multi-thread decoding: ~25-30% faster

To be enabled with the flag WEBP_USE_THREAD.
For now it's only available on unix (pthread), when using Makefile.unix
Will be switched on more generally later.

In-loop filtering and output (=rescaling/yuv->rgb conversion)
is done in parallel to bitstream decoding, lagging 1 row behind.

Example:
examples/dwebp bryce.webp -v
Time to decode picture: 0.680s

examples/dwebp bryce.webp -v -mt
Time to decode picture: 0.515s

Change-Id: Ic30a897423137a3bdace9c4e30465ef758fe53f2
This commit is contained in:
Pascal Massimino
2011-07-22 13:09:10 -07:00
parent acd8ba4229
commit fc7815d692
16 changed files with 519 additions and 120 deletions

View File

@ -36,12 +36,16 @@ endif
# Extra flags to enable experimental features and code
# EXTRA_FLAGS += -DWEBP_EXPERIMENTAL_FEATURES
# Extra flags to enable multi-threading
EXTRA_FLAGS += -DWEBP_USE_THREAD
EXTRA_LIBS += -lpthread
# Extra flags to emulate C89 strictness with the full ANSI
EXTRA_FLAGS += -Wextra -Wold-style-definition
EXTRA_FLAGS += -Wmissing-prototypes
EXTRA_FLAGS += -Wmissing-declarations
EXTRA_FLAGS += -Wdeclaration-after-statement
# EXTRA_FLAGS += -Wvla
EXTRA_FLAGS += -Wvla
#### Nothing should normally be changed below this line ####
@ -58,7 +62,7 @@ OBJS = src/enc/webpenc.o src/enc/bit_writer.o src/enc/syntax.o \
src/dec/bits.o src/dec/dsp.o src/dec/dsp_sse2.o src/dec/frame.o \
src/dec/webp.o src/dec/quant.o src/dec/tree.o src/dec/vp8.o \
src/dec/yuv.o src/dec/idec.o src/dec/alpha.o src/dec/layer.o \
src/dec/io.o src/dec/io_sse2.o src/dec/buffer.o
src/dec/io.o src/dec/io_sse2.o src/dec/buffer.o src/dec/thread.o
HDRS = src/webp/encode.h src/enc/vp8enci.h src/enc/bit_writer.h \
src/enc/cost.h src/dec/bits.h src/dec/vp8i.h src/dec/yuv.h
OUTPUT = examples/cwebp examples/dwebp src/libwebp.a