Merge "Improve MV prediction in vp8_pick_inter_mode() for speed>3"
diff --git a/build/make/configure.sh b/build/make/configure.sh
index f8329aa..88698fd 100755
--- a/build/make/configure.sh
+++ b/build/make/configure.sh
@@ -664,7 +664,7 @@
elif enabled armv7
then
check_add_cflags -march=armv7-a -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=softfp #-ftree-vectorize
- check_add_asflags -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=softfp #-march=armv7-a
+ check_add_asflags -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=softfp #-march=armv7-a
else
check_add_cflags -march=${tgt_isa}
check_add_asflags -march=${tgt_isa}
@@ -695,10 +695,17 @@
arch_int=${arch_int%%te}
check_add_asflags --pd "\"ARCHITECTURE SETA ${arch_int}\""
enabled debug && add_asflags -g
+ add_cflags --gnu
+ add_cflags --enum_is_int
+ add_cflags --wchar32
;;
esac
case ${tgt_os} in
+ none*)
+ disable multithread
+ disable os_support
+ ;;
darwin*)
SDK_PATH=/Developer/Platforms/iPhoneOS.platform/Developer
TOOLCHAIN_PATH=${SDK_PATH}/usr/bin
@@ -747,13 +754,9 @@
|| die "Must supply --libc when targetting *-linux-rvct"
# Set up compiler
- add_cflags --gnu
- add_cflags --enum_is_int
add_cflags --library_interface=aeabi_glibc
add_cflags --no_hide_all
- add_cflags --wchar32
add_cflags --dwarf2
- add_cflags --gnu
# Set up linker
add_ldflags --sysv --no_startup --no_ref_cpp_init
diff --git a/configure b/configure
index cb568d7..ed19906 100755
--- a/configure
+++ b/configure
@@ -79,11 +79,13 @@
# alphabetically by architecture, generic-gnu last.
all_platforms="${all_platforms} armv5te-linux-rvct"
all_platforms="${all_platforms} armv5te-linux-gcc"
+all_platforms="${all_platforms} armv5te-none-rvct"
all_platforms="${all_platforms} armv5te-symbian-gcc"
all_platforms="${all_platforms} armv5te-wince-vs8"
all_platforms="${all_platforms} armv6-darwin-gcc"
all_platforms="${all_platforms} armv6-linux-rvct"
all_platforms="${all_platforms} armv6-linux-gcc"
+all_platforms="${all_platforms} armv6-none-rvct"
all_platforms="${all_platforms} armv6-symbian-gcc"
all_platforms="${all_platforms} armv6-wince-vs8"
all_platforms="${all_platforms} iwmmxt-linux-rvct"
@@ -95,6 +97,7 @@
all_platforms="${all_platforms} armv7-darwin-gcc" #neon Cortex-A8
all_platforms="${all_platforms} armv7-linux-rvct" #neon Cortex-A8
all_platforms="${all_platforms} armv7-linux-gcc" #neon Cortex-A8
+all_platforms="${all_platforms} armv7-none-rvct" #neon Cortex-A8
all_platforms="${all_platforms} mips32-linux-gcc"
all_platforms="${all_platforms} ppc32-darwin8-gcc"
all_platforms="${all_platforms} ppc32-darwin9-gcc"
@@ -159,6 +162,7 @@
enable md5
enable spatial_resampling
enable multithread
+enable os_support
[ -d ${source_path}/../include ] && enable alt_tree_layout
for d in vp8; do
@@ -253,6 +257,7 @@
small
arm_asm_detok
postproc_visualizer
+ os_support
"
CMDLINE_SELECT="
extra_warnings
diff --git a/examples.mk b/examples.mk
index a30205d..89c93db 100644
--- a/examples.mk
+++ b/examples.mk
@@ -93,8 +93,16 @@
# Handle extra library flags depending on codec configuration
-CODEC_EXTRA_LIBS-$(CONFIG_VP8) += m
+# We should not link to math library (libm) on RVCT
+# when building for bare-metal targets
+ifeq ($(CONFIG_OS_SUPPORT), yes)
+CODEC_EXTRA_LIBS-$(CONFIG_VP8) += m
+else
+ ifeq ($(CONFIG_GCC), yes)
+ CODEC_EXTRA_LIBS-$(CONFIG_VP8) += m
+ endif
+endif
#
# End of specified files. The rest of the build rules should happen
# automagically from here.
diff --git a/md5_utils.c b/md5_utils.c
index 455d9cd..9a584fa 100644
--- a/md5_utils.c
+++ b/md5_utils.c
@@ -20,8 +20,6 @@
* Still in the public domain.
*/
-#include <sys/types.h> /* for stupid systems */
-
#include <string.h> /* for memcpy() */
#include "md5_utils.h"
diff --git a/vp8/common/arm/armv6/sixtappredict8x4_v6.asm b/vp8/common/arm/armv6/sixtappredict8x4_v6.asm
index 8b99394..029e02a 100644
--- a/vp8/common/arm/armv6/sixtappredict8x4_v6.asm
+++ b/vp8/common/arm/armv6/sixtappredict8x4_v6.asm
@@ -243,8 +243,6 @@
ENDP
;-----------------
- AREA subpelfilters8_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
;One word each is reserved. Label filter_coeff can be used to access the data.
;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
_filter8_coeff_
diff --git a/vp8/common/arm/loopfilter_arm.c b/vp8/common/arm/loopfilter_arm.c
index a81c505..684a7f1 100644
--- a/vp8/common/arm/loopfilter_arm.c
+++ b/vp8/common/arm/loopfilter_arm.c
@@ -41,13 +41,13 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_horizontal_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_horizontal_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_horizontal_edge_armv6(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_horizontal_edge_armv6(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_mbloop_filter_horizontal_edge_armv6(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_horizontal_edge_armv6(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
}
void vp8_loop_filter_mbhs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -57,7 +57,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_horizontal_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
/* Vertical MB Filtering */
@@ -65,13 +65,13 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_vertical_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_vertical_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_vertical_edge_armv6(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_vertical_edge_armv6(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_mbloop_filter_vertical_edge_armv6(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_vertical_edge_armv6(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
}
void vp8_loop_filter_mbvs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -81,7 +81,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_vertical_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_vertical_edge_armv6(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
/* Horizontal B Filtering */
@@ -94,10 +94,10 @@
vp8_loop_filter_horizontal_edge_armv6(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_horizontal_edge_armv6(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_horizontal_edge_armv6(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_loop_filter_horizontal_edge_armv6(v_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_horizontal_edge_armv6(v_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
}
void vp8_loop_filter_bhs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -122,10 +122,10 @@
vp8_loop_filter_vertical_edge_armv6(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_vertical_edge_armv6(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_vertical_edge_armv6(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_loop_filter_vertical_edge_armv6(v_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_vertical_edge_armv6(v_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
}
void vp8_loop_filter_bvs_armv6(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -148,10 +148,10 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_horizontal_edge_y_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_horizontal_edge_y_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_horizontal_edge_uv_neon(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
+ vp8_mbloop_filter_horizontal_edge_uv_neon(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, v_ptr);
}
void vp8_loop_filter_mbhs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -161,7 +161,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_horizontal_edge_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_horizontal_edge_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
/* Vertical MB Filtering */
@@ -169,10 +169,10 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_vertical_edge_y_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_vertical_edge_y_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_vertical_edge_uv_neon(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
+ vp8_mbloop_filter_vertical_edge_uv_neon(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, v_ptr);
}
void vp8_loop_filter_mbvs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -182,7 +182,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_vertical_edge_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_vertical_edge_neon(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
/* Horizontal B Filtering */
@@ -195,7 +195,7 @@
vp8_loop_filter_horizontal_edge_y_neon(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_horizontal_edge_uv_neon(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4 * uv_stride);
+ vp8_loop_filter_horizontal_edge_uv_neon(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, v_ptr + 4 * uv_stride);
}
void vp8_loop_filter_bhs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -220,7 +220,7 @@
vp8_loop_filter_vertical_edge_y_neon(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_vertical_edge_uv_neon(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4);
+ vp8_loop_filter_vertical_edge_uv_neon(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, v_ptr + 4);
}
void vp8_loop_filter_bvs_neon(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
diff --git a/vp8/common/arm/neon/bilinearpredict16x16_neon.asm b/vp8/common/arm/neon/bilinearpredict16x16_neon.asm
index bb72bad..79e1a69 100644
--- a/vp8/common/arm/neon/bilinearpredict16x16_neon.asm
+++ b/vp8/common/arm/neon/bilinearpredict16x16_neon.asm
@@ -350,10 +350,7 @@
ENDP
;-----------------
- AREA bifilters16_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_bifilter16_coeff_
DCD bifilter16_coeff
bifilter16_coeff
diff --git a/vp8/common/arm/neon/bilinearpredict4x4_neon.asm b/vp8/common/arm/neon/bilinearpredict4x4_neon.asm
index 6d4820b..10cd1b8 100644
--- a/vp8/common/arm/neon/bilinearpredict4x4_neon.asm
+++ b/vp8/common/arm/neon/bilinearpredict4x4_neon.asm
@@ -123,10 +123,7 @@
ENDP
;-----------------
- AREA bilinearfilters4_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_bifilter4_coeff_
DCD bifilter4_coeff
bifilter4_coeff
diff --git a/vp8/common/arm/neon/bilinearpredict8x4_neon.asm b/vp8/common/arm/neon/bilinearpredict8x4_neon.asm
index b9f3ce0..bf37bb0 100644
--- a/vp8/common/arm/neon/bilinearpredict8x4_neon.asm
+++ b/vp8/common/arm/neon/bilinearpredict8x4_neon.asm
@@ -128,10 +128,7 @@
ENDP
;-----------------
- AREA bifilters8x4_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_bifilter8x4_coeff_
DCD bifilter8x4_coeff
bifilter8x4_coeff
diff --git a/vp8/common/arm/neon/bilinearpredict8x8_neon.asm b/vp8/common/arm/neon/bilinearpredict8x8_neon.asm
index f7a7d14..9b29df6 100644
--- a/vp8/common/arm/neon/bilinearpredict8x8_neon.asm
+++ b/vp8/common/arm/neon/bilinearpredict8x8_neon.asm
@@ -176,10 +176,7 @@
ENDP
;-----------------
- AREA bifilters8_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_bifilter8_coeff_
DCD bifilter8_coeff
bifilter8_coeff
diff --git a/vp8/common/arm/neon/loopfilter_neon.asm b/vp8/common/arm/neon/loopfilter_neon.asm
index bf0c357..d3a79f6 100644
--- a/vp8/common/arm/neon/loopfilter_neon.asm
+++ b/vp8/common/arm/neon/loopfilter_neon.asm
@@ -397,7 +397,8 @@
bx lr
ENDP ; |vp8_loop_filter_horizontal_edge_y_neon|
- AREA loopfilter_dat, DATA, READONLY
+;-----------------
+
_lf_coeff_
DCD lf_coeff
lf_coeff
diff --git a/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm b/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
index 0b84dc7..5fe7e7e 100644
--- a/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
+++ b/vp8/common/arm/neon/loopfiltersimplehorizontaledge_neon.asm
@@ -104,10 +104,7 @@
ENDP ; |vp8_loop_filter_simple_horizontal_edge_neon|
;-----------------
- AREA hloopfiltery_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 16 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_lfhy_coeff_
DCD lfhy_coeff
lfhy_coeff
diff --git a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
index a793d09..c30378b 100644
--- a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
+++ b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
@@ -145,10 +145,7 @@
ENDP ; |vp8_loop_filter_simple_vertical_edge_neon|
;-----------------
- AREA vloopfiltery_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 16 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_vlfy_coeff_
DCD vlfy_coeff
vlfy_coeff
diff --git a/vp8/common/arm/neon/mbloopfilter_neon.asm b/vp8/common/arm/neon/mbloopfilter_neon.asm
index 255dd56..981adff 100644
--- a/vp8/common/arm/neon/mbloopfilter_neon.asm
+++ b/vp8/common/arm/neon/mbloopfilter_neon.asm
@@ -505,7 +505,8 @@
bx lr
ENDP ; |vp8_mbloop_filter_neon|
- AREA mbloopfilter_dat, DATA, READONLY
+;-----------------
+
_mblf_coeff_
DCD mblf_coeff
mblf_coeff
diff --git a/vp8/common/arm/neon/shortidct4x4llm_neon.asm b/vp8/common/arm/neon/shortidct4x4llm_neon.asm
index d77a287..16c4d2d 100644
--- a/vp8/common/arm/neon/shortidct4x4llm_neon.asm
+++ b/vp8/common/arm/neon/shortidct4x4llm_neon.asm
@@ -113,10 +113,7 @@
ENDP
;-----------------
- AREA idct4x4_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_idct_coeff_
DCD idct_coeff
idct_coeff
diff --git a/vp8/common/arm/neon/sixtappredict16x16_neon.asm b/vp8/common/arm/neon/sixtappredict16x16_neon.asm
index e434a70..3c22fa1 100644
--- a/vp8/common/arm/neon/sixtappredict16x16_neon.asm
+++ b/vp8/common/arm/neon/sixtappredict16x16_neon.asm
@@ -476,10 +476,7 @@
ENDP
;-----------------
- AREA subpelfilters16_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_filter16_coeff_
DCD filter16_coeff
filter16_coeff
diff --git a/vp8/common/arm/neon/sixtappredict4x4_neon.asm b/vp8/common/arm/neon/sixtappredict4x4_neon.asm
index 3d22d77..2dc3f59 100644
--- a/vp8/common/arm/neon/sixtappredict4x4_neon.asm
+++ b/vp8/common/arm/neon/sixtappredict4x4_neon.asm
@@ -407,10 +407,7 @@
ENDP
;-----------------
- AREA subpelfilters4_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_filter4_coeff_
DCD filter4_coeff
filter4_coeff
diff --git a/vp8/common/arm/neon/sixtappredict8x4_neon.asm b/vp8/common/arm/neon/sixtappredict8x4_neon.asm
index 1dd6b1b..0904f52 100644
--- a/vp8/common/arm/neon/sixtappredict8x4_neon.asm
+++ b/vp8/common/arm/neon/sixtappredict8x4_neon.asm
@@ -458,10 +458,7 @@
ENDP
;-----------------
- AREA subpelfilters8_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_filter8_coeff_
DCD filter8_coeff
filter8_coeff
diff --git a/vp8/common/arm/neon/sixtappredict8x8_neon.asm b/vp8/common/arm/neon/sixtappredict8x8_neon.asm
index 37255c7..33af86f 100644
--- a/vp8/common/arm/neon/sixtappredict8x8_neon.asm
+++ b/vp8/common/arm/neon/sixtappredict8x8_neon.asm
@@ -509,10 +509,7 @@
ENDP
;-----------------
- AREA subpelfilters8_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_filter8_coeff_
DCD filter8_coeff
filter8_coeff
diff --git a/vp8/common/loopfilter.c b/vp8/common/loopfilter.c
index f9d0823..37c5b77 100644
--- a/vp8/common/loopfilter.c
+++ b/vp8/common/loopfilter.c
@@ -28,13 +28,13 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_horizontal_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_horizontal_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_horizontal_edge_c(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_horizontal_edge_c(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
}
void vp8_loop_filter_mbhs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -44,7 +44,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_horizontal_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_horizontal_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
/* Vertical MB Filtering */
@@ -52,13 +52,13 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_vertical_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_vertical_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_vertical_edge_c(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_vertical_edge_c(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
}
void vp8_loop_filter_mbvs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -68,7 +68,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_vertical_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_vertical_edge_c(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
/* Horizontal B Filtering */
@@ -81,10 +81,10 @@
vp8_loop_filter_horizontal_edge_c(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_horizontal_edge_c(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_horizontal_edge_c(v_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
}
void vp8_loop_filter_bhs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -109,10 +109,10 @@
vp8_loop_filter_vertical_edge_c(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_vertical_edge_c(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_vertical_edge_c(v_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
}
void vp8_loop_filter_bvs_c(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -137,8 +137,6 @@
int block_inside_limit = 0;
int HEVThresh;
- const int yhedge_boost = 2;
- const int uvhedge_boost = 2;
/* For each possible value for the loop filter fill out a "loop_filter_info" entry. */
for (i = 0; i <= MAX_LOOP_FILTER; i++)
@@ -182,15 +180,9 @@
for (j = 0; j < 16; j++)
{
lfi[i].lim[j] = block_inside_limit;
- lfi[i].mbflim[j] = filt_lvl + yhedge_boost;
- lfi[i].mbthr[j] = HEVThresh;
+ lfi[i].mbflim[j] = filt_lvl + 2;
lfi[i].flim[j] = filt_lvl;
lfi[i].thr[j] = HEVThresh;
- lfi[i].uvlim[j] = block_inside_limit;
- lfi[i].uvmbflim[j] = filt_lvl + uvhedge_boost;
- lfi[i].uvmbthr[j] = HEVThresh;
- lfi[i].uvflim[j] = filt_lvl;
- lfi[i].uvthr[j] = HEVThresh;
}
}
@@ -249,57 +241,52 @@
for (j = 0; j < 16; j++)
{
/*lfi[i].lim[j] = block_inside_limit;
- lfi[i].mbflim[j] = filt_lvl+yhedge_boost;*/
- lfi[i].mbthr[j] = HEVThresh;
+ lfi[i].mbflim[j] = filt_lvl+2;*/
/*lfi[i].flim[j] = filt_lvl;*/
lfi[i].thr[j] = HEVThresh;
- /*lfi[i].uvlim[j] = block_inside_limit;
- lfi[i].uvmbflim[j] = filt_lvl+uvhedge_boost;*/
- lfi[i].uvmbthr[j] = HEVThresh;
- /*lfi[i].uvflim[j] = filt_lvl;*/
- lfi[i].uvthr[j] = HEVThresh;
}
}
}
-void vp8_adjust_mb_lf_value(MACROBLOCKD *mbd, int *filter_level)
+int vp8_adjust_mb_lf_value(MACROBLOCKD *mbd, int filter_level)
{
MB_MODE_INFO *mbmi = &mbd->mode_info_context->mbmi;
if (mbd->mode_ref_lf_delta_enabled)
{
/* Apply delta for reference frame */
- *filter_level += mbd->ref_lf_deltas[mbmi->ref_frame];
+ filter_level += mbd->ref_lf_deltas[mbmi->ref_frame];
/* Apply delta for mode */
if (mbmi->ref_frame == INTRA_FRAME)
{
/* Only the split mode BPRED has a further special case */
if (mbmi->mode == B_PRED)
- *filter_level += mbd->mode_lf_deltas[0];
+ filter_level += mbd->mode_lf_deltas[0];
}
else
{
/* Zero motion mode */
if (mbmi->mode == ZEROMV)
- *filter_level += mbd->mode_lf_deltas[1];
+ filter_level += mbd->mode_lf_deltas[1];
/* Split MB motion mode */
else if (mbmi->mode == SPLITMV)
- *filter_level += mbd->mode_lf_deltas[3];
+ filter_level += mbd->mode_lf_deltas[3];
/* All other inter motion modes (Nearest, Near, New) */
else
- *filter_level += mbd->mode_lf_deltas[2];
+ filter_level += mbd->mode_lf_deltas[2];
}
/* Range check */
- if (*filter_level > MAX_LOOP_FILTER)
- *filter_level = MAX_LOOP_FILTER;
- else if (*filter_level < 0)
- *filter_level = 0;
+ if (filter_level > MAX_LOOP_FILTER)
+ filter_level = MAX_LOOP_FILTER;
+ else if (filter_level < 0)
+ filter_level = 0;
}
+ return filter_level;
}
@@ -373,7 +360,7 @@
* These specified to 8th pel as they are always compared to values that are in 1/8th pel units
* Apply any context driven MB level adjustment
*/
- vp8_adjust_mb_lf_value(mbd, &filter_level);
+ filter_level = vp8_adjust_mb_lf_value(mbd, filter_level);
if (filter_level)
{
@@ -473,7 +460,7 @@
filter_level = baseline_filter_level[Segment];
/* Apply any context driven MB level adjustment */
- vp8_adjust_mb_lf_value(mbd, &filter_level);
+ filter_level = vp8_adjust_mb_lf_value(mbd, filter_level);
if (filter_level)
{
diff --git a/vp8/common/loopfilter.h b/vp8/common/loopfilter.h
index e456834..2e5997c 100644
--- a/vp8/common/loopfilter.h
+++ b/vp8/common/loopfilter.h
@@ -32,12 +32,6 @@
DECLARE_ALIGNED(16, signed char, flim[16]);
DECLARE_ALIGNED(16, signed char, thr[16]);
DECLARE_ALIGNED(16, signed char, mbflim[16]);
- DECLARE_ALIGNED(16, signed char, mbthr[16]);
- DECLARE_ALIGNED(16, signed char, uvlim[16]);
- DECLARE_ALIGNED(16, signed char, uvflim[16]);
- DECLARE_ALIGNED(16, signed char, uvthr[16]);
- DECLARE_ALIGNED(16, signed char, uvmbflim[16]);
- DECLARE_ALIGNED(16, signed char, uvmbthr[16]);
} loop_filter_info;
diff --git a/vp8/common/onyxc_int.h b/vp8/common/onyxc_int.h
index e011ec9..c8c2277 100644
--- a/vp8/common/onyxc_int.h
+++ b/vp8/common/onyxc_int.h
@@ -200,7 +200,7 @@
} VP8_COMMON;
-void vp8_adjust_mb_lf_value(MACROBLOCKD *mbd, int *filter_level);
+int vp8_adjust_mb_lf_value(MACROBLOCKD *mbd, int filter_level);
void vp8_init_loop_filter(VP8_COMMON *cm);
void vp8_frame_init_loop_filter(loop_filter_info *lfi, int frame_type);
extern void vp8_loop_filter_frame(VP8_COMMON *cm, MACROBLOCKD *mbd, int filt_val);
diff --git a/vp8/common/ppc/loopfilter_altivec.c b/vp8/common/ppc/loopfilter_altivec.c
index bad3cf3..e602fee 100644
--- a/vp8/common/ppc/loopfilter_altivec.c
+++ b/vp8/common/ppc/loopfilter_altivec.c
@@ -56,10 +56,10 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void)simpler_lpf;
- mbloop_filter_horizontal_edge_y_ppc(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr);
+ mbloop_filter_horizontal_edge_y_ppc(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr);
if (u_ptr)
- mbloop_filter_horizontal_edge_uv_ppc(u_ptr, v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr);
+ mbloop_filter_horizontal_edge_uv_ppc(u_ptr, v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr);
}
void loop_filter_mbhs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -77,10 +77,10 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void)simpler_lpf;
- mbloop_filter_vertical_edge_y_ppc(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr);
+ mbloop_filter_vertical_edge_y_ppc(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr);
if (u_ptr)
- mbloop_filter_vertical_edge_uv_ppc(u_ptr, v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr);
+ mbloop_filter_vertical_edge_uv_ppc(u_ptr, v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr);
}
void loop_filter_mbvs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -104,7 +104,7 @@
loop_filter_horizontal_edge_y_ppc(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr);
if (u_ptr)
- loop_filter_horizontal_edge_uv_ppc(u_ptr + 4 * uv_stride, v_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr);
+ loop_filter_horizontal_edge_uv_ppc(u_ptr + 4 * uv_stride, v_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr);
}
void loop_filter_bhs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
@@ -127,7 +127,7 @@
loop_filter_vertical_edge_y_ppc(y_ptr, y_stride, lfi->flim, lfi->lim, lfi->thr);
if (u_ptr)
- loop_filter_vertical_edge_uv_ppc(u_ptr + 4, v_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr);
+ loop_filter_vertical_edge_uv_ppc(u_ptr + 4, v_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr);
}
void loop_filter_bvs_ppc(unsigned char *y_ptr, unsigned char *u_ptr, unsigned char *v_ptr,
diff --git a/vp8/common/threading.h b/vp8/common/threading.h
index 1929f7c..0dd7bbc 100644
--- a/vp8/common/threading.h
+++ b/vp8/common/threading.h
@@ -14,6 +14,8 @@
#define VPXINFINITE 10000 /* 10second. */
+#if CONFIG_OS_SUPPORT
+
/* Thread management macros */
#ifdef _WIN32
/* Win32 */
@@ -88,4 +90,8 @@
#define x86_pause_hint()
#endif
+#else /* CONFIG_OS_SUPPORT = 0 */
+#define THREAD_FUNCTION void *
+#endif /* CONFIG_OS_SUPPORT */
+
#endif
diff --git a/vp8/common/x86/loopfilter_x86.c b/vp8/common/x86/loopfilter_x86.c
index 93107e1..3621fe1 100644
--- a/vp8/common/x86/loopfilter_x86.c
+++ b/vp8/common/x86/loopfilter_x86.c
@@ -45,13 +45,13 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_horizontal_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_horizontal_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_horizontal_edge_mmx(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_horizontal_edge_mmx(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_mbloop_filter_horizontal_edge_mmx(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_horizontal_edge_mmx(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
}
@@ -62,7 +62,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_horizontal_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
@@ -71,13 +71,13 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_vertical_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_vertical_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_vertical_edge_mmx(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_vertical_edge_mmx(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_mbloop_filter_vertical_edge_mmx(v_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, 1);
+ vp8_mbloop_filter_vertical_edge_mmx(v_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, 1);
}
@@ -88,7 +88,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_vertical_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_vertical_edge_mmx(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
@@ -102,10 +102,10 @@
vp8_loop_filter_horizontal_edge_mmx(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_horizontal_edge_mmx(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_horizontal_edge_mmx(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_loop_filter_horizontal_edge_mmx(v_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_horizontal_edge_mmx(v_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
}
@@ -132,10 +132,10 @@
vp8_loop_filter_vertical_edge_mmx(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_vertical_edge_mmx(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_vertical_edge_mmx(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
if (v_ptr)
- vp8_loop_filter_vertical_edge_mmx(v_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, 1);
+ vp8_loop_filter_vertical_edge_mmx(v_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, 1);
}
@@ -159,10 +159,10 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_horizontal_edge_uv_sse2(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
+ vp8_mbloop_filter_horizontal_edge_uv_sse2(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, v_ptr);
}
@@ -173,7 +173,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_horizontal_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
@@ -182,10 +182,10 @@
int y_stride, int uv_stride, loop_filter_info *lfi, int simpler_lpf)
{
(void) simpler_lpf;
- vp8_mbloop_filter_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_mbloop_filter_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_mbloop_filter_vertical_edge_uv_sse2(u_ptr, uv_stride, lfi->uvmbflim, lfi->uvlim, lfi->uvmbthr, v_ptr);
+ vp8_mbloop_filter_vertical_edge_uv_sse2(u_ptr, uv_stride, lfi->mbflim, lfi->lim, lfi->thr, v_ptr);
}
@@ -196,7 +196,7 @@
(void) v_ptr;
(void) uv_stride;
(void) simpler_lpf;
- vp8_loop_filter_simple_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->mbthr, 2);
+ vp8_loop_filter_simple_vertical_edge_sse2(y_ptr, y_stride, lfi->mbflim, lfi->lim, lfi->thr, 2);
}
@@ -210,7 +210,7 @@
vp8_loop_filter_horizontal_edge_sse2(y_ptr + 12 * y_stride, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_horizontal_edge_uv_sse2(u_ptr + 4 * uv_stride, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4 * uv_stride);
+ vp8_loop_filter_horizontal_edge_uv_sse2(u_ptr + 4 * uv_stride, uv_stride, lfi->flim, lfi->lim, lfi->thr, v_ptr + 4 * uv_stride);
}
@@ -237,7 +237,7 @@
vp8_loop_filter_vertical_edge_sse2(y_ptr + 12, y_stride, lfi->flim, lfi->lim, lfi->thr, 2);
if (u_ptr)
- vp8_loop_filter_vertical_edge_uv_sse2(u_ptr + 4, uv_stride, lfi->uvflim, lfi->uvlim, lfi->uvthr, v_ptr + 4);
+ vp8_loop_filter_vertical_edge_uv_sse2(u_ptr + 4, uv_stride, lfi->flim, lfi->lim, lfi->thr, v_ptr + 4);
}
diff --git a/vp8/decoder/threading.c b/vp8/decoder/threading.c
index fc8406e..ec2cb2b 100644
--- a/vp8/decoder/threading.c
+++ b/vp8/decoder/threading.c
@@ -9,7 +9,7 @@
*/
-#ifndef WIN32
+#if !defined(WIN32) && CONFIG_OS_SUPPORT == 1
# include <unistd.h>
#endif
#ifdef __APPLE__
@@ -320,7 +320,7 @@
* These are specified to 8th pel as they are always compared to values that are in 1/8th pel units
* Apply any context driven MB level adjustment
*/
- vp8_adjust_mb_lf_value(xd, &filter_level);
+ filter_level = vp8_adjust_mb_lf_value(xd, filter_level);
}
/* Distance of Mb to the various image edges.
@@ -863,7 +863,7 @@
* These are specified to 8th pel as they are always compared to values that are in 1/8th pel units
* Apply any context driven MB level adjustment
*/
- vp8_adjust_mb_lf_value(xd, &filter_level);
+ filter_level = vp8_adjust_mb_lf_value(xd, filter_level);
}
/* Distance of Mb to the various image edges.
diff --git a/vp8/encoder/arm/neon/fastfdct4x4_neon.asm b/vp8/encoder/arm/neon/fastfdct4x4_neon.asm
index 8c191a7..1cc0bd7 100644
--- a/vp8/encoder/arm/neon/fastfdct4x4_neon.asm
+++ b/vp8/encoder/arm/neon/fastfdct4x4_neon.asm
@@ -112,10 +112,7 @@
ENDP
;-----------------
- AREA fastfdct_dat, DATA, READONLY
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_ffdct_coeff_
DCD ffdct_coeff
ffdct_coeff
diff --git a/vp8/encoder/arm/neon/fastfdct8x4_neon.asm b/vp8/encoder/arm/neon/fastfdct8x4_neon.asm
index ca351a1..f6e8bbb 100644
--- a/vp8/encoder/arm/neon/fastfdct8x4_neon.asm
+++ b/vp8/encoder/arm/neon/fastfdct8x4_neon.asm
@@ -165,10 +165,7 @@
ENDP
;-----------------
- AREA fastfdct8x4_dat, DATA, READONLY
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_ffdct8_coeff_
DCD ffdct8_coeff
ffdct8_coeff
diff --git a/vp8/encoder/arm/neon/shortfdct_neon.asm b/vp8/encoder/arm/neon/shortfdct_neon.asm
index 5af5cb8..1b7f362 100644
--- a/vp8/encoder/arm/neon/shortfdct_neon.asm
+++ b/vp8/encoder/arm/neon/shortfdct_neon.asm
@@ -122,10 +122,7 @@
ENDP
;-----------------
- AREA dct4x4_dat, DATA, READONLY
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_dct_matrix_
DCD dct_matrix
dct_matrix
diff --git a/vp8/encoder/arm/neon/vp8_subpixelvariance16x16_neon.asm b/vp8/encoder/arm/neon/vp8_subpixelvariance16x16_neon.asm
index 1475f76..5107d8b 100644
--- a/vp8/encoder/arm/neon/vp8_subpixelvariance16x16_neon.asm
+++ b/vp8/encoder/arm/neon/vp8_subpixelvariance16x16_neon.asm
@@ -416,10 +416,7 @@
ENDP
;-----------------
- AREA vp8e_bilinear_taps_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_BilinearTaps_coeff_
DCD bilinear_taps_coeff
bilinear_taps_coeff
diff --git a/vp8/encoder/arm/neon/vp8_subpixelvariance8x8_neon.asm b/vp8/encoder/arm/neon/vp8_subpixelvariance8x8_neon.asm
index cf4da62..38b5878 100644
--- a/vp8/encoder/arm/neon/vp8_subpixelvariance8x8_neon.asm
+++ b/vp8/encoder/arm/neon/vp8_subpixelvariance8x8_neon.asm
@@ -215,10 +215,7 @@
ENDP
;-----------------
- AREA bilinear_taps_dat, DATA, READWRITE ;read/write by default
-;Data section with name data_area is specified. DCD reserves space in memory for 48 data.
-;One word each is reserved. Label filter_coeff can be used to access the data.
-;Data address: filter_coeff, filter_coeff+4, filter_coeff+8 ...
+
_BilinearTaps_coeff_
DCD bilinear_taps_coeff
bilinear_taps_coeff
diff --git a/vp8/encoder/encodeframe.c b/vp8/encoder/encodeframe.c
index 1689b43..f06a4ca 100644
--- a/vp8/encoder/encodeframe.c
+++ b/vp8/encoder/encodeframe.c
@@ -399,13 +399,6 @@
// Clear Zbin mode boost for default case
cpi->zbin_mode_boost = 0;
- // vp8cx_init_quantizer() is first called in vp8_create_compressor(). A check is added here so that vp8cx_init_quantizer() is only called
- // when these values are not all zero.
- if (cpi->common.y1dc_delta_q | cpi->common.y2dc_delta_q | cpi->common.uvdc_delta_q | cpi->common.y2ac_delta_q | cpi->common.uvac_delta_q)
- {
- vp8cx_init_quantizer(cpi);
- }
-
// MB level quantizer setup
vp8cx_mb_init_quantizer(cpi, &cpi->mb);
}
@@ -487,6 +480,16 @@
int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
int seg_map_index = (mb_row * cpi->common.mb_cols);
+#if CONFIG_MULTITHREAD
+ const int nsync = cpi->mt_sync_range;
+ const int rightmost_col = cm->mb_cols - 1;
+ volatile const int *last_row_current_mb_col;
+
+ if ((cpi->b_multi_threaded != 0) && (mb_row != 0))
+ last_row_current_mb_col = &cpi->mt_current_mb_col[mb_row - 1];
+ else
+ last_row_current_mb_col = &rightmost_col;
+#endif
// reset above block coeffs
xd->above_context = cm->above_context;
@@ -532,6 +535,21 @@
x->rddiv = cpi->RDDIV;
x->rdmult = cpi->RDMULT;
+#if CONFIG_MULTITHREAD
+ if ((cpi->b_multi_threaded != 0) && (mb_row != 0))
+ {
+ if ((mb_col & (nsync - 1)) == 0)
+ {
+ while (mb_col > (*last_row_current_mb_col - nsync)
+ && (*last_row_current_mb_col) != (cm->mb_cols - 1))
+ {
+ x86_pause_hint();
+ thread_sleep(0);
+ }
+ }
+ }
+#endif
+
if(cpi->oxcf.tuning == VP8_TUNE_SSIM)
activity_sum += vp8_activity_masking(cpi, x);
@@ -628,7 +646,12 @@
x->partition_info++;
xd->above_context++;
- cpi->current_mb_col_main = mb_col;
+#if CONFIG_MULTITHREAD
+ if (cpi->b_multi_threaded != 0)
+ {
+ cpi->mt_current_mb_col[mb_row] = mb_col;
+ }
+#endif
}
//extend the recon for intra prediction
@@ -642,12 +665,15 @@
xd->mode_info_context++;
x->partition_info++;
x->activity_sum += activity_sum;
+
+#if CONFIG_MULTITHREAD
+ if ((cpi->b_multi_threaded != 0) && (mb_row == cm->mb_rows - 1))
+ {
+ sem_post(&cpi->h_event_end_encoding); /* signal frame encoding end */
+ }
+#endif
}
-
-
-
-
void vp8_encode_frame(VP8_COMP *cpi)
{
int mb_row;
@@ -800,22 +826,16 @@
vp8cx_init_mbrthread_data(cpi, x, cpi->mb_row_ei, 1, cpi->encoding_thread_count);
+ for (i = 0; i < cm->mb_rows; i++)
+ cpi->mt_current_mb_col[i] = 0;
+
+ for (i = 0; i < cpi->encoding_thread_count; i++)
+ {
+ sem_post(&cpi->h_event_start_encoding[i]);
+ }
+
for (mb_row = 0; mb_row < cm->mb_rows; mb_row += (cpi->encoding_thread_count + 1))
{
- cpi->current_mb_col_main = -1;
-
- for (i = 0; i < cpi->encoding_thread_count; i++)
- {
- if ((mb_row + i + 1) >= cm->mb_rows)
- break;
-
- cpi->mb_row_ei[i].mb_row = mb_row + i + 1;
- cpi->mb_row_ei[i].tp = cpi->tok + (mb_row + i + 1) * (cm->mb_cols * 16 * 24);
- cpi->mb_row_ei[i].current_mb_col = -1;
- //SetEvent(cpi->h_event_mbrencoding[i]);
- sem_post(&cpi->h_event_mbrencoding[i]);
- }
-
vp8_zero(cm->left_context)
tp = cpi->tok + mb_row * (cm->mb_cols * 16 * 24);
@@ -830,26 +850,10 @@
xd->mode_info_context += xd->mode_info_stride * cpi->encoding_thread_count;
x->partition_info += xd->mode_info_stride * cpi->encoding_thread_count;
- if (mb_row < cm->mb_rows - 1)
- //WaitForSingleObject(cpi->h_event_main, INFINITE);
- sem_wait(&cpi->h_event_main);
}
- /*
- for( ;mb_row<cm->mb_rows; mb_row ++)
- {
- vp8_zero( cm->left_context)
+ sem_wait(&cpi->h_event_end_encoding); /* wait for other threads to finish */
- tp = cpi->tok + mb_row * (cm->mb_cols * 16 * 24);
-
- encode_mb_row(cpi, cm, mb_row, x, xd, &tp, segment_counts, &totalrate);
- // adjust to the next row of mbs
- x->src.y_buffer += 16 * x->src.y_stride - 16 * cm->mb_cols;
- x->src.u_buffer += 8 * x->src.uv_stride - 8 * cm->mb_cols;
- x->src.v_buffer += 8 * x->src.uv_stride - 8 * cm->mb_cols;
-
- }
- */
cpi->tok_count = 0;
for (mb_row = 0; mb_row < cm->mb_rows; mb_row ++)
@@ -859,7 +863,6 @@
if (xd->segmentation_enabled)
{
-
int i, j;
if (xd->segmentation_enabled)
@@ -871,7 +874,6 @@
segment_counts[j] += cpi->mb_row_ei[i].segment_counts[j];
}
}
-
}
for (i = 0; i < cpi->encoding_thread_count; i++)
@@ -1147,77 +1149,41 @@
int vp8cx_encode_intra_macro_block(VP8_COMP *cpi, MACROBLOCK *x, TOKENEXTRA **t)
{
int Error4x4, Error16x16, error_uv;
- B_PREDICTION_MODE intra_bmodes[16];
int rate4x4, rate16x16, rateuv;
int dist4x4, dist16x16, distuv;
int rate = 0;
int rate4x4_tokenonly = 0;
int rate16x16_tokenonly = 0;
int rateuv_tokenonly = 0;
- int i;
x->e_mbd.mode_info_context->mbmi.ref_frame = INTRA_FRAME;
#if !(CONFIG_REALTIME_ONLY)
-
- if (cpi->sf.RD || cpi->compressor_speed != 2)
+ if (cpi->sf.RD && cpi->compressor_speed != 2)
{
- Error4x4 = vp8_rd_pick_intra4x4mby_modes(cpi, x, &rate4x4, &rate4x4_tokenonly, &dist4x4);
-
- //save the b modes for possible later use
- for (i = 0; i < 16; i++)
- intra_bmodes[i] = x->e_mbd.block[i].bmi.mode;
+ error_uv = vp8_rd_pick_intra_mbuv_mode(cpi, x, &rateuv, &rateuv_tokenonly, &distuv);
+ rate += rateuv;
Error16x16 = vp8_rd_pick_intra16x16mby_mode(cpi, x, &rate16x16, &rate16x16_tokenonly, &dist16x16);
- error_uv = vp8_rd_pick_intra_mbuv_mode(cpi, x, &rateuv, &rateuv_tokenonly, &distuv);
+ Error4x4 = vp8_rd_pick_intra4x4mby_modes(cpi, x, &rate4x4, &rate4x4_tokenonly, &dist4x4);
- vp8_encode_intra16x16mbuv(IF_RTCD(&cpi->rtcd), x);
- rate += rateuv;
-
- if (Error4x4 < Error16x16)
- {
- rate += rate4x4;
- x->e_mbd.mode_info_context->mbmi.mode = B_PRED;
-
- // get back the intra block modes
- for (i = 0; i < 16; i++)
- x->e_mbd.block[i].bmi.mode = intra_bmodes[i];
-
- vp8_encode_intra4x4mby(IF_RTCD(&cpi->rtcd), x);
- cpi->prediction_error += Error4x4 ;
-#if 0
- // Experimental RD code
- cpi->frame_distortion += dist4x4;
-#endif
- }
- else
- {
- vp8_encode_intra16x16mby(IF_RTCD(&cpi->rtcd), x);
- rate += rate16x16;
-
-#if 0
- // Experimental RD code
- cpi->prediction_error += Error16x16;
- cpi->frame_distortion += dist16x16;
-#endif
- }
-
- sum_intra_stats(cpi, x);
-
- vp8_tokenize_mb(cpi, &x->e_mbd, t);
+ rate += (Error4x4 < Error16x16) ? rate4x4 : rate16x16;
}
else
#endif
{
-
- int rate2, distortion2;
+ int rate2, best_distortion;
MB_PREDICTION_MODE mode, best_mode = DC_PRED;
int this_rd;
Error16x16 = INT_MAX;
+ vp8_pick_intra_mbuv_mode(x);
+
for (mode = DC_PRED; mode <= TM_PRED; mode ++)
{
+ int distortion2;
+
x->e_mbd.mode_info_context->mbmi.mode = mode;
vp8_build_intra_predictors_mby_ptr(&x->e_mbd);
distortion2 = VARIANCE_INVOKE(&cpi->rtcd.variance, get16x16prederror)(x->src.y_buffer, x->src.y_stride, x->e_mbd.predictor, 16, 0x7fffffff);
@@ -1228,35 +1194,28 @@
{
Error16x16 = this_rd;
best_mode = mode;
+ best_distortion = distortion2;
}
}
+ x->e_mbd.mode_info_context->mbmi.mode = best_mode;
- vp8_pick_intra4x4mby_modes(IF_RTCD(&cpi->rtcd), x, &rate2, &distortion2);
-
- if (distortion2 == INT_MAX)
- Error4x4 = INT_MAX;
- else
- Error4x4 = RD_ESTIMATE(x->rdmult, x->rddiv, rate2, distortion2);
-
- if (Error4x4 < Error16x16)
- {
- x->e_mbd.mode_info_context->mbmi.mode = B_PRED;
- vp8_encode_intra4x4mby(IF_RTCD(&cpi->rtcd), x);
- cpi->prediction_error += Error4x4;
- }
- else
- {
- x->e_mbd.mode_info_context->mbmi.mode = best_mode;
- vp8_encode_intra16x16mby(IF_RTCD(&cpi->rtcd), x);
- cpi->prediction_error += Error16x16;
- }
-
- vp8_pick_intra_mbuv_mode(x);
- vp8_encode_intra16x16mbuv(IF_RTCD(&cpi->rtcd), x);
- sum_intra_stats(cpi, x);
- vp8_tokenize_mb(cpi, &x->e_mbd, t);
+ Error4x4 = vp8_pick_intra4x4mby_modes(IF_RTCD(&cpi->rtcd), x, &rate2, &best_distortion);
}
+ if (Error4x4 < Error16x16)
+ {
+ x->e_mbd.mode_info_context->mbmi.mode = B_PRED;
+ vp8_encode_intra4x4mby(IF_RTCD(&cpi->rtcd), x);
+ }
+ else
+ {
+ vp8_encode_intra16x16mby(IF_RTCD(&cpi->rtcd), x);
+ }
+
+ vp8_encode_intra16x16mbuv(IF_RTCD(&cpi->rtcd), x);
+ sum_intra_stats(cpi, x);
+ vp8_tokenize_mb(cpi, &x->e_mbd, t);
+
return rate;
}
#ifdef SPEEDSTATS
diff --git a/vp8/encoder/ethreading.c b/vp8/encoder/ethreading.c
index 7b5cfe0..8da93a6 100644
--- a/vp8/encoder/ethreading.c
+++ b/vp8/encoder/ethreading.c
@@ -8,15 +8,16 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#include "onyx_int.h"
#include "threading.h"
#include "common.h"
#include "extend.h"
-
-extern int vp8cx_encode_inter_macroblock(VP8_COMP *cpi, MACROBLOCK *x, TOKENEXTRA **t, int recon_yoffset, int recon_uvoffset);
-extern int vp8cx_encode_intra_macro_block(VP8_COMP *cpi, MACROBLOCK *x, TOKENEXTRA **t);
+extern int vp8cx_encode_inter_macroblock(VP8_COMP *cpi, MACROBLOCK *x,
+ TOKENEXTRA **t, int recon_yoffset,
+ int recon_uvoffset);
+extern int vp8cx_encode_intra_macro_block(VP8_COMP *cpi, MACROBLOCK *x,
+ TOKENEXTRA **t);
extern void vp8cx_mb_init_quantizer(VP8_COMP *cpi, MACROBLOCK *x);
extern void vp8_build_block_offsets(MACROBLOCK *x);
extern void vp8_setup_block_ptrs(MACROBLOCK *x);
@@ -26,10 +27,11 @@
{
#if CONFIG_MULTITHREAD
int ithread = ((ENCODETHREAD_DATA *)p_data)->ithread;
- VP8_COMP *cpi = (VP8_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr1);
+ VP8_COMP *cpi = (VP8_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr1);
MB_ROW_COMP *mbri = (MB_ROW_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr2);
ENTROPY_CONTEXT_PLANES mb_row_left_context;
+ const int nsync = cpi->mt_sync_range;
//printf("Started thread %d\n", ithread);
while (1)
@@ -38,210 +40,209 @@
break;
//if(WaitForSingleObject(cpi->h_event_mbrencoding[ithread], INFINITE) == WAIT_OBJECT_0)
- if (sem_wait(&cpi->h_event_mbrencoding[ithread]) == 0)
+ if (sem_wait(&cpi->h_event_start_encoding[ithread]) == 0)
{
+ VP8_COMMON *cm = &cpi->common;
+ int mb_row;
+ MACROBLOCK *x = &mbri->mb;
+ MACROBLOCKD *xd = &x->e_mbd;
+ TOKENEXTRA *tp ;
+
+ int *segment_counts = mbri->segment_counts;
+ int *totalrate = &mbri->totalrate;
+
if (cpi->b_multi_threaded == FALSE) // we're shutting down
break;
- else
+
+ for (mb_row = ithread + 1; mb_row < cm->mb_rows; mb_row += (cpi->encoding_thread_count + 1))
{
- VP8_COMMON *cm = &cpi->common;
- int mb_row = mbri->mb_row;
- MACROBLOCK *x = &mbri->mb;
- MACROBLOCKD *xd = &x->e_mbd;
- TOKENEXTRA **tp = &mbri->tp;
- int *segment_counts = mbri->segment_counts;
- int *totalrate = &mbri->totalrate;
+ int i;
+ int recon_yoffset, recon_uvoffset;
+ int mb_col;
+ int ref_fb_idx = cm->lst_fb_idx;
+ int dst_fb_idx = cm->new_fb_idx;
+ int recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride;
+ int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
+ volatile int *last_row_current_mb_col;
+ INT64 activity_sum = 0;
+
+ tp = cpi->tok + (mb_row * (cm->mb_cols * 16 * 24));
+
+ last_row_current_mb_col = &cpi->mt_current_mb_col[mb_row - 1];
+
+ // reset above block coeffs
+ xd->above_context = cm->above_context;
+ xd->left_context = &mb_row_left_context;
+
+ vp8_zero(mb_row_left_context);
+
+ xd->up_available = (mb_row != 0);
+ recon_yoffset = (mb_row * recon_y_stride * 16);
+ recon_uvoffset = (mb_row * recon_uv_stride * 8);
+
+ cpi->tplist[mb_row].start = tp;
+
+ //printf("Thread mb_row = %d\n", mb_row);
+
+ // for each macroblock col in image
+ for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
- int i;
- int recon_yoffset, recon_uvoffset;
- int mb_col;
- int ref_fb_idx = cm->lst_fb_idx;
- int dst_fb_idx = cm->new_fb_idx;
- int recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride;
- int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
- volatile int *last_row_current_mb_col;
- INT64 activity_sum = 0;
+ int seg_map_index = (mb_row * cm->mb_cols);
- if (ithread > 0)
- last_row_current_mb_col = &cpi->mb_row_ei[ithread-1].current_mb_col;
- else
- last_row_current_mb_col = &cpi->current_mb_col_main;
-
- // reset above block coeffs
- xd->above_context = cm->above_context;
- xd->left_context = &mb_row_left_context;
-
- vp8_zero(mb_row_left_context);
-
- xd->up_available = (mb_row != 0);
- recon_yoffset = (mb_row * recon_y_stride * 16);
- recon_uvoffset = (mb_row * recon_uv_stride * 8);
-
-
- cpi->tplist[mb_row].start = *tp;
-
- //printf("Thread mb_row = %d\n", mb_row);
-
- // for each macroblock col in image
- for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
+ if ((mb_col & (nsync - 1)) == 0)
{
- int seg_map_index = (mb_row * cm->mb_cols);
-
- while (mb_col > (*last_row_current_mb_col - 1) && *last_row_current_mb_col != cm->mb_cols - 1)
+ while (mb_col > (*last_row_current_mb_col - nsync) && *last_row_current_mb_col != cm->mb_cols - 1)
{
x86_pause_hint();
thread_sleep(0);
}
-
- // Distance of Mb to the various image edges.
- // These specified to 8th pel as they are always compared to values that are in 1/8th pel units
- xd->mb_to_left_edge = -((mb_col * 16) << 3);
- xd->mb_to_right_edge = ((cm->mb_cols - 1 - mb_col) * 16) << 3;
- xd->mb_to_top_edge = -((mb_row * 16) << 3);
- xd->mb_to_bottom_edge = ((cm->mb_rows - 1 - mb_row) * 16) << 3;
-
- // Set up limit values for motion vectors used to prevent them extending outside the UMV borders
- x->mv_col_min = -((mb_col * 16) + (VP8BORDERINPIXELS - 16));
- x->mv_col_max = ((cm->mb_cols - 1 - mb_col) * 16) + (VP8BORDERINPIXELS - 16);
- x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
- x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
-
- xd->dst.y_buffer = cm->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
- xd->dst.u_buffer = cm->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
- xd->dst.v_buffer = cm->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
- xd->left_available = (mb_col != 0);
-
- x->rddiv = cpi->RDDIV;
- x->rdmult = cpi->RDMULT;
-
- if(cpi->oxcf.tuning == VP8_TUNE_SSIM)
- activity_sum += vp8_activity_masking(cpi, x);
-
- // Is segmentation enabled
- // MB level adjutment to quantizer
- if (xd->segmentation_enabled)
- {
- // Code to set segment id in xd->mbmi.segment_id for current MB (with range checking)
- if (cpi->segmentation_map[seg_map_index+mb_col] <= 3)
- xd->mode_info_context->mbmi.segment_id = cpi->segmentation_map[seg_map_index+mb_col];
- else
- xd->mode_info_context->mbmi.segment_id = 0;
-
- vp8cx_mb_init_quantizer(cpi, x);
- }
- else
- xd->mode_info_context->mbmi.segment_id = 0; // Set to Segment 0 by default
-
- x->active_ptr = cpi->active_map + seg_map_index + mb_col;
-
- if (cm->frame_type == KEY_FRAME)
- {
- *totalrate += vp8cx_encode_intra_macro_block(cpi, x, tp);
-#ifdef MODE_STATS
- y_modes[xd->mbmi.mode] ++;
-#endif
- }
- else
- {
- *totalrate += vp8cx_encode_inter_macroblock(cpi, x, tp, recon_yoffset, recon_uvoffset);
-
-#ifdef MODE_STATS
- inter_y_modes[xd->mbmi.mode] ++;
-
- if (xd->mbmi.mode == SPLITMV)
- {
- int b;
-
- for (b = 0; b < xd->mbmi.partition_count; b++)
- {
- inter_b_modes[x->partition->bmi[b].mode] ++;
- }
- }
-
-#endif
-
- // Count of last ref frame 0,0 useage
- if ((xd->mode_info_context->mbmi.mode == ZEROMV) && (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME))
- cpi->inter_zz_count ++;
-
- // Special case code for cyclic refresh
- // If cyclic update enabled then copy xd->mbmi.segment_id; (which may have been updated based on mode
- // during vp8cx_encode_inter_macroblock()) back into the global sgmentation map
- if (cpi->cyclic_refresh_mode_enabled && xd->segmentation_enabled)
- {
- cpi->segmentation_map[seg_map_index+mb_col] = xd->mode_info_context->mbmi.segment_id;
-
- // If the block has been refreshed mark it as clean (the magnitude of the -ve influences how long it will be before we consider another refresh):
- // Else if it was coded (last frame 0,0) and has not already been refreshed then mark it as a candidate for cleanup next time (marked 0)
- // else mark it as dirty (1).
- if (xd->mode_info_context->mbmi.segment_id)
- cpi->cyclic_refresh_map[seg_map_index+mb_col] = -1;
- else if ((xd->mode_info_context->mbmi.mode == ZEROMV) && (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME))
- {
- if (cpi->cyclic_refresh_map[seg_map_index+mb_col] == 1)
- cpi->cyclic_refresh_map[seg_map_index+mb_col] = 0;
- }
- else
- cpi->cyclic_refresh_map[seg_map_index+mb_col] = 1;
-
- }
- }
- cpi->tplist[mb_row].stop = *tp;
-
- x->gf_active_ptr++; // Increment pointer into gf useage flags structure for next mb
-
- for (i = 0; i < 16; i++)
- vpx_memcpy(&xd->mode_info_context->bmi[i], &xd->block[i].bmi, sizeof(xd->block[i].bmi));
-
- // adjust to the next column of macroblocks
- x->src.y_buffer += 16;
- x->src.u_buffer += 8;
- x->src.v_buffer += 8;
-
- recon_yoffset += 16;
- recon_uvoffset += 8;
-
- // Keep track of segment useage
- segment_counts[xd->mode_info_context->mbmi.segment_id] ++;
-
- // skip to next mb
- xd->mode_info_context++;
- x->partition_info++;
-
- xd->above_context++;
-
- cpi->mb_row_ei[ithread].current_mb_col = mb_col;
-
}
- //extend the recon for intra prediction
- vp8_extend_mb_row(
- &cm->yv12_fb[dst_fb_idx],
- xd->dst.y_buffer + 16,
- xd->dst.u_buffer + 8,
- xd->dst.v_buffer + 8);
+ // Distance of Mb to the various image edges.
+ // These specified to 8th pel as they are always compared to values that are in 1/8th pel units
+ xd->mb_to_left_edge = -((mb_col * 16) << 3);
+ xd->mb_to_right_edge = ((cm->mb_cols - 1 - mb_col) * 16) << 3;
+ xd->mb_to_top_edge = -((mb_row * 16) << 3);
+ xd->mb_to_bottom_edge = ((cm->mb_rows - 1 - mb_row) * 16) << 3;
- // this is to account for the border
+ // Set up limit values for motion vectors used to prevent them extending outside the UMV borders
+ x->mv_col_min = -((mb_col * 16) + (VP8BORDERINPIXELS - 16));
+ x->mv_col_max = ((cm->mb_cols - 1 - mb_col) * 16) + (VP8BORDERINPIXELS - 16);
+ x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
+ x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
+
+ xd->dst.y_buffer = cm->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
+ xd->dst.u_buffer = cm->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
+ xd->dst.v_buffer = cm->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
+ xd->left_available = (mb_col != 0);
+
+ x->rddiv = cpi->RDDIV;
+ x->rdmult = cpi->RDMULT;
+
+ if (cpi->oxcf.tuning == VP8_TUNE_SSIM)
+ activity_sum += vp8_activity_masking(cpi, x);
+
+ // Is segmentation enabled
+ // MB level adjutment to quantizer
+ if (xd->segmentation_enabled)
+ {
+ // Code to set segment id in xd->mbmi.segment_id for current MB (with range checking)
+ if (cpi->segmentation_map[seg_map_index + mb_col] <= 3)
+ xd->mode_info_context->mbmi.segment_id = cpi->segmentation_map[seg_map_index + mb_col];
+ else
+ xd->mode_info_context->mbmi.segment_id = 0;
+
+ vp8cx_mb_init_quantizer(cpi, x);
+ }
+ else
+ xd->mode_info_context->mbmi.segment_id = 0; // Set to Segment 0 by default
+
+ x->active_ptr = cpi->active_map + seg_map_index + mb_col;
+
+ if (cm->frame_type == KEY_FRAME)
+ {
+ *totalrate += vp8cx_encode_intra_macro_block(cpi, x, &tp);
+#ifdef MODE_STATS
+ y_modes[xd->mbmi.mode] ++;
+#endif
+ }
+ else
+ {
+ *totalrate += vp8cx_encode_inter_macroblock(cpi, x, &tp, recon_yoffset, recon_uvoffset);
+
+#ifdef MODE_STATS
+ inter_y_modes[xd->mbmi.mode] ++;
+
+ if (xd->mbmi.mode == SPLITMV)
+ {
+ int b;
+
+ for (b = 0; b < xd->mbmi.partition_count; b++)
+ {
+ inter_b_modes[x->partition->bmi[b].mode] ++;
+ }
+ }
+
+#endif
+
+ // Count of last ref frame 0,0 useage
+ if ((xd->mode_info_context->mbmi.mode == ZEROMV) && (xd->mode_info_context->mbmi.ref_frame == LAST_FRAME))
+ cpi->inter_zz_count++;
+
+ // Special case code for cyclic refresh
+ // If cyclic update enabled then copy xd->mbmi.segment_id; (which may have been updated based on mode
+ // during vp8cx_encode_inter_macroblock()) back into the global sgmentation map
+ if (cpi->cyclic_refresh_mode_enabled && xd->segmentation_enabled)
+ {
+ const MB_MODE_INFO * mbmi = &xd->mode_info_context->mbmi;
+ cpi->segmentation_map[seg_map_index + mb_col] = mbmi->segment_id;
+
+ // If the block has been refreshed mark it as clean (the magnitude of the -ve influences how long it will be before we consider another refresh):
+ // Else if it was coded (last frame 0,0) and has not already been refreshed then mark it as a candidate for cleanup next time (marked 0)
+ // else mark it as dirty (1).
+ if (mbmi->segment_id)
+ cpi->cyclic_refresh_map[seg_map_index + mb_col] = -1;
+ else if ((mbmi->mode == ZEROMV) && (mbmi->ref_frame == LAST_FRAME))
+ {
+ if (cpi->cyclic_refresh_map[seg_map_index + mb_col] == 1)
+ cpi->cyclic_refresh_map[seg_map_index + mb_col] = 0;
+ }
+ else
+ cpi->cyclic_refresh_map[seg_map_index + mb_col] = 1;
+
+ }
+ }
+ cpi->tplist[mb_row].stop = tp;
+
+ x->gf_active_ptr++; // Increment pointer into gf useage flags structure for next mb
+
+ for (i = 0; i < 16; i++)
+ vpx_memcpy(&xd->mode_info_context->bmi[i], &xd->block[i].bmi, sizeof(xd->block[i].bmi));
+
+ // adjust to the next column of macroblocks
+ x->src.y_buffer += 16;
+ x->src.u_buffer += 8;
+ x->src.v_buffer += 8;
+
+ recon_yoffset += 16;
+ recon_uvoffset += 8;
+
+ // Keep track of segment useage
+ segment_counts[xd->mode_info_context->mbmi.segment_id]++;
+
+ // skip to next mb
xd->mode_info_context++;
x->partition_info++;
- x->activity_sum += activity_sum;
+ xd->above_context++;
- x->src.y_buffer += 16 * x->src.y_stride * (cpi->encoding_thread_count + 1) - 16 * cm->mb_cols;
- x->src.u_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
- x->src.v_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
-
- xd->mode_info_context += xd->mode_info_stride * cpi->encoding_thread_count;
- x->partition_info += xd->mode_info_stride * cpi->encoding_thread_count;
-
- if (ithread == (cpi->encoding_thread_count - 1) || mb_row == cm->mb_rows - 1)
- {
- //SetEvent(cpi->h_event_main);
- sem_post(&cpi->h_event_main);
- }
-
+ cpi->mt_current_mb_col[mb_row] = mb_col;
}
+ //extend the recon for intra prediction
+ vp8_extend_mb_row(
+ &cm->yv12_fb[dst_fb_idx],
+ xd->dst.y_buffer + 16,
+ xd->dst.u_buffer + 8,
+ xd->dst.v_buffer + 8);
+
+ // this is to account for the border
+ xd->mode_info_context++;
+ x->partition_info++;
+ x->activity_sum += activity_sum;
+
+ x->src.y_buffer += 16 * x->src.y_stride * (cpi->encoding_thread_count + 1) - 16 * cm->mb_cols;
+ x->src.u_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
+ x->src.v_buffer += 8 * x->src.uv_stride * (cpi->encoding_thread_count + 1) - 8 * cm->mb_cols;
+
+ xd->mode_info_context += xd->mode_info_stride * cpi->encoding_thread_count;
+ x->partition_info += xd->mode_info_stride * cpi->encoding_thread_count;
+
+ if (mb_row == cm->mb_rows - 1)
+ {
+ //SetEvent(cpi->h_event_main);
+ sem_post(&cpi->h_event_end_encoding); /* signal frame encoding end */
+ }
}
}
}
@@ -363,7 +364,6 @@
}
}
-
void vp8cx_init_mbrthread_data(VP8_COMP *cpi,
MACROBLOCK *x,
MB_ROW_COMP *mbr_ei,
@@ -414,7 +414,6 @@
mb->src.u_buffer += 8 * x->src.uv_stride * (i + 1);
mb->src.v_buffer += 8 * x->src.uv_stride * (i + 1);
-
vp8_build_block_offsets(mb);
vp8_setup_block_dptrs(mbd);
@@ -431,7 +430,6 @@
}
}
-
void vp8cx_create_encoder_threads(VP8_COMP *cpi)
{
cpi->b_multi_threaded = 0;
@@ -451,14 +449,15 @@
else
cpi->encoding_thread_count = cpi->oxcf.multi_threaded - 1;
-
CHECK_MEM_ERROR(cpi->h_encoding_thread, vpx_malloc(sizeof(pthread_t) * cpi->encoding_thread_count));
- CHECK_MEM_ERROR(cpi->h_event_mbrencoding, vpx_malloc(sizeof(sem_t) * cpi->encoding_thread_count));
+ CHECK_MEM_ERROR(cpi->h_event_start_encoding, vpx_malloc(sizeof(sem_t) * cpi->encoding_thread_count));
CHECK_MEM_ERROR(cpi->mb_row_ei, vpx_memalign(32, sizeof(MB_ROW_COMP) * cpi->encoding_thread_count));
vpx_memset(cpi->mb_row_ei, 0, sizeof(MB_ROW_COMP) * cpi->encoding_thread_count);
CHECK_MEM_ERROR(cpi->en_thread_data, vpx_malloc(sizeof(ENCODETHREAD_DATA) * cpi->encoding_thread_count));
+ CHECK_MEM_ERROR(cpi->mt_current_mb_col, vpx_malloc(sizeof(*cpi->mt_current_mb_col) * cpi->common.mb_rows));
+
//cpi->h_event_main = CreateEvent(NULL, FALSE, FALSE, NULL);
- sem_init(&cpi->h_event_main, 0, 0);
+ sem_init(&cpi->h_event_end_encoding, 0, 0);
cpi->b_multi_threaded = 1;
@@ -466,11 +465,13 @@
for (ithread = 0; ithread < cpi->encoding_thread_count; ithread++)
{
+ ENCODETHREAD_DATA * ethd = &cpi->en_thread_data[ithread];
+
//cpi->h_event_mbrencoding[ithread] = CreateEvent(NULL, FALSE, FALSE, NULL);
- sem_init(&cpi->h_event_mbrencoding[ithread], 0, 0);
- cpi->en_thread_data[ithread].ithread = ithread;
- cpi->en_thread_data[ithread].ptr1 = (void *)cpi;
- cpi->en_thread_data[ithread].ptr2 = (void *)&cpi->mb_row_ei[ithread];
+ sem_init(&cpi->h_event_start_encoding[ithread], 0, 0);
+ ethd->ithread = ithread;
+ ethd->ptr1 = (void *)cpi;
+ ethd->ptr2 = (void *)&cpi->mb_row_ei[ithread];
//printf(" call begin thread %d \n", ithread);
@@ -482,8 +483,7 @@
// 0,
// NULL);
- pthread_create(&cpi->h_encoding_thread[ithread], 0, thread_encoding_proc, (&cpi->en_thread_data[ithread]));
-
+ pthread_create(&cpi->h_encoding_thread[ithread], 0, thread_encoding_proc, ethd);
}
}
@@ -505,18 +505,21 @@
for (i = 0; i < cpi->encoding_thread_count; i++)
{
//SetEvent(cpi->h_event_mbrencoding[i]);
- sem_post(&cpi->h_event_mbrencoding[i]);
+ sem_post(&cpi->h_event_start_encoding[i]);
pthread_join(cpi->h_encoding_thread[i], 0);
- }
- for (i = 0; i < cpi->encoding_thread_count; i++)
- sem_destroy(&cpi->h_event_mbrencoding[i]);
+ sem_destroy(&cpi->h_event_start_encoding[i]);
+ }
}
+
+ sem_destroy(&cpi->h_event_end_encoding);
+
//free thread related resources
- vpx_free(cpi->h_event_mbrencoding);
+ vpx_free(cpi->h_event_start_encoding);
vpx_free(cpi->h_encoding_thread);
vpx_free(cpi->mb_row_ei);
vpx_free(cpi->en_thread_data);
+ vpx_free(cpi->mt_current_mb_col);
}
#endif
diff --git a/vp8/encoder/firstpass.c b/vp8/encoder/firstpass.c
index 06e26be..fc6f043 100644
--- a/vp8/encoder/firstpass.c
+++ b/vp8/encoder/firstpass.c
@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
-
#include "math.h"
#include "limits.h"
#include "block.h"
@@ -178,40 +177,68 @@
return modified_err;
}
+static const double weight_table[256] = {
+0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000,
+0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000,
+0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000,
+0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000, 0.020000,
+0.020000, 0.031250, 0.062500, 0.093750, 0.125000, 0.156250, 0.187500, 0.218750,
+0.250000, 0.281250, 0.312500, 0.343750, 0.375000, 0.406250, 0.437500, 0.468750,
+0.500000, 0.531250, 0.562500, 0.593750, 0.625000, 0.656250, 0.687500, 0.718750,
+0.750000, 0.781250, 0.812500, 0.843750, 0.875000, 0.906250, 0.937500, 0.968750,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000,
+1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000, 1.000000
+};
+
double vp8_simple_weight(YV12_BUFFER_CONFIG *source)
{
int i, j;
unsigned char *src = source->y_buffer;
- unsigned char value;
double sum_weights = 0.0;
- double Weight;
// Loop throught the Y plane raw examining levels and creating a weight for the image
- for (i = 0; i < source->y_height; i++)
+ i = source->y_height;
+ do
{
- for (j = 0; j < source->y_width; j++)
+ j = source->y_width;
+ do
{
- value = src[j];
-
- if (value >= 64)
- Weight = 1.0;
- else if (value > 32)
- Weight = (value - 32.0f) / 32.0f;
- else
- Weight = 0.02;
-
- sum_weights += Weight;
- }
-
+ sum_weights += weight_table[ *src];
+ src++;
+ }while(--j);
+ src -= source->y_width;
src += source->y_stride;
- }
+ }while(--i);
sum_weights /= (source->y_height * source->y_width);
return sum_weights;
}
+
// This function returns the current per frame maximum bitrate target
int frame_max_bits(VP8_COMP *cpi)
{
@@ -440,7 +467,6 @@
vp8_output_stats(cpi, cpi->output_pkt_list, cpi->total_stats);
}
-
void vp8_zz_motion_search( VP8_COMP *cpi, MACROBLOCK * x, YV12_BUFFER_CONFIG * recon_buffer, int * best_motion_err, int recon_yoffset )
{
MACROBLOCKD * const xd = & x->e_mbd;
@@ -460,7 +486,6 @@
VARIANCE_INVOKE(IF_RTCD(&cpi->rtcd.variance), mse16x16) ( src_ptr, src_stride, ref_ptr, ref_stride, (unsigned int *)(best_motion_err));
}
-
void vp8_first_pass_motion_search(VP8_COMP *cpi, MACROBLOCK *x, MV *ref_mv, MV *best_mv, YV12_BUFFER_CONFIG *recon_buffer, int *best_motion_err, int recon_yoffset )
{
MACROBLOCKD *const xd = & x->e_mbd;
@@ -548,7 +573,6 @@
int sum_in_vectors = 0;
- MV best_ref_mv = {0, 0};
MV zero_ref_mv = {0, 0};
unsigned char *fp_motion_map_ptr = cpi->fp_motion_map;
@@ -586,13 +610,20 @@
// for each macroblock row in image
for (mb_row = 0; mb_row < cm->mb_rows; mb_row++)
{
- MV best_ref_mv = {0, 0};
+ int_mv best_ref_mv;
+
+ best_ref_mv.as_int = 0;
// reset above block coeffs
xd->up_available = (mb_row != 0);
recon_yoffset = (mb_row * recon_y_stride * 16);
recon_uvoffset = (mb_row * recon_uv_stride * 8);
+ // Set up limit values for motion vectors to prevent them extending outside the UMV borders
+ x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
+ x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
+
+
// for each macroblock col in image
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
@@ -625,8 +656,6 @@
// Set up limit values for motion vectors to prevent them extending outside the UMV borders
x->mv_col_min = -((mb_col * 16) + (VP8BORDERINPIXELS - 16));
x->mv_col_max = ((cm->mb_cols - 1 - mb_col) * 16) + (VP8BORDERINPIXELS - 16);
- x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
- x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
// Other than for the first frame do a motion search
if (cm->current_video_frame > 0)
@@ -647,12 +676,12 @@
// Test last reference frame using the previous best mv as the
// starting point (best reference) for the search
- vp8_first_pass_motion_search(cpi, x, &best_ref_mv,
+ vp8_first_pass_motion_search(cpi, x, &best_ref_mv.as_mv,
&d->bmi.mv.as_mv, lst_yv12,
&motion_error, recon_yoffset);
// If the current best reference mv is not centred on 0,0 then do a 0,0 based search as well
- if ((best_ref_mv.col != 0) || (best_ref_mv.row != 0))
+ if (best_ref_mv.as_int)
{
tmp_err = INT_MAX;
vp8_first_pass_motion_search(cpi, x, &zero_ref_mv, &tmp_mv,
@@ -664,7 +693,6 @@
d->bmi.mv.as_mv.row = tmp_mv.row;
d->bmi.mv.as_mv.col = tmp_mv.col;
}
-
}
// Experimental search in a second reference frame ((0,0) based only)
@@ -693,6 +721,9 @@
xd->pre.v_buffer = lst_yv12->v_buffer + recon_uvoffset;
}
+ /* Intra assumed best */
+ best_ref_mv.as_int = 0;
+
if (motion_error <= this_error)
{
d->bmi.mv.as_mv.row <<= 3;
@@ -708,13 +739,10 @@
sum_mvcs += d->bmi.mv.as_mv.col * d->bmi.mv.as_mv.col;
intercount++;
- best_ref_mv.row = d->bmi.mv.as_mv.row;
- best_ref_mv.col = d->bmi.mv.as_mv.col;
- //best_ref_mv.row = 0;
- //best_ref_mv.col = 0;
+ best_ref_mv.as_int = d->bmi.mv.as_int;
// Was the vector non-zero
- if (d->bmi.mv.as_mv.row || d->bmi.mv.as_mv.col)
+ if (d->bmi.mv.as_int)
{
mvcount++;
@@ -770,12 +798,6 @@
*fp_motion_map_ptr = 1;
}
}
- else
- {
- // Intra was best
- best_ref_mv.row = 0;
- best_ref_mv.col = 0;
- }
}
coded_error += this_error;
@@ -813,6 +835,7 @@
fps.coded_error = coded_error >> 8;
weight = vp8_simple_weight(cpi->Source);
+
if (weight < 0.1)
weight = 0.1;
diff --git a/vp8/encoder/onyx_if.c b/vp8/encoder/onyx_if.c
index 92b2f6f..912d641 100644
--- a/vp8/encoder/onyx_if.c
+++ b/vp8/encoder/onyx_if.c
@@ -1465,6 +1465,17 @@
vpx_internal_error(&cpi->common.error, VPX_CODEC_MEM_ERROR,
"Failed to allocate firstpass stats");
#endif
+
+#if CONFIG_MULTITHREAD
+ if (width < 640)
+ cpi->mt_sync_range = 1;
+ else if (width <= 1280)
+ cpi->mt_sync_range = 4;
+ else if (width <= 2560)
+ cpi->mt_sync_range = 8;
+ else
+ cpi->mt_sync_range = 16;
+#endif
}
@@ -3112,11 +3123,14 @@
return 1;
}
+
static void set_quantizer(VP8_COMP *cpi, int Q)
{
VP8_COMMON *cm = &cpi->common;
MACROBLOCKD *mbd = &cpi->mb.e_mbd;
+ int update = 0;
+ update |= cm->base_qindex != Q;
cm->base_qindex = Q;
cm->y1dc_delta_q = 0;
@@ -3127,13 +3141,19 @@
if(Q<4)
{
+ update |= cm->y2dc_delta_q != 4-Q;
cm->y2dc_delta_q = 4-Q;
}
+
// Set Segment specific quatizers
mbd->segment_feature_data[MB_LVL_ALT_Q][0] = cpi->segment_feature_data[MB_LVL_ALT_Q][0];
mbd->segment_feature_data[MB_LVL_ALT_Q][1] = cpi->segment_feature_data[MB_LVL_ALT_Q][1];
mbd->segment_feature_data[MB_LVL_ALT_Q][2] = cpi->segment_feature_data[MB_LVL_ALT_Q][2];
mbd->segment_feature_data[MB_LVL_ALT_Q][3] = cpi->segment_feature_data[MB_LVL_ALT_Q][3];
+
+ if(update)
+ vp8cx_init_quantizer(cpi);
+
}
static void update_alt_ref_frame_and_stats(VP8_COMP *cpi)
@@ -3603,6 +3623,17 @@
// Test code for segmentation of gf/arf (0,0)
//segmentation_test_function((VP8_PTR) cpi);
+#if CONFIG_REALTIME_ONLY
+ if(cpi->oxcf.auto_key && cm->frame_type != KEY_FRAME)
+ {
+ if(cpi->force_next_frame_intra)
+ {
+ cm->frame_type = KEY_FRAME; /* delayed intra frame */
+ }
+ }
+ cpi->force_next_frame_intra = 0;
+#endif
+
// For an alt ref frame in 2 pass we skip the call to the second pass function that sets the target bandwidth
#if !(CONFIG_REALTIME_ONLY)
@@ -4113,6 +4144,14 @@
// (assuming that we didn't)!
if (cpi->pass != 2 && cpi->oxcf.auto_key && cm->frame_type != KEY_FRAME)
{
+
+#if CONFIG_REALTIME_ONLY
+ {
+ /* we don't do re-encoding in realtime mode
+ * if key frame is decided than we force it on next frame */
+ cpi->force_next_frame_intra = decide_key_frame(cpi);
+ }
+#else
if (decide_key_frame(cpi))
{
vp8_calc_auto_iframe_target_size(cpi);
@@ -4151,6 +4190,7 @@
resize_key_frame(cpi);
continue;
}
+#endif
}
vp8_clear_system_state();
@@ -4413,7 +4453,7 @@
}
// Update the GF useage maps.
- // Update the GF useage maps.
+ // This is done after completing the compression of a frame when all modes etc. are finalized but before loop filter
// This is done after completing the compression of a frame when all modes etc. are finalized but before loop filter
vp8_update_gf_useage_maps(cpi, cm, &cpi->mb);
diff --git a/vp8/encoder/onyx_int.h b/vp8/encoder/onyx_int.h
index 0d01c2f..e8a452d 100644
--- a/vp8/encoder/onyx_int.h
+++ b/vp8/encoder/onyx_int.h
@@ -191,11 +191,8 @@
typedef struct
{
MACROBLOCK mb;
- int mb_row;
- TOKENEXTRA *tp;
int segment_counts[MAX_MB_SEGMENTS];
int totalrate;
- int current_mb_col;
} MB_ROW_COMP;
typedef struct
@@ -343,7 +340,6 @@
long long last_prediction_error;
long long intra_error;
long long last_intra_error;
- long long last_auto_filter_prediction_error;
#if 0
// Experimental RD code
@@ -594,7 +590,8 @@
signed char *cyclic_refresh_map;
// multithread data
- int current_mb_col_main;
+ int * mt_current_mb_col;
+ int mt_sync_range;
int processor_core_count;
int b_multi_threaded;
int encoding_thread_count;
@@ -607,8 +604,8 @@
#if CONFIG_MULTITHREAD
//events
- sem_t *h_event_mbrencoding;
- sem_t h_event_main;
+ sem_t *h_event_start_encoding;
+ sem_t h_event_end_encoding;
#endif
TOKENLIST *tplist;
@@ -680,6 +677,9 @@
int *lf_ref_frame_sign_bias;
int *lf_ref_frame;
+#if CONFIG_REALTIME_ONLY
+ int force_next_frame_intra; /* force next frame to intra when kf_auto says so */
+#endif
} VP8_COMP;
void control_data_rate(VP8_COMP *cpi);
diff --git a/vp8/encoder/pickinter.c b/vp8/encoder/pickinter.c
index d60a554..287ca61 100644
--- a/vp8/encoder/pickinter.c
+++ b/vp8/encoder/pickinter.c
@@ -168,8 +168,6 @@
B_PREDICTION_MODE *best_mode,
B_PREDICTION_MODE above,
B_PREDICTION_MODE left,
- ENTROPY_CONTEXT *a,
- ENTROPY_CONTEXT *l,
int *bestrate,
int *bestdistortion)
@@ -179,8 +177,6 @@
int rate;
int distortion;
unsigned int *mode_costs;
- (void) l;
- (void) a;
if (x->e_mbd.frame_type == KEY_FRAME)
{
@@ -211,6 +207,7 @@
b->bmi.mode = (B_PREDICTION_MODE)(*best_mode);
vp8_encode_intra4x4block(rtcd, x, be, b, b->bmi.mode);
+
return best_rd;
}
@@ -220,17 +217,8 @@
MACROBLOCKD *const xd = &mb->e_mbd;
int i;
int cost = mb->mbmode_cost [xd->frame_type] [B_PRED];
- int error = RD_ESTIMATE(mb->rdmult, mb->rddiv, cost, 0); // Rd estimate for the cost of the block prediction mode
+ int error;
int distortion = 0;
- ENTROPY_CONTEXT_PLANES t_above, t_left;
- ENTROPY_CONTEXT *ta;
- ENTROPY_CONTEXT *tl;
-
- vpx_memcpy(&t_above, mb->e_mbd.above_context, sizeof(ENTROPY_CONTEXT_PLANES));
- vpx_memcpy(&t_left, mb->e_mbd.left_context, sizeof(ENTROPY_CONTEXT_PLANES));
-
- ta = (ENTROPY_CONTEXT *)&t_above;
- tl = (ENTROPY_CONTEXT *)&t_left;
vp8_intra_prediction_down_copy(xd);
@@ -243,10 +231,8 @@
B_PREDICTION_MODE UNINITIALIZED_IS_SAFE(best_mode);
int UNINITIALIZED_IS_SAFE(r), UNINITIALIZED_IS_SAFE(d);
- error += pick_intra4x4block(rtcd,
- mb, mb->block + i, xd->block + i, &best_mode, A, L,
- ta + vp8_block2above[i],
- tl + vp8_block2left[i], &r, &d);
+ pick_intra4x4block(rtcd, mb, mb->block + i, xd->block + i,
+ &best_mode, A, L, &r, &d);
cost += r;
distortion += d;
@@ -264,10 +250,15 @@
*Rate = cost;
if (i == 16)
+ {
*best_dist = distortion;
+ error = RD_ESTIMATE(mb->rdmult, mb->rddiv, cost, distortion);
+ }
else
+ {
*best_dist = INT_MAX;
-
+ error = INT_MAX;
+ }
return error;
}
diff --git a/vp8/encoder/picklpf.c b/vp8/encoder/picklpf.c
index 09e8b54..78aa866 100644
--- a/vp8/encoder/picklpf.c
+++ b/vp8/encoder/picklpf.c
@@ -296,7 +296,6 @@
int filt_err = 0;
int min_filter_level;
int max_filter_level;
- int prediction_difference = (int)(100 * abs((int)(cpi->last_auto_filter_prediction_error - cpi->prediction_error)) / (1 + cpi->prediction_error));
int filter_step;
int filt_high = 0;
@@ -478,6 +477,5 @@
cpi->last_auto_filt_val = filt_best;
cpi->last_auto_filt_q = cm->base_qindex;
- cpi->last_auto_filter_prediction_error = cpi->prediction_error;
cpi->frames_since_auto_filter = 0;
}
diff --git a/vp8/encoder/rdopt.c b/vp8/encoder/rdopt.c
index ac563c0..1ece1bf 100644
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -711,16 +711,10 @@
int distortion;
int best_rd = INT_MAX;
int this_rd;
- int i;
//Y Search for 16x16 intra prediction mode
for (mode = DC_PRED; mode <= TM_PRED; mode++)
{
- for (i = 0; i < 16; i++)
- {
- vpx_memset(&x->e_mbd.block[i].bmi, 0, sizeof(B_MODE_INFO));
- }
-
x->e_mbd.mode_info_context->mbmi.mode = mode;
vp8_build_intra_predictors_mby_ptr(&x->e_mbd);
@@ -1964,15 +1958,9 @@
switch (this_mode)
{
case B_PRED:
-
- for (i = 0; i < 16; i++)
- {
- vpx_memset(&x->e_mbd.block[i].bmi, 0, sizeof(B_MODE_INFO));
- }
// Note the rate value returned here includes the cost of coding the BPRED mode : x->mbmode_cost[x->e_mbd.frame_type][BPRED];
vp8_rd_pick_intra4x4mby_modes(cpi, x, &rate, &rate_y, &distortion);
rate2 += rate;
-
distortion2 += distortion;
rate2 += uv_intra_rate;
rate_uv = uv_intra_rate_tokenonly;
@@ -2014,22 +2002,16 @@
case V_PRED:
case H_PRED:
case TM_PRED:
- for (i = 0; i < 16; i++)
- {
- vpx_memset(&x->e_mbd.block[i].bmi, 0, sizeof(B_MODE_INFO));
- }
x->e_mbd.mode_info_context->mbmi.ref_frame = INTRA_FRAME;
vp8_build_intra_predictors_mby_ptr(&x->e_mbd);
- {
- macro_block_yrd(x, &rate_y, &distortion, IF_RTCD(&cpi->rtcd.encodemb)) ;
- rate2 += rate_y;
- distortion2 += distortion;
- rate2 += x->mbmode_cost[x->e_mbd.frame_type][x->e_mbd.mode_info_context->mbmi.mode];
- rate2 += uv_intra_rate;
- rate_uv = uv_intra_rate_tokenonly;
- distortion2 += uv_intra_distortion;
- distortion_uv = uv_intra_distortion;
- }
+ macro_block_yrd(x, &rate_y, &distortion, IF_RTCD(&cpi->rtcd.encodemb)) ;
+ rate2 += rate_y;
+ distortion2 += distortion;
+ rate2 += x->mbmode_cost[x->e_mbd.frame_type][x->e_mbd.mode_info_context->mbmi.mode];
+ rate2 += uv_intra_rate;
+ rate_uv = uv_intra_rate_tokenonly;
+ distortion2 += uv_intra_distortion;
+ distortion_uv = uv_intra_distortion;
break;
case NEWMV:
@@ -2234,22 +2216,28 @@
else if (x->encode_breakout)
{
int sum, sse;
+ int threshold = (xd->block[0].dequant[1]
+ * xd->block[0].dequant[1] >>4);
+
+ if(threshold < x->encode_breakout)
+ threshold = x->encode_breakout;
VARIANCE_INVOKE(&cpi->rtcd.variance, get16x16var)
(x->src.y_buffer, x->src.y_stride,
x->e_mbd.predictor, 16, (unsigned int *)(&sse), &sum);
- if (sse < x->encode_breakout)
+ if (sse < threshold)
{
// Check u and v to make sure skip is ok
int sse2 = 0;
-
- // add dc check
- if (abs(sum) < (cpi->common.Y2dequant[0][0] << 2))
+ /* If theres is no codeable 2nd order dc
+ or a very small uniform pixel change change */
+ if (abs(sum) < (xd->block[24].dequant[0]<<2)||
+ ((sum * sum>>8) > sse && abs(sum) <128))
{
sse2 = VP8_UVSSE(x, IF_RTCD(&cpi->rtcd.variance));
- if (sse2 * 2 < x->encode_breakout)
+ if (sse2 * 2 < threshold)
{
x->skip = 1;
distortion2 = sse + sse2;
@@ -2395,6 +2383,7 @@
if (x->skip)
break;
+
}
// Reduce the activation RD thresholds for the best choice mode
@@ -2465,6 +2454,15 @@
}
+ if(best_mbmode.mode <= B_PRED)
+ {
+ int i;
+ for (i = 0; i < 16; i++)
+ {
+ best_bmodes[i].mv.as_int = 0;
+ }
+ }
+
// macroblock modes
vpx_memcpy(&x->e_mbd.mode_info_context->mbmi, &best_mbmode, sizeof(MB_MODE_INFO));
vpx_memcpy(x->partition_info, &best_partition, sizeof(PARTITION_INFO));
diff --git a/vp8/vp8_cx_iface.c b/vp8/vp8_cx_iface.c
index 903c56c..b23bd95 100644
--- a/vp8/vp8_cx_iface.c
+++ b/vp8/vp8_cx_iface.c
@@ -142,8 +142,8 @@
RANGE_CHECK(cfg, g_timebase.den, 1, 1000000000);
RANGE_CHECK(cfg, g_timebase.num, 1, cfg->g_timebase.den);
RANGE_CHECK_HI(cfg, g_profile, 3);
- RANGE_CHECK_HI(cfg, rc_min_quantizer, 63);
RANGE_CHECK_HI(cfg, rc_max_quantizer, 63);
+ RANGE_CHECK_HI(cfg, rc_min_quantizer, cfg->rc_max_quantizer);
RANGE_CHECK_HI(cfg, g_threads, 64);
#if !(CONFIG_REALTIME_ONLY)
RANGE_CHECK_HI(cfg, g_lag_in_frames, 25);
diff --git a/vpx/internal/vpx_codec_internal.h b/vpx/internal/vpx_codec_internal.h
index dcb451d..9cde426 100644
--- a/vpx/internal/vpx_codec_internal.h
+++ b/vpx/internal/vpx_codec_internal.h
@@ -321,7 +321,7 @@
{
vpx_codec_put_frame_cb_fn_t put_frame;
vpx_codec_put_slice_cb_fn_t put_slice;
- };
+ } u;
void *user_priv;
} vpx_codec_priv_cb_pair_t;
diff --git a/vpx/src/vpx_decoder.c b/vpx/src/vpx_decoder.c
index 27049a5..7b78e5c 100644
--- a/vpx/src/vpx_decoder.c
+++ b/vpx/src/vpx_decoder.c
@@ -160,7 +160,7 @@
res = VPX_CODEC_ERROR;
else
{
- ctx->priv->dec.put_frame_cb.put_frame = cb;
+ ctx->priv->dec.put_frame_cb.u.put_frame = cb;
ctx->priv->dec.put_frame_cb.user_priv = user_priv;
res = VPX_CODEC_OK;
}
@@ -182,7 +182,7 @@
res = VPX_CODEC_ERROR;
else
{
- ctx->priv->dec.put_slice_cb.put_slice = cb;
+ ctx->priv->dec.put_slice_cb.u.put_slice = cb;
ctx->priv->dec.put_slice_cb.user_priv = user_priv;
res = VPX_CODEC_OK;
}
diff --git a/vpx_ports/vpx_timer.h b/vpx_ports/vpx_timer.h
index 37a0c7c..c8335a0 100644
--- a/vpx_ports/vpx_timer.h
+++ b/vpx_ports/vpx_timer.h
@@ -12,6 +12,8 @@
#ifndef VPX_TIMER_H
#define VPX_TIMER_H
+#if CONFIG_OS_SUPPORT
+
#if defined(_WIN32)
/*
* Win32 specific includes
@@ -93,5 +95,27 @@
#endif
}
+#else /* CONFIG_OS_SUPPORT = 0*/
+
+/* Empty timer functions if CONFIG_OS_SUPPORT = 0 */
+#ifndef timersub
+#define timersub(a, b, result)
+#endif
+
+struct vpx_usec_timer
+{
+ void *dummy;
+};
+
+static void
+vpx_usec_timer_start(struct vpx_usec_timer *t) { }
+
+static void
+vpx_usec_timer_mark(struct vpx_usec_timer *t) { }
+
+static long
+vpx_usec_timer_elapsed(struct vpx_usec_timer *t) { return 0; }
+
+#endif /* CONFIG_OS_SUPPORT */
#endif
diff --git a/vpxdec.c b/vpxdec.c
index dd17963..ca9af1e 100644
--- a/vpxdec.c
+++ b/vpxdec.c
@@ -17,14 +17,7 @@
#include <stdarg.h>
#include <string.h>
#include <limits.h>
-#if defined(_WIN32)
-#include <io.h>
-#define snprintf _snprintf
-#define isatty _isatty
-#define fileno _fileno
-#else
-#include <unistd.h>
-#endif
+
#define VPX_CODEC_DISABLE_COMPAT 1
#include "vpx_config.h"
#include "vpx/vpx_decoder.h"
@@ -38,6 +31,17 @@
#include "tools_common.h"
#include "nestegg/include/nestegg/nestegg.h"
+#if CONFIG_OS_SUPPORT
+#if defined(_WIN32)
+#include <io.h>
+#define snprintf _snprintf
+#define isatty _isatty
+#define fileno _fileno
+#else
+#include <unistd.h>
+#endif
+#endif
+
#ifndef PATH_MAX
#define PATH_MAX 256
#endif
@@ -866,7 +870,7 @@
strcmp(fn, "-") ? fn : "stdin");
return EXIT_FAILURE;
}
-
+#if CONFIG_OS_SUPPORT
/* Make sure we don't dump to the terminal, unless forced to with -o - */
if(!outfile_pattern && isatty(fileno(stdout)) && !do_md5 && !noblit)
{
@@ -875,7 +879,7 @@
"override.\n");
return EXIT_FAILURE;
}
-
+#endif
input.infile = infile;
if(file_is_ivf(infile, &fourcc, &width, &height, &fps_den,
&fps_num))
diff --git a/vpxenc.c b/vpxenc.c
index 2c133e7..d0f4c23 100755
--- a/vpxenc.c
+++ b/vpxenc.c
@@ -12,7 +12,7 @@
/* This is a simple program that encodes YV12 files and generates ivf
* files using the new interface.
*/
-#if defined(_WIN32)
+#if defined(_WIN32) || !CONFIG_OS_SUPPORT
#define USE_POSIX_MMAP 0
#else
#define USE_POSIX_MMAP 1
@@ -58,6 +58,14 @@
#define LITERALU64(n) n##LLU
#endif
+/* We should use 32-bit file operations in WebM file format
+ * when building ARM executable file (.axf) with RVCT */
+#if !CONFIG_OS_SUPPORT
+typedef long off_t;
+#define fseeko fseek
+#define ftello ftell
+#endif
+
static const char *exec_name;
static const struct codec_item
