx_codec_ctx_t enc;
vpx_codec_enc_cfg_t cfg;
EXPECT_EQ(VPX_CODEC_OK, vpx_codec_enc_config_default(codec, &cfg, 0));
cfg.rc_target_bitrate = 100;
EXPECT_EQ(VPX_CODEC_OK, vpx_codec_enc_init(&enc, codec, &cfg, 0));
for (int level = 0; level <= 256; ++level) {
if (level == 10 || level == 11 || level == 20 || level == 21 ||
level == 30 || level == 31 || level == 40 || level == 41 ||
level == 50 || level == 51 || level == 52 || level == 60 ||
level == 61 || level == 62 || level == 0 || level == 255)
level == 61 || level == 62 || level == 0 || level == 1 || level == 255)
EXPECT_EQ(VPX_CODEC_OK,
vpx_codec_control(&enc, VP9E_SET_TARGET_LEVEL, level));
else

View File

@ -114,6 +114,18 @@ void InitInput(Pixel *s, Pixel *ref_s, ACMRandom *rnd, const uint8_t limit,
}
}
uint8_t GetOuterThresh(ACMRandom *rnd) {
return static_cast<uint8_t>(rnd->RandRange(3 * MAX_LOOP_FILTER + 5));
}
uint8_t GetInnerThresh(ACMRandom *rnd) {
return static_cast<uint8_t>(rnd->RandRange(MAX_LOOP_FILTER + 1));
}
uint8_t GetHevThresh(ACMRandom *rnd) {
return static_cast<uint8_t>(rnd->RandRange(MAX_LOOP_FILTER + 1) >> 4);
}
class Loop8Test6Param : public ::testing::TestWithParam<loop8_param_t> {
public:
virtual ~Loop8Test6Param() {}
@ -162,15 +174,15 @@ TEST_P(Loop8Test6Param, OperationCheck) {
int first_failure = -1;
for (int i = 0; i < count_test_block; ++i) {
int err_count = 0;
uint8_t tmp = static_cast<uint8_t>(rnd(3 * MAX_LOOP_FILTER + 4));
uint8_t tmp = GetOuterThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
blimit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = static_cast<uint8_t>(rnd(MAX_LOOP_FILTER));
tmp = GetInnerThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
limit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = rnd.Rand8();
tmp = GetHevThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
thresh[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
@ -221,15 +233,15 @@ TEST_P(Loop8Test6Param, ValueCheck) {
for (int i = 0; i < count_test_block; ++i) {
int err_count = 0;
uint8_t tmp = static_cast<uint8_t>(rnd(3 * MAX_LOOP_FILTER + 4));
uint8_t tmp = GetOuterThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
blimit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = static_cast<uint8_t>(rnd(MAX_LOOP_FILTER));
tmp = GetInnerThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
limit[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = rnd.Rand8();
tmp = GetHevThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
thresh[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
@ -271,27 +283,27 @@ TEST_P(Loop8Test9Param, OperationCheck) {
int first_failure = -1;
for (int i = 0; i < count_test_block; ++i) {
int err_count = 0;
uint8_t tmp = static_cast<uint8_t>(rnd(3 * MAX_LOOP_FILTER + 4));
uint8_t tmp = GetOuterThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
blimit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = static_cast<uint8_t>(rnd(MAX_LOOP_FILTER));
tmp = GetInnerThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
limit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = rnd.Rand8();
tmp = GetHevThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
thresh0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = static_cast<uint8_t>(rnd(3 * MAX_LOOP_FILTER + 4));
tmp = GetOuterThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
blimit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = static_cast<uint8_t>(rnd(MAX_LOOP_FILTER));
tmp = GetInnerThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
limit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = rnd.Rand8();
tmp = GetHevThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
thresh1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
@ -334,27 +346,27 @@ TEST_P(Loop8Test9Param, ValueCheck) {
int first_failure = -1;
for (int i = 0; i < count_test_block; ++i) {
int err_count = 0;
uint8_t tmp = static_cast<uint8_t>(rnd(3 * MAX_LOOP_FILTER + 4));
uint8_t tmp = GetOuterThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
blimit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = static_cast<uint8_t>(rnd(MAX_LOOP_FILTER));
tmp = GetInnerThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
limit0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = rnd.Rand8();
tmp = GetHevThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
thresh0[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = static_cast<uint8_t>(rnd(3 * MAX_LOOP_FILTER + 4));
tmp = GetOuterThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
blimit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = static_cast<uint8_t>(rnd(MAX_LOOP_FILTER));
tmp = GetInnerThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
limit1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
tmp = rnd.Rand8();
tmp = GetHevThresh(&rnd);
DECLARE_ALIGNED(16, const uint8_t,
thresh1[16]) = { tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp,
tmp, tmp, tmp, tmp, tmp, tmp, tmp, tmp };
@ -402,10 +414,10 @@ INSTANTIATE_TEST_CASE_P(
&vpx_highbd_lpf_vertical_4_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_8_sse2,
&vpx_highbd_lpf_horizontal_8_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_edge_8_sse2,
&vpx_highbd_lpf_horizontal_edge_8_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_edge_16_sse2,
&vpx_highbd_lpf_horizontal_edge_16_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
&vpx_highbd_lpf_horizontal_16_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_16_dual_sse2,
&vpx_highbd_lpf_horizontal_16_dual_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_8_sse2,
&vpx_highbd_lpf_vertical_8_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_16_sse2,
@ -416,10 +428,10 @@ INSTANTIATE_TEST_CASE_P(
&vpx_highbd_lpf_vertical_4_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_8_sse2,
&vpx_highbd_lpf_horizontal_8_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_edge_8_sse2,
&vpx_highbd_lpf_horizontal_edge_8_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_edge_16_sse2,
&vpx_highbd_lpf_horizontal_edge_16_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
&vpx_highbd_lpf_horizontal_16_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_16_dual_sse2,
&vpx_highbd_lpf_horizontal_16_dual_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_8_sse2,
&vpx_highbd_lpf_vertical_8_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_16_sse2,
@ -430,10 +442,10 @@ INSTANTIATE_TEST_CASE_P(
&vpx_highbd_lpf_vertical_4_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_8_sse2,
&vpx_highbd_lpf_horizontal_8_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_edge_8_sse2,
&vpx_highbd_lpf_horizontal_edge_8_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_edge_16_sse2,
&vpx_highbd_lpf_horizontal_edge_16_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_16_sse2,
&vpx_highbd_lpf_horizontal_16_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_16_dual_sse2,
&vpx_highbd_lpf_horizontal_16_dual_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_8_sse2,
&vpx_highbd_lpf_vertical_8_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_16_sse2,
@ -450,10 +462,9 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_sse2, &vpx_lpf_horizontal_4_c, 8),
make_tuple(&vpx_lpf_horizontal_8_sse2, &vpx_lpf_horizontal_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_8_sse2,
&vpx_lpf_horizontal_edge_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_16_sse2,
&vpx_lpf_horizontal_edge_16_c, 8),
make_tuple(&vpx_lpf_horizontal_16_sse2, &vpx_lpf_horizontal_16_c, 8),
make_tuple(&vpx_lpf_horizontal_16_dual_sse2,
&vpx_lpf_horizontal_16_dual_c, 8),
make_tuple(&vpx_lpf_vertical_4_sse2, &vpx_lpf_vertical_4_c, 8),
make_tuple(&vpx_lpf_vertical_8_sse2, &vpx_lpf_vertical_8_c, 8),
make_tuple(&vpx_lpf_vertical_16_sse2, &vpx_lpf_vertical_16_c, 8),
@ -465,10 +476,10 @@ INSTANTIATE_TEST_CASE_P(
#if HAVE_AVX2 && (!CONFIG_VP9_HIGHBITDEPTH)
INSTANTIATE_TEST_CASE_P(
AVX2, Loop8Test6Param,
::testing::Values(make_tuple(&vpx_lpf_horizontal_edge_8_avx2,
&vpx_lpf_horizontal_edge_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_16_avx2,
&vpx_lpf_horizontal_edge_16_c, 8)));
::testing::Values(make_tuple(&vpx_lpf_horizontal_16_avx2,
&vpx_lpf_horizontal_16_c, 8),
make_tuple(&vpx_lpf_horizontal_16_dual_avx2,
&vpx_lpf_horizontal_16_dual_c, 8)));
#endif
#if HAVE_SSE2
@ -515,15 +526,89 @@ INSTANTIATE_TEST_CASE_P(
#if HAVE_NEON
#if CONFIG_VP9_HIGHBITDEPTH
// No neon high bitdepth functions.
INSTANTIATE_TEST_CASE_P(
NEON, Loop8Test6Param,
::testing::Values(make_tuple(&vpx_highbd_lpf_horizontal_4_neon,
&vpx_highbd_lpf_horizontal_4_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_4_neon,
&vpx_highbd_lpf_horizontal_4_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_4_neon,
&vpx_highbd_lpf_horizontal_4_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_8_neon,
&vpx_highbd_lpf_horizontal_8_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_8_neon,
&vpx_highbd_lpf_horizontal_8_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_8_neon,
&vpx_highbd_lpf_horizontal_8_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_16_neon,
&vpx_highbd_lpf_horizontal_16_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_16_neon,
&vpx_highbd_lpf_horizontal_16_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_16_neon,
&vpx_highbd_lpf_horizontal_16_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_16_dual_neon,
&vpx_highbd_lpf_horizontal_16_dual_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_16_dual_neon,
&vpx_highbd_lpf_horizontal_16_dual_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_16_dual_neon,
&vpx_highbd_lpf_horizontal_16_dual_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_4_neon,
&vpx_highbd_lpf_vertical_4_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_4_neon,
&vpx_highbd_lpf_vertical_4_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_4_neon,
&vpx_highbd_lpf_vertical_4_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_8_neon,
&vpx_highbd_lpf_vertical_8_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_8_neon,
&vpx_highbd_lpf_vertical_8_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_8_neon,
&vpx_highbd_lpf_vertical_8_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_16_neon,
&vpx_highbd_lpf_vertical_16_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_16_neon,
&vpx_highbd_lpf_vertical_16_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_16_neon,
&vpx_highbd_lpf_vertical_16_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_16_dual_neon,
&vpx_highbd_lpf_vertical_16_dual_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_16_dual_neon,
&vpx_highbd_lpf_vertical_16_dual_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_16_dual_neon,
&vpx_highbd_lpf_vertical_16_dual_c, 12)));
INSTANTIATE_TEST_CASE_P(
NEON, Loop8Test9Param,
::testing::Values(make_tuple(&vpx_highbd_lpf_horizontal_4_dual_neon,
&vpx_highbd_lpf_horizontal_4_dual_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_4_dual_neon,
&vpx_highbd_lpf_horizontal_4_dual_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_4_dual_neon,
&vpx_highbd_lpf_horizontal_4_dual_c, 12),
make_tuple(&vpx_highbd_lpf_horizontal_8_dual_neon,
&vpx_highbd_lpf_horizontal_8_dual_c, 8),
make_tuple(&vpx_highbd_lpf_horizontal_8_dual_neon,
&vpx_highbd_lpf_horizontal_8_dual_c, 10),
make_tuple(&vpx_highbd_lpf_horizontal_8_dual_neon,
&vpx_highbd_lpf_horizontal_8_dual_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_4_dual_neon,
&vpx_highbd_lpf_vertical_4_dual_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_4_dual_neon,
&vpx_highbd_lpf_vertical_4_dual_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_4_dual_neon,
&vpx_highbd_lpf_vertical_4_dual_c, 12),
make_tuple(&vpx_highbd_lpf_vertical_8_dual_neon,
&vpx_highbd_lpf_vertical_8_dual_c, 8),
make_tuple(&vpx_highbd_lpf_vertical_8_dual_neon,
&vpx_highbd_lpf_vertical_8_dual_c, 10),
make_tuple(&vpx_highbd_lpf_vertical_8_dual_neon,
&vpx_highbd_lpf_vertical_8_dual_c, 12)));
#else
INSTANTIATE_TEST_CASE_P(
NEON, Loop8Test6Param,
::testing::Values(
make_tuple(&vpx_lpf_horizontal_edge_8_neon,
&vpx_lpf_horizontal_edge_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_16_neon,
&vpx_lpf_horizontal_edge_16_c, 8),
make_tuple(&vpx_lpf_horizontal_16_neon, &vpx_lpf_horizontal_16_c, 8),
make_tuple(&vpx_lpf_horizontal_16_dual_neon,
&vpx_lpf_horizontal_16_dual_c, 8),
make_tuple(&vpx_lpf_vertical_16_neon, &vpx_lpf_vertical_16_c, 8),
make_tuple(&vpx_lpf_vertical_16_dual_neon, &vpx_lpf_vertical_16_dual_c,
8),
@ -550,8 +635,9 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_dspr2, &vpx_lpf_horizontal_4_c, 8),
make_tuple(&vpx_lpf_horizontal_8_dspr2, &vpx_lpf_horizontal_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_8, &vpx_lpf_horizontal_edge_8, 8),
make_tuple(&vpx_lpf_horizontal_edge_16, &vpx_lpf_horizontal_edge_16, 8),
make_tuple(&vpx_lpf_horizontal_16_dspr2, &vpx_lpf_horizontal_16_c, 8),
make_tuple(&vpx_lpf_horizontal_16_dual_dspr2,
&vpx_lpf_horizontal_16_dual_c, 8),
make_tuple(&vpx_lpf_vertical_4_dspr2, &vpx_lpf_vertical_4_c, 8),
make_tuple(&vpx_lpf_vertical_8_dspr2, &vpx_lpf_vertical_8_c, 8),
make_tuple(&vpx_lpf_vertical_16_dspr2, &vpx_lpf_vertical_16_c, 8),
@ -576,10 +662,9 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values(
make_tuple(&vpx_lpf_horizontal_4_msa, &vpx_lpf_horizontal_4_c, 8),
make_tuple(&vpx_lpf_horizontal_8_msa, &vpx_lpf_horizontal_8_c, 8),
make_tuple(&vpx_lpf_horizontal_edge_8_msa, &vpx_lpf_horizontal_edge_8_c,
8),
make_tuple(&vpx_lpf_horizontal_edge_16_msa,
&vpx_lpf_horizontal_edge_16_c, 8),
make_tuple(&vpx_lpf_horizontal_16_msa, &vpx_lpf_horizontal_16_c, 8),
make_tuple(&vpx_lpf_horizontal_16_dual_msa,
&vpx_lpf_horizontal_16_dual_c, 8),
make_tuple(&vpx_lpf_vertical_4_msa, &vpx_lpf_vertical_4_c, 8),
make_tuple(&vpx_lpf_vertical_8_msa, &vpx_lpf_vertical_8_c, 8),
make_tuple(&vpx_lpf_vertical_16_msa, &vpx_lpf_vertical_16_c, 8)));

View File

@ -107,10 +107,10 @@ TEST_P(MinMaxTest, CompareReferenceAndVaryStride) {
int min_ref, max_ref, min, max;
reference_minmax(a, a_stride, b, b_stride, &min_ref, &max_ref);
ASM_REGISTER_STATE_CHECK(mm_func_(a, a_stride, b, b_stride, &min, &max));
EXPECT_EQ(max_ref, max) << "when a_stride = " << a_stride
<< " and b_stride = " << b_stride;
EXPECT_EQ(min_ref, min) << "when a_stride = " << a_stride
<< " and b_stride = " << b_stride;
EXPECT_EQ(max_ref, max)
<< "when a_stride = " << a_stride << " and b_stride = " << b_stride;
EXPECT_EQ(min_ref, min)
<< "when a_stride = " << a_stride << " and b_stride = " << b_stride;
}
}
}
@ -127,4 +127,9 @@ INSTANTIATE_TEST_CASE_P(NEON, MinMaxTest,
::testing::Values(&vpx_minmax_8x8_neon));
#endif
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(MSA, MinMaxTest,
::testing::Values(&vpx_minmax_8x8_msa));
#endif
} // namespace

File diff suppressed because it is too large Load Diff

View File

@ -7,22 +7,42 @@
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <limits.h>
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"
#include "test/acm_random.h"
#include "test/buffer.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"
#include "vpx/vpx_integer.h"
#include "vpx_mem/vpx_mem.h"
typedef void (*PostProcFunc)(unsigned char *src_ptr, unsigned char *dst_ptr,
int src_pixels_per_line, int dst_pixels_per_line,
int cols, unsigned char *flimit, int size);
using libvpx_test::ACMRandom;
using libvpx_test::Buffer;
typedef void (*VpxPostProcDownAndAcrossMbRowFunc)(
unsigned char *src_ptr, unsigned char *dst_ptr, int src_pixels_per_line,
int dst_pixels_per_line, int cols, unsigned char *flimit, int size);
typedef void (*VpxMbPostProcAcrossIpFunc)(unsigned char *src, int pitch,
int rows, int cols, int flimit);
typedef void (*VpxMbPostProcDownFunc)(unsigned char *dst, int pitch, int rows,
int cols, int flimit);
namespace {
class VPxPostProcessingFilterTest
: public ::testing::TestWithParam<PostProcFunc> {
// Compute the filter level used in post proc from the loop filter strength
int q2mbl(int x) {
if (x < 20) x = 20;
x = 50 + (x - 50) * 10 / 8;
return x * x / 3;
}
class VpxPostProcDownAndAcrossMbRowTest
: public ::testing::TestWithParam<VpxPostProcDownAndAcrossMbRowFunc> {
public:
virtual void TearDown() { libvpx_test::ClearSystemState(); }
};
@ -30,31 +50,22 @@ class VPxPostProcessingFilterTest
// Test routine for the VPx post-processing function
// vpx_post_proc_down_and_across_mb_row_c.
TEST_P(VPxPostProcessingFilterTest, FilterOutputCheck) {
TEST_P(VpxPostProcDownAndAcrossMbRowTest, CheckFilterOutput) {
// Size of the underlying data block that will be filtered.
const int block_width = 16;
const int block_height = 16;
// 5-tap filter needs 2 padding rows above and below the block in the input.
const int input_width = block_width;
const int input_height = block_height + 4;
const int input_stride = input_width;
const int input_size = input_width * input_height;
Buffer<uint8_t> src_image = Buffer<uint8_t>(block_width, block_height, 2);
ASSERT_TRUE(src_image.Init());
// Filter extends output block by 8 samples at left and right edges.
const int output_width = block_width + 16;
const int output_height = block_height;
const int output_stride = output_width;
const int output_size = output_width * output_height;
// Though the left padding is only 8 bytes, the assembly code tries to
// read 16 bytes before the pointer.
Buffer<uint8_t> dst_image =
Buffer<uint8_t>(block_width, block_height, 8, 16, 8, 8);
ASSERT_TRUE(dst_image.Init());
uint8_t *const src_image =
reinterpret_cast<uint8_t *>(vpx_calloc(input_size, 1));
uint8_t *const dst_image =
reinterpret_cast<uint8_t *>(vpx_calloc(output_size, 1));
// Pointers to top-left pixel of block in the input and output images.
uint8_t *const src_image_ptr = src_image + (input_stride << 1);
uint8_t *const dst_image_ptr = dst_image + 8;
uint8_t *const flimits =
reinterpret_cast<uint8_t *>(vpx_memalign(16, block_width));
(void)memset(flimits, 255, block_width);
@ -62,53 +73,412 @@ TEST_P(VPxPostProcessingFilterTest, FilterOutputCheck) {
// Initialize pixels in the input:
// block pixels to value 1,
// border pixels to value 10.
(void)memset(src_image, 10, input_size);
uint8_t *pixel_ptr = src_image_ptr;
for (int i = 0; i < block_height; ++i) {
for (int j = 0; j < block_width; ++j) {
pixel_ptr[j] = 1;
}
pixel_ptr += input_stride;
}
src_image.SetPadding(10);
src_image.Set(1);
// Initialize pixels in the output to 99.
(void)memset(dst_image, 99, output_size);
dst_image.Set(99);
ASM_REGISTER_STATE_CHECK(GetParam()(src_image_ptr, dst_image_ptr,
input_stride, output_stride, block_width,
flimits, 16));
ASM_REGISTER_STATE_CHECK(GetParam()(
src_image.TopLeftPixel(), dst_image.TopLeftPixel(), src_image.stride(),
dst_image.stride(), block_width, flimits, 16));
static const uint8_t expected_data[block_height] = { 4, 3, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 3, 4 };
static const uint8_t kExpectedOutput[block_height] = {
4, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3, 4
};
pixel_ptr = dst_image_ptr;
uint8_t *pixel_ptr = dst_image.TopLeftPixel();
for (int i = 0; i < block_height; ++i) {
for (int j = 0; j < block_width; ++j) {
EXPECT_EQ(expected_data[i], pixel_ptr[j])
<< "VPxPostProcessingFilterTest failed with invalid filter output";
ASSERT_EQ(kExpectedOutput[i], pixel_ptr[j])
<< "at (" << i << ", " << j << ")";
}
pixel_ptr += output_stride;
pixel_ptr += dst_image.stride();
}
vpx_free(src_image);
vpx_free(dst_image);
vpx_free(flimits);
};
TEST_P(VpxPostProcDownAndAcrossMbRowTest, CheckCvsAssembly) {
// Size of the underlying data block that will be filtered.
// Y blocks are always a multiple of 16 wide and exactly 16 high. U and V
// blocks are always a multiple of 8 wide and exactly 8 high.
const int block_width = 136;
const int block_height = 16;
// 5-tap filter needs 2 padding rows above and below the block in the input.
// SSE2 reads in blocks of 16. Pad an extra 8 in case the width is not %16.
Buffer<uint8_t> src_image =
Buffer<uint8_t>(block_width, block_height, 2, 2, 10, 2);
ASSERT_TRUE(src_image.Init());
// Filter extends output block by 8 samples at left and right edges.
// Though the left padding is only 8 bytes, there is 'above' padding as well
// so when the assembly code tries to read 16 bytes before the pointer it is
// not a problem.
// SSE2 reads in blocks of 16. Pad an extra 8 in case the width is not %16.
Buffer<uint8_t> dst_image =
Buffer<uint8_t>(block_width, block_height, 8, 8, 16, 8);
ASSERT_TRUE(dst_image.Init());
Buffer<uint8_t> dst_image_ref = Buffer<uint8_t>(block_width, block_height, 8);
ASSERT_TRUE(dst_image_ref.Init());
// Filter values are set in blocks of 16 for Y and 8 for U/V. Each macroblock
// can have a different filter. SSE2 assembly reads flimits in blocks of 16 so
// it must be padded out.
const int flimits_width = block_width % 16 ? block_width + 8 : block_width;
uint8_t *const flimits =
reinterpret_cast<uint8_t *>(vpx_memalign(16, flimits_width));
ACMRandom rnd;
rnd.Reset(ACMRandom::DeterministicSeed());
// Initialize pixels in the input:
// block pixels to random values.
// border pixels to value 10.
src_image.SetPadding(10);
src_image.Set(&rnd, &ACMRandom::Rand8);
for (int blocks = 0; blocks < block_width; blocks += 8) {
(void)memset(flimits, 0, sizeof(*flimits) * flimits_width);
for (int f = 0; f < 255; f++) {
(void)memset(flimits + blocks, f, sizeof(*flimits) * 8);
dst_image.Set(0);
dst_image_ref.Set(0);
vpx_post_proc_down_and_across_mb_row_c(
src_image.TopLeftPixel(), dst_image_ref.TopLeftPixel(),
src_image.stride(), dst_image_ref.stride(), block_width, flimits,
block_height);
ASM_REGISTER_STATE_CHECK(
GetParam()(src_image.TopLeftPixel(), dst_image.TopLeftPixel(),
src_image.stride(), dst_image.stride(), block_width,
flimits, block_height));
ASSERT_TRUE(dst_image.CheckValues(dst_image_ref));
}
}
vpx_free(flimits);
}
class VpxMbPostProcAcrossIpTest
: public ::testing::TestWithParam<VpxMbPostProcAcrossIpFunc> {
public:
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
void SetCols(unsigned char *s, int rows, int cols, int src_width) {
for (int r = 0; r < rows; r++) {
for (int c = 0; c < cols; c++) {
s[c] = c;
}
s += src_width;
}
}
void RunComparison(const unsigned char *expected_output, unsigned char *src_c,
int rows, int cols, int src_pitch) {
for (int r = 0; r < rows; r++) {
for (int c = 0; c < cols; c++) {
ASSERT_EQ(expected_output[c], src_c[c])
<< "at (" << r << ", " << c << ")";
}
src_c += src_pitch;
}
}
void RunFilterLevel(unsigned char *s, int rows, int cols, int src_width,
int filter_level, const unsigned char *expected_output) {
ASM_REGISTER_STATE_CHECK(
GetParam()(s, src_width, rows, cols, filter_level));
RunComparison(expected_output, s, rows, cols, src_width);
}
};
TEST_P(VpxMbPostProcAcrossIpTest, CheckLowFilterOutput) {
const int rows = 16;
const int cols = 16;
Buffer<uint8_t> src = Buffer<uint8_t>(cols, rows, 8, 8, 17, 8);
ASSERT_TRUE(src.Init());
src.SetPadding(10);
SetCols(src.TopLeftPixel(), rows, cols, src.stride());
Buffer<uint8_t> expected_output = Buffer<uint8_t>(cols, rows, 0);
ASSERT_TRUE(expected_output.Init());
SetCols(expected_output.TopLeftPixel(), rows, cols, expected_output.stride());
RunFilterLevel(src.TopLeftPixel(), rows, cols, src.stride(), q2mbl(0),
expected_output.TopLeftPixel());
}
TEST_P(VpxMbPostProcAcrossIpTest, CheckMediumFilterOutput) {
const int rows = 16;
const int cols = 16;
Buffer<uint8_t> src = Buffer<uint8_t>(cols, rows, 8, 8, 17, 8);
ASSERT_TRUE(src.Init());
src.SetPadding(10);
SetCols(src.TopLeftPixel(), rows, cols, src.stride());
static const unsigned char kExpectedOutput[cols] = {
2, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 13
};
RunFilterLevel(src.TopLeftPixel(), rows, cols, src.stride(), q2mbl(70),
kExpectedOutput);
}
TEST_P(VpxMbPostProcAcrossIpTest, CheckHighFilterOutput) {
const int rows = 16;
const int cols = 16;
Buffer<uint8_t> src = Buffer<uint8_t>(cols, rows, 8, 8, 17, 8);
ASSERT_TRUE(src.Init());
src.SetPadding(10);
SetCols(src.TopLeftPixel(), rows, cols, src.stride());
static const unsigned char kExpectedOutput[cols] = {
2, 2, 3, 4, 4, 5, 6, 7, 8, 9, 10, 11, 11, 12, 13, 13
};
RunFilterLevel(src.TopLeftPixel(), rows, cols, src.stride(), INT_MAX,
kExpectedOutput);
SetCols(src.TopLeftPixel(), rows, cols, src.stride());
RunFilterLevel(src.TopLeftPixel(), rows, cols, src.stride(), q2mbl(100),
kExpectedOutput);
}
TEST_P(VpxMbPostProcAcrossIpTest, CheckCvsAssembly) {
const int rows = 16;
const int cols = 16;
Buffer<uint8_t> c_mem = Buffer<uint8_t>(cols, rows, 8, 8, 17, 8);
ASSERT_TRUE(c_mem.Init());
Buffer<uint8_t> asm_mem = Buffer<uint8_t>(cols, rows, 8, 8, 17, 8);
ASSERT_TRUE(asm_mem.Init());
// When level >= 100, the filter behaves the same as the level = INT_MAX
// When level < 20, it behaves the same as the level = 0
for (int level = 0; level < 100; level++) {
c_mem.SetPadding(10);
asm_mem.SetPadding(10);
SetCols(c_mem.TopLeftPixel(), rows, cols, c_mem.stride());
SetCols(asm_mem.TopLeftPixel(), rows, cols, asm_mem.stride());
vpx_mbpost_proc_across_ip_c(c_mem.TopLeftPixel(), c_mem.stride(), rows,
cols, q2mbl(level));
ASM_REGISTER_STATE_CHECK(GetParam()(
asm_mem.TopLeftPixel(), asm_mem.stride(), rows, cols, q2mbl(level)));
ASSERT_TRUE(asm_mem.CheckValues(c_mem));
}
}
class VpxMbPostProcDownTest
: public ::testing::TestWithParam<VpxMbPostProcDownFunc> {
public:
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
void SetRows(unsigned char *src_c, int rows, int cols, int src_width) {
for (int r = 0; r < rows; r++) {
memset(src_c, r, cols);
src_c += src_width;
}
}
void RunComparison(const unsigned char *expected_output, unsigned char *src_c,
int rows, int cols, int src_pitch) {
for (int r = 0; r < rows; r++) {
for (int c = 0; c < cols; c++) {
ASSERT_EQ(expected_output[r * rows + c], src_c[c])
<< "at (" << r << ", " << c << ")";
}
src_c += src_pitch;
}
}
void RunFilterLevel(unsigned char *s, int rows, int cols, int src_width,
int filter_level, const unsigned char *expected_output) {
ASM_REGISTER_STATE_CHECK(
GetParam()(s, src_width, rows, cols, filter_level));
RunComparison(expected_output, s, rows, cols, src_width);
}
};
TEST_P(VpxMbPostProcDownTest, CheckHighFilterOutput) {
const int rows = 16;
const int cols = 16;
Buffer<uint8_t> src_c = Buffer<uint8_t>(cols, rows, 8, 8, 8, 17);
ASSERT_TRUE(src_c.Init());
src_c.SetPadding(10);
SetRows(src_c.TopLeftPixel(), rows, cols, src_c.stride());
static const unsigned char kExpectedOutput[rows * cols] = {
2, 2, 1, 1, 2, 2, 2, 2, 2, 2, 1, 1, 2, 2, 2, 2, 2, 2, 2,
2, 3, 2, 2, 2, 2, 2, 2, 2, 3, 2, 2, 2, 3, 3, 3, 3, 3, 3,
3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 3, 4, 4, 3, 3, 3,
4, 4, 3, 4, 4, 3, 3, 4, 5, 4, 4, 4, 4, 4, 4, 4, 5, 4, 4,
4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7,
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8,
8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 9, 8, 9, 9, 8, 8, 8, 9,
9, 8, 9, 9, 8, 8, 8, 9, 9, 10, 10, 9, 9, 9, 10, 10, 9, 10, 10,
9, 9, 9, 10, 10, 10, 11, 10, 10, 10, 11, 10, 11, 10, 11, 10, 10, 10, 11,
10, 11, 11, 11, 11, 11, 11, 11, 12, 11, 11, 11, 11, 11, 11, 11, 12, 11, 12,
12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 13, 12,
13, 12, 13, 12, 12, 12, 13, 12, 13, 12, 13, 12, 13, 13, 13, 14, 13, 13, 13,
13, 13, 13, 13, 14, 13, 13, 13, 13
};
RunFilterLevel(src_c.TopLeftPixel(), rows, cols, src_c.stride(), INT_MAX,
kExpectedOutput);
src_c.SetPadding(10);
SetRows(src_c.TopLeftPixel(), rows, cols, src_c.stride());
RunFilterLevel(src_c.TopLeftPixel(), rows, cols, src_c.stride(), q2mbl(100),
kExpectedOutput);
}
TEST_P(VpxMbPostProcDownTest, CheckMediumFilterOutput) {
const int rows = 16;
const int cols = 16;
Buffer<uint8_t> src_c = Buffer<uint8_t>(cols, rows, 8, 8, 8, 17);
ASSERT_TRUE(src_c.Init());
src_c.SetPadding(10);
SetRows(src_c.TopLeftPixel(), rows, cols, src_c.stride());
static const unsigned char kExpectedOutput[rows * cols] = {
2, 2, 1, 1, 2, 2, 2, 2, 2, 2, 1, 1, 2, 2, 2, 2, 2, 2, 2,
2, 3, 2, 2, 2, 2, 2, 2, 2, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3,
3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 7, 7,
7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8,
8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 9, 9,
9, 9, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10, 10,
10, 10, 10, 10, 10, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11, 11,
11, 11, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 13,
13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 13, 12, 12, 13, 12,
13, 12, 13, 12, 12, 12, 13, 12, 13, 12, 13, 12, 13, 13, 13, 14, 13, 13, 13,
13, 13, 13, 13, 14, 13, 13, 13, 13
};
RunFilterLevel(src_c.TopLeftPixel(), rows, cols, src_c.stride(), q2mbl(70),
kExpectedOutput);
}
TEST_P(VpxMbPostProcDownTest, CheckLowFilterOutput) {
const int rows = 16;
const int cols = 16;
Buffer<uint8_t> src_c = Buffer<uint8_t>(cols, rows, 8, 8, 8, 17);
ASSERT_TRUE(src_c.Init());
src_c.SetPadding(10);
SetRows(src_c.TopLeftPixel(), rows, cols, src_c.stride());
unsigned char *expected_output = new unsigned char[rows * cols];
ASSERT_TRUE(expected_output != NULL);
SetRows(expected_output, rows, cols, cols);
RunFilterLevel(src_c.TopLeftPixel(), rows, cols, src_c.stride(), q2mbl(0),
expected_output);
delete[] expected_output;
}
TEST_P(VpxMbPostProcDownTest, CheckCvsAssembly) {
const int rows = 16;
const int cols = 16;
ACMRandom rnd;
rnd.Reset(ACMRandom::DeterministicSeed());
Buffer<uint8_t> src_c = Buffer<uint8_t>(cols, rows, 8, 8, 8, 17);
ASSERT_TRUE(src_c.Init());
Buffer<uint8_t> src_asm = Buffer<uint8_t>(cols, rows, 8, 8, 8, 17);
ASSERT_TRUE(src_asm.Init());
for (int level = 0; level < 100; level++) {
src_c.SetPadding(10);
src_asm.SetPadding(10);
src_c.Set(&rnd, &ACMRandom::Rand8);
src_asm.CopyFrom(src_c);
vpx_mbpost_proc_down_c(src_c.TopLeftPixel(), src_c.stride(), rows, cols,
q2mbl(level));
ASM_REGISTER_STATE_CHECK(GetParam()(
src_asm.TopLeftPixel(), src_asm.stride(), rows, cols, q2mbl(level)));
ASSERT_TRUE(src_asm.CheckValues(src_c));
src_c.SetPadding(10);
src_asm.SetPadding(10);
src_c.Set(&rnd, &ACMRandom::Rand8Extremes);
src_asm.CopyFrom(src_c);
vpx_mbpost_proc_down_c(src_c.TopLeftPixel(), src_c.stride(), rows, cols,
q2mbl(level));
ASM_REGISTER_STATE_CHECK(GetParam()(
src_asm.TopLeftPixel(), src_asm.stride(), rows, cols, q2mbl(level)));
ASSERT_TRUE(src_asm.CheckValues(src_c));
}
}
INSTANTIATE_TEST_CASE_P(
C, VPxPostProcessingFilterTest,
C, VpxPostProcDownAndAcrossMbRowTest,
::testing::Values(vpx_post_proc_down_and_across_mb_row_c));
INSTANTIATE_TEST_CASE_P(C, VpxMbPostProcAcrossIpTest,
::testing::Values(vpx_mbpost_proc_across_ip_c));
INSTANTIATE_TEST_CASE_P(C, VpxMbPostProcDownTest,
::testing::Values(vpx_mbpost_proc_down_c));
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(
SSE2, VPxPostProcessingFilterTest,
SSE2, VpxPostProcDownAndAcrossMbRowTest,
::testing::Values(vpx_post_proc_down_and_across_mb_row_sse2));
#endif
INSTANTIATE_TEST_CASE_P(SSE2, VpxMbPostProcAcrossIpTest,
::testing::Values(vpx_mbpost_proc_across_ip_sse2));
INSTANTIATE_TEST_CASE_P(SSE2, VpxMbPostProcDownTest,
::testing::Values(vpx_mbpost_proc_down_sse2));
#endif // HAVE_SSE2
#if HAVE_NEON
INSTANTIATE_TEST_CASE_P(
NEON, VpxPostProcDownAndAcrossMbRowTest,
::testing::Values(vpx_post_proc_down_and_across_mb_row_neon));
INSTANTIATE_TEST_CASE_P(NEON, VpxMbPostProcAcrossIpTest,
::testing::Values(vpx_mbpost_proc_across_ip_neon));
INSTANTIATE_TEST_CASE_P(NEON, VpxMbPostProcDownTest,
::testing::Values(vpx_mbpost_proc_down_neon));
#endif // HAVE_NEON
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(
MSA, VPxPostProcessingFilterTest,
MSA, VpxPostProcDownAndAcrossMbRowTest,
::testing::Values(vpx_post_proc_down_and_across_mb_row_msa));
#endif
INSTANTIATE_TEST_CASE_P(MSA, VpxMbPostProcAcrossIpTest,
::testing::Values(vpx_mbpost_proc_across_ip_msa));
INSTANTIATE_TEST_CASE_P(MSA, VpxMbPostProcDownTest,
::testing::Values(vpx_mbpost_proc_down_msa));
#endif // HAVE_MSA
} // namespace

View File

@ -292,15 +292,13 @@ INSTANTIATE_TEST_CASE_P(
NEON, SixtapPredictTest,
::testing::Values(make_tuple(16, 16, &vp8_sixtap_predict16x16_neon),
make_tuple(8, 8, &vp8_sixtap_predict8x8_neon),
make_tuple(8, 4, &vp8_sixtap_predict8x4_neon)));
make_tuple(8, 4, &vp8_sixtap_predict8x4_neon),
make_tuple(4, 4, &vp8_sixtap_predict4x4_neon)));
#endif
#if HAVE_MMX
INSTANTIATE_TEST_CASE_P(
MMX, SixtapPredictTest,
::testing::Values(make_tuple(16, 16, &vp8_sixtap_predict16x16_mmx),
make_tuple(8, 8, &vp8_sixtap_predict8x8_mmx),
make_tuple(8, 4, &vp8_sixtap_predict8x4_mmx),
make_tuple(4, 4, &vp8_sixtap_predict4x4_mmx)));
::testing::Values(make_tuple(4, 4, &vp8_sixtap_predict4x4_mmx)));
#endif
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(
@ -326,6 +324,15 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(4, 4, &vp8_sixtap_predict4x4_msa)));
#endif
#if HAVE_MMI
INSTANTIATE_TEST_CASE_P(
MMI, SixtapPredictTest,
::testing::Values(make_tuple(16, 16, &vp8_sixtap_predict16x16_mmi),
make_tuple(8, 8, &vp8_sixtap_predict8x8_mmi),
make_tuple(8, 4, &vp8_sixtap_predict8x4_mmi),
make_tuple(4, 4, &vp8_sixtap_predict4x4_mmi)));
#endif
class BilinearPredictTest : public PredictTestBase {};
TEST_P(BilinearPredictTest, TestWithRandomData) {
@ -352,9 +359,7 @@ INSTANTIATE_TEST_CASE_P(
#if HAVE_MMX
INSTANTIATE_TEST_CASE_P(
MMX, BilinearPredictTest,
::testing::Values(make_tuple(16, 16, &vp8_bilinear_predict16x16_mmx),
make_tuple(8, 8, &vp8_bilinear_predict8x8_mmx),
make_tuple(8, 4, &vp8_bilinear_predict8x4_mmx),
::testing::Values(make_tuple(8, 4, &vp8_bilinear_predict8x4_mmx),
make_tuple(4, 4, &vp8_bilinear_predict4x4_mmx)));
#endif
#if HAVE_SSE2

View File

@ -200,4 +200,12 @@ INSTANTIATE_TEST_CASE_P(
make_tuple(&vp8_fast_quantize_b_msa, &vp8_fast_quantize_b_c),
make_tuple(&vp8_regular_quantize_b_msa, &vp8_regular_quantize_b_c)));
#endif // HAVE_MSA
#if HAVE_MMI
INSTANTIATE_TEST_CASE_P(
MMI, QuantizeTest,
::testing::Values(
make_tuple(&vp8_fast_quantize_b_mmi, &vp8_fast_quantize_b_c),
make_tuple(&vp8_regular_quantize_b_mmi, &vp8_regular_quantize_b_c)));
#endif // HAVE_MMI
} // namespace

View File

@ -32,7 +32,9 @@
#undef NOMINMAX
#define NOMINMAX
#ifndef WIN32_LEAN_AND_MEAN
#define WIN32_LEAN_AND_MEAN
#endif
#include <windows.h>
#include <winnt.h>
@ -111,8 +113,8 @@ class RegisterStateCheck {
int64_t post_store[8];
vpx_push_neon(post_store);
for (int i = 0; i < 8; ++i) {
EXPECT_EQ(pre_store_[i], post_store[i]) << "d" << i + 8
<< " has been modified";
EXPECT_EQ(pre_store_[i], post_store[i])
<< "d" << i + 8 << " has been modified";
}
}

View File

@ -298,10 +298,10 @@ TEST_P(ResizeTest, TestExternalResizeWorks) {
unsigned int expected_h;
ScaleForFrameNumber(frame, kInitialWidth, kInitialHeight, &expected_w,
&expected_h, 0);
EXPECT_EQ(expected_w, info->w) << "Frame " << frame
<< " had unexpected width";
EXPECT_EQ(expected_h, info->h) << "Frame " << frame
<< " had unexpected height";
EXPECT_EQ(expected_w, info->w)
<< "Frame " << frame << " had unexpected width";
EXPECT_EQ(expected_h, info->h)
<< "Frame " << frame << " had unexpected height";
}
}
@ -513,10 +513,10 @@ TEST_P(ResizeRealtimeTest, TestExternalResizeWorks) {
unsigned int expected_h;
ScaleForFrameNumber(frame, kInitialWidth, kInitialHeight, &expected_w,
&expected_h, 1);
EXPECT_EQ(expected_w, info->w) << "Frame " << frame
<< " had unexpected width";
EXPECT_EQ(expected_h, info->h) << "Frame " << frame
<< " had unexpected height";
EXPECT_EQ(expected_w, info->w)
<< "Frame " << frame << " had unexpected width";
EXPECT_EQ(expected_h, info->h)
<< "Frame " << frame << " had unexpected height";
EXPECT_EQ(static_cast<unsigned int>(0), GetMismatchFrames());
}
}

View File

@ -644,19 +644,50 @@ INSTANTIATE_TEST_CASE_P(C, SADx4Test, ::testing::ValuesIn(x4d_c_tests));
#if HAVE_NEON
const SadMxNParam neon_tests[] = {
SadMxNParam(64, 64, &vpx_sad64x64_neon),
SadMxNParam(64, 32, &vpx_sad64x32_neon),
SadMxNParam(32, 32, &vpx_sad32x32_neon),
SadMxNParam(16, 32, &vpx_sad16x32_neon),
SadMxNParam(16, 16, &vpx_sad16x16_neon),
SadMxNParam(16, 8, &vpx_sad16x8_neon),
SadMxNParam(8, 16, &vpx_sad8x16_neon),
SadMxNParam(8, 8, &vpx_sad8x8_neon),
SadMxNParam(8, 4, &vpx_sad8x4_neon),
SadMxNParam(4, 8, &vpx_sad4x8_neon),
SadMxNParam(4, 4, &vpx_sad4x4_neon),
};
INSTANTIATE_TEST_CASE_P(NEON, SADTest, ::testing::ValuesIn(neon_tests));
const SadMxNAvgParam avg_neon_tests[] = {
SadMxNAvgParam(64, 64, &vpx_sad64x64_avg_neon),
SadMxNAvgParam(64, 32, &vpx_sad64x32_avg_neon),
SadMxNAvgParam(32, 64, &vpx_sad32x64_avg_neon),
SadMxNAvgParam(32, 32, &vpx_sad32x32_avg_neon),
SadMxNAvgParam(32, 16, &vpx_sad32x16_avg_neon),
SadMxNAvgParam(16, 32, &vpx_sad16x32_avg_neon),
SadMxNAvgParam(16, 16, &vpx_sad16x16_avg_neon),
SadMxNAvgParam(16, 8, &vpx_sad16x8_avg_neon),
SadMxNAvgParam(8, 16, &vpx_sad8x16_avg_neon),
SadMxNAvgParam(8, 8, &vpx_sad8x8_avg_neon),
SadMxNAvgParam(8, 4, &vpx_sad8x4_avg_neon),
SadMxNAvgParam(4, 8, &vpx_sad4x8_avg_neon),
SadMxNAvgParam(4, 4, &vpx_sad4x4_avg_neon),
};
INSTANTIATE_TEST_CASE_P(NEON, SADavgTest, ::testing::ValuesIn(avg_neon_tests));
const SadMxNx4Param x4d_neon_tests[] = {
SadMxNx4Param(64, 64, &vpx_sad64x64x4d_neon),
SadMxNx4Param(64, 32, &vpx_sad64x32x4d_neon),
SadMxNx4Param(32, 64, &vpx_sad32x64x4d_neon),
SadMxNx4Param(32, 32, &vpx_sad32x32x4d_neon),
SadMxNx4Param(32, 16, &vpx_sad32x16x4d_neon),
SadMxNx4Param(16, 32, &vpx_sad16x32x4d_neon),
SadMxNx4Param(16, 16, &vpx_sad16x16x4d_neon),
SadMxNx4Param(16, 8, &vpx_sad16x8x4d_neon),
SadMxNx4Param(8, 16, &vpx_sad8x16x4d_neon),
SadMxNx4Param(8, 8, &vpx_sad8x8x4d_neon),
SadMxNx4Param(8, 4, &vpx_sad8x4x4d_neon),
SadMxNx4Param(4, 8, &vpx_sad4x8x4d_neon),
SadMxNx4Param(4, 4, &vpx_sad4x4x4d_neon),
};
INSTANTIATE_TEST_CASE_P(NEON, SADx4Test, ::testing::ValuesIn(x4d_neon_tests));
#endif // HAVE_NEON
@ -865,6 +896,14 @@ const SadMxNx4Param x4d_avx2_tests[] = {
INSTANTIATE_TEST_CASE_P(AVX2, SADx4Test, ::testing::ValuesIn(x4d_avx2_tests));
#endif // HAVE_AVX2
#if HAVE_AVX512
const SadMxNx4Param x4d_avx512_tests[] = {
SadMxNx4Param(64, 64, &vpx_sad64x64x4d_avx512),
};
INSTANTIATE_TEST_CASE_P(AVX512, SADx4Test,
::testing::ValuesIn(x4d_avx512_tests));
#endif // HAVE_AVX512
//------------------------------------------------------------------------------
// MIPS functions
#if HAVE_MSA
@ -920,4 +959,98 @@ const SadMxNx4Param x4d_msa_tests[] = {
INSTANTIATE_TEST_CASE_P(MSA, SADx4Test, ::testing::ValuesIn(x4d_msa_tests));
#endif // HAVE_MSA
//------------------------------------------------------------------------------
// VSX functions
#if HAVE_VSX
const SadMxNParam vsx_tests[] = {
SadMxNParam(64, 64, &vpx_sad64x64_vsx),
SadMxNParam(64, 32, &vpx_sad64x32_vsx),
SadMxNParam(32, 64, &vpx_sad32x64_vsx),
SadMxNParam(32, 32, &vpx_sad32x32_vsx),
SadMxNParam(32, 16, &vpx_sad32x16_vsx),
SadMxNParam(16, 32, &vpx_sad16x32_vsx),
SadMxNParam(16, 16, &vpx_sad16x16_vsx),
SadMxNParam(16, 8, &vpx_sad16x8_vsx),
};
INSTANTIATE_TEST_CASE_P(VSX, SADTest, ::testing::ValuesIn(vsx_tests));
const SadMxNAvgParam avg_vsx_tests[] = {
SadMxNAvgParam(64, 64, &vpx_sad64x64_avg_vsx),
SadMxNAvgParam(64, 32, &vpx_sad64x32_avg_vsx),
SadMxNAvgParam(32, 64, &vpx_sad32x64_avg_vsx),
SadMxNAvgParam(32, 32, &vpx_sad32x32_avg_vsx),
SadMxNAvgParam(32, 16, &vpx_sad32x16_avg_vsx),
SadMxNAvgParam(16, 32, &vpx_sad16x32_avg_vsx),
SadMxNAvgParam(16, 16, &vpx_sad16x16_avg_vsx),
SadMxNAvgParam(16, 8, &vpx_sad16x8_avg_vsx),
};
INSTANTIATE_TEST_CASE_P(VSX, SADavgTest, ::testing::ValuesIn(avg_vsx_tests));
const SadMxNx4Param x4d_vsx_tests[] = {
SadMxNx4Param(64, 64, &vpx_sad64x64x4d_vsx),
SadMxNx4Param(64, 32, &vpx_sad64x32x4d_vsx),
SadMxNx4Param(32, 64, &vpx_sad32x64x4d_vsx),
SadMxNx4Param(32, 32, &vpx_sad32x32x4d_vsx),
SadMxNx4Param(32, 16, &vpx_sad32x16x4d_vsx),
SadMxNx4Param(16, 32, &vpx_sad16x32x4d_vsx),
SadMxNx4Param(16, 16, &vpx_sad16x16x4d_vsx),
SadMxNx4Param(16, 8, &vpx_sad16x8x4d_vsx),
};
INSTANTIATE_TEST_CASE_P(VSX, SADx4Test, ::testing::ValuesIn(x4d_vsx_tests));
#endif // HAVE_VSX
//------------------------------------------------------------------------------
// Loongson functions
#if HAVE_MMI
const SadMxNParam mmi_tests[] = {
SadMxNParam(64, 64, &vpx_sad64x64_mmi),
SadMxNParam(64, 32, &vpx_sad64x32_mmi),
SadMxNParam(32, 64, &vpx_sad32x64_mmi),
SadMxNParam(32, 32, &vpx_sad32x32_mmi),
SadMxNParam(32, 16, &vpx_sad32x16_mmi),
SadMxNParam(16, 32, &vpx_sad16x32_mmi),
SadMxNParam(16, 16, &vpx_sad16x16_mmi),
SadMxNParam(16, 8, &vpx_sad16x8_mmi),
SadMxNParam(8, 16, &vpx_sad8x16_mmi),
SadMxNParam(8, 8, &vpx_sad8x8_mmi),
SadMxNParam(8, 4, &vpx_sad8x4_mmi),
SadMxNParam(4, 8, &vpx_sad4x8_mmi),
SadMxNParam(4, 4, &vpx_sad4x4_mmi),
};
INSTANTIATE_TEST_CASE_P(MMI, SADTest, ::testing::ValuesIn(mmi_tests));
const SadMxNAvgParam avg_mmi_tests[] = {
SadMxNAvgParam(64, 64, &vpx_sad64x64_avg_mmi),
SadMxNAvgParam(64, 32, &vpx_sad64x32_avg_mmi),
SadMxNAvgParam(32, 64, &vpx_sad32x64_avg_mmi),
SadMxNAvgParam(32, 32, &vpx_sad32x32_avg_mmi),
SadMxNAvgParam(32, 16, &vpx_sad32x16_avg_mmi),
SadMxNAvgParam(16, 32, &vpx_sad16x32_avg_mmi),
SadMxNAvgParam(16, 16, &vpx_sad16x16_avg_mmi),
SadMxNAvgParam(16, 8, &vpx_sad16x8_avg_mmi),
SadMxNAvgParam(8, 16, &vpx_sad8x16_avg_mmi),
SadMxNAvgParam(8, 8, &vpx_sad8x8_avg_mmi),
SadMxNAvgParam(8, 4, &vpx_sad8x4_avg_mmi),
SadMxNAvgParam(4, 8, &vpx_sad4x8_avg_mmi),
SadMxNAvgParam(4, 4, &vpx_sad4x4_avg_mmi),
};
INSTANTIATE_TEST_CASE_P(MMI, SADavgTest, ::testing::ValuesIn(avg_mmi_tests));
const SadMxNx4Param x4d_mmi_tests[] = {
SadMxNx4Param(64, 64, &vpx_sad64x64x4d_mmi),
SadMxNx4Param(64, 32, &vpx_sad64x32x4d_mmi),
SadMxNx4Param(32, 64, &vpx_sad32x64x4d_mmi),
SadMxNx4Param(32, 32, &vpx_sad32x32x4d_mmi),
SadMxNx4Param(32, 16, &vpx_sad32x16x4d_mmi),
SadMxNx4Param(16, 32, &vpx_sad16x32x4d_mmi),
SadMxNx4Param(16, 16, &vpx_sad16x16x4d_mmi),
SadMxNx4Param(16, 8, &vpx_sad16x8x4d_mmi),
SadMxNx4Param(8, 16, &vpx_sad8x16x4d_mmi),
SadMxNx4Param(8, 8, &vpx_sad8x8x4d_mmi),
SadMxNx4Param(8, 4, &vpx_sad8x4x4d_mmi),
SadMxNx4Param(4, 8, &vpx_sad4x8x4d_mmi),
SadMxNx4Param(4, 4, &vpx_sad4x4x4d_mmi),
};
INSTANTIATE_TEST_CASE_P(MMI, SADx4Test, ::testing::ValuesIn(x4d_mmi_tests));
#endif // HAVE_MMI
} // namespace

View File

@ -146,14 +146,6 @@ TEST(VP8RoiMapTest, ParameterCheck) {
if (deltas_valid != roi_retval) break;
}
// Test that we report and error if cyclic refresh is enabled.
cpi.cyclic_refresh_mode_enabled = 1;
roi_retval =
vp8_set_roimap(&cpi, roi_map, cpi.common.mb_rows, cpi.common.mb_cols,
delta_q, delta_lf, threshold);
EXPECT_EQ(-1, roi_retval) << "cyclic refresh check error";
cpi.cyclic_refresh_mode_enabled = 0;
// Test invalid number of rows or colums.
roi_retval =
vp8_set_roimap(&cpi, roi_map, cpi.common.mb_rows + 1,

169
libs/libvpx/test/stress.sh Executable file
View File

@ -0,0 +1,169 @@
#!/bin/sh
##
## Copyright (c) 2016 The WebM project authors. All Rights Reserved.
##
## Use of this source code is governed by a BSD-style license
## that can be found in the LICENSE file in the root of the source
## tree. An additional intellectual property rights grant can be found
## in the file PATENTS. All contributing project authors may
## be found in the AUTHORS file in the root of the source tree.
##
## This file performs a stress test. It runs (STRESS_ONEPASS_MAX_JOBS,
## default=5) one, (STRESS_TWOPASS_MAX_JOBS, default=5) two pass &
## (STRESS_RT_MAX_JOBS, default=5) encodes and (STRESS_<codec>_DECODE_MAX_JOBS,
## default=30) decodes in parallel.
. $(dirname $0)/tools_common.sh
YUV="${LIBVPX_TEST_DATA_PATH}/niklas_1280_720_30.yuv"
VP8="${LIBVPX_TEST_DATA_PATH}/tos_vp8.webm"
VP9="${LIBVPX_TEST_DATA_PATH}/vp90-2-sintel_1920x818_tile_1x4_fpm_2279kbps.webm"
DATA_URL="http://downloads.webmproject.org/test_data/libvpx/"
SHA1_FILE="$(dirname $0)/test-data.sha1"
# Set sha1sum to proper sha program (sha1sum, shasum, sha1). This code is
# cribbed from libs.mk.
[ -x "$(which sha1sum)" ] && sha1sum=sha1sum
[ -x "$(which shasum)" ] && sha1sum=shasum
[ -x "$(which sha1)" ] && sha1sum=sha1
# Download a file from the url and check its sha1sum.
download_and_check_file() {
# Get the file from the file path.
local readonly root="${1#${LIBVPX_TEST_DATA_PATH}/}"
# Download the file using curl. Trap to insure non partial file.
(trap "rm -f $1" INT TERM \
&& eval "curl --retry 1 -L -o $1 ${DATA_URL}${root} ${devnull}")
# Check the sha1 sum of the file.
if [ -n "${sha1sum}" ]; then
set -e
grep ${root} ${SHA1_FILE} \
| (cd ${LIBVPX_TEST_DATA_PATH}; ${sha1sum} -c);
fi
}
# Environment check: Make sure input is available.
stress_verify_environment() {
if [ ! -e "${SHA1_FILE}" ] ; then
echo "Missing ${SHA1_FILE}"
return 1
fi
for file in "${YUV}" "${VP8}" "${VP9}"; do
if [ ! -e "${file}" ] ; then
download_and_check_file "${file}"
fi
done
if [ ! -e "${YUV}" ] || [ ! -e "${VP8}" ] || [ ! -e "${VP9}" ] ; then
elog "Libvpx test data must exist in LIBVPX_TEST_DATA_PATH."
return 1
fi
if [ -z "$(vpx_tool_path vpxenc)" ]; then
elog "vpxenc not found. It must exist in LIBVPX_BIN_PATH or its parent."
return 1
fi
if [ -z "$(vpx_tool_path vpxdec)" ]; then
elog "vpxdec not found. It must exist in LIBVPX_BIN_PATH or its parent."
return 1
fi
}
# This function runs tests on libvpx that run multiple encodes and decodes
# in parallel in hopes of catching synchronization and/or threading issues.
stress() {
local readonly decoder="$(vpx_tool_path vpxdec)"
local readonly encoder="$(vpx_tool_path vpxenc)"
local readonly codec="$1"
local readonly webm="$2"
local readonly decode_count="$3"
local readonly threads="$4"
local readonly enc_args="$5"
local pids=""
local rt_max_jobs=${STRESS_RT_MAX_JOBS:-5}
local onepass_max_jobs=${STRESS_ONEPASS_MAX_JOBS:-5}
local twopass_max_jobs=${STRESS_TWOPASS_MAX_JOBS:-5}
# Enable job control, so we can run multiple processes.
set -m
# Start $onepass_max_jobs encode jobs in parallel.
for i in $(seq ${onepass_max_jobs}); do
bitrate=$(($i * 20 + 300))
eval "${VPX_TEST_PREFIX}" "${encoder}" "--codec=${codec} -w 1280 -h 720" \
"${YUV}" "-t ${threads} --limit=150 --test-decode=fatal --passes=1" \
"--target-bitrate=${bitrate} -o ${VPX_TEST_OUTPUT_DIR}/${i}.1pass.webm" \
"${enc_args}" ${devnull} &
pids="${pids} $!"
done
# Start $twopass_max_jobs encode jobs in parallel.
for i in $(seq ${twopass_max_jobs}); do
bitrate=$(($i * 20 + 300))
eval "${VPX_TEST_PREFIX}" "${encoder}" "--codec=${codec} -w 1280 -h 720" \
"${YUV}" "-t ${threads} --limit=150 --test-decode=fatal --passes=2" \
"--target-bitrate=${bitrate} -o ${VPX_TEST_OUTPUT_DIR}/${i}.2pass.webm" \
"${enc_args}" ${devnull} &
pids="${pids} $!"
done
# Start $rt_max_jobs rt encode jobs in parallel.
for i in $(seq ${rt_max_jobs}); do
bitrate=$(($i * 20 + 300))
eval "${VPX_TEST_PREFIX}" "${encoder}" "--codec=${codec} -w 1280 -h 720" \
"${YUV}" "-t ${threads} --limit=150 --test-decode=fatal " \
"--target-bitrate=${bitrate} --lag-in-frames=0 --error-resilient=1" \
"--kf-min-dist=3000 --kf-max-dist=3000 --cpu-used=-6 --static-thresh=1" \
"--end-usage=cbr --min-q=2 --max-q=56 --undershoot-pct=100" \
"--overshoot-pct=15 --buf-sz=1000 --buf-initial-sz=500" \
"--buf-optimal-sz=600 --max-intra-rate=900 --resize-allowed=0" \
"--drop-frame=0 --passes=1 --rt --noise-sensitivity=4" \
"-o ${VPX_TEST_OUTPUT_DIR}/${i}.rt.webm" ${devnull} &
pids="${pids} $!"
done
# Start $decode_count decode jobs in parallel.
for i in $(seq "${decode_count}"); do
eval "${decoder}" "-t ${threads}" "${webm}" "--noblit" ${devnull} &
pids="${pids} $!"
done
# Wait for all parallel jobs to finish.
fail=0
for job in "${pids}"; do
wait $job || fail=$(($fail + 1))
done
return $fail
}
vp8_stress_test() {
local vp8_max_jobs=${STRESS_VP8_DECODE_MAX_JOBS:-40}
if [ "$(vp8_decode_available)" = "yes" -a \
"$(vp8_encode_available)" = "yes" ]; then
stress vp8 "${VP8}" "${vp8_max_jobs}" 4
fi
}
vp9_stress() {
local vp9_max_jobs=${STRESS_VP9_DECODE_MAX_JOBS:-25}
if [ "$(vp9_decode_available)" = "yes" -a \
"$(vp9_encode_available)" = "yes" ]; then
stress vp9 "${VP9}" "${vp9_max_jobs}" "$@"
fi
}
vp9_stress_test() {
for threads in 4 8 100; do
vp9_stress "$threads" "--row-mt=0"
done
}
vp9_stress_test_row_mt() {
for threads in 4 8 100; do
vp9_stress "$threads" "--row-mt=1"
done
}
run_tests stress_verify_environment \
"vp8_stress_test vp9_stress_test vp9_stress_test_row_mt"

View File

@ -110,4 +110,10 @@ INSTANTIATE_TEST_CASE_P(
::testing::Values(make_tuple(&vpx_sum_squares_2d_i16_c,
&vpx_sum_squares_2d_i16_sse2)));
#endif // HAVE_SSE2
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(MSA, SumSquaresTest, ::testing::Values(make_tuple(
&vpx_sum_squares_2d_i16_c,
&vpx_sum_squares_2d_i16_msa)));
#endif // HAVE_MSA
} // namespace

View File

@ -438,7 +438,7 @@ TEST_F(SvcTest, SetAutoAltRefOption) {
// Test that decoder can handle an SVC frame as the first frame in a sequence.
TEST_F(SvcTest, OnePassEncodeOneFrame) {
codec_enc_.g_pass = VPX_RC_ONE_PASS;
vpx_fixed_buf output = { 0 };
vpx_fixed_buf output = vpx_fixed_buf();
Pass2EncodeNFrames(NULL, 1, 2, &output);
DecodeNFrames(&output, 1);
FreeBitstreamBuffers(&output, 1);

View File

@ -0,0 +1,277 @@
/*
* Copyright (c) 2016 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <limits>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vp9_rtcd.h"
#include "test/acm_random.h"
#include "test/buffer.h"
#include "test/register_state_check.h"
#include "vpx_ports/vpx_timer.h"
namespace {
using ::libvpx_test::ACMRandom;
using ::libvpx_test::Buffer;
typedef void (*TemporalFilterFunc)(const uint8_t *a, unsigned int stride,
const uint8_t *b, unsigned int w,
unsigned int h, int filter_strength,
int filter_weight, unsigned int *accumulator,
uint16_t *count);
// Calculate the difference between 'a' and 'b', sum in blocks of 9, and apply
// filter based on strength and weight. Store the resulting filter amount in
// 'count' and apply it to 'b' and store it in 'accumulator'.
void reference_filter(const Buffer<uint8_t> &a, const Buffer<uint8_t> &b, int w,
int h, int filter_strength, int filter_weight,
Buffer<unsigned int> *accumulator,
Buffer<uint16_t> *count) {
Buffer<int> diff_sq = Buffer<int>(w, h, 0);
ASSERT_TRUE(diff_sq.Init());
diff_sq.Set(0);
int rounding = 0;
if (filter_strength > 0) {
rounding = 1 << (filter_strength - 1);
}
// Calculate all the differences. Avoids re-calculating a bunch of extra
// values.
for (int height = 0; height < h; ++height) {
for (int width = 0; width < w; ++width) {
int diff = a.TopLeftPixel()[height * a.stride() + width] -
b.TopLeftPixel()[height * b.stride() + width];
diff_sq.TopLeftPixel()[height * diff_sq.stride() + width] = diff * diff;
}
}
// For any given point, sum the neighboring values and calculate the
// modifier.
for (int height = 0; height < h; ++height) {
for (int width = 0; width < w; ++width) {
// Determine how many values are being summed.
int summed_values = 9;
if (height == 0 || height == (h - 1)) {
summed_values -= 3;
}
if (width == 0 || width == (w - 1)) {
if (summed_values == 6) { // corner
summed_values -= 2;
} else {
summed_values -= 3;
}
}
// Sum the diff_sq of the surrounding values.
int sum = 0;
for (int idy = -1; idy <= 1; ++idy) {
for (int idx = -1; idx <= 1; ++idx) {
const int y = height + idy;
const int x = width + idx;
// If inside the border.
if (y >= 0 && y < h && x >= 0 && x < w) {
sum += diff_sq.TopLeftPixel()[y * diff_sq.stride() + x];
}
}
}
sum *= 3;
sum /= summed_values;
sum += rounding;
sum >>= filter_strength;
// Clamp the value and invert it.
if (sum > 16) sum = 16;
sum = 16 - sum;
sum *= filter_weight;
count->TopLeftPixel()[height * count->stride() + width] += sum;
accumulator->TopLeftPixel()[height * accumulator->stride() + width] +=
sum * b.TopLeftPixel()[height * b.stride() + width];
}
}
}
class TemporalFilterTest : public ::testing::TestWithParam<TemporalFilterFunc> {
public:
virtual void SetUp() {
filter_func_ = GetParam();
rnd_.Reset(ACMRandom::DeterministicSeed());
}
protected:
TemporalFilterFunc filter_func_;
ACMRandom rnd_;
};
TEST_P(TemporalFilterTest, SizeCombinations) {
// Depending on subsampling this function may be called with values of 8 or 16
// for width and height, in any combination.
Buffer<uint8_t> a = Buffer<uint8_t>(16, 16, 8);
ASSERT_TRUE(a.Init());
const int filter_weight = 2;
const int filter_strength = 6;
for (int width = 8; width <= 16; width += 8) {
for (int height = 8; height <= 16; height += 8) {
// The second buffer must not have any border.
Buffer<uint8_t> b = Buffer<uint8_t>(width, height, 0);
ASSERT_TRUE(b.Init());
Buffer<unsigned int> accum_ref = Buffer<unsigned int>(width, height, 0);
ASSERT_TRUE(accum_ref.Init());
Buffer<unsigned int> accum_chk = Buffer<unsigned int>(width, height, 0);
ASSERT_TRUE(accum_chk.Init());
Buffer<uint16_t> count_ref = Buffer<uint16_t>(width, height, 0);
ASSERT_TRUE(count_ref.Init());
Buffer<uint16_t> count_chk = Buffer<uint16_t>(width, height, 0);
ASSERT_TRUE(count_chk.Init());
// The difference between the buffers must be small to pass the threshold
// to apply the filter.
a.Set(&rnd_, 0, 7);
b.Set(&rnd_, 0, 7);
accum_ref.Set(rnd_.Rand8());
accum_chk.CopyFrom(accum_ref);
count_ref.Set(rnd_.Rand8());
count_chk.CopyFrom(count_ref);
reference_filter(a, b, width, height, filter_strength, filter_weight,
&accum_ref, &count_ref);
ASM_REGISTER_STATE_CHECK(
filter_func_(a.TopLeftPixel(), a.stride(), b.TopLeftPixel(), width,
height, filter_strength, filter_weight,
accum_chk.TopLeftPixel(), count_chk.TopLeftPixel()));
EXPECT_TRUE(accum_chk.CheckValues(accum_ref));
EXPECT_TRUE(count_chk.CheckValues(count_ref));
if (HasFailure()) {
printf("Width: %d Height: %d\n", width, height);
count_chk.PrintDifference(count_ref);
accum_chk.PrintDifference(accum_ref);
return;
}
}
}
}
TEST_P(TemporalFilterTest, CompareReferenceRandom) {
for (int width = 8; width <= 16; width += 8) {
for (int height = 8; height <= 16; height += 8) {
Buffer<uint8_t> a = Buffer<uint8_t>(width, height, 8);
ASSERT_TRUE(a.Init());
// The second buffer must not have any border.
Buffer<uint8_t> b = Buffer<uint8_t>(width, height, 0);
ASSERT_TRUE(b.Init());
Buffer<unsigned int> accum_ref = Buffer<unsigned int>(width, height, 0);
ASSERT_TRUE(accum_ref.Init());
Buffer<unsigned int> accum_chk = Buffer<unsigned int>(width, height, 0);
ASSERT_TRUE(accum_chk.Init());
Buffer<uint16_t> count_ref = Buffer<uint16_t>(width, height, 0);
ASSERT_TRUE(count_ref.Init());
Buffer<uint16_t> count_chk = Buffer<uint16_t>(width, height, 0);
ASSERT_TRUE(count_chk.Init());
for (int filter_strength = 0; filter_strength <= 6; ++filter_strength) {
for (int filter_weight = 0; filter_weight <= 2; ++filter_weight) {
for (int repeat = 0; repeat < 100; ++repeat) {
if (repeat < 50) {
a.Set(&rnd_, 0, 7);
b.Set(&rnd_, 0, 7);
} else {
// Check large (but close) values as well.
a.Set(&rnd_, std::numeric_limits<uint8_t>::max() - 7,
std::numeric_limits<uint8_t>::max());
b.Set(&rnd_, std::numeric_limits<uint8_t>::max() - 7,
std::numeric_limits<uint8_t>::max());
}
accum_ref.Set(rnd_.Rand8());
accum_chk.CopyFrom(accum_ref);
count_ref.Set(rnd_.Rand8());
count_chk.CopyFrom(count_ref);
reference_filter(a, b, width, height, filter_strength,
filter_weight, &accum_ref, &count_ref);
ASM_REGISTER_STATE_CHECK(filter_func_(
a.TopLeftPixel(), a.stride(), b.TopLeftPixel(), width, height,
filter_strength, filter_weight, accum_chk.TopLeftPixel(),
count_chk.TopLeftPixel()));
EXPECT_TRUE(accum_chk.CheckValues(accum_ref));
EXPECT_TRUE(count_chk.CheckValues(count_ref));
if (HasFailure()) {
printf("Weight: %d Strength: %d\n", filter_weight,
filter_strength);
count_chk.PrintDifference(count_ref);
accum_chk.PrintDifference(accum_ref);
return;
}
}
}
}
}
}
}
TEST_P(TemporalFilterTest, DISABLED_Speed) {
Buffer<uint8_t> a = Buffer<uint8_t>(16, 16, 8);
ASSERT_TRUE(a.Init());
const int filter_weight = 2;
const int filter_strength = 6;
for (int width = 8; width <= 16; width += 8) {
for (int height = 8; height <= 16; height += 8) {
// The second buffer must not have any border.
Buffer<uint8_t> b = Buffer<uint8_t>(width, height, 0);
ASSERT_TRUE(b.Init());
Buffer<unsigned int> accum_ref = Buffer<unsigned int>(width, height, 0);
ASSERT_TRUE(accum_ref.Init());
Buffer<unsigned int> accum_chk = Buffer<unsigned int>(width, height, 0);
ASSERT_TRUE(accum_chk.Init());
Buffer<uint16_t> count_ref = Buffer<uint16_t>(width, height, 0);
ASSERT_TRUE(count_ref.Init());
Buffer<uint16_t> count_chk = Buffer<uint16_t>(width, height, 0);
ASSERT_TRUE(count_chk.Init());
a.Set(&rnd_, 0, 7);
b.Set(&rnd_, 0, 7);
accum_chk.Set(0);
count_chk.Set(0);
vpx_usec_timer timer;
vpx_usec_timer_start(&timer);
for (int i = 0; i < 10000; ++i) {
filter_func_(a.TopLeftPixel(), a.stride(), b.TopLeftPixel(), width,
height, filter_strength, filter_weight,
accum_chk.TopLeftPixel(), count_chk.TopLeftPixel());
}
vpx_usec_timer_mark(&timer);
const int elapsed_time = static_cast<int>(vpx_usec_timer_elapsed(&timer));
printf("Temporal filter %dx%d time: %5d us\n", width, height,
elapsed_time);
}
}
}
INSTANTIATE_TEST_CASE_P(C, TemporalFilterTest,
::testing::Values(&vp9_temporal_filter_apply_c));
#if HAVE_SSE4_1
INSTANTIATE_TEST_CASE_P(SSE4_1, TemporalFilterTest,
::testing::Values(&vp9_temporal_filter_apply_sse4_1));
#endif // HAVE_SSE4_1
} // namespace

View File

@ -20,8 +20,10 @@ LIBVPX_TEST_DATA-$(CONFIG_ENCODERS) += park_joy_90p_8_440.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += desktop_credits.y4m
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += niklas_1280_720_30.y4m
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += noisy_clip_640_360.y4m
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += rush_hour_444.y4m
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += screendata.y4m
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += niklas_640_480_30.yuv
# Test vectors
LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += vp80-00-comprehensive-001.ivf
@ -730,6 +732,8 @@ LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp93-2-20-12bit-yuv444.webm.md5
endif # CONFIG_VP9_HIGHBITDEPTH
# Invalid files for testing libvpx error checking.
LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += invalid-bug-1443.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += invalid-bug-1443.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += invalid-vp80-00-comprehensive-018.ivf.2kf_0x6.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP8_DECODER) += invalid-vp80-00-comprehensive-018.ivf.2kf_0x6.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-01-v3.webm
@ -764,15 +768,23 @@ LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s195
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s19552_r01-05_b6-.v2.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s20492_r01-05_b6-.v2.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-09-subpixel-00.ivf.s20492_r01-05_b6-.v2.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-10-show-existing-frame.webm.ivf.s180315_r01-05_b6-.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-10-show-existing-frame.webm.ivf.s180315_r01-05_b6-.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s3676_r01-05_b6-.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s73804_r01-05_b6-.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-12-droppable_1.ivf.s73804_r01-05_b6-.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-21-resize_inter_320x180_5_3-4.webm.ivf.s45551_r01-05_b6-.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-21-resize_inter_320x180_5_3-4.webm.ivf.s45551_r01-05_b6-.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp91-2-mixedrefcsp-444to420.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp91-2-mixedrefcsp-444to420.ivf.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-07-frame_parallel-1.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-07-frame_parallel-2.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-vp90-2-07-frame_parallel-3.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-crbug-629481.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-crbug-629481.webm.res
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-crbug-667044.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += invalid-crbug-667044.webm.res
ifeq ($(CONFIG_DECODE_PERF_TESTS),yes)
# Encode / Decode test
@ -807,7 +819,6 @@ LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += kirland_640_480_30.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += macmarcomoving_640_480_30.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += macmarcostationary_640_480_30.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += niklas_1280_720_30.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += niklas_640_480_30.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += tacomanarrows_640_480_30.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += tacomasmallcameramovement_640_480_30.yuv
LIBVPX_TEST_DATA-$(CONFIG_VP9_ENCODER) += thaloundeskmtg_640_480_30.yuv
@ -865,3 +876,7 @@ LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-21-resize_inter_1920x1080_7_1-2
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-21-resize_inter_1920x1080_7_1-2.webm.md5
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-21-resize_inter_1920x1080_7_3-4.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-21-resize_inter_1920x1080_7_3-4.webm.md5
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-22-svc_1280x720_3.ivf
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-22-svc_1280x720_3.ivf.md5
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-22-svc_1280x720_1.webm
LIBVPX_TEST_DATA-$(CONFIG_VP9_DECODER) += vp90-2-22-svc_1280x720_1.webm.md5

View File

@ -6,6 +6,8 @@ b87815bf86020c592ccc7a846ba2e28ec8043902 *hantro_odd.yuv
456d1493e52d32a5c30edf44a27debc1fa6b253a *invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-.ivf.res
c123d1f9f02fb4143abb5e271916e3a3080de8f6 *invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-z.ivf
456d1493e52d32a5c30edf44a27debc1fa6b253a *invalid-vp90-2-00-quantizer-11.webm.ivf.s52984_r01-05_b6-z.ivf.res
efafb92b7567bc04c3f1432ea6c268c1c31affd5 *invalid-vp90-2-21-resize_inter_320x180_5_3-4.webm.ivf.s45551_r01-05_b6-.ivf
5d9474c0309b7ca09a182d888f73b37a8fe1362c *invalid-vp90-2-21-resize_inter_320x180_5_3-4.webm.ivf.s45551_r01-05_b6-.ivf.res
fe346136b9b8c1e6f6084cc106485706915795e4 *invalid-vp90-01-v3.webm
5d9474c0309b7ca09a182d888f73b37a8fe1362c *invalid-vp90-01-v3.webm.res
d78e2fceba5ac942246503ec8366f879c4775ca5 *invalid-vp90-02-v2.webm
@ -14,6 +16,7 @@ df1a1453feb3c00d7d89746c7003b4163523bff3 *invalid-vp90-03-v3.webm
4935c62becc68c13642a03db1e6d3e2331c1c612 *invalid-vp90-03-v3.webm.res
d637297561dd904eb2c97a9015deeb31c4a1e8d2 *invalid-vp90-2-08-tile_1x4_frame_parallel_all_key.webm
3a204bdbeaa3c6458b77bcebb8366d107267f55d *invalid-vp90-2-08-tile_1x4_frame_parallel_all_key.webm.res
9aa21d8b2cb9d39abe8a7bb6032dc66955fb4342 *noisy_clip_640_360.y4m
a432f96ff0a787268e2f94a8092ab161a18d1b06 *park_joy_90p_10_420.y4m
0b194cc312c3a2e84d156a221b0a5eb615dfddc5 *park_joy_90p_10_422.y4m
ff0e0a21dc2adc95b8c1b37902713700655ced17 *park_joy_90p_10_444.y4m
@ -838,3 +841,16 @@ a000d568431d07379dd5a8ec066061c07e560b47 *invalid-vp90-2-00-quantizer-63.ivf.kf_
1e75aad3433c5c21c194a7b53fc393970f0a8d7f *invalid-vp90-2-00-quantizer-63.ivf.kf_65527x61446.ivf.res
235182f9a1c5c8841552510dd4288487447bfc40 *invalid-vp80-00-comprehensive-018.ivf.2kf_0x6.ivf
787f04f0483320d536894282f3358a4f8cac1cf9 *invalid-vp80-00-comprehensive-018.ivf.2kf_0x6.ivf.res
91d3cefd0deb98f3b0caf3a2d900ec7a7605e53a *invalid-vp90-2-10-show-existing-frame.webm.ivf.s180315_r01-05_b6-.ivf
1e472baaf5f6113459f0399a38a5a5e68d17799d *invalid-vp90-2-10-show-existing-frame.webm.ivf.s180315_r01-05_b6-.ivf.res
70057835bf29d14e66699ce5f022df2551fb6b37 *invalid-crbug-629481.webm
5d9474c0309b7ca09a182d888f73b37a8fe1362c *invalid-crbug-629481.webm.res
7602e00378161ca36ae93cc6ee12dd30b5ba1e1d *vp90-2-22-svc_1280x720_3.ivf
02e53e3eefbf25ec0929047fe50876acdeb040bd *vp90-2-22-svc_1280x720_3.ivf.md5
6fa3d3ac306a3d9ce1d610b78441dc00d2c2d4b9 *tos_vp8.webm
e402cbbf9e550ae017a1e9f1f73931c1d18474e8 *invalid-crbug-667044.webm
d3964f9dad9f60363c81b688324d95b4ec7c8038 *invalid-crbug-667044.webm.res
fd9df7f3f6992af1d7a9dde975c9a0d6f28c053d *invalid-bug-1443.ivf
fd3020fa6e9ca5966206738654c97dec313b0a95 *invalid-bug-1443.ivf.res
17696cd21e875f1d6e5d418cbf89feab02c8850a *vp90-2-22-svc_1280x720_1.webm
e2f9e1e47a791b4e939a9bdc50bf7a25b3761f77 *vp90-2-22-svc_1280x720_1.webm.md5

View File

@ -1,4 +1,5 @@
LIBVPX_TEST_SRCS-yes += acm_random.h
LIBVPX_TEST_SRCS-yes += buffer.h
LIBVPX_TEST_SRCS-yes += clear_system_state.h
LIBVPX_TEST_SRCS-yes += codec_factory.h
LIBVPX_TEST_SRCS-yes += md5_helper.h
@ -35,9 +36,9 @@ LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += cq_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += keyframe_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += byte_alignment_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += decode_svc_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += external_frame_buffer_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += user_priv_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += vp9_frame_parallel_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += active_map_refresh_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += active_map_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += borders_test.cc
@ -46,6 +47,7 @@ LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += frame_size_tests.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_lossless_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_end_to_end_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_ethread_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_motion_vector_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += level_test.cc
LIBVPX_TEST_SRCS-yes += decode_test_driver.cc
@ -121,6 +123,7 @@ LIBVPX_TEST_SRCS-$(CONFIG_VP8_ENCODER) += vp8_fdct4x4_test.cc
LIBVPX_TEST_SRCS-yes += idct_test.cc
LIBVPX_TEST_SRCS-yes += predict_test.cc
LIBVPX_TEST_SRCS-yes += vpx_scale_test.cc
LIBVPX_TEST_SRCS-yes += vpx_scale_test.h
ifeq ($(CONFIG_VP8_ENCODER)$(CONFIG_TEMPORAL_DENOISING),yesyes)
LIBVPX_TEST_SRCS-$(HAVE_SSE2) += vp8_denoiser_sse2_test.cc
@ -148,14 +151,20 @@ LIBVPX_TEST_SRCS-yes += vp9_intrapred_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += vp9_decrypt_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_DECODER) += vp9_thread_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += avg_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += comp_avg_pred_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct16x16_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct32x32_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += fdct4x4_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct_partial_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += dct_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += fdct8x8_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += hadamard_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += minmax_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_scale_test.cc
ifneq ($(CONFIG_REALTIME_ONLY),yes)
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += temporal_filter_test.cc
endif
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += variance_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_error_block_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_block_error_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_quantize_test.cc
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_subtract_test.cc
@ -166,7 +175,7 @@ LIBVPX_TEST_SRCS-$(CONFIG_INTERNAL_STATS) += consistency_test.cc
endif
ifeq ($(CONFIG_VP9_ENCODER)$(CONFIG_VP9_TEMPORAL_DENOISING),yesyes)
LIBVPX_TEST_SRCS-$(HAVE_SSE2) += vp9_denoiser_sse2_test.cc
LIBVPX_TEST_SRCS-yes += vp9_denoiser_test.cc
endif
LIBVPX_TEST_SRCS-$(CONFIG_VP9_ENCODER) += vp9_arf_freq_test.cc

View File

@ -29,6 +29,8 @@ namespace {
typedef void (*VpxPredFunc)(uint8_t *dst, ptrdiff_t y_stride,
const uint8_t *above, const uint8_t *left);
const int kBPS = 32;
const int kTotalPixels = 32 * kBPS;
const int kNumVp9IntraPredFuncs = 13;
const char *kVp9IntraPredNames[kNumVp9IntraPredFuncs] = {
"DC_PRED", "DC_LEFT_PRED", "DC_TOP_PRED", "DC_128_PRED", "V_PRED",
@ -36,107 +38,121 @@ const char *kVp9IntraPredNames[kNumVp9IntraPredFuncs] = {
"D207_PRED", "D63_PRED", "TM_PRED"
};
template <typename Pixel>
struct IntraPredTestMem {
void Init(int block_size, int bd) {
libvpx_test::ACMRandom rnd(libvpx_test::ACMRandom::DeterministicSeed());
Pixel *const above = above_mem + 16;
const int mask = (1 << bd) - 1;
for (int i = 0; i < kTotalPixels; ++i) ref_src[i] = rnd.Rand16() & mask;
for (int i = 0; i < kBPS; ++i) left[i] = rnd.Rand16() & mask;
for (int i = -1; i < kBPS; ++i) above[i] = rnd.Rand16() & mask;
// some code assumes the top row has been extended:
// d45/d63 C-code, for instance, but not the assembly.
// TODO(jzern): this style of extension isn't strictly necessary.
ASSERT_LE(block_size, kBPS);
for (int i = block_size; i < 2 * kBPS; ++i) {
above[i] = above[block_size - 1];
}
}
DECLARE_ALIGNED(16, Pixel, src[kTotalPixels]);
DECLARE_ALIGNED(16, Pixel, ref_src[kTotalPixels]);
DECLARE_ALIGNED(16, Pixel, left[kBPS]);
DECLARE_ALIGNED(16, Pixel, above_mem[2 * kBPS + 16]);
};
typedef IntraPredTestMem<uint8_t> Vp9IntraPredTestMem;
void CheckMd5Signature(const char name[], const char *const signatures[],
const void *data, size_t data_size, int elapsed_time,
int idx) {
libvpx_test::MD5 md5;
md5.Add(reinterpret_cast<const uint8_t *>(data), data_size);
printf("Mode %s[%12s]: %5d ms MD5: %s\n", name, kVp9IntraPredNames[idx],
elapsed_time, md5.Get());
EXPECT_STREQ(signatures[idx], md5.Get());
}
void TestIntraPred(const char name[], VpxPredFunc const *pred_funcs,
const char *const pred_func_names[], int num_funcs,
const char *const signatures[], int block_size,
int num_pixels_per_test) {
libvpx_test::ACMRandom rnd(libvpx_test::ACMRandom::DeterministicSeed());
const int kBPS = 32;
const int kTotalPixels = 32 * kBPS;
DECLARE_ALIGNED(16, uint8_t, src[kTotalPixels]);
DECLARE_ALIGNED(16, uint8_t, ref_src[kTotalPixels]);
DECLARE_ALIGNED(16, uint8_t, left[kBPS]);
DECLARE_ALIGNED(16, uint8_t, above_mem[2 * kBPS + 16]);
uint8_t *const above = above_mem + 16;
for (int i = 0; i < kTotalPixels; ++i) ref_src[i] = rnd.Rand8();
for (int i = 0; i < kBPS; ++i) left[i] = rnd.Rand8();
for (int i = -1; i < kBPS; ++i) above[i] = rnd.Rand8();
const int kNumTests = static_cast<int>(2.e10 / num_pixels_per_test);
const char *const signatures[], int block_size) {
const int kNumTests = static_cast<int>(
2.e10 / (block_size * block_size * kNumVp9IntraPredFuncs));
Vp9IntraPredTestMem intra_pred_test_mem;
const uint8_t *const above = intra_pred_test_mem.above_mem + 16;
// some code assumes the top row has been extended:
// d45/d63 C-code, for instance, but not the assembly.
// TODO(jzern): this style of extension isn't strictly necessary.
ASSERT_LE(block_size, kBPS);
memset(above + block_size, above[block_size - 1], 2 * kBPS - block_size);
intra_pred_test_mem.Init(block_size, 8);
for (int k = 0; k < num_funcs; ++k) {
for (int k = 0; k < kNumVp9IntraPredFuncs; ++k) {
if (pred_funcs[k] == NULL) continue;
memcpy(src, ref_src, sizeof(src));
memcpy(intra_pred_test_mem.src, intra_pred_test_mem.ref_src,
sizeof(intra_pred_test_mem.src));
vpx_usec_timer timer;
vpx_usec_timer_start(&timer);
for (int num_tests = 0; num_tests < kNumTests; ++num_tests) {
pred_funcs[k](src, kBPS, above, left);
pred_funcs[k](intra_pred_test_mem.src, kBPS, above,
intra_pred_test_mem.left);
}
libvpx_test::ClearSystemState();
vpx_usec_timer_mark(&timer);
const int elapsed_time =
static_cast<int>(vpx_usec_timer_elapsed(&timer) / 1000);
libvpx_test::MD5 md5;
md5.Add(src, sizeof(src));
printf("Mode %s[%12s]: %5d ms MD5: %s\n", name, pred_func_names[k],
elapsed_time, md5.Get());
EXPECT_STREQ(signatures[k], md5.Get());
CheckMd5Signature(name, signatures, intra_pred_test_mem.src,
sizeof(intra_pred_test_mem.src), elapsed_time, k);
}
}
void TestIntraPred4(VpxPredFunc const *pred_funcs) {
static const int kNumVp9IntraFuncs = 13;
static const char *const kSignatures[kNumVp9IntraFuncs] = {
"4334156168b34ab599d9b5b30f522fe9", "bc4649d5ba47c7ff178d92e475960fb0",
"8d316e5933326dcac24e1064794b5d12", "a27270fed024eafd762c95de85f4da51",
"c33dff000d4256c2b8f3bf9e9bab14d2", "44d8cddc2ad8f79b8ed3306051722b4f",
"eb54839b2bad6699d8946f01ec041cd0", "ecb0d56ae5f677ea45127ce9d5c058e4",
"0b7936841f6813da818275944895b574", "9117972ef64f91a58ff73e1731c81db2",
"c56d5e8c729e46825f46dd5d3b5d508a", "c0889e2039bcf7bcb5d2f33cdca69adc",
"309a618577b27c648f9c5ee45252bc8f",
static const char *const kSignatures[kNumVp9IntraPredFuncs] = {
"e7ed7353c3383fff942e500e9bfe82fe", "2a4a26fcc6ce005eadc08354d196c8a9",
"269d92eff86f315d9c38fe7640d85b15", "ae2960eea9f71ee3dabe08b282ec1773",
"6c1abcc44e90148998b51acd11144e9c", "f7bb3186e1ef8a2b326037ff898cad8e",
"364c1f3fb2f445f935aec2a70a67eaa4", "141624072a4a56773f68fadbdd07c4a7",
"7be49b08687a5f24df3a2c612fca3876", "459bb5d9fd5b238348179c9a22108cd6",
"73edb8831bf1bdfce21ae8eaa43b1234", "2e2457f2009c701a355a8b25eb74fcda",
"52ae4e8bdbe41494c1f43051d4dd7f0b"
};
TestIntraPred("Intra4", pred_funcs, kVp9IntraPredNames, kNumVp9IntraFuncs,
kSignatures, 4, 4 * 4 * kNumVp9IntraFuncs);
TestIntraPred("Intra4", pred_funcs, kSignatures, 4);
}
void TestIntraPred8(VpxPredFunc const *pred_funcs) {
static const int kNumVp9IntraFuncs = 13;
static const char *const kSignatures[kNumVp9IntraFuncs] = {
"7694ddeeefed887faf9d339d18850928", "7d726b1213591b99f736be6dec65065b",
"19c5711281357a485591aaf9c96c0a67", "ba6b66877a089e71cd938e3b8c40caac",
"802440c93317e0f8ba93fab02ef74265", "9e09a47a15deb0b9d8372824f9805080",
"b7c2d8c662268c0c427da412d7b0311d", "78339c1c60bb1d67d248ab8c4da08b7f",
"5c97d70f7d47de1882a6cd86c165c8a9", "8182bf60688b42205acd95e59e967157",
"08323400005a297f16d7e57e7fe1eaac", "95f7bfc262329a5849eda66d8f7c68ce",
"815b75c8e0d91cc1ae766dc5d3e445a3",
static const char *const kSignatures[kNumVp9IntraPredFuncs] = {
"d8bbae5d6547cfc17e4f5f44c8730e88", "373bab6d931868d41a601d9d88ce9ac3",
"6fdd5ff4ff79656c14747598ca9e3706", "d9661c2811d6a73674f40ffb2b841847",
"7c722d10b19ccff0b8c171868e747385", "f81dd986eb2b50f750d3a7da716b7e27",
"d500f2c8fc78f46a4c74e4dcf51f14fb", "0e3523f9cab2142dd37fd07ec0760bce",
"79ac4efe907f0a0f1885d43066cfedee", "19ecf2432ac305057de3b6578474eec6",
"4f985b61acc6dd5d2d2585fa89ea2e2d", "f1bb25a9060dd262f405f15a38f5f674",
"209ea00801584829e9a0f7be7d4a74ba"
};
TestIntraPred("Intra8", pred_funcs, kVp9IntraPredNames, kNumVp9IntraFuncs,
kSignatures, 8, 8 * 8 * kNumVp9IntraFuncs);
TestIntraPred("Intra8", pred_funcs, kSignatures, 8);
}
void TestIntraPred16(VpxPredFunc const *pred_funcs) {
static const int kNumVp9IntraFuncs = 13;
static const char *const kSignatures[kNumVp9IntraFuncs] = {
"b40dbb555d5d16a043dc361e6694fe53", "fb08118cee3b6405d64c1fd68be878c6",
"6c190f341475c837cc38c2e566b64875", "db5c34ccbe2c7f595d9b08b0dc2c698c",
"a62cbfd153a1f0b9fed13e62b8408a7a", "143df5b4c89335e281103f610f5052e4",
"d87feb124107cdf2cfb147655aa0bb3c", "7841fae7d4d47b519322e6a03eeed9dc",
"f6ebed3f71cbcf8d6d0516ce87e11093", "3cc480297dbfeed01a1c2d78dd03d0c5",
"b9f69fa6532b372c545397dcb78ef311", "a8fe1c70432f09d0c20c67bdb6432c4d",
"b8a41aa968ec108af447af4217cba91b",
static const char *const kSignatures[kNumVp9IntraPredFuncs] = {
"50971c07ce26977d30298538fffec619", "527a6b9e0dc5b21b98cf276305432bef",
"7eff2868f80ebc2c43a4f367281d80f7", "67cd60512b54964ef6aff1bd4816d922",
"48371c87dc95c08a33b2048f89cf6468", "b0acf2872ee411d7530af6d2625a7084",
"f32aafed4d8d3776ed58bcb6188756d5", "dae208f3dca583529cff49b73f7c4183",
"7af66a2f4c8e0b4908e40f047e60c47c", "125e3ab6ab9bc961f183ec366a7afa88",
"6b90f25b23983c35386b9fd704427622", "f8d6b11d710edc136a7c62c917435f93",
"ed308f18614a362917f411c218aee532"
};
TestIntraPred("Intra16", pred_funcs, kVp9IntraPredNames, kNumVp9IntraFuncs,
kSignatures, 16, 16 * 16 * kNumVp9IntraFuncs);
TestIntraPred("Intra16", pred_funcs, kSignatures, 16);
}
void TestIntraPred32(VpxPredFunc const *pred_funcs) {
static const int kNumVp9IntraFuncs = 13;
static const char *const kSignatures[kNumVp9IntraFuncs] = {
"558541656d84f9ae7896db655826febe", "b3587a1f9a01495fa38c8cd3c8e2a1bf",
"4c6501e64f25aacc55a2a16c7e8f0255", "b3b01379ba08916ef6b1b35f7d9ad51c",
"0f1eb38b6cbddb3d496199ef9f329071", "911c06efb9ed1c3b4c104b232b55812f",
"9225beb0ddfa7a1d24eaa1be430a6654", "0a6d584a44f8db9aa7ade2e2fdb9fc9e",
"b01c9076525216925f3456f034fb6eee", "d267e20ad9e5cd2915d1a47254d3d149",
"ed012a4a5da71f36c2393023184a0e59", "f162b51ed618d28b936974cff4391da5",
"9e1370c6d42e08d357d9612c93a71cfc",
static const char *const kSignatures[kNumVp9IntraPredFuncs] = {
"a0a618c900e65ae521ccc8af789729f2", "985aaa7c72b4a6c2fb431d32100cf13a",
"10662d09febc3ca13ee4e700120daeb5", "b3b01379ba08916ef6b1b35f7d9ad51c",
"9f4261755795af97e34679c333ec7004", "bc2c9da91ad97ef0d1610fb0a9041657",
"75c79b1362ad18abfcdb1aa0aacfc21d", "4039bb7da0f6860090d3c57b5c85468f",
"b29fff7b61804e68383e3a609b33da58", "e1aa5e49067fd8dba66c2eb8d07b7a89",
"4e042822909c1c06d3b10a88281df1eb", "72eb9d9e0e67c93f4c66b70348e9fef7",
"a22d102bcb51ca798aac12ca4ae8f2e8"
};
TestIntraPred("Intra32", pred_funcs, kVp9IntraPredNames, kNumVp9IntraFuncs,
kSignatures, 32, 32 * 32 * kNumVp9IntraFuncs);
TestIntraPred("Intra32", pred_funcs, kSignatures, 32);
}
} // namespace
@ -153,7 +169,6 @@ void TestIntraPred32(VpxPredFunc const *pred_funcs) {
}
// -----------------------------------------------------------------------------
// 4x4
INTRA_PRED_TEST(C, TestIntraPred4, vpx_dc_predictor_4x4_c,
vpx_dc_left_predictor_4x4_c, vpx_dc_top_predictor_4x4_c,
@ -163,47 +178,6 @@ INTRA_PRED_TEST(C, TestIntraPred4, vpx_dc_predictor_4x4_c,
vpx_d153_predictor_4x4_c, vpx_d207_predictor_4x4_c,
vpx_d63_predictor_4x4_c, vpx_tm_predictor_4x4_c)
#if HAVE_SSE2
INTRA_PRED_TEST(SSE2, TestIntraPred4, vpx_dc_predictor_4x4_sse2,
vpx_dc_left_predictor_4x4_sse2, vpx_dc_top_predictor_4x4_sse2,
vpx_dc_128_predictor_4x4_sse2, vpx_v_predictor_4x4_sse2,
vpx_h_predictor_4x4_sse2, vpx_d45_predictor_4x4_sse2, NULL,
NULL, NULL, vpx_d207_predictor_4x4_sse2, NULL,
vpx_tm_predictor_4x4_sse2)
#endif // HAVE_SSE2
#if HAVE_SSSE3
INTRA_PRED_TEST(SSSE3, TestIntraPred4, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_d153_predictor_4x4_ssse3, NULL,
vpx_d63_predictor_4x4_ssse3, NULL)
#endif // HAVE_SSSE3
#if HAVE_DSPR2
INTRA_PRED_TEST(DSPR2, TestIntraPred4, vpx_dc_predictor_4x4_dspr2, NULL, NULL,
NULL, NULL, vpx_h_predictor_4x4_dspr2, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_tm_predictor_4x4_dspr2)
#endif // HAVE_DSPR2
#if HAVE_NEON
INTRA_PRED_TEST(NEON, TestIntraPred4, vpx_dc_predictor_4x4_neon,
vpx_dc_left_predictor_4x4_neon, vpx_dc_top_predictor_4x4_neon,
vpx_dc_128_predictor_4x4_neon, vpx_v_predictor_4x4_neon,
vpx_h_predictor_4x4_neon, vpx_d45_predictor_4x4_neon,
vpx_d135_predictor_4x4_neon, NULL, NULL, NULL, NULL,
vpx_tm_predictor_4x4_neon)
#endif // HAVE_NEON
#if HAVE_MSA
INTRA_PRED_TEST(MSA, TestIntraPred4, vpx_dc_predictor_4x4_msa,
vpx_dc_left_predictor_4x4_msa, vpx_dc_top_predictor_4x4_msa,
vpx_dc_128_predictor_4x4_msa, vpx_v_predictor_4x4_msa,
vpx_h_predictor_4x4_msa, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_4x4_msa)
#endif // HAVE_MSA
// -----------------------------------------------------------------------------
// 8x8
INTRA_PRED_TEST(C, TestIntraPred8, vpx_dc_predictor_8x8_c,
vpx_dc_left_predictor_8x8_c, vpx_dc_top_predictor_8x8_c,
vpx_dc_128_predictor_8x8_c, vpx_v_predictor_8x8_c,
@ -212,46 +186,6 @@ INTRA_PRED_TEST(C, TestIntraPred8, vpx_dc_predictor_8x8_c,
vpx_d153_predictor_8x8_c, vpx_d207_predictor_8x8_c,
vpx_d63_predictor_8x8_c, vpx_tm_predictor_8x8_c)
#if HAVE_SSE2
INTRA_PRED_TEST(SSE2, TestIntraPred8, vpx_dc_predictor_8x8_sse2,
vpx_dc_left_predictor_8x8_sse2, vpx_dc_top_predictor_8x8_sse2,
vpx_dc_128_predictor_8x8_sse2, vpx_v_predictor_8x8_sse2,
vpx_h_predictor_8x8_sse2, vpx_d45_predictor_8x8_sse2, NULL,
NULL, NULL, NULL, NULL, vpx_tm_predictor_8x8_sse2)
#endif // HAVE_SSE2
#if HAVE_SSSE3
INTRA_PRED_TEST(SSSE3, TestIntraPred8, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_d153_predictor_8x8_ssse3,
vpx_d207_predictor_8x8_ssse3, vpx_d63_predictor_8x8_ssse3, NULL)
#endif // HAVE_SSSE3
#if HAVE_DSPR2
INTRA_PRED_TEST(DSPR2, TestIntraPred8, vpx_dc_predictor_8x8_dspr2, NULL, NULL,
NULL, NULL, vpx_h_predictor_8x8_dspr2, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_tm_predictor_8x8_c)
#endif // HAVE_DSPR2
#if HAVE_NEON
INTRA_PRED_TEST(NEON, TestIntraPred8, vpx_dc_predictor_8x8_neon,
vpx_dc_left_predictor_8x8_neon, vpx_dc_top_predictor_8x8_neon,
vpx_dc_128_predictor_8x8_neon, vpx_v_predictor_8x8_neon,
vpx_h_predictor_8x8_neon, vpx_d45_predictor_8x8_neon, NULL,
NULL, NULL, NULL, NULL, vpx_tm_predictor_8x8_neon)
#endif // HAVE_NEON
#if HAVE_MSA
INTRA_PRED_TEST(MSA, TestIntraPred8, vpx_dc_predictor_8x8_msa,
vpx_dc_left_predictor_8x8_msa, vpx_dc_top_predictor_8x8_msa,
vpx_dc_128_predictor_8x8_msa, vpx_v_predictor_8x8_msa,
vpx_h_predictor_8x8_msa, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_8x8_msa)
#endif // HAVE_MSA
// -----------------------------------------------------------------------------
// 16x16
INTRA_PRED_TEST(C, TestIntraPred16, vpx_dc_predictor_16x16_c,
vpx_dc_left_predictor_16x16_c, vpx_dc_top_predictor_16x16_c,
vpx_dc_128_predictor_16x16_c, vpx_v_predictor_16x16_c,
@ -260,48 +194,6 @@ INTRA_PRED_TEST(C, TestIntraPred16, vpx_dc_predictor_16x16_c,
vpx_d153_predictor_16x16_c, vpx_d207_predictor_16x16_c,
vpx_d63_predictor_16x16_c, vpx_tm_predictor_16x16_c)
#if HAVE_SSE2
INTRA_PRED_TEST(SSE2, TestIntraPred16, vpx_dc_predictor_16x16_sse2,
vpx_dc_left_predictor_16x16_sse2,
vpx_dc_top_predictor_16x16_sse2,
vpx_dc_128_predictor_16x16_sse2, vpx_v_predictor_16x16_sse2,
vpx_h_predictor_16x16_sse2, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_16x16_sse2)
#endif // HAVE_SSE2
#if HAVE_SSSE3
INTRA_PRED_TEST(SSSE3, TestIntraPred16, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_d45_predictor_16x16_ssse3, NULL, NULL,
vpx_d153_predictor_16x16_ssse3, vpx_d207_predictor_16x16_ssse3,
vpx_d63_predictor_16x16_ssse3, NULL)
#endif // HAVE_SSSE3
#if HAVE_DSPR2
INTRA_PRED_TEST(DSPR2, TestIntraPred16, vpx_dc_predictor_16x16_dspr2, NULL,
NULL, NULL, NULL, vpx_h_predictor_16x16_dspr2, NULL, NULL, NULL,
NULL, NULL, NULL, NULL)
#endif // HAVE_DSPR2
#if HAVE_NEON
INTRA_PRED_TEST(NEON, TestIntraPred16, vpx_dc_predictor_16x16_neon,
vpx_dc_left_predictor_16x16_neon,
vpx_dc_top_predictor_16x16_neon,
vpx_dc_128_predictor_16x16_neon, vpx_v_predictor_16x16_neon,
vpx_h_predictor_16x16_neon, vpx_d45_predictor_16x16_neon, NULL,
NULL, NULL, NULL, NULL, vpx_tm_predictor_16x16_neon)
#endif // HAVE_NEON
#if HAVE_MSA
INTRA_PRED_TEST(MSA, TestIntraPred16, vpx_dc_predictor_16x16_msa,
vpx_dc_left_predictor_16x16_msa, vpx_dc_top_predictor_16x16_msa,
vpx_dc_128_predictor_16x16_msa, vpx_v_predictor_16x16_msa,
vpx_h_predictor_16x16_msa, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_16x16_msa)
#endif // HAVE_MSA
// -----------------------------------------------------------------------------
// 32x32
INTRA_PRED_TEST(C, TestIntraPred32, vpx_dc_predictor_32x32_c,
vpx_dc_left_predictor_32x32_c, vpx_dc_top_predictor_32x32_c,
vpx_dc_128_predictor_32x32_c, vpx_v_predictor_32x32_c,
@ -311,6 +203,26 @@ INTRA_PRED_TEST(C, TestIntraPred32, vpx_dc_predictor_32x32_c,
vpx_d63_predictor_32x32_c, vpx_tm_predictor_32x32_c)
#if HAVE_SSE2
INTRA_PRED_TEST(SSE2, TestIntraPred4, vpx_dc_predictor_4x4_sse2,
vpx_dc_left_predictor_4x4_sse2, vpx_dc_top_predictor_4x4_sse2,
vpx_dc_128_predictor_4x4_sse2, vpx_v_predictor_4x4_sse2,
vpx_h_predictor_4x4_sse2, vpx_d45_predictor_4x4_sse2, NULL,
NULL, NULL, vpx_d207_predictor_4x4_sse2, NULL,
vpx_tm_predictor_4x4_sse2)
INTRA_PRED_TEST(SSE2, TestIntraPred8, vpx_dc_predictor_8x8_sse2,
vpx_dc_left_predictor_8x8_sse2, vpx_dc_top_predictor_8x8_sse2,
vpx_dc_128_predictor_8x8_sse2, vpx_v_predictor_8x8_sse2,
vpx_h_predictor_8x8_sse2, vpx_d45_predictor_8x8_sse2, NULL,
NULL, NULL, NULL, NULL, vpx_tm_predictor_8x8_sse2)
INTRA_PRED_TEST(SSE2, TestIntraPred16, vpx_dc_predictor_16x16_sse2,
vpx_dc_left_predictor_16x16_sse2,
vpx_dc_top_predictor_16x16_sse2,
vpx_dc_128_predictor_16x16_sse2, vpx_v_predictor_16x16_sse2,
vpx_h_predictor_16x16_sse2, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_16x16_sse2)
INTRA_PRED_TEST(SSE2, TestIntraPred32, vpx_dc_predictor_32x32_sse2,
vpx_dc_left_predictor_32x32_sse2,
vpx_dc_top_predictor_32x32_sse2,
@ -320,22 +232,79 @@ INTRA_PRED_TEST(SSE2, TestIntraPred32, vpx_dc_predictor_32x32_sse2,
#endif // HAVE_SSE2
#if HAVE_SSSE3
INTRA_PRED_TEST(SSSE3, TestIntraPred4, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_d153_predictor_4x4_ssse3, NULL,
vpx_d63_predictor_4x4_ssse3, NULL)
INTRA_PRED_TEST(SSSE3, TestIntraPred8, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_d153_predictor_8x8_ssse3,
vpx_d207_predictor_8x8_ssse3, vpx_d63_predictor_8x8_ssse3, NULL)
INTRA_PRED_TEST(SSSE3, TestIntraPred16, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_d45_predictor_16x16_ssse3, NULL, NULL,
vpx_d153_predictor_16x16_ssse3, vpx_d207_predictor_16x16_ssse3,
vpx_d63_predictor_16x16_ssse3, NULL)
INTRA_PRED_TEST(SSSE3, TestIntraPred32, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_d45_predictor_32x32_ssse3, NULL, NULL,
vpx_d153_predictor_32x32_ssse3, vpx_d207_predictor_32x32_ssse3,
vpx_d63_predictor_32x32_ssse3, NULL)
#endif // HAVE_SSSE3
#if HAVE_DSPR2
INTRA_PRED_TEST(DSPR2, TestIntraPred4, vpx_dc_predictor_4x4_dspr2, NULL, NULL,
NULL, NULL, vpx_h_predictor_4x4_dspr2, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_tm_predictor_4x4_dspr2)
INTRA_PRED_TEST(DSPR2, TestIntraPred8, vpx_dc_predictor_8x8_dspr2, NULL, NULL,
NULL, NULL, vpx_h_predictor_8x8_dspr2, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_tm_predictor_8x8_c)
INTRA_PRED_TEST(DSPR2, TestIntraPred16, vpx_dc_predictor_16x16_dspr2, NULL,
NULL, NULL, NULL, vpx_h_predictor_16x16_dspr2, NULL, NULL, NULL,
NULL, NULL, NULL, NULL)
#endif // HAVE_DSPR2
#if HAVE_NEON
INTRA_PRED_TEST(NEON, TestIntraPred4, vpx_dc_predictor_4x4_neon,
vpx_dc_left_predictor_4x4_neon, vpx_dc_top_predictor_4x4_neon,
vpx_dc_128_predictor_4x4_neon, vpx_v_predictor_4x4_neon,
vpx_h_predictor_4x4_neon, vpx_d45_predictor_4x4_neon,
vpx_d135_predictor_4x4_neon, NULL, NULL, NULL, NULL,
vpx_tm_predictor_4x4_neon)
INTRA_PRED_TEST(NEON, TestIntraPred8, vpx_dc_predictor_8x8_neon,
vpx_dc_left_predictor_8x8_neon, vpx_dc_top_predictor_8x8_neon,
vpx_dc_128_predictor_8x8_neon, vpx_v_predictor_8x8_neon,
vpx_h_predictor_8x8_neon, vpx_d45_predictor_8x8_neon,
vpx_d135_predictor_8x8_neon, NULL, NULL, NULL, NULL,
vpx_tm_predictor_8x8_neon)
INTRA_PRED_TEST(NEON, TestIntraPred16, vpx_dc_predictor_16x16_neon,
vpx_dc_left_predictor_16x16_neon,
vpx_dc_top_predictor_16x16_neon,
vpx_dc_128_predictor_16x16_neon, vpx_v_predictor_16x16_neon,
vpx_h_predictor_16x16_neon, vpx_d45_predictor_16x16_neon,
vpx_d135_predictor_16x16_neon, NULL, NULL, NULL, NULL,
vpx_tm_predictor_16x16_neon)
INTRA_PRED_TEST(NEON, TestIntraPred32, vpx_dc_predictor_32x32_neon,
vpx_dc_left_predictor_32x32_neon,
vpx_dc_top_predictor_32x32_neon,
vpx_dc_128_predictor_32x32_neon, vpx_v_predictor_32x32_neon,
vpx_h_predictor_32x32_neon, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_h_predictor_32x32_neon, vpx_d45_predictor_32x32_neon,
vpx_d135_predictor_32x32_neon, NULL, NULL, NULL, NULL,
vpx_tm_predictor_32x32_neon)
#endif // HAVE_NEON
#if HAVE_MSA
INTRA_PRED_TEST(MSA, TestIntraPred4, vpx_dc_predictor_4x4_msa,
vpx_dc_left_predictor_4x4_msa, vpx_dc_top_predictor_4x4_msa,
vpx_dc_128_predictor_4x4_msa, vpx_v_predictor_4x4_msa,
vpx_h_predictor_4x4_msa, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_4x4_msa)
INTRA_PRED_TEST(MSA, TestIntraPred8, vpx_dc_predictor_8x8_msa,
vpx_dc_left_predictor_8x8_msa, vpx_dc_top_predictor_8x8_msa,
vpx_dc_128_predictor_8x8_msa, vpx_v_predictor_8x8_msa,
vpx_h_predictor_8x8_msa, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_8x8_msa)
INTRA_PRED_TEST(MSA, TestIntraPred16, vpx_dc_predictor_16x16_msa,
vpx_dc_left_predictor_16x16_msa, vpx_dc_top_predictor_16x16_msa,
vpx_dc_128_predictor_16x16_msa, vpx_v_predictor_16x16_msa,
vpx_h_predictor_16x16_msa, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_16x16_msa)
INTRA_PRED_TEST(MSA, TestIntraPred32, vpx_dc_predictor_32x32_msa,
vpx_dc_left_predictor_32x32_msa, vpx_dc_top_predictor_32x32_msa,
vpx_dc_128_predictor_32x32_msa, vpx_v_predictor_32x32_msa,
@ -343,4 +312,275 @@ INTRA_PRED_TEST(MSA, TestIntraPred32, vpx_dc_predictor_32x32_msa,
vpx_tm_predictor_32x32_msa)
#endif // HAVE_MSA
#if HAVE_VSX
INTRA_PRED_TEST(VSX, TestIntraPred4, NULL, NULL, NULL, NULL, NULL,
vpx_h_predictor_4x4_vsx, NULL, NULL, NULL, NULL, NULL, NULL,
vpx_tm_predictor_4x4_vsx)
INTRA_PRED_TEST(VSX, TestIntraPred8, vpx_dc_predictor_8x8_vsx, NULL, NULL, NULL,
NULL, vpx_h_predictor_8x8_vsx, vpx_d45_predictor_8x8_vsx, NULL,
NULL, NULL, NULL, vpx_d63_predictor_8x8_vsx,
vpx_tm_predictor_8x8_vsx)
INTRA_PRED_TEST(VSX, TestIntraPred16, vpx_dc_predictor_16x16_vsx,
vpx_dc_left_predictor_16x16_vsx, vpx_dc_top_predictor_16x16_vsx,
vpx_dc_128_predictor_16x16_vsx, vpx_v_predictor_16x16_vsx,
vpx_h_predictor_16x16_vsx, vpx_d45_predictor_16x16_vsx, NULL,
NULL, NULL, NULL, vpx_d63_predictor_16x16_vsx,
vpx_tm_predictor_16x16_vsx)
INTRA_PRED_TEST(VSX, TestIntraPred32, vpx_dc_predictor_32x32_vsx,
vpx_dc_left_predictor_32x32_vsx, vpx_dc_top_predictor_32x32_vsx,
vpx_dc_128_predictor_32x32_vsx, vpx_v_predictor_32x32_vsx,
vpx_h_predictor_32x32_vsx, vpx_d45_predictor_32x32_vsx, NULL,
NULL, NULL, NULL, vpx_d63_predictor_32x32_vsx,
vpx_tm_predictor_32x32_vsx)
#endif // HAVE_VSX
// -----------------------------------------------------------------------------
#if CONFIG_VP9_HIGHBITDEPTH
namespace {
typedef void (*VpxHighbdPredFunc)(uint16_t *dst, ptrdiff_t y_stride,
const uint16_t *above, const uint16_t *left,
int bd);
typedef IntraPredTestMem<uint16_t> Vp9HighbdIntraPredTestMem;
void TestHighbdIntraPred(const char name[], VpxHighbdPredFunc const *pred_funcs,
const char *const signatures[], int block_size) {
const int kNumTests = static_cast<int>(
2.e10 / (block_size * block_size * kNumVp9IntraPredFuncs));
Vp9HighbdIntraPredTestMem intra_pred_test_mem;
const uint16_t *const above = intra_pred_test_mem.above_mem + 16;
intra_pred_test_mem.Init(block_size, 12);
for (int k = 0; k < kNumVp9IntraPredFuncs; ++k) {
if (pred_funcs[k] == NULL) continue;
memcpy(intra_pred_test_mem.src, intra_pred_test_mem.ref_src,
sizeof(intra_pred_test_mem.src));
vpx_usec_timer timer;
vpx_usec_timer_start(&timer);
for (int num_tests = 0; num_tests < kNumTests; ++num_tests) {
pred_funcs[k](intra_pred_test_mem.src, kBPS, above,
intra_pred_test_mem.left, 12);
}
libvpx_test::ClearSystemState();
vpx_usec_timer_mark(&timer);
const int elapsed_time =
static_cast<int>(vpx_usec_timer_elapsed(&timer) / 1000);
CheckMd5Signature(name, signatures, intra_pred_test_mem.src,
sizeof(intra_pred_test_mem.src), elapsed_time, k);
}
}
void TestHighbdIntraPred4(VpxHighbdPredFunc const *pred_funcs) {
static const char *const kSignatures[kNumVp9IntraPredFuncs] = {
"11f74af6c5737df472f3275cbde062fa", "51bea056b6447c93f6eb8f6b7e8f6f71",
"27e97f946766331795886f4de04c5594", "53ab15974b049111fb596c5168ec7e3f",
"f0b640bb176fbe4584cf3d32a9b0320a", "729783ca909e03afd4b47111c80d967b",
"fbf1c30793d9f32812e4d9f905d53530", "293fc903254a33754133314c6cdba81f",
"f8074d704233e73dfd35b458c6092374", "aa6363d08544a1ec4da33d7a0be5640d",
"462abcfdfa3d087bb33c9a88f2aec491", "863eab65d22550dd44a2397277c1ec71",
"23d61df1574d0fa308f9731811047c4b"
};
TestHighbdIntraPred("Intra4", pred_funcs, kSignatures, 4);
}
void TestHighbdIntraPred8(VpxHighbdPredFunc const *pred_funcs) {
static const char *const kSignatures[kNumVp9IntraPredFuncs] = {
"03da8829fe94663047fd108c5fcaa71d", "ecdb37b8120a2d3a4c706b016bd1bfd7",
"1d4543ed8d2b9368cb96898095fe8a75", "f791c9a67b913cbd82d9da8ecede30e2",
"065c70646f4dbaff913282f55a45a441", "51f87123616662ef7c35691497dfd0ba",
"2a5b0131ef4716f098ee65e6df01e3dd", "9ffe186a6bc7db95275f1bbddd6f7aba",
"a3258a2eae2e2bd55cb8f71351b22998", "8d909f0a2066e39b3216092c6289ece4",
"d183abb30b9f24c886a0517e991b22c7", "702a42fe4c7d665dc561b2aeeb60f311",
"7b5dbbbe7ae3a4ac2948731600bde5d6"
};
TestHighbdIntraPred("Intra8", pred_funcs, kSignatures, 8);
}
void TestHighbdIntraPred16(VpxHighbdPredFunc const *pred_funcs) {
static const char *const kSignatures[kNumVp9IntraPredFuncs] = {
"e33cb3f56a878e2fddb1b2fc51cdd275", "c7bff6f04b6052c8ab335d726dbbd52d",
"d0b0b47b654a9bcc5c6008110a44589b", "78f5da7b10b2b9ab39f114a33b6254e9",
"c78e31d23831abb40d6271a318fdd6f3", "90d1347f4ec9198a0320daecb6ff90b8",
"d2c623746cbb64a0c9e29c10f2c57041", "cf28bd387b81ad3e5f1a1c779a4b70a0",
"24c304330431ddeaf630f6ce94af2eac", "91a329798036bf64e8e00a87b131b8b1",
"d39111f22885307f920796a42084c872", "e2e702f7250ece98dd8f3f2854c31eeb",
"e2fb05b01eb8b88549e85641d8ce5b59"
};
TestHighbdIntraPred("Intra16", pred_funcs, kSignatures, 16);
}
void TestHighbdIntraPred32(VpxHighbdPredFunc const *pred_funcs) {
static const char *const kSignatures[kNumVp9IntraPredFuncs] = {
"a3e8056ba7e36628cce4917cd956fedd", "cc7d3024fe8748b512407edee045377e",
"2aab0a0f330a1d3e19b8ecb8f06387a3", "a547bc3fb7b06910bf3973122a426661",
"26f712514da95042f93d6e8dc8e431dc", "bb08c6e16177081daa3d936538dbc2e3",
"8f031af3e2650e89620d8d2c3a843d8b", "42867c8553285e94ee8e4df7abafbda8",
"6496bdee96100667833f546e1be3d640", "2ebfa25bf981377e682e580208504300",
"3e8ae52fd1f607f348aa4cb436c71ab7", "3d4efe797ca82193613696753ea624c4",
"cb8aab6d372278f3131e8d99efde02d9"
};
TestHighbdIntraPred("Intra32", pred_funcs, kSignatures, 32);
}
} // namespace
// Defines a test case for |arch| (e.g., C, SSE2, ...) passing the predictors
// to |test_func|. The test name is 'arch.test_func', e.g., C.TestIntraPred4.
#define HIGHBD_INTRA_PRED_TEST(arch, test_func, dc, dc_left, dc_top, dc_128, \
v, h, d45, d135, d117, d153, d207, d63, tm) \
TEST(arch, test_func) { \
static const VpxHighbdPredFunc vpx_intra_pred[] = { \
dc, dc_left, dc_top, dc_128, v, h, d45, d135, d117, d153, d207, d63, tm \
}; \
test_func(vpx_intra_pred); \
}
// -----------------------------------------------------------------------------
HIGHBD_INTRA_PRED_TEST(
C, TestHighbdIntraPred4, vpx_highbd_dc_predictor_4x4_c,
vpx_highbd_dc_left_predictor_4x4_c, vpx_highbd_dc_top_predictor_4x4_c,
vpx_highbd_dc_128_predictor_4x4_c, vpx_highbd_v_predictor_4x4_c,
vpx_highbd_h_predictor_4x4_c, vpx_highbd_d45_predictor_4x4_c,
vpx_highbd_d135_predictor_4x4_c, vpx_highbd_d117_predictor_4x4_c,
vpx_highbd_d153_predictor_4x4_c, vpx_highbd_d207_predictor_4x4_c,
vpx_highbd_d63_predictor_4x4_c, vpx_highbd_tm_predictor_4x4_c)
HIGHBD_INTRA_PRED_TEST(
C, TestHighbdIntraPred8, vpx_highbd_dc_predictor_8x8_c,
vpx_highbd_dc_left_predictor_8x8_c, vpx_highbd_dc_top_predictor_8x8_c,
vpx_highbd_dc_128_predictor_8x8_c, vpx_highbd_v_predictor_8x8_c,
vpx_highbd_h_predictor_8x8_c, vpx_highbd_d45_predictor_8x8_c,
vpx_highbd_d135_predictor_8x8_c, vpx_highbd_d117_predictor_8x8_c,
vpx_highbd_d153_predictor_8x8_c, vpx_highbd_d207_predictor_8x8_c,
vpx_highbd_d63_predictor_8x8_c, vpx_highbd_tm_predictor_8x8_c)
HIGHBD_INTRA_PRED_TEST(
C, TestHighbdIntraPred16, vpx_highbd_dc_predictor_16x16_c,
vpx_highbd_dc_left_predictor_16x16_c, vpx_highbd_dc_top_predictor_16x16_c,
vpx_highbd_dc_128_predictor_16x16_c, vpx_highbd_v_predictor_16x16_c,
vpx_highbd_h_predictor_16x16_c, vpx_highbd_d45_predictor_16x16_c,
vpx_highbd_d135_predictor_16x16_c, vpx_highbd_d117_predictor_16x16_c,
vpx_highbd_d153_predictor_16x16_c, vpx_highbd_d207_predictor_16x16_c,
vpx_highbd_d63_predictor_16x16_c, vpx_highbd_tm_predictor_16x16_c)
HIGHBD_INTRA_PRED_TEST(
C, TestHighbdIntraPred32, vpx_highbd_dc_predictor_32x32_c,
vpx_highbd_dc_left_predictor_32x32_c, vpx_highbd_dc_top_predictor_32x32_c,
vpx_highbd_dc_128_predictor_32x32_c, vpx_highbd_v_predictor_32x32_c,
vpx_highbd_h_predictor_32x32_c, vpx_highbd_d45_predictor_32x32_c,
vpx_highbd_d135_predictor_32x32_c, vpx_highbd_d117_predictor_32x32_c,
vpx_highbd_d153_predictor_32x32_c, vpx_highbd_d207_predictor_32x32_c,
vpx_highbd_d63_predictor_32x32_c, vpx_highbd_tm_predictor_32x32_c)
#if HAVE_SSE2
HIGHBD_INTRA_PRED_TEST(
SSE2, TestHighbdIntraPred4, vpx_highbd_dc_predictor_4x4_sse2,
vpx_highbd_dc_left_predictor_4x4_sse2, vpx_highbd_dc_top_predictor_4x4_sse2,
vpx_highbd_dc_128_predictor_4x4_sse2, vpx_highbd_v_predictor_4x4_sse2,
vpx_highbd_h_predictor_4x4_sse2, NULL, vpx_highbd_d135_predictor_4x4_sse2,
vpx_highbd_d117_predictor_4x4_sse2, vpx_highbd_d153_predictor_4x4_sse2,
vpx_highbd_d207_predictor_4x4_sse2, vpx_highbd_d63_predictor_4x4_sse2,
vpx_highbd_tm_predictor_4x4_c)
HIGHBD_INTRA_PRED_TEST(SSE2, TestHighbdIntraPred8,
vpx_highbd_dc_predictor_8x8_sse2,
vpx_highbd_dc_left_predictor_8x8_sse2,
vpx_highbd_dc_top_predictor_8x8_sse2,
vpx_highbd_dc_128_predictor_8x8_sse2,
vpx_highbd_v_predictor_8x8_sse2,
vpx_highbd_h_predictor_8x8_sse2, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_highbd_tm_predictor_8x8_sse2)
HIGHBD_INTRA_PRED_TEST(SSE2, TestHighbdIntraPred16,
vpx_highbd_dc_predictor_16x16_sse2,
vpx_highbd_dc_left_predictor_16x16_sse2,
vpx_highbd_dc_top_predictor_16x16_sse2,
vpx_highbd_dc_128_predictor_16x16_sse2,
vpx_highbd_v_predictor_16x16_sse2,
vpx_highbd_h_predictor_16x16_sse2, NULL, NULL, NULL,
NULL, NULL, NULL, vpx_highbd_tm_predictor_16x16_sse2)
HIGHBD_INTRA_PRED_TEST(SSE2, TestHighbdIntraPred32,
vpx_highbd_dc_predictor_32x32_sse2,
vpx_highbd_dc_left_predictor_32x32_sse2,
vpx_highbd_dc_top_predictor_32x32_sse2,
vpx_highbd_dc_128_predictor_32x32_sse2,
vpx_highbd_v_predictor_32x32_sse2,
vpx_highbd_h_predictor_32x32_sse2, NULL, NULL, NULL,
NULL, NULL, NULL, vpx_highbd_tm_predictor_32x32_sse2)
#endif // HAVE_SSE2
#if HAVE_SSSE3
HIGHBD_INTRA_PRED_TEST(SSSE3, TestHighbdIntraPred4, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_highbd_d45_predictor_4x4_ssse3, NULL,
NULL, NULL, NULL, NULL, NULL)
HIGHBD_INTRA_PRED_TEST(SSSE3, TestHighbdIntraPred8, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_highbd_d45_predictor_8x8_ssse3,
vpx_highbd_d135_predictor_8x8_ssse3,
vpx_highbd_d117_predictor_8x8_ssse3,
vpx_highbd_d153_predictor_8x8_ssse3,
vpx_highbd_d207_predictor_8x8_ssse3,
vpx_highbd_d63_predictor_8x8_ssse3, NULL)
HIGHBD_INTRA_PRED_TEST(SSSE3, TestHighbdIntraPred16, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_highbd_d45_predictor_16x16_ssse3,
vpx_highbd_d135_predictor_16x16_ssse3,
vpx_highbd_d117_predictor_16x16_ssse3,
vpx_highbd_d153_predictor_16x16_ssse3,
vpx_highbd_d207_predictor_16x16_ssse3,
vpx_highbd_d63_predictor_16x16_ssse3, NULL)
HIGHBD_INTRA_PRED_TEST(SSSE3, TestHighbdIntraPred32, NULL, NULL, NULL, NULL,
NULL, NULL, vpx_highbd_d45_predictor_32x32_ssse3,
vpx_highbd_d135_predictor_32x32_ssse3,
vpx_highbd_d117_predictor_32x32_ssse3,
vpx_highbd_d153_predictor_32x32_ssse3,
vpx_highbd_d207_predictor_32x32_ssse3,
vpx_highbd_d63_predictor_32x32_ssse3, NULL)
#endif // HAVE_SSSE3
#if HAVE_NEON
HIGHBD_INTRA_PRED_TEST(
NEON, TestHighbdIntraPred4, vpx_highbd_dc_predictor_4x4_neon,
vpx_highbd_dc_left_predictor_4x4_neon, vpx_highbd_dc_top_predictor_4x4_neon,
vpx_highbd_dc_128_predictor_4x4_neon, vpx_highbd_v_predictor_4x4_neon,
vpx_highbd_h_predictor_4x4_neon, vpx_highbd_d45_predictor_4x4_neon,
vpx_highbd_d135_predictor_4x4_neon, NULL, NULL, NULL, NULL,
vpx_highbd_tm_predictor_4x4_neon)
HIGHBD_INTRA_PRED_TEST(
NEON, TestHighbdIntraPred8, vpx_highbd_dc_predictor_8x8_neon,
vpx_highbd_dc_left_predictor_8x8_neon, vpx_highbd_dc_top_predictor_8x8_neon,
vpx_highbd_dc_128_predictor_8x8_neon, vpx_highbd_v_predictor_8x8_neon,
vpx_highbd_h_predictor_8x8_neon, vpx_highbd_d45_predictor_8x8_neon,
vpx_highbd_d135_predictor_8x8_neon, NULL, NULL, NULL, NULL,
vpx_highbd_tm_predictor_8x8_neon)
HIGHBD_INTRA_PRED_TEST(NEON, TestHighbdIntraPred16,
vpx_highbd_dc_predictor_16x16_neon,
vpx_highbd_dc_left_predictor_16x16_neon,
vpx_highbd_dc_top_predictor_16x16_neon,
vpx_highbd_dc_128_predictor_16x16_neon,
vpx_highbd_v_predictor_16x16_neon,
vpx_highbd_h_predictor_16x16_neon,
vpx_highbd_d45_predictor_16x16_neon,
vpx_highbd_d135_predictor_16x16_neon, NULL, NULL, NULL,
NULL, vpx_highbd_tm_predictor_16x16_neon)
HIGHBD_INTRA_PRED_TEST(NEON, TestHighbdIntraPred32,
vpx_highbd_dc_predictor_32x32_neon,
vpx_highbd_dc_left_predictor_32x32_neon,
vpx_highbd_dc_top_predictor_32x32_neon,
vpx_highbd_dc_128_predictor_32x32_neon,
vpx_highbd_v_predictor_32x32_neon,
vpx_highbd_h_predictor_32x32_neon,
vpx_highbd_d45_predictor_32x32_neon,
vpx_highbd_d135_predictor_32x32_neon, NULL, NULL, NULL,
NULL, vpx_highbd_tm_predictor_32x32_neon)
#endif // HAVE_NEON
#endif // CONFIG_VP9_HIGHBITDEPTH
#include "test/test_libvpx.cc"

View File

@ -53,6 +53,9 @@ int main(int argc, char **argv) {
}
if (!(simd_caps & HAS_AVX)) append_negative_gtest_filter(":AVX.*:AVX/*");
if (!(simd_caps & HAS_AVX2)) append_negative_gtest_filter(":AVX2.*:AVX2/*");
if (!(simd_caps & HAS_AVX512)) {
append_negative_gtest_filter(":AVX512.*:AVX512/*");
}
#endif // ARCH_X86 || ARCH_X86_64
#if !CONFIG_SHARED

View File

@ -28,13 +28,10 @@
namespace {
enum DecodeMode { kSerialMode, kFrameParallelMode };
const int kThreads = 0;
const int kFileName = 1;
const int kDecodeMode = 0;
const int kThreads = 1;
const int kFileName = 2;
typedef std::tr1::tuple<int, int, const char *> DecodeParam;
typedef std::tr1::tuple<int, const char *> DecodeParam;
class TestVectorTest : public ::libvpx_test::DecoderTest,
public ::libvpx_test::CodecTestWithParam<DecodeParam> {
@ -53,8 +50,8 @@ class TestVectorTest : public ::libvpx_test::DecoderTest,
void OpenMD5File(const std::string &md5_file_name_) {
md5_file_ = libvpx_test::OpenTestDataFile(md5_file_name_);
ASSERT_TRUE(md5_file_ != NULL) << "Md5 file open failed. Filename: "
<< md5_file_name_;
ASSERT_TRUE(md5_file_ != NULL)
<< "Md5 file open failed. Filename: " << md5_file_name_;
}
virtual void DecompressedFrameHook(const vpx_image_t &img,
@ -92,29 +89,14 @@ class TestVectorTest : public ::libvpx_test::DecoderTest,
TEST_P(TestVectorTest, MD5Match) {
const DecodeParam input = GET_PARAM(1);
const std::string filename = std::tr1::get<kFileName>(input);
const int threads = std::tr1::get<kThreads>(input);
const int mode = std::tr1::get<kDecodeMode>(input);
vpx_codec_flags_t flags = 0;
vpx_codec_dec_cfg_t cfg = vpx_codec_dec_cfg_t();
char str[256];
if (mode == kFrameParallelMode) {
flags |= VPX_CODEC_USE_FRAME_THREADING;
#if CONFIG_VP9_DECODER
// TODO(hkuang): Fix frame parallel decode bug. See issue 1086.
if (resize_clips_.find(filename) != resize_clips_.end()) {
printf("Skipping the test file: %s, due to frame parallel decode bug.\n",
filename.c_str());
return;
}
#endif
}
cfg.threads = std::tr1::get<kThreads>(input);
cfg.threads = threads;
snprintf(str, sizeof(str) / sizeof(str[0]) - 1,
"file: %s mode: %s threads: %d", filename.c_str(),
mode == 0 ? "Serial" : "Parallel", threads);
snprintf(str, sizeof(str) / sizeof(str[0]) - 1, "file: %s threads: %d",
filename.c_str(), cfg.threads);
SCOPED_TRACE(str);
// Open compressed video file.
@ -145,38 +127,44 @@ TEST_P(TestVectorTest, MD5Match) {
ASSERT_NO_FATAL_FAILURE(RunLoop(video.get(), cfg));
}
// Test VP8 decode in serial mode with single thread.
// NOTE: VP8 only support serial mode.
#if CONFIG_VP8_DECODER
VP8_INSTANTIATE_TEST_CASE(
TestVectorTest,
::testing::Combine(
::testing::Values(0), // Serial Mode.
::testing::Values(1), // Single thread.
::testing::ValuesIn(libvpx_test::kVP8TestVectors,
libvpx_test::kVP8TestVectors +
libvpx_test::kNumVP8TestVectors)));
// Test VP8 decode in with different numbers of threads.
INSTANTIATE_TEST_CASE_P(
VP8MultiThreaded, TestVectorTest,
::testing::Combine(
::testing::Values(
static_cast<const libvpx_test::CodecFactory *>(&libvpx_test::kVP8)),
::testing::Combine(
::testing::Range(2, 9), // With 2 ~ 8 threads.
::testing::ValuesIn(libvpx_test::kVP8TestVectors,
libvpx_test::kVP8TestVectors +
libvpx_test::kNumVP8TestVectors))));
#endif // CONFIG_VP8_DECODER
// Test VP9 decode in serial mode with single thread.
#if CONFIG_VP9_DECODER
VP9_INSTANTIATE_TEST_CASE(
TestVectorTest,
::testing::Combine(
::testing::Values(0), // Serial Mode.
::testing::Values(1), // Single thread.
::testing::ValuesIn(libvpx_test::kVP9TestVectors,
libvpx_test::kVP9TestVectors +
libvpx_test::kNumVP9TestVectors)));
// Test VP9 decode in frame parallel mode with different number of threads.
INSTANTIATE_TEST_CASE_P(
VP9MultiThreadedFrameParallel, TestVectorTest,
VP9MultiThreaded, TestVectorTest,
::testing::Combine(
::testing::Values(
static_cast<const libvpx_test::CodecFactory *>(&libvpx_test::kVP9)),
::testing::Combine(
::testing::Values(1), // Frame Parallel mode.
::testing::Range(2, 9), // With 2 ~ 8 threads.
::testing::ValuesIn(libvpx_test::kVP9TestVectors,
libvpx_test::kVP9TestVectors +

View File

@ -371,9 +371,12 @@ const char *const kVP9TestVectors[] = {
#endif // CONFIG_VP9_HIGHBITDEPTH
"vp90-2-20-big_superframe-01.webm",
"vp90-2-20-big_superframe-02.webm",
"vp90-2-22-svc_1280x720_1.webm",
RESIZE_TEST_VECTORS
};
const char *const kVP9TestVectorsSvc[] = { "vp90-2-22-svc_1280x720_3.ivf" };
const int kNumVP9TestVectors = NELEMENTS(kVP9TestVectors);
const int kNumVP9TestVectorsSvc = NELEMENTS(kVP9TestVectorsSvc);
const char *const kVP9TestVectorsResize[] = { RESIZE_TEST_VECTORS };
const int kNumVP9TestVectorsResize = NELEMENTS(kVP9TestVectorsResize);
#undef RESIZE_TEST_VECTORS

View File

@ -23,6 +23,8 @@ extern const char *const kVP8TestVectors[];
#if CONFIG_VP9_DECODER
extern const int kNumVP9TestVectors;
extern const char *const kVP9TestVectors[];
extern const int kNumVP9TestVectorsSvc;
extern const char *const kVP9TestVectorsSvc[];
extern const int kNumVP9TestVectorsResize;
extern const char *const kVP9TestVectorsResize[];
#endif // CONFIG_VP9_DECODER

View File

@ -54,7 +54,10 @@ twopass_encoder_vp9() {
fi
}
twopass_encoder_tests="twopass_encoder_vp8
twopass_encoder_vp9"
run_tests twopass_encoder_verify_environment "${twopass_encoder_tests}"
if [ "$(vpx_config_option_enabled CONFIG_REALTIME_ONLY)" != "yes" ]; then
twopass_encoder_tests="twopass_encoder_vp8
twopass_encoder_vp9"
run_tests twopass_encoder_verify_environment "${twopass_encoder_tests}"
fi

File diff suppressed because it is too large Load Diff

View File

@ -13,7 +13,9 @@
#if defined(_WIN32)
#undef NOMINMAX
#define NOMINMAX
#ifndef WIN32_LEAN_AND_MEAN
#define WIN32_LEAN_AND_MEAN
#endif
#include <windows.h>
#endif
#include <cstdio>

View File

@ -17,12 +17,16 @@
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vpx_config.h"
#include "./vp8_rtcd.h"
#include "test/acm_random.h"
#include "vpx/vpx_integer.h"
#include "vpx_ports/mem.h"
namespace {
typedef void (*FdctFunc)(int16_t *a, int16_t *b, int a_stride);
const int cospi8sqrt2minus1 = 20091;
const int sinpi8sqrt2 = 35468;
@ -68,10 +72,21 @@ void reference_idct4x4(const int16_t *input, int16_t *output) {
using libvpx_test::ACMRandom;
TEST(VP8FdctTest, SignBiasCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
class FdctTest : public ::testing::TestWithParam<FdctFunc> {
public:
virtual void SetUp() {
fdct_func_ = GetParam();
rnd_.Reset(ACMRandom::DeterministicSeed());
}
protected:
FdctFunc fdct_func_;
ACMRandom rnd_;
};
TEST_P(FdctTest, SignBiasCheck) {
int16_t test_input_block[16];
int16_t test_output_block[16];
DECLARE_ALIGNED(16, int16_t, test_output_block[16]);
const int pitch = 8;
int count_sign_block[16][2];
const int count_test_block = 1000000;
@ -81,10 +96,10 @@ TEST(VP8FdctTest, SignBiasCheck) {
for (int i = 0; i < count_test_block; ++i) {
// Initialize a test block with input range [-255, 255].
for (int j = 0; j < 16; ++j) {
test_input_block[j] = rnd.Rand8() - rnd.Rand8();
test_input_block[j] = rnd_.Rand8() - rnd_.Rand8();
}
vp8_short_fdct4x4_c(test_input_block, test_output_block, pitch);
fdct_func_(test_input_block, test_output_block, pitch);
for (int j = 0; j < 16; ++j) {
if (test_output_block[j] < 0) {
@ -110,10 +125,10 @@ TEST(VP8FdctTest, SignBiasCheck) {
for (int i = 0; i < count_test_block; ++i) {
// Initialize a test block with input range [-15, 15].
for (int j = 0; j < 16; ++j) {
test_input_block[j] = (rnd.Rand8() >> 4) - (rnd.Rand8() >> 4);
test_input_block[j] = (rnd_.Rand8() >> 4) - (rnd_.Rand8() >> 4);
}
vp8_short_fdct4x4_c(test_input_block, test_output_block, pitch);
fdct_func_(test_input_block, test_output_block, pitch);
for (int j = 0; j < 16; ++j) {
if (test_output_block[j] < 0) {
@ -135,23 +150,22 @@ TEST(VP8FdctTest, SignBiasCheck) {
<< "Error: 4x4 FDCT has a sign bias > 10% for input range [-15, 15]";
};
TEST(VP8FdctTest, RoundTripErrorCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
TEST_P(FdctTest, RoundTripErrorCheck) {
int max_error = 0;
double total_error = 0;
const int count_test_block = 1000000;
for (int i = 0; i < count_test_block; ++i) {
int16_t test_input_block[16];
int16_t test_temp_block[16];
int16_t test_output_block[16];
DECLARE_ALIGNED(16, int16_t, test_temp_block[16]);
// Initialize a test block with input range [-255, 255].
for (int j = 0; j < 16; ++j) {
test_input_block[j] = rnd.Rand8() - rnd.Rand8();
test_input_block[j] = rnd_.Rand8() - rnd_.Rand8();
}
const int pitch = 8;
vp8_short_fdct4x4_c(test_input_block, test_temp_block, pitch);
fdct_func_(test_input_block, test_temp_block, pitch);
reference_idct4x4(test_temp_block, test_output_block);
for (int j = 0; j < 16; ++j) {
@ -169,4 +183,24 @@ TEST(VP8FdctTest, RoundTripErrorCheck) {
<< "Error: FDCT/IDCT has average roundtrip error > 1 per block";
};
INSTANTIATE_TEST_CASE_P(C, FdctTest, ::testing::Values(vp8_short_fdct4x4_c));
#if HAVE_NEON
INSTANTIATE_TEST_CASE_P(NEON, FdctTest,
::testing::Values(vp8_short_fdct4x4_neon));
#endif // HAVE_NEON
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(SSE2, FdctTest,
::testing::Values(vp8_short_fdct4x4_sse2));
#endif // HAVE_SSE2
#if HAVE_MSA
INSTANTIATE_TEST_CASE_P(MSA, FdctTest,
::testing::Values(vp8_short_fdct4x4_msa));
#endif // HAVE_MSA
#if HAVE_MMI
INSTANTIATE_TEST_CASE_P(MMI, FdctTest,
::testing::Values(vp8_short_fdct4x4_mmi));
#endif // HAVE_MMI
} // namespace

View File

@ -23,36 +23,36 @@
#include "vp9/common/vp9_entropy.h"
#include "vpx/vpx_codec.h"
#include "vpx/vpx_integer.h"
#include "vpx_dsp/vpx_dsp_common.h"
using libvpx_test::ACMRandom;
namespace {
#if CONFIG_VP9_HIGHBITDEPTH
const int kNumIterations = 1000;
typedef int64_t (*ErrorBlockFunc)(const tran_low_t *coeff,
typedef int64_t (*HBDBlockErrorFunc)(const tran_low_t *coeff,
const tran_low_t *dqcoeff,
intptr_t block_size, int64_t *ssz,
int bps);
typedef std::tr1::tuple<HBDBlockErrorFunc, HBDBlockErrorFunc, vpx_bit_depth_t>
BlockErrorParam;
typedef int64_t (*BlockErrorFunc)(const tran_low_t *coeff,
const tran_low_t *dqcoeff,
intptr_t block_size, int64_t *ssz, int bps);
intptr_t block_size, int64_t *ssz);
typedef std::tr1::tuple<ErrorBlockFunc, ErrorBlockFunc, vpx_bit_depth_t>
ErrorBlockParam;
// wrapper for 8-bit block error functions without a 'bps' param.
typedef int64_t (*HighBdBlockError8bit)(const tran_low_t *coeff,
const tran_low_t *dqcoeff,
intptr_t block_size, int64_t *ssz);
template <HighBdBlockError8bit fn>
int64_t HighBdBlockError8bitWrapper(const tran_low_t *coeff,
const tran_low_t *dqcoeff,
intptr_t block_size, int64_t *ssz,
int bps) {
EXPECT_EQ(8, bps);
template <BlockErrorFunc fn>
int64_t BlockError8BitWrapper(const tran_low_t *coeff,
const tran_low_t *dqcoeff, intptr_t block_size,
int64_t *ssz, int bps) {
EXPECT_EQ(bps, 8);
return fn(coeff, dqcoeff, block_size, ssz);
}
class ErrorBlockTest : public ::testing::TestWithParam<ErrorBlockParam> {
class BlockErrorTest : public ::testing::TestWithParam<BlockErrorParam> {
public:
virtual ~ErrorBlockTest() {}
virtual ~BlockErrorTest() {}
virtual void SetUp() {
error_block_op_ = GET_PARAM(0);
ref_error_block_op_ = GET_PARAM(1);
@ -63,11 +63,11 @@ class ErrorBlockTest : public ::testing::TestWithParam<ErrorBlockParam> {
protected:
vpx_bit_depth_t bit_depth_;
ErrorBlockFunc error_block_op_;
ErrorBlockFunc ref_error_block_op_;
HBDBlockErrorFunc error_block_op_;
HBDBlockErrorFunc ref_error_block_op_;
};
TEST_P(ErrorBlockTest, OperationCheck) {
TEST_P(BlockErrorTest, OperationCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
DECLARE_ALIGNED(16, tran_low_t, coeff[4096]);
DECLARE_ALIGNED(16, tran_low_t, dqcoeff[4096]);
@ -110,7 +110,7 @@ TEST_P(ErrorBlockTest, OperationCheck) {
<< "First failed at test case " << first_failure;
}
TEST_P(ErrorBlockTest, ExtremeValues) {
TEST_P(BlockErrorTest, ExtremeValues) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
DECLARE_ALIGNED(16, tran_low_t, coeff[4096]);
DECLARE_ALIGNED(16, tran_low_t, dqcoeff[4096]);
@ -171,29 +171,28 @@ TEST_P(ErrorBlockTest, ExtremeValues) {
using std::tr1::make_tuple;
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(
SSE2, ErrorBlockTest,
::testing::Values(
make_tuple(&vp9_highbd_block_error_sse2, &vp9_highbd_block_error_c,
VPX_BITS_10),
make_tuple(&vp9_highbd_block_error_sse2, &vp9_highbd_block_error_c,
VPX_BITS_12),
make_tuple(&vp9_highbd_block_error_sse2, &vp9_highbd_block_error_c,
VPX_BITS_8),
make_tuple(
&HighBdBlockError8bitWrapper<vp9_highbd_block_error_8bit_sse2>,
&HighBdBlockError8bitWrapper<vp9_highbd_block_error_8bit_c>,
VPX_BITS_8)));
const BlockErrorParam sse2_block_error_tests[] = {
#if CONFIG_VP9_HIGHBITDEPTH
make_tuple(&vp9_highbd_block_error_sse2, &vp9_highbd_block_error_c,
VPX_BITS_10),
make_tuple(&vp9_highbd_block_error_sse2, &vp9_highbd_block_error_c,
VPX_BITS_12),
make_tuple(&vp9_highbd_block_error_sse2, &vp9_highbd_block_error_c,
VPX_BITS_8),
#endif // CONFIG_VP9_HIGHBITDEPTH
make_tuple(&BlockError8BitWrapper<vp9_block_error_sse2>,
&BlockError8BitWrapper<vp9_block_error_c>, VPX_BITS_8)
};
INSTANTIATE_TEST_CASE_P(SSE2, BlockErrorTest,
::testing::ValuesIn(sse2_block_error_tests));
#endif // HAVE_SSE2
#if HAVE_AVX
#if HAVE_AVX2
INSTANTIATE_TEST_CASE_P(
AVX, ErrorBlockTest,
::testing::Values(make_tuple(
&HighBdBlockError8bitWrapper<vp9_highbd_block_error_8bit_avx>,
&HighBdBlockError8bitWrapper<vp9_highbd_block_error_8bit_c>,
VPX_BITS_8)));
#endif // HAVE_AVX
#endif // CONFIG_VP9_HIGHBITDEPTH
AVX2, BlockErrorTest,
::testing::Values(make_tuple(&BlockError8BitWrapper<vp9_block_error_avx2>,
&BlockError8BitWrapper<vp9_block_error_c>,
VPX_BITS_8)));
#endif // HAVE_AVX2
} // namespace

View File

@ -29,11 +29,21 @@ using libvpx_test::ACMRandom;
namespace {
const int kNumPixels = 64 * 64;
class VP9DenoiserTest : public ::testing::TestWithParam<BLOCK_SIZE> {
typedef int (*Vp9DenoiserFilterFunc)(const uint8_t *sig, int sig_stride,
const uint8_t *mc_avg, int mc_avg_stride,
uint8_t *avg, int avg_stride,
int increase_denoising, BLOCK_SIZE bs,
int motion_magnitude);
typedef std::tr1::tuple<Vp9DenoiserFilterFunc, BLOCK_SIZE> VP9DenoiserTestParam;
class VP9DenoiserTest
: public ::testing::Test,
public ::testing::WithParamInterface<VP9DenoiserTestParam> {
public:
virtual ~VP9DenoiserTest() {}
virtual void SetUp() { bs_ = GetParam(); }
virtual void SetUp() { bs_ = GET_PARAM(1); }
virtual void TearDown() { libvpx_test::ClearSystemState(); }
@ -76,9 +86,9 @@ TEST_P(VP9DenoiserTest, BitexactCheck) {
64, avg_block_c, 64, 0, bs_,
motion_magnitude_random));
ASM_REGISTER_STATE_CHECK(vp9_denoiser_filter_sse2(
sig_block, 64, mc_avg_block, 64, avg_block_sse2, 64, 0, bs_,
motion_magnitude_random));
ASM_REGISTER_STATE_CHECK(GET_PARAM(0)(sig_block, 64, mc_avg_block, 64,
avg_block_sse2, 64, 0, bs_,
motion_magnitude_random));
// Test bitexactness.
for (int h = 0; h < (4 << b_height_log2_lookup[bs_]); ++h) {
@ -89,10 +99,36 @@ TEST_P(VP9DenoiserTest, BitexactCheck) {
}
}
using std::tr1::make_tuple;
// Test for all block size.
INSTANTIATE_TEST_CASE_P(SSE2, VP9DenoiserTest,
::testing::Values(BLOCK_8X8, BLOCK_8X16, BLOCK_16X8,
BLOCK_16X16, BLOCK_16X32, BLOCK_32X16,
BLOCK_32X32, BLOCK_32X64, BLOCK_64X32,
BLOCK_64X64));
#if HAVE_SSE2
INSTANTIATE_TEST_CASE_P(
SSE2, VP9DenoiserTest,
::testing::Values(make_tuple(&vp9_denoiser_filter_sse2, BLOCK_8X8),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_8X16),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_16X8),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_16X16),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_16X32),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_32X16),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_32X32),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_32X64),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_64X32),
make_tuple(&vp9_denoiser_filter_sse2, BLOCK_64X64)));
#endif // HAVE_SSE2
#if HAVE_NEON
INSTANTIATE_TEST_CASE_P(
NEON, VP9DenoiserTest,
::testing::Values(make_tuple(&vp9_denoiser_filter_neon, BLOCK_8X8),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_8X16),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_16X8),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_16X16),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_16X32),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_32X16),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_32X32),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_32X64),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_64X32),
make_tuple(&vp9_denoiser_filter_neon, BLOCK_64X64)));
#endif
} // namespace

View File

@ -99,9 +99,7 @@ class VpxEncoderParmsGetToDecoder
vpx_codec_ctx_t *const vp9_decoder = decoder->GetDecoder();
vpx_codec_alg_priv_t *const priv =
reinterpret_cast<vpx_codec_alg_priv_t *>(vp9_decoder->priv);
FrameWorkerData *const worker_data =
reinterpret_cast<FrameWorkerData *>(priv->frame_workers[0].data1);
VP9_COMMON *const common = &worker_data->pbi->common;
VP9_COMMON *const common = &priv->pbi->common;
if (encode_parms.lossless) {
EXPECT_EQ(0, common->base_qindex);

View File

@ -16,17 +16,221 @@
#include "test/md5_helper.h"
#include "test/util.h"
#include "test/y4m_video_source.h"
#include "vp9/encoder/vp9_firstpass.h"
namespace {
class VPxEncoderThreadTest
// FIRSTPASS_STATS struct:
// {
// 25 double members;
// 1 int64_t member;
// }
// Whenever FIRSTPASS_STATS struct is modified, the following constants need to
// be revisited.
const int kDbl = 25;
const int kInt = 1;
const size_t kFirstPassStatsSz = kDbl * sizeof(double) + kInt * sizeof(int64_t);
class VPxFirstPassEncoderThreadTest
: public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWith2Params<libvpx_test::TestMode, int> {
protected:
VPxFirstPassEncoderThreadTest()
: EncoderTest(GET_PARAM(0)), encoder_initialized_(false), tiles_(0),
encoding_mode_(GET_PARAM(1)), set_cpu_used_(GET_PARAM(2)) {
init_flags_ = VPX_CODEC_USE_PSNR;
row_mt_mode_ = 1;
first_pass_only_ = true;
firstpass_stats_.buf = NULL;
firstpass_stats_.sz = 0;
}
virtual ~VPxFirstPassEncoderThreadTest() { free(firstpass_stats_.buf); }
virtual void SetUp() {
InitializeConfig();
SetMode(encoding_mode_);
cfg_.rc_end_usage = VPX_VBR;
cfg_.rc_2pass_vbr_minsection_pct = 5;
cfg_.rc_2pass_vbr_maxsection_pct = 2000;
cfg_.rc_max_quantizer = 56;
cfg_.rc_min_quantizer = 0;
}
virtual void BeginPassHook(unsigned int /*pass*/) {
encoder_initialized_ = false;
abort_ = false;
}
virtual void EndPassHook() {
// For first pass stats test, only run first pass encoder.
if (first_pass_only_ && cfg_.g_pass == VPX_RC_FIRST_PASS)
abort_ |= first_pass_only_;
}
virtual void PreEncodeFrameHook(::libvpx_test::VideoSource * /*video*/,
::libvpx_test::Encoder *encoder) {
if (!encoder_initialized_) {
// Encode in 2-pass mode.
encoder->Control(VP9E_SET_TILE_COLUMNS, tiles_);
encoder->Control(VP8E_SET_CPUUSED, set_cpu_used_);
encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 1);
encoder->Control(VP8E_SET_ARNR_MAXFRAMES, 7);
encoder->Control(VP8E_SET_ARNR_STRENGTH, 5);
encoder->Control(VP8E_SET_ARNR_TYPE, 3);
encoder->Control(VP9E_SET_FRAME_PARALLEL_DECODING, 0);
if (encoding_mode_ == ::libvpx_test::kTwoPassGood)
encoder->Control(VP9E_SET_ROW_MT, row_mt_mode_);
encoder_initialized_ = true;
}
}
virtual void StatsPktHook(const vpx_codec_cx_pkt_t *pkt) {
const uint8_t *const pkt_buf =
reinterpret_cast<uint8_t *>(pkt->data.twopass_stats.buf);
const size_t pkt_size = pkt->data.twopass_stats.sz;
// First pass stats size equals sizeof(FIRSTPASS_STATS)
EXPECT_EQ(pkt_size, kFirstPassStatsSz)
<< "Error: First pass stats size doesn't equal kFirstPassStatsSz";
firstpass_stats_.buf =
realloc(firstpass_stats_.buf, firstpass_stats_.sz + pkt_size);
memcpy((uint8_t *)firstpass_stats_.buf + firstpass_stats_.sz, pkt_buf,
pkt_size);
firstpass_stats_.sz += pkt_size;
}
bool encoder_initialized_;
int tiles_;
::libvpx_test::TestMode encoding_mode_;
int set_cpu_used_;
int row_mt_mode_;
bool first_pass_only_;
vpx_fixed_buf_t firstpass_stats_;
};
static void compare_fp_stats(vpx_fixed_buf_t *fp_stats, double factor) {
// fp_stats consists of 2 set of first pass encoding stats. These 2 set of
// stats are compared to check if the stats match or at least are very close.
FIRSTPASS_STATS *stats1 = reinterpret_cast<FIRSTPASS_STATS *>(fp_stats->buf);
int nframes_ = (int)(fp_stats->sz / sizeof(FIRSTPASS_STATS));
FIRSTPASS_STATS *stats2 = stats1 + nframes_ / 2;
int i, j;
// The total stats are also output and included in the first pass stats. Here
// ignore that in the comparison.
for (i = 0; i < (nframes_ / 2 - 1); ++i) {
const double *frame_stats1 = reinterpret_cast<double *>(stats1);
const double *frame_stats2 = reinterpret_cast<double *>(stats2);
for (j = 0; j < kDbl; ++j) {
ASSERT_LE(fabs(*frame_stats1 - *frame_stats2),
fabs(*frame_stats1) / factor)
<< "First failure @ frame #" << i << " stat #" << j << " ("
<< *frame_stats1 << " vs. " << *frame_stats2 << ")";
frame_stats1++;
frame_stats2++;
}
stats1++;
stats2++;
}
// Reset firstpass_stats_ to 0.
memset((uint8_t *)fp_stats->buf, 0, fp_stats->sz);
fp_stats->sz = 0;
}
static void compare_fp_stats_md5(vpx_fixed_buf_t *fp_stats) {
// fp_stats consists of 2 set of first pass encoding stats. These 2 set of
// stats are compared to check if the stats match.
uint8_t *stats1 = reinterpret_cast<uint8_t *>(fp_stats->buf);
uint8_t *stats2 = stats1 + fp_stats->sz / 2;
::libvpx_test::MD5 md5_row_mt_0, md5_row_mt_1;
md5_row_mt_0.Add(stats1, fp_stats->sz / 2);
const char *md5_row_mt_0_str = md5_row_mt_0.Get();
md5_row_mt_1.Add(stats2, fp_stats->sz / 2);
const char *md5_row_mt_1_str = md5_row_mt_1.Get();
// Check md5 match.
ASSERT_STREQ(md5_row_mt_0_str, md5_row_mt_1_str)
<< "MD5 checksums don't match";
// Reset firstpass_stats_ to 0.
memset((uint8_t *)fp_stats->buf, 0, fp_stats->sz);
fp_stats->sz = 0;
}
TEST_P(VPxFirstPassEncoderThreadTest, FirstPassStatsTest) {
::libvpx_test::Y4mVideoSource video("niklas_1280_720_30.y4m", 0, 60);
first_pass_only_ = true;
cfg_.rc_target_bitrate = 1000;
// Test row_mt_mode: 0 vs 1 at single thread case(threads = 1, tiles_ = 0)
tiles_ = 0;
cfg_.g_threads = 1;
row_mt_mode_ = 0;
init_flags_ = VPX_CODEC_USE_PSNR;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
row_mt_mode_ = 1;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
// Compare to check if using or not using row-mt generates close stats.
ASSERT_NO_FATAL_FAILURE(compare_fp_stats(&firstpass_stats_, 1000.0));
// Test single thread vs multiple threads
row_mt_mode_ = 1;
tiles_ = 0;
cfg_.g_threads = 1;
init_flags_ = VPX_CODEC_USE_PSNR;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
cfg_.g_threads = 4;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
// Compare to check if single-thread and multi-thread stats are close enough.
ASSERT_NO_FATAL_FAILURE(compare_fp_stats(&firstpass_stats_, 1000.0));
// Bit exact test in row_mt mode.
// When row_mt_mode_=1 and using >1 threads, the encoder generates bit exact
// result.
row_mt_mode_ = 1;
tiles_ = 2;
cfg_.g_threads = 2;
init_flags_ = VPX_CODEC_USE_PSNR;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
cfg_.g_threads = 8;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
// Compare to check if stats match with row-mt=0/1.
compare_fp_stats_md5(&firstpass_stats_);
}
class VPxEncoderThreadTest
: public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWith4Params<libvpx_test::TestMode, int,
int, int> {
protected:
VPxEncoderThreadTest()
: EncoderTest(GET_PARAM(0)), encoder_initialized_(false), tiles_(2),
: EncoderTest(GET_PARAM(0)), encoder_initialized_(false),
tiles_(GET_PARAM(3)), threads_(GET_PARAM(4)),
encoding_mode_(GET_PARAM(1)), set_cpu_used_(GET_PARAM(2)) {
init_flags_ = VPX_CODEC_USE_PSNR;
md5_.clear();
row_mt_mode_ = 1;
psnr_ = 0.0;
nframes_ = 0;
}
virtual ~VPxEncoderThreadTest() {}
@ -35,7 +239,6 @@ class VPxEncoderThreadTest
SetMode(encoding_mode_);
if (encoding_mode_ != ::libvpx_test::kRealTime) {
cfg_.g_lag_in_frames = 3;
cfg_.rc_end_usage = VPX_VBR;
cfg_.rc_2pass_vbr_minsection_pct = 5;
cfg_.rc_2pass_vbr_maxsection_pct = 2000;
@ -50,6 +253,8 @@ class VPxEncoderThreadTest
virtual void BeginPassHook(unsigned int /*pass*/) {
encoder_initialized_ = false;
psnr_ = 0.0;
nframes_ = 0;
}
virtual void PreEncodeFrameHook(::libvpx_test::VideoSource * /*video*/,
@ -63,14 +268,22 @@ class VPxEncoderThreadTest
encoder->Control(VP8E_SET_ARNR_MAXFRAMES, 7);
encoder->Control(VP8E_SET_ARNR_STRENGTH, 5);
encoder->Control(VP8E_SET_ARNR_TYPE, 3);
encoder->Control(VP9E_SET_FRAME_PARALLEL_DECODING, 0);
} else {
encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 0);
encoder->Control(VP9E_SET_AQ_MODE, 3);
}
encoder->Control(VP9E_SET_ROW_MT, row_mt_mode_);
encoder_initialized_ = true;
}
}
virtual void PSNRPktHook(const vpx_codec_cx_pkt_t *pkt) {
psnr_ += pkt->data.psnr.psnr[0];
nframes_++;
}
virtual void DecompressedFrameHook(const vpx_image_t &img,
vpx_codec_pts_t /*pts*/) {
::libvpx_test::MD5 md5_res;
@ -89,40 +302,127 @@ class VPxEncoderThreadTest
return true;
}
double GetAveragePsnr() const { return nframes_ ? (psnr_ / nframes_) : 0.0; }
bool encoder_initialized_;
int tiles_;
int threads_;
::libvpx_test::TestMode encoding_mode_;
int set_cpu_used_;
int row_mt_mode_;
double psnr_;
unsigned int nframes_;
std::vector<std::string> md5_;
};
TEST_P(VPxEncoderThreadTest, EncoderResultTest) {
std::vector<std::string> single_thr_md5, multi_thr_md5;
::libvpx_test::Y4mVideoSource video("niklas_1280_720_30.y4m", 15, 20);
cfg_.rc_target_bitrate = 1000;
// Part 1: Bit exact test for row_mt_mode_ = 0.
// This part keeps original unit tests done before row-mt code is checked in.
row_mt_mode_ = 0;
// Encode using single thread.
cfg_.g_threads = 1;
init_flags_ = VPX_CODEC_USE_PSNR;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
single_thr_md5 = md5_;
const std::vector<std::string> single_thr_md5 = md5_;
md5_.clear();
// Encode using multiple threads.
cfg_.g_threads = 4;
cfg_.g_threads = threads_;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
multi_thr_md5 = md5_;
const std::vector<std::string> multi_thr_md5 = md5_;
md5_.clear();
// Compare to check if two vectors are equal.
ASSERT_EQ(single_thr_md5, multi_thr_md5);
// Part 2: row_mt_mode_ = 0 vs row_mt_mode_ = 1 single thread bit exact test.
row_mt_mode_ = 1;
// Encode using single thread
cfg_.g_threads = 1;
init_flags_ = VPX_CODEC_USE_PSNR;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
std::vector<std::string> row_mt_single_thr_md5 = md5_;
md5_.clear();
ASSERT_EQ(single_thr_md5, row_mt_single_thr_md5);
// Part 3: Bit exact test with row-mt on
// When row_mt_mode_=1 and using >1 threads, the encoder generates bit exact
// result.
row_mt_mode_ = 1;
row_mt_single_thr_md5.clear();
// Encode using 2 threads.
cfg_.g_threads = 2;
init_flags_ = VPX_CODEC_USE_PSNR;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
row_mt_single_thr_md5 = md5_;
md5_.clear();
// Encode using multiple threads.
cfg_.g_threads = threads_;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
const std::vector<std::string> row_mt_multi_thr_md5 = md5_;
md5_.clear();
// Compare to check if two vectors are equal.
ASSERT_EQ(row_mt_single_thr_md5, row_mt_multi_thr_md5);
// Part 4: PSNR test with bit_match_mode_ = 0
row_mt_mode_ = 1;
// Encode using single thread.
cfg_.g_threads = 1;
init_flags_ = VPX_CODEC_USE_PSNR;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
const double single_thr_psnr = GetAveragePsnr();
// Encode using multiple threads.
cfg_.g_threads = threads_;
ASSERT_NO_FATAL_FAILURE(RunLoop(&video));
const double multi_thr_psnr = GetAveragePsnr();
EXPECT_NEAR(single_thr_psnr, multi_thr_psnr, 0.1);
}
VP9_INSTANTIATE_TEST_CASE(VPxEncoderThreadTest,
::testing::Values(::libvpx_test::kTwoPassGood,
::libvpx_test::kOnePassGood,
::libvpx_test::kRealTime),
::testing::Range(1, 9));
INSTANTIATE_TEST_CASE_P(
VP9, VPxFirstPassEncoderThreadTest,
::testing::Combine(
::testing::Values(
static_cast<const libvpx_test::CodecFactory *>(&libvpx_test::kVP9)),
::testing::Values(::libvpx_test::kTwoPassGood),
::testing::Range(0, 4))); // cpu_used
// Split this into two instantiations so that we can distinguish
// between very slow runs ( ie cpu_speed 0 ) vs ones that can be
// run nightly by adding Large to the title.
INSTANTIATE_TEST_CASE_P(
VP9, VPxEncoderThreadTest,
::testing::Combine(
::testing::Values(
static_cast<const libvpx_test::CodecFactory *>(&libvpx_test::kVP9)),
::testing::Values(::libvpx_test::kTwoPassGood,
::libvpx_test::kOnePassGood,
::libvpx_test::kRealTime),
::testing::Range(3, 9), // cpu_used
::testing::Range(0, 3), // tile_columns
::testing::Range(2, 5))); // threads
INSTANTIATE_TEST_CASE_P(
VP9Large, VPxEncoderThreadTest,
::testing::Combine(
::testing::Values(
static_cast<const libvpx_test::CodecFactory *>(&libvpx_test::kVP9)),
::testing::Values(::libvpx_test::kTwoPassGood,
::libvpx_test::kOnePassGood,
::libvpx_test::kRealTime),
::testing::Range(0, 3), // cpu_used
::testing::Range(0, 3), // tile_columns
::testing::Range(2, 5))); // threads
} // namespace

View File

@ -1,217 +0,0 @@
/*
* Copyright (c) 2014 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <cstdio>
#include <cstdlib>
#include <string>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vpx_config.h"
#include "test/codec_factory.h"
#include "test/decode_test_driver.h"
#include "test/ivf_video_source.h"
#include "test/md5_helper.h"
#include "test/util.h"
#if CONFIG_WEBM_IO
#include "test/webm_video_source.h"
#endif
#include "vpx_mem/vpx_mem.h"
namespace {
using std::string;
#if CONFIG_WEBM_IO
struct PauseFileList {
const char *name;
// md5 sum for decoded frames which does not include skipped frames.
const char *expected_md5;
const int pause_frame_num;
};
// Decodes |filename| with |num_threads|. Pause at the specified frame_num,
// seek to next key frame and then continue decoding until the end. Return
// the md5 of the decoded frames which does not include skipped frames.
string DecodeFileWithPause(const string &filename, int num_threads,
int pause_num) {
libvpx_test::WebMVideoSource video(filename);
video.Init();
int in_frames = 0;
int out_frames = 0;
vpx_codec_dec_cfg_t cfg = vpx_codec_dec_cfg_t();
cfg.threads = num_threads;
vpx_codec_flags_t flags = 0;
flags |= VPX_CODEC_USE_FRAME_THREADING;
libvpx_test::VP9Decoder decoder(cfg, flags);
libvpx_test::MD5 md5;
video.Begin();
do {
++in_frames;
const vpx_codec_err_t res =
decoder.DecodeFrame(video.cxdata(), video.frame_size());
if (res != VPX_CODEC_OK) {
EXPECT_EQ(VPX_CODEC_OK, res) << decoder.DecodeError();
break;
}
// Pause at specified frame number.
if (in_frames == pause_num) {
// Flush the decoder and then seek to next key frame.
decoder.DecodeFrame(NULL, 0);
video.SeekToNextKeyFrame();
} else {
video.Next();
}
// Flush the decoder at the end of the video.
if (!video.cxdata()) decoder.DecodeFrame(NULL, 0);
libvpx_test::DxDataIterator dec_iter = decoder.GetDxData();
const vpx_image_t *img;
// Get decompressed data
while ((img = dec_iter.Next())) {
++out_frames;
md5.Add(img);
}
} while (video.cxdata() != NULL);
EXPECT_EQ(in_frames, out_frames)
<< "Input frame count does not match output frame count";
return string(md5.Get());
}
void DecodeFilesWithPause(const PauseFileList files[]) {
for (const PauseFileList *iter = files; iter->name != NULL; ++iter) {
SCOPED_TRACE(iter->name);
for (int t = 2; t <= 8; ++t) {
EXPECT_EQ(iter->expected_md5,
DecodeFileWithPause(iter->name, t, iter->pause_frame_num))
<< "threads = " << t;
}
}
}
TEST(VP9MultiThreadedFrameParallel, PauseSeekResume) {
// vp90-2-07-frame_parallel-1.webm is a 40 frame video file with
// one key frame for every ten frames.
static const PauseFileList files[] = {
{ "vp90-2-07-frame_parallel-1.webm", "6ea7c3875d67252e7caf2bc6e75b36b1",
6 },
{ "vp90-2-07-frame_parallel-1.webm", "4bb634160c7356a8d7d4299b6dc83a45",
12 },
{ "vp90-2-07-frame_parallel-1.webm", "89772591e6ef461f9fa754f916c78ed8",
26 },
{ NULL, NULL, 0 },
};
DecodeFilesWithPause(files);
}
struct FileList {
const char *name;
// md5 sum for decoded frames which does not include corrupted frames.
const char *expected_md5;
// Expected number of decoded frames which does not include corrupted frames.
const int expected_frame_count;
};
// Decodes |filename| with |num_threads|. Return the md5 of the decoded
// frames which does not include corrupted frames.
string DecodeFile(const string &filename, int num_threads,
int expected_frame_count) {
libvpx_test::WebMVideoSource video(filename);
video.Init();
vpx_codec_dec_cfg_t cfg = vpx_codec_dec_cfg_t();
cfg.threads = num_threads;
const vpx_codec_flags_t flags = VPX_CODEC_USE_FRAME_THREADING;
libvpx_test::VP9Decoder decoder(cfg, flags);
libvpx_test::MD5 md5;
video.Begin();
int out_frames = 0;
do {
const vpx_codec_err_t res =
decoder.DecodeFrame(video.cxdata(), video.frame_size());
// TODO(hkuang): frame parallel mode should return an error on corruption.
if (res != VPX_CODEC_OK) {
EXPECT_EQ(VPX_CODEC_OK, res) << decoder.DecodeError();
break;
}
video.Next();
// Flush the decoder at the end of the video.
if (!video.cxdata()) decoder.DecodeFrame(NULL, 0);
libvpx_test::DxDataIterator dec_iter = decoder.GetDxData();
const vpx_image_t *img;
// Get decompressed data
while ((img = dec_iter.Next())) {
++out_frames;
md5.Add(img);
}
} while (video.cxdata() != NULL);
EXPECT_EQ(expected_frame_count, out_frames)
<< "Input frame count does not match expected output frame count";
return string(md5.Get());
}
void DecodeFiles(const FileList files[]) {
for (const FileList *iter = files; iter->name != NULL; ++iter) {
SCOPED_TRACE(iter->name);
for (int t = 2; t <= 8; ++t) {
EXPECT_EQ(iter->expected_md5,
DecodeFile(iter->name, t, iter->expected_frame_count))
<< "threads = " << t;
}
}
}
TEST(VP9MultiThreadedFrameParallel, InvalidFileTest) {
static const FileList files[] = {
// invalid-vp90-2-07-frame_parallel-1.webm is a 40 frame video file with
// one key frame for every ten frames. The 11th frame has corrupted data.
{ "invalid-vp90-2-07-frame_parallel-1.webm",
"0549d0f45f60deaef8eb708e6c0eb6cb", 30 },
// invalid-vp90-2-07-frame_parallel-2.webm is a 40 frame video file with
// one key frame for every ten frames. The 1st and 31st frames have
// corrupted data.
{ "invalid-vp90-2-07-frame_parallel-2.webm",
"6a1f3cf6f9e7a364212fadb9580d525e", 20 },
// invalid-vp90-2-07-frame_parallel-3.webm is a 40 frame video file with
// one key frame for every ten frames. The 5th and 13th frames have
// corrupted data.
{ "invalid-vp90-2-07-frame_parallel-3.webm",
"8256544308de926b0681e04685b98677", 27 },
{ NULL, NULL, 0 },
};
DecodeFiles(files);
}
TEST(VP9MultiThreadedFrameParallel, ValidFileTest) {
static const FileList files[] = {
#if CONFIG_VP9_HIGHBITDEPTH
{ "vp92-2-20-10bit-yuv420.webm", "a16b99df180c584e8db2ffeda987d293", 10 },
#endif
{ NULL, NULL, 0 },
};
DecodeFiles(files);
}
#endif // CONFIG_WEBM_IO
} // namespace

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,97 @@
/*
* Copyright (c) 2017 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "test/codec_factory.h"
#include "test/encode_test_driver.h"
#include "test/util.h"
#include "test/yuv_video_source.h"
namespace {
#define MAX_EXTREME_MV 1
#define MIN_EXTREME_MV 2
// Encoding modes
const libvpx_test::TestMode kEncodingModeVectors[] = {
::libvpx_test::kTwoPassGood, ::libvpx_test::kOnePassGood,
::libvpx_test::kRealTime,
};
// Encoding speeds
const int kCpuUsedVectors[] = { 0, 1, 2, 3, 4, 5, 6 };
// MV test modes: 1 - always use maximum MV; 2 - always use minimum MV.
const int kMVTestModes[] = { MAX_EXTREME_MV, MIN_EXTREME_MV };
class MotionVectorTestLarge
: public ::libvpx_test::EncoderTest,
public ::libvpx_test::CodecTestWith3Params<libvpx_test::TestMode, int,
int> {
protected:
MotionVectorTestLarge()
: EncoderTest(GET_PARAM(0)), encoding_mode_(GET_PARAM(1)),
cpu_used_(GET_PARAM(2)), mv_test_mode_(GET_PARAM(3)) {}
virtual ~MotionVectorTestLarge() {}
virtual void SetUp() {
InitializeConfig();
SetMode(encoding_mode_);
if (encoding_mode_ != ::libvpx_test::kRealTime) {
cfg_.g_lag_in_frames = 3;
cfg_.rc_end_usage = VPX_VBR;
} else {
cfg_.g_lag_in_frames = 0;
cfg_.rc_end_usage = VPX_CBR;
cfg_.rc_buf_sz = 1000;
cfg_.rc_buf_initial_sz = 500;
cfg_.rc_buf_optimal_sz = 600;
}
}
virtual void PreEncodeFrameHook(::libvpx_test::VideoSource *video,
::libvpx_test::Encoder *encoder) {
if (video->frame() == 1) {
encoder->Control(VP8E_SET_CPUUSED, cpu_used_);
encoder->Control(VP9E_ENABLE_MOTION_VECTOR_UNIT_TEST, mv_test_mode_);
if (encoding_mode_ != ::libvpx_test::kRealTime) {
encoder->Control(VP8E_SET_ENABLEAUTOALTREF, 1);
encoder->Control(VP8E_SET_ARNR_MAXFRAMES, 7);
encoder->Control(VP8E_SET_ARNR_STRENGTH, 5);
encoder->Control(VP8E_SET_ARNR_TYPE, 3);
}
}
}
libvpx_test::TestMode encoding_mode_;
int cpu_used_;
int mv_test_mode_;
};
TEST_P(MotionVectorTestLarge, OverallTest) {
cfg_.rc_target_bitrate = 24000;
cfg_.g_profile = 0;
init_flags_ = VPX_CODEC_USE_PSNR;
testing::internal::scoped_ptr<libvpx_test::VideoSource> video;
video.reset(new libvpx_test::YUVVideoSource(
"niklas_640_480_30.yuv", VPX_IMG_FMT_I420, 3840, 2160, // 2048, 1080,
30, 1, 0, 5));
ASSERT_TRUE(video.get() != NULL);
ASSERT_NO_FATAL_FAILURE(RunLoop(video.get()));
}
VP9_INSTANTIATE_TEST_CASE(MotionVectorTestLarge,
::testing::ValuesIn(kEncodingModeVectors),
::testing::ValuesIn(kCpuUsedVectors),
::testing::ValuesIn(kMVTestModes));
} // namespace

View File

@ -14,9 +14,11 @@
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vp9_rtcd.h"
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"
#include "test/acm_random.h"
#include "test/buffer.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "test/util.h"
@ -24,11 +26,12 @@
#include "vp9/common/vp9_scan.h"
#include "vpx/vpx_codec.h"
#include "vpx/vpx_integer.h"
#include "vpx_ports/vpx_timer.h"
using libvpx_test::ACMRandom;
using libvpx_test::Buffer;
namespace {
#if CONFIG_VP9_HIGHBITDEPTH
const int number_of_iterations = 100;
typedef void (*QuantizeFunc)(const tran_low_t *coeff, intptr_t count,
@ -38,307 +41,494 @@ typedef void (*QuantizeFunc)(const tran_low_t *coeff, intptr_t count,
tran_low_t *dqcoeff, const int16_t *dequant,
uint16_t *eob, const int16_t *scan,
const int16_t *iscan);
typedef std::tr1::tuple<QuantizeFunc, QuantizeFunc, vpx_bit_depth_t>
typedef std::tr1::tuple<QuantizeFunc, QuantizeFunc, vpx_bit_depth_t,
int /*max_size*/, bool /*is_fp*/>
QuantizeParam;
class VP9QuantizeTest : public ::testing::TestWithParam<QuantizeParam> {
// Wrapper for FP version which does not use zbin or quant_shift.
typedef void (*QuantizeFPFunc)(const tran_low_t *coeff, intptr_t count,
int skip_block, const int16_t *round,
const int16_t *quant, tran_low_t *qcoeff,
tran_low_t *dqcoeff, const int16_t *dequant,
uint16_t *eob, const int16_t *scan,
const int16_t *iscan);
template <QuantizeFPFunc fn>
void QuantFPWrapper(const tran_low_t *coeff, intptr_t count, int skip_block,
const int16_t *zbin, const int16_t *round,
const int16_t *quant, const int16_t *quant_shift,
tran_low_t *qcoeff, tran_low_t *dqcoeff,
const int16_t *dequant, uint16_t *eob, const int16_t *scan,
const int16_t *iscan) {
(void)zbin;
(void)quant_shift;
fn(coeff, count, skip_block, round, quant, qcoeff, dqcoeff, dequant, eob,
scan, iscan);
}
class VP9QuantizeBase {
public:
virtual ~VP9QuantizeTest() {}
virtual void SetUp() {
quantize_op_ = GET_PARAM(0);
ref_quantize_op_ = GET_PARAM(1);
bit_depth_ = GET_PARAM(2);
mask_ = (1 << bit_depth_) - 1;
VP9QuantizeBase(vpx_bit_depth_t bit_depth, int max_size, bool is_fp)
: bit_depth_(bit_depth), max_size_(max_size), is_fp_(is_fp) {
max_value_ = (1 << bit_depth_) - 1;
zbin_ptr_ =
reinterpret_cast<int16_t *>(vpx_memalign(16, 8 * sizeof(*zbin_ptr_)));
round_fp_ptr_ = reinterpret_cast<int16_t *>(
vpx_memalign(16, 8 * sizeof(*round_fp_ptr_)));
quant_fp_ptr_ = reinterpret_cast<int16_t *>(
vpx_memalign(16, 8 * sizeof(*quant_fp_ptr_)));
round_ptr_ =
reinterpret_cast<int16_t *>(vpx_memalign(16, 8 * sizeof(*round_ptr_)));
quant_ptr_ =
reinterpret_cast<int16_t *>(vpx_memalign(16, 8 * sizeof(*quant_ptr_)));
quant_shift_ptr_ = reinterpret_cast<int16_t *>(
vpx_memalign(16, 8 * sizeof(*quant_shift_ptr_)));
dequant_ptr_ = reinterpret_cast<int16_t *>(
vpx_memalign(16, 8 * sizeof(*dequant_ptr_)));
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
vpx_bit_depth_t bit_depth_;
int mask_;
QuantizeFunc quantize_op_;
QuantizeFunc ref_quantize_op_;
};
class VP9Quantize32Test : public ::testing::TestWithParam<QuantizeParam> {
public:
virtual ~VP9Quantize32Test() {}
virtual void SetUp() {
quantize_op_ = GET_PARAM(0);
ref_quantize_op_ = GET_PARAM(1);
bit_depth_ = GET_PARAM(2);
mask_ = (1 << bit_depth_) - 1;
~VP9QuantizeBase() {
vpx_free(zbin_ptr_);
vpx_free(round_fp_ptr_);
vpx_free(quant_fp_ptr_);
vpx_free(round_ptr_);
vpx_free(quant_ptr_);
vpx_free(quant_shift_ptr_);
vpx_free(dequant_ptr_);
zbin_ptr_ = NULL;
round_fp_ptr_ = NULL;
quant_fp_ptr_ = NULL;
round_ptr_ = NULL;
quant_ptr_ = NULL;
quant_shift_ptr_ = NULL;
dequant_ptr_ = NULL;
libvpx_test::ClearSystemState();
}
virtual void TearDown() { libvpx_test::ClearSystemState(); }
protected:
int16_t *zbin_ptr_;
int16_t *round_fp_ptr_;
int16_t *quant_fp_ptr_;
int16_t *round_ptr_;
int16_t *quant_ptr_;
int16_t *quant_shift_ptr_;
int16_t *dequant_ptr_;
const vpx_bit_depth_t bit_depth_;
int max_value_;
const int max_size_;
const bool is_fp_;
};
class VP9QuantizeTest : public VP9QuantizeBase,
public ::testing::TestWithParam<QuantizeParam> {
public:
VP9QuantizeTest()
: VP9QuantizeBase(GET_PARAM(2), GET_PARAM(3), GET_PARAM(4)),
quantize_op_(GET_PARAM(0)), ref_quantize_op_(GET_PARAM(1)) {}
protected:
vpx_bit_depth_t bit_depth_;
int mask_;
QuantizeFunc quantize_op_;
QuantizeFunc ref_quantize_op_;
const QuantizeFunc quantize_op_;
const QuantizeFunc ref_quantize_op_;
};
// This quantizer compares the AC coefficients to the quantization step size to
// determine if further multiplication operations are needed.
// Based on vp9_quantize_fp_sse2().
void quantize_fp_nz_c(const tran_low_t *coeff_ptr, intptr_t n_coeffs,
int skip_block, const int16_t *round_ptr,
const int16_t *quant_ptr, tran_low_t *qcoeff_ptr,
tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr,
uint16_t *eob_ptr, const int16_t *scan,
const int16_t *iscan) {
int i, eob = -1;
const int thr = dequant_ptr[1] >> 1;
(void)iscan;
(void)skip_block;
assert(!skip_block);
// Quantization pass: All coefficients with index >= zero_flag are
// skippable. Note: zero_flag can be zero.
for (i = 0; i < n_coeffs; i += 16) {
int y;
int nzflag_cnt = 0;
int abs_coeff[16];
int coeff_sign[16];
// count nzflag for each row (16 tran_low_t)
for (y = 0; y < 16; ++y) {
const int rc = i + y;
const int coeff = coeff_ptr[rc];
coeff_sign[y] = (coeff >> 31);
abs_coeff[y] = (coeff ^ coeff_sign[y]) - coeff_sign[y];
// The first 16 are skipped in the sse2 code. Do the same here to match.
if (i >= 16 && (abs_coeff[y] <= thr)) {
nzflag_cnt++;
}
}
for (y = 0; y < 16; ++y) {
const int rc = i + y;
// If all of the AC coeffs in a row has magnitude less than the
// quantization step_size/2, quantize to zero.
if (nzflag_cnt < 16) {
int tmp =
clamp(abs_coeff[y] + round_ptr[rc != 0], INT16_MIN, INT16_MAX);
tmp = (tmp * quant_ptr[rc != 0]) >> 16;
qcoeff_ptr[rc] = (tmp ^ coeff_sign[y]) - coeff_sign[y];
dqcoeff_ptr[rc] = qcoeff_ptr[rc] * dequant_ptr[rc != 0];
} else {
qcoeff_ptr[rc] = 0;
dqcoeff_ptr[rc] = 0;
}
}
}
// Scan for eob.
for (i = 0; i < n_coeffs; i++) {
// Use the scan order to find the correct eob.
const int rc = scan[i];
if (qcoeff_ptr[rc]) {
eob = i;
}
}
*eob_ptr = eob + 1;
}
void GenerateHelperArrays(ACMRandom *rnd, int16_t *zbin, int16_t *round,
int16_t *quant, int16_t *quant_shift,
int16_t *dequant, int16_t *round_fp,
int16_t *quant_fp) {
// Max when q == 0. Otherwise, it is 48 for Y and 42 for U/V.
const int max_qrounding_factor_fp = 64;
for (int j = 0; j < 2; j++) {
// The range is 4 to 1828 in the VP9 tables.
const int qlookup = rnd->RandRange(1825) + 4;
round_fp[j] = (max_qrounding_factor_fp * qlookup) >> 7;
quant_fp[j] = (1 << 16) / qlookup;
// Values determined by deconstructing vp9_init_quantizer().
// zbin may be up to 1143 for 8 and 10 bit Y values, or 1200 for 12 bit Y
// values or U/V values of any bit depth. This is because y_delta is not
// factored into the vp9_ac_quant() call.
zbin[j] = rnd->RandRange(1200);
// round may be up to 685 for Y values or 914 for U/V.
round[j] = rnd->RandRange(914);
// quant ranges from 1 to -32703
quant[j] = static_cast<int>(rnd->RandRange(32704)) - 32703;
// quant_shift goes up to 1 << 16.
quant_shift[j] = rnd->RandRange(16384);
// dequant maxes out at 1828 for all cases.
dequant[j] = rnd->RandRange(1828);
}
for (int j = 2; j < 8; j++) {
zbin[j] = zbin[1];
round_fp[j] = round_fp[1];
quant_fp[j] = quant_fp[1];
round[j] = round[1];
quant[j] = quant[1];
quant_shift[j] = quant_shift[1];
dequant[j] = dequant[1];
}
}
TEST_P(VP9QuantizeTest, OperationCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
DECLARE_ALIGNED(16, tran_low_t, coeff_ptr[256]);
DECLARE_ALIGNED(16, int16_t, zbin_ptr[2]);
DECLARE_ALIGNED(16, int16_t, round_ptr[2]);
DECLARE_ALIGNED(16, int16_t, quant_ptr[2]);
DECLARE_ALIGNED(16, int16_t, quant_shift_ptr[2]);
DECLARE_ALIGNED(16, tran_low_t, qcoeff_ptr[256]);
DECLARE_ALIGNED(16, tran_low_t, dqcoeff_ptr[256]);
DECLARE_ALIGNED(16, tran_low_t, ref_qcoeff_ptr[256]);
DECLARE_ALIGNED(16, tran_low_t, ref_dqcoeff_ptr[256]);
DECLARE_ALIGNED(16, int16_t, dequant_ptr[2]);
DECLARE_ALIGNED(16, uint16_t, eob_ptr[1]);
DECLARE_ALIGNED(16, uint16_t, ref_eob_ptr[1]);
int err_count_total = 0;
int first_failure = -1;
for (int i = 0; i < number_of_iterations; ++i) {
const int skip_block = i == 0;
const TX_SIZE sz = (TX_SIZE)(i % 3); // TX_4X4, TX_8X8 TX_16X16
const TX_TYPE tx_type = (TX_TYPE)((i >> 2) % 3);
const scan_order *scan_order = &vp9_scan_orders[sz][tx_type];
const int count = (4 << sz) * (4 << sz); // 16, 64, 256
int err_count = 0;
*eob_ptr = rnd.Rand16();
*ref_eob_ptr = *eob_ptr;
for (int j = 0; j < count; j++) {
coeff_ptr[j] = rnd.Rand16() & mask_;
}
for (int j = 0; j < 2; j++) {
zbin_ptr[j] = rnd.Rand16() & mask_;
round_ptr[j] = rnd.Rand16();
quant_ptr[j] = rnd.Rand16();
quant_shift_ptr[j] = rnd.Rand16();
dequant_ptr[j] = rnd.Rand16();
}
ref_quantize_op_(coeff_ptr, count, skip_block, zbin_ptr, round_ptr,
quant_ptr, quant_shift_ptr, ref_qcoeff_ptr,
ref_dqcoeff_ptr, dequant_ptr, ref_eob_ptr,
scan_order->scan, scan_order->iscan);
ASM_REGISTER_STATE_CHECK(quantize_op_(
coeff_ptr, count, skip_block, zbin_ptr, round_ptr, quant_ptr,
quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr, dequant_ptr, eob_ptr,
scan_order->scan, scan_order->iscan));
for (int j = 0; j < sz; ++j) {
err_count += (ref_qcoeff_ptr[j] != qcoeff_ptr[j]) |
(ref_dqcoeff_ptr[j] != dqcoeff_ptr[j]);
}
err_count += (*ref_eob_ptr != *eob_ptr);
if (err_count && !err_count_total) {
first_failure = i;
}
err_count_total += err_count;
}
EXPECT_EQ(0, err_count_total)
<< "Error: Quantization Test, C output doesn't match SSE2 output. "
<< "First failed at test case " << first_failure;
}
Buffer<tran_low_t> coeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 16);
ASSERT_TRUE(coeff.Init());
Buffer<tran_low_t> qcoeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(qcoeff.Init());
Buffer<tran_low_t> dqcoeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(dqcoeff.Init());
Buffer<tran_low_t> ref_qcoeff =
Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(ref_qcoeff.Init());
Buffer<tran_low_t> ref_dqcoeff =
Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(ref_dqcoeff.Init());
uint16_t eob, ref_eob;
TEST_P(VP9Quantize32Test, OperationCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
DECLARE_ALIGNED(16, tran_low_t, coeff_ptr[1024]);
DECLARE_ALIGNED(16, int16_t, zbin_ptr[2]);
DECLARE_ALIGNED(16, int16_t, round_ptr[2]);
DECLARE_ALIGNED(16, int16_t, quant_ptr[2]);
DECLARE_ALIGNED(16, int16_t, quant_shift_ptr[2]);
DECLARE_ALIGNED(16, tran_low_t, qcoeff_ptr[1024]);
DECLARE_ALIGNED(16, tran_low_t, dqcoeff_ptr[1024]);
DECLARE_ALIGNED(16, tran_low_t, ref_qcoeff_ptr[1024]);
DECLARE_ALIGNED(16, tran_low_t, ref_dqcoeff_ptr[1024]);
DECLARE_ALIGNED(16, int16_t, dequant_ptr[2]);
DECLARE_ALIGNED(16, uint16_t, eob_ptr[1]);
DECLARE_ALIGNED(16, uint16_t, ref_eob_ptr[1]);
int err_count_total = 0;
int first_failure = -1;
for (int i = 0; i < number_of_iterations; ++i) {
const int skip_block = i == 0;
const TX_SIZE sz = TX_32X32;
const TX_TYPE tx_type = (TX_TYPE)(i % 4);
// Test skip block for the first three iterations to catch all the different
// sizes.
const int skip_block = 0;
TX_SIZE sz;
if (max_size_ == 16) {
sz = static_cast<TX_SIZE>(i % 3); // TX_4X4, TX_8X8 TX_16X16
} else {
sz = TX_32X32;
}
const TX_TYPE tx_type = static_cast<TX_TYPE>((i >> 2) % 3);
const scan_order *scan_order = &vp9_scan_orders[sz][tx_type];
const int count = (4 << sz) * (4 << sz); // 1024
int err_count = 0;
*eob_ptr = rnd.Rand16();
*ref_eob_ptr = *eob_ptr;
for (int j = 0; j < count; j++) {
coeff_ptr[j] = rnd.Rand16() & mask_;
}
for (int j = 0; j < 2; j++) {
zbin_ptr[j] = rnd.Rand16() & mask_;
round_ptr[j] = rnd.Rand16();
quant_ptr[j] = rnd.Rand16();
quant_shift_ptr[j] = rnd.Rand16();
dequant_ptr[j] = rnd.Rand16();
}
ref_quantize_op_(coeff_ptr, count, skip_block, zbin_ptr, round_ptr,
quant_ptr, quant_shift_ptr, ref_qcoeff_ptr,
ref_dqcoeff_ptr, dequant_ptr, ref_eob_ptr,
const int count = (4 << sz) * (4 << sz);
coeff.Set(&rnd, -max_value_, max_value_);
GenerateHelperArrays(&rnd, zbin_ptr_, round_ptr_, quant_ptr_,
quant_shift_ptr_, dequant_ptr_, round_fp_ptr_,
quant_fp_ptr_);
int16_t *r_ptr = (is_fp_) ? round_fp_ptr_ : round_ptr_;
int16_t *q_ptr = (is_fp_) ? quant_fp_ptr_ : quant_ptr_;
ref_quantize_op_(coeff.TopLeftPixel(), count, skip_block, zbin_ptr_, r_ptr,
q_ptr, quant_shift_ptr_, ref_qcoeff.TopLeftPixel(),
ref_dqcoeff.TopLeftPixel(), dequant_ptr_, &ref_eob,
scan_order->scan, scan_order->iscan);
ASM_REGISTER_STATE_CHECK(quantize_op_(
coeff_ptr, count, skip_block, zbin_ptr, round_ptr, quant_ptr,
quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr, dequant_ptr, eob_ptr,
scan_order->scan, scan_order->iscan));
for (int j = 0; j < sz; ++j) {
err_count += (ref_qcoeff_ptr[j] != qcoeff_ptr[j]) |
(ref_dqcoeff_ptr[j] != dqcoeff_ptr[j]);
coeff.TopLeftPixel(), count, skip_block, zbin_ptr_, r_ptr, q_ptr,
quant_shift_ptr_, qcoeff.TopLeftPixel(), dqcoeff.TopLeftPixel(),
dequant_ptr_, &eob, scan_order->scan, scan_order->iscan));
EXPECT_TRUE(qcoeff.CheckValues(ref_qcoeff));
EXPECT_TRUE(dqcoeff.CheckValues(ref_dqcoeff));
EXPECT_EQ(eob, ref_eob);
if (HasFailure()) {
printf("Failure on iteration %d.\n", i);
qcoeff.PrintDifference(ref_qcoeff);
dqcoeff.PrintDifference(ref_dqcoeff);
return;
}
err_count += (*ref_eob_ptr != *eob_ptr);
if (err_count && !err_count_total) {
first_failure = i;
}
err_count_total += err_count;
}
EXPECT_EQ(0, err_count_total)
<< "Error: Quantization Test, C output doesn't match SSE2 output. "
<< "First failed at test case " << first_failure;
}
TEST_P(VP9QuantizeTest, EOBCheck) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
DECLARE_ALIGNED(16, tran_low_t, coeff_ptr[256]);
DECLARE_ALIGNED(16, int16_t, zbin_ptr[2]);
DECLARE_ALIGNED(16, int16_t, round_ptr[2]);
DECLARE_ALIGNED(16, int16_t, quant_ptr[2]);
DECLARE_ALIGNED(16, int16_t, quant_shift_ptr[2]);
DECLARE_ALIGNED(16, tran_low_t, qcoeff_ptr[256]);
DECLARE_ALIGNED(16, tran_low_t, dqcoeff_ptr[256]);
DECLARE_ALIGNED(16, tran_low_t, ref_qcoeff_ptr[256]);
DECLARE_ALIGNED(16, tran_low_t, ref_dqcoeff_ptr[256]);
DECLARE_ALIGNED(16, int16_t, dequant_ptr[2]);
DECLARE_ALIGNED(16, uint16_t, eob_ptr[1]);
DECLARE_ALIGNED(16, uint16_t, ref_eob_ptr[1]);
int err_count_total = 0;
int first_failure = -1;
Buffer<tran_low_t> coeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 16);
ASSERT_TRUE(coeff.Init());
Buffer<tran_low_t> qcoeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(qcoeff.Init());
Buffer<tran_low_t> dqcoeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(dqcoeff.Init());
Buffer<tran_low_t> ref_qcoeff =
Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(ref_qcoeff.Init());
Buffer<tran_low_t> ref_dqcoeff =
Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(ref_dqcoeff.Init());
uint16_t eob, ref_eob;
for (int i = 0; i < number_of_iterations; ++i) {
int skip_block = i == 0;
TX_SIZE sz = (TX_SIZE)(i % 3); // TX_4X4, TX_8X8 TX_16X16
TX_TYPE tx_type = (TX_TYPE)((i >> 2) % 3);
const int skip_block = 0;
TX_SIZE sz;
if (max_size_ == 16) {
sz = static_cast<TX_SIZE>(i % 3); // TX_4X4, TX_8X8 TX_16X16
} else {
sz = TX_32X32;
}
const TX_TYPE tx_type = static_cast<TX_TYPE>((i >> 2) % 3);
const scan_order *scan_order = &vp9_scan_orders[sz][tx_type];
int count = (4 << sz) * (4 << sz); // 16, 64, 256
int err_count = 0;
*eob_ptr = rnd.Rand16();
*ref_eob_ptr = *eob_ptr;
int count = (4 << sz) * (4 << sz);
// Two random entries
for (int j = 0; j < count; j++) {
coeff_ptr[j] = 0;
}
coeff_ptr[rnd(count)] = rnd.Rand16() & mask_;
coeff_ptr[rnd(count)] = rnd.Rand16() & mask_;
for (int j = 0; j < 2; j++) {
zbin_ptr[j] = rnd.Rand16() & mask_;
round_ptr[j] = rnd.Rand16();
quant_ptr[j] = rnd.Rand16();
quant_shift_ptr[j] = rnd.Rand16();
dequant_ptr[j] = rnd.Rand16();
}
ref_quantize_op_(coeff_ptr, count, skip_block, zbin_ptr, round_ptr,
quant_ptr, quant_shift_ptr, ref_qcoeff_ptr,
ref_dqcoeff_ptr, dequant_ptr, ref_eob_ptr,
coeff.Set(0);
coeff.TopLeftPixel()[rnd(count)] =
static_cast<int>(rnd.RandRange(max_value_ * 2)) - max_value_;
coeff.TopLeftPixel()[rnd(count)] =
static_cast<int>(rnd.RandRange(max_value_ * 2)) - max_value_;
GenerateHelperArrays(&rnd, zbin_ptr_, round_ptr_, quant_ptr_,
quant_shift_ptr_, dequant_ptr_, round_fp_ptr_,
quant_fp_ptr_);
int16_t *r_ptr = (is_fp_) ? round_fp_ptr_ : round_ptr_;
int16_t *q_ptr = (is_fp_) ? quant_fp_ptr_ : quant_ptr_;
ref_quantize_op_(coeff.TopLeftPixel(), count, skip_block, zbin_ptr_, r_ptr,
q_ptr, quant_shift_ptr_, ref_qcoeff.TopLeftPixel(),
ref_dqcoeff.TopLeftPixel(), dequant_ptr_, &ref_eob,
scan_order->scan, scan_order->iscan);
ASM_REGISTER_STATE_CHECK(quantize_op_(
coeff_ptr, count, skip_block, zbin_ptr, round_ptr, quant_ptr,
quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr, dequant_ptr, eob_ptr,
scan_order->scan, scan_order->iscan));
for (int j = 0; j < sz; ++j) {
err_count += (ref_qcoeff_ptr[j] != qcoeff_ptr[j]) |
(ref_dqcoeff_ptr[j] != dqcoeff_ptr[j]);
ASM_REGISTER_STATE_CHECK(quantize_op_(
coeff.TopLeftPixel(), count, skip_block, zbin_ptr_, r_ptr, q_ptr,
quant_shift_ptr_, qcoeff.TopLeftPixel(), dqcoeff.TopLeftPixel(),
dequant_ptr_, &eob, scan_order->scan, scan_order->iscan));
EXPECT_TRUE(qcoeff.CheckValues(ref_qcoeff));
EXPECT_TRUE(dqcoeff.CheckValues(ref_dqcoeff));
EXPECT_EQ(eob, ref_eob);
if (HasFailure()) {
printf("Failure on iteration %d.\n", i);
qcoeff.PrintDifference(ref_qcoeff);
dqcoeff.PrintDifference(ref_dqcoeff);
return;
}
err_count += (*ref_eob_ptr != *eob_ptr);
if (err_count && !err_count_total) {
first_failure = i;
}
err_count_total += err_count;
}
EXPECT_EQ(0, err_count_total)
<< "Error: Quantization Test, C output doesn't match SSE2 output. "
<< "First failed at test case " << first_failure;
}
TEST_P(VP9Quantize32Test, EOBCheck) {
TEST_P(VP9QuantizeTest, DISABLED_Speed) {
ACMRandom rnd(ACMRandom::DeterministicSeed());
DECLARE_ALIGNED(16, tran_low_t, coeff_ptr[1024]);
DECLARE_ALIGNED(16, int16_t, zbin_ptr[2]);
DECLARE_ALIGNED(16, int16_t, round_ptr[2]);
DECLARE_ALIGNED(16, int16_t, quant_ptr[2]);
DECLARE_ALIGNED(16, int16_t, quant_shift_ptr[2]);
DECLARE_ALIGNED(16, tran_low_t, qcoeff_ptr[1024]);
DECLARE_ALIGNED(16, tran_low_t, dqcoeff_ptr[1024]);
DECLARE_ALIGNED(16, tran_low_t, ref_qcoeff_ptr[1024]);
DECLARE_ALIGNED(16, tran_low_t, ref_dqcoeff_ptr[1024]);
DECLARE_ALIGNED(16, int16_t, dequant_ptr[2]);
DECLARE_ALIGNED(16, uint16_t, eob_ptr[1]);
DECLARE_ALIGNED(16, uint16_t, ref_eob_ptr[1]);
int err_count_total = 0;
int first_failure = -1;
for (int i = 0; i < number_of_iterations; ++i) {
int skip_block = i == 0;
TX_SIZE sz = TX_32X32;
TX_TYPE tx_type = (TX_TYPE)(i % 4);
const scan_order *scan_order = &vp9_scan_orders[sz][tx_type];
int count = (4 << sz) * (4 << sz); // 1024
int err_count = 0;
*eob_ptr = rnd.Rand16();
*ref_eob_ptr = *eob_ptr;
for (int j = 0; j < count; j++) {
coeff_ptr[j] = 0;
}
// Two random entries
coeff_ptr[rnd(count)] = rnd.Rand16() & mask_;
coeff_ptr[rnd(count)] = rnd.Rand16() & mask_;
for (int j = 0; j < 2; j++) {
zbin_ptr[j] = rnd.Rand16() & mask_;
round_ptr[j] = rnd.Rand16();
quant_ptr[j] = rnd.Rand16();
quant_shift_ptr[j] = rnd.Rand16();
dequant_ptr[j] = rnd.Rand16();
}
Buffer<tran_low_t> coeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 16);
ASSERT_TRUE(coeff.Init());
Buffer<tran_low_t> qcoeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(qcoeff.Init());
Buffer<tran_low_t> dqcoeff = Buffer<tran_low_t>(max_size_, max_size_, 0, 32);
ASSERT_TRUE(dqcoeff.Init());
uint16_t eob;
TX_SIZE starting_sz, ending_sz;
ref_quantize_op_(coeff_ptr, count, skip_block, zbin_ptr, round_ptr,
quant_ptr, quant_shift_ptr, ref_qcoeff_ptr,
ref_dqcoeff_ptr, dequant_ptr, ref_eob_ptr,
scan_order->scan, scan_order->iscan);
ASM_REGISTER_STATE_CHECK(quantize_op_(
coeff_ptr, count, skip_block, zbin_ptr, round_ptr, quant_ptr,
quant_shift_ptr, qcoeff_ptr, dqcoeff_ptr, dequant_ptr, eob_ptr,
scan_order->scan, scan_order->iscan));
for (int j = 0; j < sz; ++j) {
err_count += (ref_qcoeff_ptr[j] != qcoeff_ptr[j]) |
(ref_dqcoeff_ptr[j] != dqcoeff_ptr[j]);
}
err_count += (*ref_eob_ptr != *eob_ptr);
if (err_count && !err_count_total) {
first_failure = i;
}
err_count_total += err_count;
if (max_size_ == 16) {
starting_sz = TX_4X4;
ending_sz = TX_16X16;
} else {
starting_sz = TX_32X32;
ending_sz = TX_32X32;
}
for (TX_SIZE sz = starting_sz; sz <= ending_sz; ++sz) {
// zbin > coeff, zbin < coeff.
for (int i = 0; i < 2; ++i) {
const int skip_block = 0;
// TX_TYPE defines the scan order. That is not relevant to the speed test.
// Pick the first one.
const TX_TYPE tx_type = DCT_DCT;
const scan_order *scan_order = &vp9_scan_orders[sz][tx_type];
const int count = (4 << sz) * (4 << sz);
GenerateHelperArrays(&rnd, zbin_ptr_, round_ptr_, quant_ptr_,
quant_shift_ptr_, dequant_ptr_, round_fp_ptr_,
quant_fp_ptr_);
int16_t *r_ptr = (is_fp_) ? round_fp_ptr_ : round_ptr_;
int16_t *q_ptr = (is_fp_) ? quant_fp_ptr_ : quant_ptr_;
if (i == 0) {
// When |coeff values| are less than zbin the results are 0.
int threshold = 100;
if (max_size_ == 32) {
// For 32x32, the threshold is halved. Double it to keep the values
// from clearing it.
threshold = 200;
}
for (int j = 0; j < 8; ++j) zbin_ptr_[j] = threshold;
coeff.Set(&rnd, -99, 99);
} else if (i == 1) {
for (int j = 0; j < 8; ++j) zbin_ptr_[j] = 50;
coeff.Set(&rnd, -500, 500);
}
vpx_usec_timer timer;
vpx_usec_timer_start(&timer);
for (int j = 0; j < 100000000 / count; ++j) {
quantize_op_(coeff.TopLeftPixel(), count, skip_block, zbin_ptr_, r_ptr,
q_ptr, quant_shift_ptr_, qcoeff.TopLeftPixel(),
dqcoeff.TopLeftPixel(), dequant_ptr_, &eob,
scan_order->scan, scan_order->iscan);
}
vpx_usec_timer_mark(&timer);
const int elapsed_time = static_cast<int>(vpx_usec_timer_elapsed(&timer));
if (i == 0) printf("Bypass calculations.\n");
if (i == 1) printf("Full calculations.\n");
printf("Quantize %dx%d time: %5d ms\n", 4 << sz, 4 << sz,
elapsed_time / 1000);
}
printf("\n");
}
EXPECT_EQ(0, err_count_total)
<< "Error: Quantization Test, C output doesn't match SSE2 output. "
<< "First failed at test case " << first_failure;
}
using std::tr1::make_tuple;
#if HAVE_SSE2
#if CONFIG_VP9_HIGHBITDEPTH
// TODO(johannkoenig): Fix vpx_quantize_b_sse2 in highbitdepth builds.
// make_tuple(&vpx_quantize_b_sse2, &vpx_highbd_quantize_b_c, VPX_BITS_8),
INSTANTIATE_TEST_CASE_P(
SSE2, VP9QuantizeTest,
::testing::Values(make_tuple(&vpx_highbd_quantize_b_sse2,
&vpx_highbd_quantize_b_c, VPX_BITS_8),
make_tuple(&vpx_highbd_quantize_b_sse2,
&vpx_highbd_quantize_b_c, VPX_BITS_10),
make_tuple(&vpx_highbd_quantize_b_sse2,
&vpx_highbd_quantize_b_c, VPX_BITS_12)));
::testing::Values(
make_tuple(&vpx_highbd_quantize_b_sse2, &vpx_highbd_quantize_b_c,
VPX_BITS_8, 16, false),
make_tuple(&vpx_highbd_quantize_b_sse2, &vpx_highbd_quantize_b_c,
VPX_BITS_10, 16, false),
make_tuple(&vpx_highbd_quantize_b_sse2, &vpx_highbd_quantize_b_c,
VPX_BITS_12, 16, false),
make_tuple(&vpx_highbd_quantize_b_32x32_sse2,
&vpx_highbd_quantize_b_32x32_c, VPX_BITS_8, 32, false),
make_tuple(&vpx_highbd_quantize_b_32x32_sse2,
&vpx_highbd_quantize_b_32x32_c, VPX_BITS_10, 32, false),
make_tuple(&vpx_highbd_quantize_b_32x32_sse2,
&vpx_highbd_quantize_b_32x32_c, VPX_BITS_12, 32, false)));
#else
INSTANTIATE_TEST_CASE_P(
SSE2, VP9Quantize32Test,
::testing::Values(make_tuple(&vpx_highbd_quantize_b_32x32_sse2,
&vpx_highbd_quantize_b_32x32_c, VPX_BITS_8),
make_tuple(&vpx_highbd_quantize_b_32x32_sse2,
&vpx_highbd_quantize_b_32x32_c, VPX_BITS_10),
make_tuple(&vpx_highbd_quantize_b_32x32_sse2,
&vpx_highbd_quantize_b_32x32_c, VPX_BITS_12)));
#endif // HAVE_SSE2
SSE2, VP9QuantizeTest,
::testing::Values(make_tuple(&vpx_quantize_b_sse2, &vpx_quantize_b_c,
VPX_BITS_8, 16, false),
make_tuple(&QuantFPWrapper<vp9_quantize_fp_sse2>,
&QuantFPWrapper<quantize_fp_nz_c>, VPX_BITS_8,
16, true)));
#endif // CONFIG_VP9_HIGHBITDEPTH
#endif // HAVE_SSE2
#if HAVE_SSSE3 && !CONFIG_VP9_HIGHBITDEPTH
#if ARCH_X86_64
INSTANTIATE_TEST_CASE_P(
SSSE3, VP9QuantizeTest,
::testing::Values(make_tuple(&vpx_quantize_b_ssse3, &vpx_quantize_b_c,
VPX_BITS_8, 16, false),
make_tuple(&QuantFPWrapper<vp9_quantize_fp_ssse3>,
&QuantFPWrapper<quantize_fp_nz_c>, VPX_BITS_8,
16, true)));
#else
INSTANTIATE_TEST_CASE_P(SSSE3, VP9QuantizeTest,
::testing::Values(make_tuple(&vpx_quantize_b_ssse3,
&vpx_quantize_b_c,
VPX_BITS_8, 16, false)));
#endif
#if ARCH_X86_64
// TODO(johannkoenig): SSSE3 optimizations do not yet pass this test.
INSTANTIATE_TEST_CASE_P(
DISABLED_SSSE3, VP9QuantizeTest,
::testing::Values(make_tuple(&vpx_quantize_b_32x32_ssse3,
&vpx_quantize_b_32x32_c, VPX_BITS_8, 32,
false),
make_tuple(&QuantFPWrapper<vp9_quantize_fp_32x32_ssse3>,
&QuantFPWrapper<vp9_quantize_fp_32x32_c>,
VPX_BITS_8, 32, true)));
#endif // ARCH_X86_64
#endif // HAVE_SSSE3 && !CONFIG_VP9_HIGHBITDEPTH
// TODO(johannkoenig): AVX optimizations do not yet pass the 32x32 test or
// highbitdepth configurations.
#if HAVE_AVX && !CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
AVX, VP9QuantizeTest,
::testing::Values(make_tuple(&vpx_quantize_b_avx, &vpx_quantize_b_c,
VPX_BITS_8, 16, false),
// Even though SSSE3 and AVX do not match the reference
// code, we can keep them in sync with each other.
make_tuple(&vpx_quantize_b_32x32_avx,
&vpx_quantize_b_32x32_ssse3, VPX_BITS_8, 32,
false)));
#endif // HAVE_AVX && !CONFIG_VP9_HIGHBITDEPTH
// TODO(webm:1448): dqcoeff is not handled correctly in HBD builds.
#if HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH
INSTANTIATE_TEST_CASE_P(
NEON, VP9QuantizeTest,
::testing::Values(make_tuple(&vpx_quantize_b_neon, &vpx_quantize_b_c,
VPX_BITS_8, 16, false),
make_tuple(&vpx_quantize_b_32x32_neon,
&vpx_quantize_b_32x32_c, VPX_BITS_8, 32,
false),
make_tuple(&QuantFPWrapper<vp9_quantize_fp_neon>,
&QuantFPWrapper<vp9_quantize_fp_c>, VPX_BITS_8,
16, true),
make_tuple(&QuantFPWrapper<vp9_quantize_fp_32x32_neon>,
&QuantFPWrapper<vp9_quantize_fp_32x32_c>,
VPX_BITS_8, 32, true)));
#endif // HAVE_NEON && !CONFIG_VP9_HIGHBITDEPTH
// Only useful to compare "Speed" test results.
INSTANTIATE_TEST_CASE_P(
DISABLED_C, VP9QuantizeTest,
::testing::Values(
make_tuple(&vpx_quantize_b_c, &vpx_quantize_b_c, VPX_BITS_8, 16, false),
make_tuple(&vpx_quantize_b_32x32_c, &vpx_quantize_b_32x32_c, VPX_BITS_8,
32, false),
make_tuple(&QuantFPWrapper<vp9_quantize_fp_c>,
&QuantFPWrapper<vp9_quantize_fp_c>, VPX_BITS_8, 16, true),
make_tuple(&QuantFPWrapper<quantize_fp_nz_c>,
&QuantFPWrapper<quantize_fp_nz_c>, VPX_BITS_8, 16, true),
make_tuple(&QuantFPWrapper<vp9_quantize_fp_32x32_c>,
&QuantFPWrapper<vp9_quantize_fp_32x32_c>, VPX_BITS_8, 32,
true)));
} // namespace

View File

@ -0,0 +1,214 @@
/*
* Copyright (c) 2017 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <assert.h>
#include <stdio.h>
#include <string.h>
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vp9_rtcd.h"
#include "./vpx_config.h"
#include "./vpx_scale_rtcd.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "test/vpx_scale_test.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/vpx_timer.h"
#include "vpx_scale/yv12config.h"
namespace libvpx_test {
typedef void (*ScaleFrameFunc)(const YV12_BUFFER_CONFIG *src,
YV12_BUFFER_CONFIG *dst,
INTERP_FILTER filter_type, int phase_scaler);
class ScaleTest : public VpxScaleBase,
public ::testing::TestWithParam<ScaleFrameFunc> {
public:
virtual ~ScaleTest() {}
protected:
virtual void SetUp() { scale_fn_ = GetParam(); }
void ReferenceScaleFrame(INTERP_FILTER filter_type, int phase_scaler) {
vp9_scale_and_extend_frame_c(&img_, &ref_img_, filter_type, phase_scaler);
}
void ScaleFrame(INTERP_FILTER filter_type, int phase_scaler) {
ASM_REGISTER_STATE_CHECK(
scale_fn_(&img_, &dst_img_, filter_type, phase_scaler));
}
void RunTest() {
static const int kNumSizesToTest = 20;
static const int kNumScaleFactorsToTest = 4;
static const int kSizesToTest[] = {
2, 4, 6, 8, 10, 12, 14, 16, 18, 20,
22, 24, 26, 28, 30, 32, 34, 68, 128, 134
};
static const int kScaleFactors[] = { 1, 2, 3, 4 };
for (INTERP_FILTER filter_type = 0; filter_type < 4; ++filter_type) {
for (int phase_scaler = 0; phase_scaler < 16; ++phase_scaler) {
for (int h = 0; h < kNumSizesToTest; ++h) {
const int src_height = kSizesToTest[h];
for (int w = 0; w < kNumSizesToTest; ++w) {
const int src_width = kSizesToTest[w];
for (int sf_up_idx = 0; sf_up_idx < kNumScaleFactorsToTest;
++sf_up_idx) {
const int sf_up = kScaleFactors[sf_up_idx];
for (int sf_down_idx = 0; sf_down_idx < kNumScaleFactorsToTest;
++sf_down_idx) {
const int sf_down = kScaleFactors[sf_down_idx];
const int dst_width = src_width * sf_up / sf_down;
const int dst_height = src_height * sf_up / sf_down;
if (sf_up == sf_down && sf_up != 1) {
continue;
}
// I420 frame width and height must be even.
if (!dst_width || !dst_height || dst_width & 1 ||
dst_height & 1) {
continue;
}
// vpx_convolve8_c() has restriction on the step which cannot
// exceed 64 (ratio 1 to 4).
if (src_width > 4 * dst_width || src_height > 4 * dst_height) {
continue;
}
ASSERT_NO_FATAL_FAILURE(ResetScaleImages(
src_width, src_height, dst_width, dst_height));
ReferenceScaleFrame(filter_type, phase_scaler);
ScaleFrame(filter_type, phase_scaler);
if (memcmp(dst_img_.buffer_alloc, ref_img_.buffer_alloc,
ref_img_.frame_size)) {
printf(
"filter_type = %d, phase_scaler = %d, src_width = %4d, "
"src_height = %4d, dst_width = %4d, dst_height = %4d, "
"scale factor = %d:%d\n",
filter_type, phase_scaler, src_width, src_height,
dst_width, dst_height, sf_down, sf_up);
PrintDiff();
}
CompareImages(dst_img_);
DeallocScaleImages();
}
}
}
}
}
}
}
void PrintDiffComponent(const uint8_t *const ref, const uint8_t *const opt,
const int stride, const int width, const int height,
const int plane_idx) const {
for (int y = 0; y < height; y++) {
for (int x = 0; x < width; x++) {
if (ref[y * stride + x] != opt[y * stride + x]) {
printf("Plane %d pixel[%d][%d] diff:%6d (ref),%6d (opt)\n", plane_idx,
y, x, ref[y * stride + x], opt[y * stride + x]);
break;
}
}
}
}
void PrintDiff() const {
assert(ref_img_.y_stride == dst_img_.y_stride);
assert(ref_img_.y_width == dst_img_.y_width);
assert(ref_img_.y_height == dst_img_.y_height);
assert(ref_img_.uv_stride == dst_img_.uv_stride);
assert(ref_img_.uv_width == dst_img_.uv_width);
assert(ref_img_.uv_height == dst_img_.uv_height);
if (memcmp(dst_img_.buffer_alloc, ref_img_.buffer_alloc,
ref_img_.frame_size)) {
PrintDiffComponent(ref_img_.y_buffer, dst_img_.y_buffer,
ref_img_.y_stride, ref_img_.y_width, ref_img_.y_height,
0);
PrintDiffComponent(ref_img_.u_buffer, dst_img_.u_buffer,
ref_img_.uv_stride, ref_img_.uv_width,
ref_img_.uv_height, 1);
PrintDiffComponent(ref_img_.v_buffer, dst_img_.v_buffer,
ref_img_.uv_stride, ref_img_.uv_width,
ref_img_.uv_height, 2);
}
}
ScaleFrameFunc scale_fn_;
};
TEST_P(ScaleTest, ScaleFrame) { ASSERT_NO_FATAL_FAILURE(RunTest()); }
TEST_P(ScaleTest, DISABLED_Speed) {
static const int kCountSpeedTestBlock = 100;
static const int kNumScaleFactorsToTest = 4;
static const int kScaleFactors[] = { 1, 2, 3, 4 };
const int src_width = 1280;
const int src_height = 720;
for (INTERP_FILTER filter_type = 2; filter_type < 4; ++filter_type) {
for (int phase_scaler = 0; phase_scaler < 2; ++phase_scaler) {
for (int sf_up_idx = 0; sf_up_idx < kNumScaleFactorsToTest; ++sf_up_idx) {
const int sf_up = kScaleFactors[sf_up_idx];
for (int sf_down_idx = 0; sf_down_idx < kNumScaleFactorsToTest;
++sf_down_idx) {
const int sf_down = kScaleFactors[sf_down_idx];
const int dst_width = src_width * sf_up / sf_down;
const int dst_height = src_height * sf_up / sf_down;
if (sf_up == sf_down && sf_up != 1) {
continue;
}
// I420 frame width and height must be even.
if (dst_width & 1 || dst_height & 1) {
continue;
}
ASSERT_NO_FATAL_FAILURE(
ResetScaleImages(src_width, src_height, dst_width, dst_height));
ASM_REGISTER_STATE_CHECK(
ReferenceScaleFrame(filter_type, phase_scaler));
vpx_usec_timer timer;
vpx_usec_timer_start(&timer);
for (int i = 0; i < kCountSpeedTestBlock; ++i) {
ScaleFrame(filter_type, phase_scaler);
}
libvpx_test::ClearSystemState();
vpx_usec_timer_mark(&timer);
const int elapsed_time =
static_cast<int>(vpx_usec_timer_elapsed(&timer) / 1000);
CompareImages(dst_img_);
DeallocScaleImages();
printf(
"filter_type = %d, phase_scaler = %d, src_width = %4d, "
"src_height = %4d, dst_width = %4d, dst_height = %4d, "
"scale factor = %d:%d, scale time: %5d ms\n",
filter_type, phase_scaler, src_width, src_height, dst_width,
dst_height, sf_down, sf_up, elapsed_time);
}
}
}
}
}
INSTANTIATE_TEST_CASE_P(C, ScaleTest,
::testing::Values(vp9_scale_and_extend_frame_c));
#if HAVE_SSSE3
INSTANTIATE_TEST_CASE_P(SSSE3, ScaleTest,
::testing::Values(vp9_scale_and_extend_frame_ssse3));
#endif // HAVE_SSSE3
#if HAVE_NEON
INSTANTIATE_TEST_CASE_P(NEON, ScaleTest,
::testing::Values(vp9_scale_and_extend_frame_neon));
#endif // HAVE_NEON
} // namespace libvpx_test

View File

@ -85,8 +85,8 @@ class SkipLoopFilterTest {
// TODO(fgalligan): Move the MD5 testing code into another class.
void OpenMd5File(const std::string &md5_file_name) {
md5_file_ = libvpx_test::OpenTestDataFile(md5_file_name);
ASSERT_TRUE(md5_file_ != NULL) << "MD5 file open failed. Filename: "
<< md5_file_name;
ASSERT_TRUE(md5_file_ != NULL)
<< "MD5 file open failed. Filename: " << md5_file_name;
}
// Reads the next line of the MD5 file.

View File

@ -101,4 +101,9 @@ INSTANTIATE_TEST_CASE_P(MSA, VP9SubtractBlockTest,
::testing::Values(vpx_subtract_block_msa));
#endif
#if HAVE_MMI
INSTANTIATE_TEST_CASE_P(MMI, VP9SubtractBlockTest,
::testing::Values(vpx_subtract_block_mmi));
#endif
} // namespace vp9

View File

@ -187,8 +187,8 @@ void DecodeFiles(const FileList files[]) {
for (const FileList *iter = files; iter->name != NULL; ++iter) {
SCOPED_TRACE(iter->name);
for (int t = 1; t <= 8; ++t) {
EXPECT_EQ(iter->expected_md5, DecodeFile(iter->name, t)) << "threads = "
<< t;
EXPECT_EQ(iter->expected_md5, DecodeFile(iter->name, t))
<< "threads = " << t;
}
}
}

View File

@ -14,149 +14,17 @@
#include "./vpx_scale_rtcd.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "test/vpx_scale_test.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/vpx_timer.h"
#include "vpx_scale/yv12config.h"
namespace {
namespace libvpx_test {
typedef void (*ExtendFrameBorderFunc)(YV12_BUFFER_CONFIG *ybf);
typedef void (*CopyFrameFunc)(const YV12_BUFFER_CONFIG *src_ybf,
YV12_BUFFER_CONFIG *dst_ybf);
class VpxScaleBase {
public:
virtual ~VpxScaleBase() { libvpx_test::ClearSystemState(); }
void ResetImage(int width, int height) {
width_ = width;
height_ = height;
memset(&img_, 0, sizeof(img_));
ASSERT_EQ(0, vp8_yv12_alloc_frame_buffer(&img_, width_, height_,
VP8BORDERINPIXELS));
memset(img_.buffer_alloc, kBufFiller, img_.frame_size);
FillPlane(img_.y_buffer, img_.y_crop_width, img_.y_crop_height,
img_.y_stride);
FillPlane(img_.u_buffer, img_.uv_crop_width, img_.uv_crop_height,
img_.uv_stride);
FillPlane(img_.v_buffer, img_.uv_crop_width, img_.uv_crop_height,
img_.uv_stride);
memset(&ref_img_, 0, sizeof(ref_img_));
ASSERT_EQ(0, vp8_yv12_alloc_frame_buffer(&ref_img_, width_, height_,
VP8BORDERINPIXELS));
memset(ref_img_.buffer_alloc, kBufFiller, ref_img_.frame_size);
memset(&cpy_img_, 0, sizeof(cpy_img_));
ASSERT_EQ(0, vp8_yv12_alloc_frame_buffer(&cpy_img_, width_, height_,
VP8BORDERINPIXELS));
memset(cpy_img_.buffer_alloc, kBufFiller, cpy_img_.frame_size);
ReferenceCopyFrame();
}
void DeallocImage() {
vp8_yv12_de_alloc_frame_buffer(&img_);
vp8_yv12_de_alloc_frame_buffer(&ref_img_);
vp8_yv12_de_alloc_frame_buffer(&cpy_img_);
}
protected:
static const int kBufFiller = 123;
static const int kBufMax = kBufFiller - 1;
static void FillPlane(uint8_t *buf, int width, int height, int stride) {
for (int y = 0; y < height; ++y) {
for (int x = 0; x < width; ++x) {
buf[x + (y * stride)] = (x + (width * y)) % kBufMax;
}
}
}
static void ExtendPlane(uint8_t *buf, int crop_width, int crop_height,
int width, int height, int stride, int padding) {
// Copy the outermost visible pixel to a distance of at least 'padding.'
// The buffers are allocated such that there may be excess space outside the
// padding. As long as the minimum amount of padding is achieved it is not
// necessary to fill this space as well.
uint8_t *left = buf - padding;
uint8_t *right = buf + crop_width;
const int right_extend = padding + (width - crop_width);
const int bottom_extend = padding + (height - crop_height);
// Fill the border pixels from the nearest image pixel.
for (int y = 0; y < crop_height; ++y) {
memset(left, left[padding], padding);
memset(right, right[-1], right_extend);
left += stride;
right += stride;
}
left = buf - padding;
uint8_t *top = left - (stride * padding);
// The buffer does not always extend as far as the stride.
// Equivalent to padding + width + padding.
const int extend_width = padding + crop_width + right_extend;
// The first row was already extended to the left and right. Copy it up.
for (int y = 0; y < padding; ++y) {
memcpy(top, left, extend_width);
top += stride;
}
uint8_t *bottom = left + (crop_height * stride);
for (int y = 0; y < bottom_extend; ++y) {
memcpy(bottom, left + (crop_height - 1) * stride, extend_width);
bottom += stride;
}
}
void ReferenceExtendBorder() {
ExtendPlane(ref_img_.y_buffer, ref_img_.y_crop_width,
ref_img_.y_crop_height, ref_img_.y_width, ref_img_.y_height,
ref_img_.y_stride, ref_img_.border);
ExtendPlane(ref_img_.u_buffer, ref_img_.uv_crop_width,
ref_img_.uv_crop_height, ref_img_.uv_width, ref_img_.uv_height,
ref_img_.uv_stride, ref_img_.border / 2);
ExtendPlane(ref_img_.v_buffer, ref_img_.uv_crop_width,
ref_img_.uv_crop_height, ref_img_.uv_width, ref_img_.uv_height,
ref_img_.uv_stride, ref_img_.border / 2);
}
void ReferenceCopyFrame() {
// Copy img_ to ref_img_ and extend frame borders. This will be used for
// verifying extend_fn_ as well as copy_frame_fn_.
EXPECT_EQ(ref_img_.frame_size, img_.frame_size);
for (int y = 0; y < img_.y_crop_height; ++y) {
for (int x = 0; x < img_.y_crop_width; ++x) {
ref_img_.y_buffer[x + y * ref_img_.y_stride] =
img_.y_buffer[x + y * img_.y_stride];
}
}
for (int y = 0; y < img_.uv_crop_height; ++y) {
for (int x = 0; x < img_.uv_crop_width; ++x) {
ref_img_.u_buffer[x + y * ref_img_.uv_stride] =
img_.u_buffer[x + y * img_.uv_stride];
ref_img_.v_buffer[x + y * ref_img_.uv_stride] =
img_.v_buffer[x + y * img_.uv_stride];
}
}
ReferenceExtendBorder();
}
void CompareImages(const YV12_BUFFER_CONFIG actual) {
EXPECT_EQ(ref_img_.frame_size, actual.frame_size);
EXPECT_EQ(0, memcmp(ref_img_.buffer_alloc, actual.buffer_alloc,
ref_img_.frame_size));
}
YV12_BUFFER_CONFIG img_;
YV12_BUFFER_CONFIG ref_img_;
YV12_BUFFER_CONFIG cpy_img_;
int width_;
int height_;
};
class ExtendBorderTest
: public VpxScaleBase,
public ::testing::TestWithParam<ExtendFrameBorderFunc> {
@ -178,11 +46,11 @@ class ExtendBorderTest
static const int kSizesToTest[] = { 1, 15, 33, 145, 512, 1025, 16383 };
for (int h = 0; h < kNumSizesToTest; ++h) {
for (int w = 0; w < kNumSizesToTest; ++w) {
ResetImage(kSizesToTest[w], kSizesToTest[h]);
ASSERT_NO_FATAL_FAILURE(ResetImages(kSizesToTest[w], kSizesToTest[h]));
ReferenceCopyFrame();
ExtendBorder();
ReferenceExtendBorder();
CompareImages(img_);
DeallocImage();
DeallocImages();
}
}
}
@ -204,7 +72,7 @@ class CopyFrameTest : public VpxScaleBase,
virtual void SetUp() { copy_frame_fn_ = GetParam(); }
void CopyFrame() {
ASM_REGISTER_STATE_CHECK(copy_frame_fn_(&img_, &cpy_img_));
ASM_REGISTER_STATE_CHECK(copy_frame_fn_(&img_, &dst_img_));
}
void RunTest() {
@ -217,11 +85,11 @@ class CopyFrameTest : public VpxScaleBase,
static const int kSizesToTest[] = { 1, 15, 33, 145, 512, 1025, 16383 };
for (int h = 0; h < kNumSizesToTest; ++h) {
for (int w = 0; w < kNumSizesToTest; ++w) {
ResetImage(kSizesToTest[w], kSizesToTest[h]);
ASSERT_NO_FATAL_FAILURE(ResetImages(kSizesToTest[w], kSizesToTest[h]));
ReferenceCopyFrame();
CopyFrame();
CompareImages(cpy_img_);
DeallocImage();
CompareImages(dst_img_);
DeallocImages();
}
}
}
@ -233,4 +101,5 @@ TEST_P(CopyFrameTest, CopyFrame) { ASSERT_NO_FATAL_FAILURE(RunTest()); }
INSTANTIATE_TEST_CASE_P(C, CopyFrameTest,
::testing::Values(vp8_yv12_copy_frame_c));
} // namespace
} // namespace libvpx_test

View File

@ -0,0 +1,200 @@
/*
* Copyright (c) 2014 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#ifndef TEST_VPX_SCALE_TEST_H_
#define TEST_VPX_SCALE_TEST_H_
#include "third_party/googletest/src/include/gtest/gtest.h"
#include "./vpx_config.h"
#include "./vpx_scale_rtcd.h"
#include "test/acm_random.h"
#include "test/clear_system_state.h"
#include "test/register_state_check.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_scale/yv12config.h"
using libvpx_test::ACMRandom;
namespace libvpx_test {
class VpxScaleBase {
public:
virtual ~VpxScaleBase() { libvpx_test::ClearSystemState(); }
void ResetImage(YV12_BUFFER_CONFIG *const img, const int width,
const int height) {
memset(img, 0, sizeof(*img));
ASSERT_EQ(
0, vp8_yv12_alloc_frame_buffer(img, width, height, VP8BORDERINPIXELS));
memset(img->buffer_alloc, kBufFiller, img->frame_size);
}
void ResetImages(const int width, const int height) {
ResetImage(&img_, width, height);
ResetImage(&ref_img_, width, height);
ResetImage(&dst_img_, width, height);
FillPlane(img_.y_buffer, img_.y_crop_width, img_.y_crop_height,
img_.y_stride);
FillPlane(img_.u_buffer, img_.uv_crop_width, img_.uv_crop_height,
img_.uv_stride);
FillPlane(img_.v_buffer, img_.uv_crop_width, img_.uv_crop_height,
img_.uv_stride);
}
void ResetScaleImage(YV12_BUFFER_CONFIG *const img, const int width,
const int height) {
memset(img, 0, sizeof(*img));
#if CONFIG_VP9_HIGHBITDEPTH
ASSERT_EQ(0, vpx_alloc_frame_buffer(img, width, height, 1, 1, 0,
VP9_ENC_BORDER_IN_PIXELS, 0));
#else
ASSERT_EQ(0, vpx_alloc_frame_buffer(img, width, height, 1, 1,
VP9_ENC_BORDER_IN_PIXELS, 0));
#endif
memset(img->buffer_alloc, kBufFiller, img->frame_size);
}
void ResetScaleImages(const int src_width, const int src_height,
const int dst_width, const int dst_height) {
ResetScaleImage(&img_, src_width, src_height);
ResetScaleImage(&ref_img_, dst_width, dst_height);
ResetScaleImage(&dst_img_, dst_width, dst_height);
FillPlaneExtreme(img_.y_buffer, img_.y_crop_width, img_.y_crop_height,
img_.y_stride);
FillPlaneExtreme(img_.u_buffer, img_.uv_crop_width, img_.uv_crop_height,
img_.uv_stride);
FillPlaneExtreme(img_.v_buffer, img_.uv_crop_width, img_.uv_crop_height,
img_.uv_stride);
}
void DeallocImages() {
vp8_yv12_de_alloc_frame_buffer(&img_);
vp8_yv12_de_alloc_frame_buffer(&ref_img_);
vp8_yv12_de_alloc_frame_buffer(&dst_img_);
}
void DeallocScaleImages() {
vpx_free_frame_buffer(&img_);
vpx_free_frame_buffer(&ref_img_);
vpx_free_frame_buffer(&dst_img_);
}
protected:
static const int kBufFiller = 123;
static const int kBufMax = kBufFiller - 1;
static void FillPlane(uint8_t *const buf, const int width, const int height,
const int stride) {
for (int y = 0; y < height; ++y) {
for (int x = 0; x < width; ++x) {
buf[x + (y * stride)] = (x + (width * y)) % kBufMax;
}
}
}
static void FillPlaneExtreme(uint8_t *const buf, const int width,
const int height, const int stride) {
ACMRandom rnd;
for (int y = 0; y < height; ++y) {
for (int x = 0; x < width; ++x) {
buf[x + (y * stride)] = rnd.Rand8() % 2 ? 255 : 0;
}
}
}
static void ExtendPlane(uint8_t *buf, int crop_width, int crop_height,
int width, int height, int stride, int padding) {
// Copy the outermost visible pixel to a distance of at least 'padding.'
// The buffers are allocated such that there may be excess space outside the
// padding. As long as the minimum amount of padding is achieved it is not
// necessary to fill this space as well.
uint8_t *left = buf - padding;
uint8_t *right = buf + crop_width;
const int right_extend = padding + (width - crop_width);
const int bottom_extend = padding + (height - crop_height);
// Fill the border pixels from the nearest image pixel.
for (int y = 0; y < crop_height; ++y) {
memset(left, left[padding], padding);
memset(right, right[-1], right_extend);
left += stride;
right += stride;
}
left = buf - padding;
uint8_t *top = left - (stride * padding);
// The buffer does not always extend as far as the stride.
// Equivalent to padding + width + padding.
const int extend_width = padding + crop_width + right_extend;
// The first row was already extended to the left and right. Copy it up.
for (int y = 0; y < padding; ++y) {
memcpy(top, left, extend_width);
top += stride;
}
uint8_t *bottom = left + (crop_height * stride);
for (int y = 0; y < bottom_extend; ++y) {
memcpy(bottom, left + (crop_height - 1) * stride, extend_width);
bottom += stride;
}
}
void ReferenceExtendBorder() {
ExtendPlane(ref_img_.y_buffer, ref_img_.y_crop_width,
ref_img_.y_crop_height, ref_img_.y_width, ref_img_.y_height,
ref_img_.y_stride, ref_img_.border);
ExtendPlane(ref_img_.u_buffer, ref_img_.uv_crop_width,
ref_img_.uv_crop_height, ref_img_.uv_width, ref_img_.uv_height,
ref_img_.uv_stride, ref_img_.border / 2);
ExtendPlane(ref_img_.v_buffer, ref_img_.uv_crop_width,
ref_img_.uv_crop_height, ref_img_.uv_width, ref_img_.uv_height,
ref_img_.uv_stride, ref_img_.border / 2);
}
void ReferenceCopyFrame() {
// Copy img_ to ref_img_ and extend frame borders. This will be used for
// verifying extend_fn_ as well as copy_frame_fn_.
EXPECT_EQ(ref_img_.frame_size, img_.frame_size);
for (int y = 0; y < img_.y_crop_height; ++y) {
for (int x = 0; x < img_.y_crop_width; ++x) {
ref_img_.y_buffer[x + y * ref_img_.y_stride] =
img_.y_buffer[x + y * img_.y_stride];
}
}
for (int y = 0; y < img_.uv_crop_height; ++y) {
for (int x = 0; x < img_.uv_crop_width; ++x) {
ref_img_.u_buffer[x + y * ref_img_.uv_stride] =
img_.u_buffer[x + y * img_.uv_stride];
ref_img_.v_buffer[x + y * ref_img_.uv_stride] =
img_.v_buffer[x + y * img_.uv_stride];
}
}
ReferenceExtendBorder();
}
void CompareImages(const YV12_BUFFER_CONFIG actual) {
EXPECT_EQ(ref_img_.frame_size, actual.frame_size);
EXPECT_EQ(0, memcmp(ref_img_.buffer_alloc, actual.buffer_alloc,
ref_img_.frame_size));
}
YV12_BUFFER_CONFIG img_;
YV12_BUFFER_CONFIG ref_img_;
YV12_BUFFER_CONFIG dst_img_;
};
} // namespace libvpx_test
#endif // TEST_VPX_SCALE_TEST_H_

View File

@ -40,6 +40,8 @@ vpx_tsvc_encoder() {
local timebase_den="1000"
local speed="6"
local frame_drop_thresh="30"
local max_threads="4"
local error_resilient="1"
shift 2
@ -48,11 +50,22 @@ vpx_tsvc_encoder() {
return 1
fi
eval "${VPX_TEST_PREFIX}" "${encoder}" "${YUV_RAW_INPUT}" "${output_file}" \
"${codec}" "${YUV_RAW_INPUT_WIDTH}" "${YUV_RAW_INPUT_HEIGHT}" \
"${timebase_num}" "${timebase_den}" "${speed}" "${frame_drop_thresh}" \
"$@" \
${devnull}
# TODO(tomfinegan): Verify file output for all thread runs.
for threads in $(seq $max_threads); do
if [ "$(vpx_config_option_enabled CONFIG_VP9_HIGHBITDEPTH)" != "yes" ]; then
eval "${VPX_TEST_PREFIX}" "${encoder}" "${YUV_RAW_INPUT}" \
"${output_file}" "${codec}" "${YUV_RAW_INPUT_WIDTH}" \
"${YUV_RAW_INPUT_HEIGHT}" "${timebase_num}" "${timebase_den}" \
"${speed}" "${frame_drop_thresh}" "${error_resilient}" "${threads}" \
"$@" ${devnull}
else
eval "${VPX_TEST_PREFIX}" "${encoder}" "${YUV_RAW_INPUT}" \
"${output_file}" "${codec}" "${YUV_RAW_INPUT_WIDTH}" \
"${YUV_RAW_INPUT_HEIGHT}" "${timebase_num}" "${timebase_den}" \
"${speed}" "${frame_drop_thresh}" "${error_resilient}" "${threads}" \
"$@" "8" ${devnull}
fi
done
}
# Confirms that all expected output files exist given the output file name
@ -72,193 +85,217 @@ files_exist() {
vpx_tsvc_encoder_vp8_mode_0() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 0 200 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_0"
vpx_tsvc_encoder vp8 "${output_basename}" 0 200 || return 1
# Mode 0 produces 1 stream
files_exist "${FUNCNAME}" 1 || return 1
files_exist "${output_basename}" 1 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_1() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 1 200 400 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_1"
vpx_tsvc_encoder vp8 "${output_basename}" 1 200 400 || return 1
# Mode 1 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
files_exist "${output_basename}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_2() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 2 200 400 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_2"
vpx_tsvc_encoder vp8 "${output_basename}" 2 200 400 || return 1
# Mode 2 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
files_exist "${output_basename}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_3() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 3 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_3"
vpx_tsvc_encoder vp8 "${output_basename}" 3 200 400 600 || return 1
# Mode 3 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_4() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 4 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_4"
vpx_tsvc_encoder vp8 "${output_basename}" 4 200 400 600 || return 1
# Mode 4 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_5() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 5 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_5"
vpx_tsvc_encoder vp8 "${output_basename}" 5 200 400 600 || return 1
# Mode 5 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_6() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 6 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_6"
vpx_tsvc_encoder vp8 "${output_basename}" 6 200 400 600 || return 1
# Mode 6 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_7() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 7 200 400 600 800 1000 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_7"
vpx_tsvc_encoder vp8 "${output_basename}" 7 200 400 600 800 1000 || return 1
# Mode 7 produces 5 streams
files_exist "${FUNCNAME}" 5 || return 1
files_exist "${output_basename}" 5 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_8() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 8 200 400 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_8"
vpx_tsvc_encoder vp8 "${output_basename}" 8 200 400 || return 1
# Mode 8 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
files_exist "${output_basename}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_9() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 9 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_9"
vpx_tsvc_encoder vp8 "${output_basename}" 9 200 400 600 || return 1
# Mode 9 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_10() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 10 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_10"
vpx_tsvc_encoder vp8 "${output_basename}" 10 200 400 600 || return 1
# Mode 10 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp8_mode_11() {
if [ "$(vp8_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp8 "${FUNCNAME}" 11 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp8_mode_11"
vpx_tsvc_encoder vp8 "${output_basename}" 11 200 400 600 || return 1
# Mode 11 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_0() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 0 200 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_0"
vpx_tsvc_encoder vp9 "${output_basename}" 0 200 || return 1
# Mode 0 produces 1 stream
files_exist "${FUNCNAME}" 1 || return 1
files_exist "${output_basename}" 1 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_1() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 1 200 400 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_1"
vpx_tsvc_encoder vp9 "${output_basename}" 1 200 400 || return 1
# Mode 1 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
files_exist "${output_basename}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_2() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 2 200 400 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_2"
vpx_tsvc_encoder vp9 "${output_basename}" 2 200 400 || return 1
# Mode 2 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
files_exist "${output_basename}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_3() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 3 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_3"
vpx_tsvc_encoder vp9 "${output_basename}" 3 200 400 600 || return 1
# Mode 3 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_4() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 4 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_4"
vpx_tsvc_encoder vp9 "${output_basename}" 4 200 400 600 || return 1
# Mode 4 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_5() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 5 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_5"
vpx_tsvc_encoder vp9 "${output_basename}" 5 200 400 600 || return 1
# Mode 5 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_6() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 6 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_6"
vpx_tsvc_encoder vp9 "${output_basename}" 6 200 400 600 || return 1
# Mode 6 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_7() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 7 200 400 600 800 1000 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_7"
vpx_tsvc_encoder vp9 "${output_basename}" 7 200 400 600 800 1000 || return 1
# Mode 7 produces 5 streams
files_exist "${FUNCNAME}" 5 || return 1
files_exist "${output_basename}" 5 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_8() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 8 200 400 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_8"
vpx_tsvc_encoder vp9 "${output_basename}" 8 200 400 || return 1
# Mode 8 produces 2 streams
files_exist "${FUNCNAME}" 2 || return 1
files_exist "${output_basename}" 2 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_9() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 9 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_9"
vpx_tsvc_encoder vp9 "${output_basename}" 9 200 400 600 || return 1
# Mode 9 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_10() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 10 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_10"
vpx_tsvc_encoder vp9 "${output_basename}" 10 200 400 600 || return 1
# Mode 10 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}
vpx_tsvc_encoder_vp9_mode_11() {
if [ "$(vp9_encode_available)" = "yes" ]; then
vpx_tsvc_encoder vp9 "${FUNCNAME}" 11 200 400 600 || return 1
local readonly output_basename="vpx_tsvc_encoder_vp9_mode_11"
vpx_tsvc_encoder vp9 "${output_basename}" 11 200 400 600 || return 1
# Mode 11 produces 3 streams
files_exist "${FUNCNAME}" 3 || return 1
files_exist "${output_basename}" 3 || return 1
fi
}

View File

@ -90,6 +90,15 @@ vpxenc_rt_params() {
--undershoot-pct=50"
}
# Forces --passes to 1 with CONFIG_REALTIME_ONLY.
vpxenc_passes_param() {
if [ "$(vpx_config_option_enabled CONFIG_REALTIME_ONLY)" = "yes" ]; then
echo "--passes=1"
else
echo "--passes=2"
fi
}
# Wrapper function for running vpxenc with pipe input. Requires that
# LIBVPX_BIN_PATH points to the directory containing vpxenc. $1 is used as the
# input file path and shifted away. All remaining parameters are passed through
@ -218,9 +227,11 @@ vpxenc_vp8_ivf_piped_input() {
vpxenc_vp9_ivf() {
if [ "$(vpxenc_can_encode_vp9)" = "yes" ]; then
local readonly output="${VPX_TEST_OUTPUT_DIR}/vp9.ivf"
local readonly passes=$(vpxenc_passes_param)
vpxenc $(yuv_input_hantro_collage) \
--codec=vp9 \
--limit="${TEST_FRAMES}" \
"${passes}" \
--ivf \
--output="${output}"
@ -235,9 +246,11 @@ vpxenc_vp9_webm() {
if [ "$(vpxenc_can_encode_vp9)" = "yes" ] && \
[ "$(webm_io_available)" = "yes" ]; then
local readonly output="${VPX_TEST_OUTPUT_DIR}/vp9.webm"
local readonly passes=$(vpxenc_passes_param)
vpxenc $(yuv_input_hantro_collage) \
--codec=vp9 \
--limit="${TEST_FRAMES}" \
"${passes}" \
--output="${output}"
if [ ! -e "${output}" ]; then
@ -339,11 +352,13 @@ vpxenc_vp9_webm_2pass() {
vpxenc_vp9_ivf_lossless() {
if [ "$(vpxenc_can_encode_vp9)" = "yes" ]; then
local readonly output="${VPX_TEST_OUTPUT_DIR}/vp9_lossless.ivf"
local readonly passes=$(vpxenc_passes_param)
vpxenc $(yuv_input_hantro_collage) \
--codec=vp9 \
--limit="${TEST_FRAMES}" \
--ivf \
--output="${output}" \
"${passes}" \
--lossless=1
if [ ! -e "${output}" ]; then
@ -356,11 +371,13 @@ vpxenc_vp9_ivf_lossless() {
vpxenc_vp9_ivf_minq0_maxq0() {
if [ "$(vpxenc_can_encode_vp9)" = "yes" ]; then
local readonly output="${VPX_TEST_OUTPUT_DIR}/vp9_lossless_minq0_maxq0.ivf"
local readonly passes=$(vpxenc_passes_param)
vpxenc $(yuv_input_hantro_collage) \
--codec=vp9 \
--limit="${TEST_FRAMES}" \
--ivf \
--output="${output}" \
"${passes}" \
--min-q=0 \
--max-q=0
@ -377,12 +394,13 @@ vpxenc_vp9_webm_lag10_frames20() {
local readonly lag_total_frames=20
local readonly lag_frames=10
local readonly output="${VPX_TEST_OUTPUT_DIR}/vp9_lag10_frames20.webm"
local readonly passes=$(vpxenc_passes_param)
vpxenc $(yuv_input_hantro_collage) \
--codec=vp9 \
--limit="${lag_total_frames}" \
--lag-in-frames="${lag_frames}" \
--output="${output}" \
--passes=2 \
"${passes}" \
--auto-alt-ref=1
if [ ! -e "${output}" ]; then
@ -397,9 +415,11 @@ vpxenc_vp9_webm_non_square_par() {
if [ "$(vpxenc_can_encode_vp9)" = "yes" ] && \
[ "$(webm_io_available)" = "yes" ]; then
local readonly output="${VPX_TEST_OUTPUT_DIR}/vp9_non_square_par.webm"
local readonly passes=$(vpxenc_passes_param)
vpxenc $(y4m_input_non_square_par) \
--codec=vp9 \
--limit="${TEST_FRAMES}" \
"${passes}" \
--output="${output}"
if [ ! -e "${output}" ]; then
@ -412,18 +432,21 @@ vpxenc_vp9_webm_non_square_par() {
vpxenc_tests="vpxenc_vp8_ivf
vpxenc_vp8_webm
vpxenc_vp8_webm_rt
vpxenc_vp8_webm_2pass
vpxenc_vp8_webm_lag10_frames20
vpxenc_vp8_ivf_piped_input
vpxenc_vp9_ivf
vpxenc_vp9_webm
vpxenc_vp9_webm_rt
vpxenc_vp9_webm_rt_multithread_tiled
vpxenc_vp9_webm_rt_multithread_tiled_frameparallel
vpxenc_vp9_webm_2pass
vpxenc_vp9_ivf_lossless
vpxenc_vp9_ivf_minq0_maxq0
vpxenc_vp9_webm_lag10_frames20
vpxenc_vp9_webm_non_square_par"
if [ "$(vpx_config_option_enabled CONFIG_REALTIME_ONLY)" != "yes" ]; then
vpxenc_tests="$vpxenc_tests
vpxenc_vp8_webm_2pass
vpxenc_vp8_webm_lag10_frames20
vpxenc_vp9_webm_2pass"
fi
run_tests vpxenc_verify_environment "${vpxenc_tests}"

View File

@ -40,8 +40,8 @@ class WebMVideoSource : public CompressedVideoSource {
virtual void Begin() {
vpx_ctx_->file = OpenTestDataFile(file_name_);
ASSERT_TRUE(vpx_ctx_->file != NULL) << "Input file open failed. Filename: "
<< file_name_;
ASSERT_TRUE(vpx_ctx_->file != NULL)
<< "Input file open failed. Filename: " << file_name_;
ASSERT_EQ(file_is_webm(webm_ctx_, vpx_ctx_), 1) << "file is not WebM";

View File

@ -34,8 +34,8 @@ class Y4mVideoSource : public VideoSource {
virtual void OpenSource() {
CloseSource();
input_file_ = OpenTestDataFile(file_name_);
ASSERT_TRUE(input_file_ != NULL) << "Input file open failed. Filename: "
<< file_name_;
ASSERT_TRUE(input_file_ != NULL)
<< "Input file open failed. Filename: " << file_name_;
}
virtual void ReadSourceToStart() {

View File

@ -43,8 +43,8 @@ class YUVVideoSource : public VideoSource {
virtual void Begin() {
if (input_file_) fclose(input_file_);
input_file_ = OpenTestDataFile(file_name_);
ASSERT_TRUE(input_file_ != NULL) << "Input file open failed. Filename: "
<< file_name_;
ASSERT_TRUE(input_file_ != NULL)
<< "Input file open failed. Filename: " << file_name_;
if (start_) {
fseek(input_file_, static_cast<unsigned>(raw_size_) * start_, SEEK_SET);
}

View File

@ -1,7 +1,7 @@
URL: http://code.google.com/p/googletest/
Version: 1.7.0
URL: https://github.com/google/googletest
Version: 1.8.0
License: BSD
License File: COPYING
License File: LICENSE
Description:
Google's framework for writing C++ tests on a variety of platforms
@ -12,10 +12,13 @@ failures, various options for running the tests, and XML test report
generation.
Local Modifications:
- Removed unused declarations of kPathSeparatorString to have warning
free build.
- Added GTEST_ATTRIBUTE_UNUSED_ to test registering dummies in TEST_P
and INSTANTIATE_TEST_CASE_P to remove warnings about unused variables
under GCC 5.
- Only define g_in_fast_death_test_child for non-Windows builds; quiets an
unused variable warning.
- Remove everything but:
googletest-release-1.8.0/googletest/
CHANGES
CONTRIBUTORS
include
LICENSE
README.md
src
- Suppress unsigned overflow instrumentation in the LCG
https://github.com/google/googletest/pull/1066

View File

@ -1,435 +0,0 @@
Google C++ Testing Framework
============================
http://code.google.com/p/googletest/
Overview
--------
Google's framework for writing C++ tests on a variety of platforms
(Linux, Mac OS X, Windows, Windows CE, Symbian, etc). Based on the
xUnit architecture. Supports automatic test discovery, a rich set of
assertions, user-defined assertions, death tests, fatal and non-fatal
failures, various options for running the tests, and XML test report
generation.
Please see the project page above for more information as well as the
mailing list for questions, discussions, and development. There is
also an IRC channel on OFTC (irc.oftc.net) #gtest available. Please
join us!
Requirements for End Users
--------------------------
Google Test is designed to have fairly minimal requirements to build
and use with your projects, but there are some. Currently, we support
Linux, Windows, Mac OS X, and Cygwin. We will also make our best
effort to support other platforms (e.g. Solaris, AIX, and z/OS).
However, since core members of the Google Test project have no access
to these platforms, Google Test may have outstanding issues there. If
you notice any problems on your platform, please notify
googletestframework@googlegroups.com. Patches for fixing them are
even more welcome!
### Linux Requirements ###
These are the base requirements to build and use Google Test from a source
package (as described below):
* GNU-compatible Make or gmake
* POSIX-standard shell
* POSIX(-2) Regular Expressions (regex.h)
* A C++98-standard-compliant compiler
### Windows Requirements ###
* Microsoft Visual C++ 7.1 or newer
### Cygwin Requirements ###
* Cygwin 1.5.25-14 or newer
### Mac OS X Requirements ###
* Mac OS X 10.4 Tiger or newer
* Developer Tools Installed
Also, you'll need CMake 2.6.4 or higher if you want to build the
samples using the provided CMake script, regardless of the platform.
Requirements for Contributors
-----------------------------
We welcome patches. If you plan to contribute a patch, you need to
build Google Test and its own tests from an SVN checkout (described
below), which has further requirements:
* Python version 2.3 or newer (for running some of the tests and
re-generating certain source files from templates)
* CMake 2.6.4 or newer
Getting the Source
------------------
There are two primary ways of getting Google Test's source code: you
can download a stable source release in your preferred archive format,
or directly check out the source from our Subversion (SVN) repositary.
The SVN checkout requires a few extra steps and some extra software
packages on your system, but lets you track the latest development and
make patches much more easily, so we highly encourage it.
### Source Package ###
Google Test is released in versioned source packages which can be
downloaded from the download page [1]. Several different archive
formats are provided, but the only difference is the tools used to
manipulate them, and the size of the resulting file. Download
whichever you are most comfortable with.
[1] http://code.google.com/p/googletest/downloads/list
Once the package is downloaded, expand it using whichever tools you
prefer for that type. This will result in a new directory with the
name "gtest-X.Y.Z" which contains all of the source code. Here are
some examples on Linux:
tar -xvzf gtest-X.Y.Z.tar.gz
tar -xvjf gtest-X.Y.Z.tar.bz2
unzip gtest-X.Y.Z.zip
### SVN Checkout ###
To check out the main branch (also known as the "trunk") of Google
Test, run the following Subversion command:
svn checkout http://googletest.googlecode.com/svn/trunk/ gtest-svn
Setting up the Build
--------------------
To build Google Test and your tests that use it, you need to tell your
build system where to find its headers and source files. The exact
way to do it depends on which build system you use, and is usually
straightforward.
### Generic Build Instructions ###
Suppose you put Google Test in directory ${GTEST_DIR}. To build it,
create a library build target (or a project as called by Visual Studio
and Xcode) to compile
${GTEST_DIR}/src/gtest-all.cc
with ${GTEST_DIR}/include in the system header search path and ${GTEST_DIR}
in the normal header search path. Assuming a Linux-like system and gcc,
something like the following will do:
g++ -isystem ${GTEST_DIR}/include -I${GTEST_DIR} \
-pthread -c ${GTEST_DIR}/src/gtest-all.cc
ar -rv libgtest.a gtest-all.o
(We need -pthread as Google Test uses threads.)
Next, you should compile your test source file with
${GTEST_DIR}/include in the system header search path, and link it
with gtest and any other necessary libraries:
g++ -isystem ${GTEST_DIR}/include -pthread path/to/your_test.cc libgtest.a \
-o your_test
As an example, the make/ directory contains a Makefile that you can
use to build Google Test on systems where GNU make is available
(e.g. Linux, Mac OS X, and Cygwin). It doesn't try to build Google
Test's own tests. Instead, it just builds the Google Test library and
a sample test. You can use it as a starting point for your own build
script.
If the default settings are correct for your environment, the
following commands should succeed:
cd ${GTEST_DIR}/make
make
./sample1_unittest
If you see errors, try to tweak the contents of make/Makefile to make
them go away. There are instructions in make/Makefile on how to do
it.
### Using CMake ###
Google Test comes with a CMake build script (CMakeLists.txt) that can
be used on a wide range of platforms ("C" stands for cross-platofrm.).
If you don't have CMake installed already, you can download it for
free from http://www.cmake.org/.
CMake works by generating native makefiles or build projects that can
be used in the compiler environment of your choice. The typical
workflow starts with:
mkdir mybuild # Create a directory to hold the build output.
cd mybuild
cmake ${GTEST_DIR} # Generate native build scripts.
If you want to build Google Test's samples, you should replace the
last command with
cmake -Dgtest_build_samples=ON ${GTEST_DIR}
If you are on a *nix system, you should now see a Makefile in the
current directory. Just type 'make' to build gtest.
If you use Windows and have Vistual Studio installed, a gtest.sln file
and several .vcproj files will be created. You can then build them
using Visual Studio.
On Mac OS X with Xcode installed, a .xcodeproj file will be generated.
### Legacy Build Scripts ###
Before settling on CMake, we have been providing hand-maintained build
projects/scripts for Visual Studio, Xcode, and Autotools. While we
continue to provide them for convenience, they are not actively
maintained any more. We highly recommend that you follow the
instructions in the previous two sections to integrate Google Test
with your existing build system.
If you still need to use the legacy build scripts, here's how:
The msvc\ folder contains two solutions with Visual C++ projects.
Open the gtest.sln or gtest-md.sln file using Visual Studio, and you
are ready to build Google Test the same way you build any Visual
Studio project. Files that have names ending with -md use DLL
versions of Microsoft runtime libraries (the /MD or the /MDd compiler
option). Files without that suffix use static versions of the runtime
libraries (the /MT or the /MTd option). Please note that one must use
the same option to compile both gtest and the test code. If you use
Visual Studio 2005 or above, we recommend the -md version as /MD is
the default for new projects in these versions of Visual Studio.
On Mac OS X, open the gtest.xcodeproj in the xcode/ folder using
Xcode. Build the "gtest" target. The universal binary framework will
end up in your selected build directory (selected in the Xcode
"Preferences..." -> "Building" pane and defaults to xcode/build).
Alternatively, at the command line, enter:
xcodebuild
This will build the "Release" configuration of gtest.framework in your
default build location. See the "xcodebuild" man page for more
information about building different configurations and building in
different locations.
If you wish to use the Google Test Xcode project with Xcode 4.x and
above, you need to either:
* update the SDK configuration options in xcode/Config/General.xconfig.
Comment options SDKROOT, MACOS_DEPLOYMENT_TARGET, and GCC_VERSION. If
you choose this route you lose the ability to target earlier versions
of MacOS X.
* Install an SDK for an earlier version. This doesn't appear to be
supported by Apple, but has been reported to work
(http://stackoverflow.com/questions/5378518).
Tweaking Google Test
--------------------
Google Test can be used in diverse environments. The default
configuration may not work (or may not work well) out of the box in
some environments. However, you can easily tweak Google Test by
defining control macros on the compiler command line. Generally,
these macros are named like GTEST_XYZ and you define them to either 1
or 0 to enable or disable a certain feature.
We list the most frequently used macros below. For a complete list,
see file include/gtest/internal/gtest-port.h.
### Choosing a TR1 Tuple Library ###
Some Google Test features require the C++ Technical Report 1 (TR1)
tuple library, which is not yet available with all compilers. The
good news is that Google Test implements a subset of TR1 tuple that's
enough for its own need, and will automatically use this when the
compiler doesn't provide TR1 tuple.
Usually you don't need to care about which tuple library Google Test
uses. However, if your project already uses TR1 tuple, you need to
tell Google Test to use the same TR1 tuple library the rest of your
project uses, or the two tuple implementations will clash. To do
that, add
-DGTEST_USE_OWN_TR1_TUPLE=0
to the compiler flags while compiling Google Test and your tests. If
you want to force Google Test to use its own tuple library, just add
-DGTEST_USE_OWN_TR1_TUPLE=1
to the compiler flags instead.
If you don't want Google Test to use tuple at all, add
-DGTEST_HAS_TR1_TUPLE=0
and all features using tuple will be disabled.
### Multi-threaded Tests ###
Google Test is thread-safe where the pthread library is available.
After #include "gtest/gtest.h", you can check the GTEST_IS_THREADSAFE
macro to see whether this is the case (yes if the macro is #defined to
1, no if it's undefined.).
If Google Test doesn't correctly detect whether pthread is available
in your environment, you can force it with
-DGTEST_HAS_PTHREAD=1
or
-DGTEST_HAS_PTHREAD=0
When Google Test uses pthread, you may need to add flags to your
compiler and/or linker to select the pthread library, or you'll get
link errors. If you use the CMake script or the deprecated Autotools
script, this is taken care of for you. If you use your own build
script, you'll need to read your compiler and linker's manual to
figure out what flags to add.
### As a Shared Library (DLL) ###
Google Test is compact, so most users can build and link it as a
static library for the simplicity. You can choose to use Google Test
as a shared library (known as a DLL on Windows) if you prefer.
To compile *gtest* as a shared library, add
-DGTEST_CREATE_SHARED_LIBRARY=1
to the compiler flags. You'll also need to tell the linker to produce
a shared library instead - consult your linker's manual for how to do
it.
To compile your *tests* that use the gtest shared library, add
-DGTEST_LINKED_AS_SHARED_LIBRARY=1
to the compiler flags.
Note: while the above steps aren't technically necessary today when
using some compilers (e.g. GCC), they may become necessary in the
future, if we decide to improve the speed of loading the library (see
http://gcc.gnu.org/wiki/Visibility for details). Therefore you are
recommended to always add the above flags when using Google Test as a
shared library. Otherwise a future release of Google Test may break
your build script.
### Avoiding Macro Name Clashes ###
In C++, macros don't obey namespaces. Therefore two libraries that
both define a macro of the same name will clash if you #include both
definitions. In case a Google Test macro clashes with another
library, you can force Google Test to rename its macro to avoid the
conflict.
Specifically, if both Google Test and some other code define macro
FOO, you can add
-DGTEST_DONT_DEFINE_FOO=1
to the compiler flags to tell Google Test to change the macro's name
from FOO to GTEST_FOO. Currently FOO can be FAIL, SUCCEED, or TEST.
For example, with -DGTEST_DONT_DEFINE_TEST=1, you'll need to write
GTEST_TEST(SomeTest, DoesThis) { ... }
instead of
TEST(SomeTest, DoesThis) { ... }
in order to define a test.
Upgrating from an Earlier Version
---------------------------------
We strive to keep Google Test releases backward compatible.
Sometimes, though, we have to make some breaking changes for the
users' long-term benefits. This section describes what you'll need to
do if you are upgrading from an earlier version of Google Test.
### Upgrading from 1.3.0 or Earlier ###
You may need to explicitly enable or disable Google Test's own TR1
tuple library. See the instructions in section "Choosing a TR1 Tuple
Library".
### Upgrading from 1.4.0 or Earlier ###
The Autotools build script (configure + make) is no longer officially
supportted. You are encouraged to migrate to your own build system or
use CMake. If you still need to use Autotools, you can find
instructions in the README file from Google Test 1.4.0.
On platforms where the pthread library is available, Google Test uses
it in order to be thread-safe. See the "Multi-threaded Tests" section
for what this means to your build script.
If you use Microsoft Visual C++ 7.1 with exceptions disabled, Google
Test will no longer compile. This should affect very few people, as a
large portion of STL (including <string>) doesn't compile in this mode
anyway. We decided to stop supporting it in order to greatly simplify
Google Test's implementation.
Developing Google Test
----------------------
This section discusses how to make your own changes to Google Test.
### Testing Google Test Itself ###
To make sure your changes work as intended and don't break existing
functionality, you'll want to compile and run Google Test's own tests.
For that you can use CMake:
mkdir mybuild
cd mybuild
cmake -Dgtest_build_tests=ON ${GTEST_DIR}
Make sure you have Python installed, as some of Google Test's tests
are written in Python. If the cmake command complains about not being
able to find Python ("Could NOT find PythonInterp (missing:
PYTHON_EXECUTABLE)"), try telling it explicitly where your Python
executable can be found:
cmake -DPYTHON_EXECUTABLE=path/to/python -Dgtest_build_tests=ON ${GTEST_DIR}
Next, you can build Google Test and all of its own tests. On *nix,
this is usually done by 'make'. To run the tests, do
make test
All tests should pass.
### Regenerating Source Files ###
Some of Google Test's source files are generated from templates (not
in the C++ sense) using a script. A template file is named FOO.pump,
where FOO is the name of the file it will generate. For example, the
file include/gtest/internal/gtest-type-util.h.pump is used to generate
gtest-type-util.h in the same directory.
Normally you don't need to worry about regenerating the source files,
unless you need to modify them. In that case, you should modify the
corresponding .pump files instead and run the pump.py Python script to
regenerate them. You can find pump.py in the scripts/ directory.
Read the Pump manual [2] for how to use it.
[2] http://code.google.com/p/googletest/wiki/PumpManual
### Contributing a Patch ###
We welcome patches. Please read the Google Test developer's guide [3]
for how you can contribute. In particular, make sure you have signed
the Contributor License Agreement, or we won't be able to accept the
patch.
[3] http://code.google.com/p/googletest/wiki/GoogleTestDevGuide
Happy testing!

View File

@ -0,0 +1,280 @@
### Generic Build Instructions ###
#### Setup ####
To build Google Test and your tests that use it, you need to tell your
build system where to find its headers and source files. The exact
way to do it depends on which build system you use, and is usually
straightforward.
#### Build ####
Suppose you put Google Test in directory `${GTEST_DIR}`. To build it,
create a library build target (or a project as called by Visual Studio
and Xcode) to compile
${GTEST_DIR}/src/gtest-all.cc
with `${GTEST_DIR}/include` in the system header search path and `${GTEST_DIR}`
in the normal header search path. Assuming a Linux-like system and gcc,
something like the following will do:
g++ -isystem ${GTEST_DIR}/include -I${GTEST_DIR} \
-pthread -c ${GTEST_DIR}/src/gtest-all.cc
ar -rv libgtest.a gtest-all.o
(We need `-pthread` as Google Test uses threads.)
Next, you should compile your test source file with
`${GTEST_DIR}/include` in the system header search path, and link it
with gtest and any other necessary libraries:
g++ -isystem ${GTEST_DIR}/include -pthread path/to/your_test.cc libgtest.a \
-o your_test
As an example, the make/ directory contains a Makefile that you can
use to build Google Test on systems where GNU make is available
(e.g. Linux, Mac OS X, and Cygwin). It doesn't try to build Google
Test's own tests. Instead, it just builds the Google Test library and
a sample test. You can use it as a starting point for your own build
script.
If the default settings are correct for your environment, the
following commands should succeed:
cd ${GTEST_DIR}/make
make
./sample1_unittest
If you see errors, try to tweak the contents of `make/Makefile` to make
them go away. There are instructions in `make/Makefile` on how to do
it.
### Using CMake ###
Google Test comes with a CMake build script (
[CMakeLists.txt](CMakeLists.txt)) that can be used on a wide range of platforms ("C" stands for
cross-platform.). If you don't have CMake installed already, you can
download it for free from <http://www.cmake.org/>.
CMake works by generating native makefiles or build projects that can
be used in the compiler environment of your choice. The typical
workflow starts with:
mkdir mybuild # Create a directory to hold the build output.
cd mybuild
cmake ${GTEST_DIR} # Generate native build scripts.
If you want to build Google Test's samples, you should replace the
last command with
cmake -Dgtest_build_samples=ON ${GTEST_DIR}
If you are on a \*nix system, you should now see a Makefile in the
current directory. Just type 'make' to build gtest.
If you use Windows and have Visual Studio installed, a `gtest.sln` file
and several `.vcproj` files will be created. You can then build them
using Visual Studio.
On Mac OS X with Xcode installed, a `.xcodeproj` file will be generated.
### Legacy Build Scripts ###
Before settling on CMake, we have been providing hand-maintained build
projects/scripts for Visual Studio, Xcode, and Autotools. While we
continue to provide them for convenience, they are not actively
maintained any more. We highly recommend that you follow the
instructions in the previous two sections to integrate Google Test
with your existing build system.
If you still need to use the legacy build scripts, here's how:
The msvc\ folder contains two solutions with Visual C++ projects.
Open the `gtest.sln` or `gtest-md.sln` file using Visual Studio, and you
are ready to build Google Test the same way you build any Visual
Studio project. Files that have names ending with -md use DLL
versions of Microsoft runtime libraries (the /MD or the /MDd compiler
option). Files without that suffix use static versions of the runtime
libraries (the /MT or the /MTd option). Please note that one must use
the same option to compile both gtest and the test code. If you use
Visual Studio 2005 or above, we recommend the -md version as /MD is
the default for new projects in these versions of Visual Studio.
On Mac OS X, open the `gtest.xcodeproj` in the `xcode/` folder using
Xcode. Build the "gtest" target. The universal binary framework will
end up in your selected build directory (selected in the Xcode
"Preferences..." -> "Building" pane and defaults to xcode/build).
Alternatively, at the command line, enter:
xcodebuild
This will build the "Release" configuration of gtest.framework in your
default build location. See the "xcodebuild" man page for more
information about building different configurations and building in
different locations.
If you wish to use the Google Test Xcode project with Xcode 4.x and
above, you need to either:
* update the SDK configuration options in xcode/Config/General.xconfig.
Comment options `SDKROOT`, `MACOS_DEPLOYMENT_TARGET`, and `GCC_VERSION`. If
you choose this route you lose the ability to target earlier versions
of MacOS X.
* Install an SDK for an earlier version. This doesn't appear to be
supported by Apple, but has been reported to work
(http://stackoverflow.com/questions/5378518).
### Tweaking Google Test ###
Google Test can be used in diverse environments. The default
configuration may not work (or may not work well) out of the box in
some environments. However, you can easily tweak Google Test by
defining control macros on the compiler command line. Generally,
these macros are named like `GTEST_XYZ` and you define them to either 1
or 0 to enable or disable a certain feature.
We list the most frequently used macros below. For a complete list,
see file [include/gtest/internal/gtest-port.h](include/gtest/internal/gtest-port.h).
### Choosing a TR1 Tuple Library ###
Some Google Test features require the C++ Technical Report 1 (TR1)
tuple library, which is not yet available with all compilers. The
good news is that Google Test implements a subset of TR1 tuple that's
enough for its own need, and will automatically use this when the
compiler doesn't provide TR1 tuple.
Usually you don't need to care about which tuple library Google Test
uses. However, if your project already uses TR1 tuple, you need to
tell Google Test to use the same TR1 tuple library the rest of your
project uses, or the two tuple implementations will clash. To do
that, add
-DGTEST_USE_OWN_TR1_TUPLE=0
to the compiler flags while compiling Google Test and your tests. If
you want to force Google Test to use its own tuple library, just add
-DGTEST_USE_OWN_TR1_TUPLE=1
to the compiler flags instead.
If you don't want Google Test to use tuple at all, add
-DGTEST_HAS_TR1_TUPLE=0
and all features using tuple will be disabled.
### Multi-threaded Tests ###
Google Test is thread-safe where the pthread library is available.
After `#include "gtest/gtest.h"`, you can check the `GTEST_IS_THREADSAFE`
macro to see whether this is the case (yes if the macro is `#defined` to
1, no if it's undefined.).
If Google Test doesn't correctly detect whether pthread is available
in your environment, you can force it with
-DGTEST_HAS_PTHREAD=1
or
-DGTEST_HAS_PTHREAD=0
When Google Test uses pthread, you may need to add flags to your
compiler and/or linker to select the pthread library, or you'll get
link errors. If you use the CMake script or the deprecated Autotools
script, this is taken care of for you. If you use your own build
script, you'll need to read your compiler and linker's manual to
figure out what flags to add.
### As a Shared Library (DLL) ###
Google Test is compact, so most users can build and link it as a
static library for the simplicity. You can choose to use Google Test
as a shared library (known as a DLL on Windows) if you prefer.
To compile *gtest* as a shared library, add
-DGTEST_CREATE_SHARED_LIBRARY=1
to the compiler flags. You'll also need to tell the linker to produce
a shared library instead - consult your linker's manual for how to do
it.
To compile your *tests* that use the gtest shared library, add
-DGTEST_LINKED_AS_SHARED_LIBRARY=1
to the compiler flags.
Note: while the above steps aren't technically necessary today when
using some compilers (e.g. GCC), they may become necessary in the
future, if we decide to improve the speed of loading the library (see
<http://gcc.gnu.org/wiki/Visibility> for details). Therefore you are
recommended to always add the above flags when using Google Test as a
shared library. Otherwise a future release of Google Test may break
your build script.
### Avoiding Macro Name Clashes ###
In C++, macros don't obey namespaces. Therefore two libraries that
both define a macro of the same name will clash if you `#include` both
definitions. In case a Google Test macro clashes with another
library, you can force Google Test to rename its macro to avoid the
conflict.
Specifically, if both Google Test and some other code define macro
FOO, you can add
-DGTEST_DONT_DEFINE_FOO=1
to the compiler flags to tell Google Test to change the macro's name
from `FOO` to `GTEST_FOO`. Currently `FOO` can be `FAIL`, `SUCCEED`,
or `TEST`. For example, with `-DGTEST_DONT_DEFINE_TEST=1`, you'll
need to write
GTEST_TEST(SomeTest, DoesThis) { ... }
instead of
TEST(SomeTest, DoesThis) { ... }
in order to define a test.
## Developing Google Test ##
This section discusses how to make your own changes to Google Test.
### Testing Google Test Itself ###
To make sure your changes work as intended and don't break existing
functionality, you'll want to compile and run Google Test's own tests.
For that you can use CMake:
mkdir mybuild
cd mybuild
cmake -Dgtest_build_tests=ON ${GTEST_DIR}
Make sure you have Python installed, as some of Google Test's tests
are written in Python. If the cmake command complains about not being
able to find Python (`Could NOT find PythonInterp (missing:
PYTHON_EXECUTABLE)`), try telling it explicitly where your Python
executable can be found:
cmake -DPYTHON_EXECUTABLE=path/to/python -Dgtest_build_tests=ON ${GTEST_DIR}
Next, you can build Google Test and all of its own tests. On \*nix,
this is usually done by 'make'. To run the tests, do
make test
All tests should pass.
Normally you don't need to worry about regenerating the source files,
unless you need to modify them. In that case, you should modify the
corresponding .pump files instead and run the pump.py Python script to
regenerate them. You can find pump.py in the [scripts/](scripts/) directory.
Read the [Pump manual](docs/PumpManual.md) for how to use it.

Some files were not shown because too many files have changed in this diff Show More