summaryrefslogtreecommitdiffstats
path: root/media/ffvpx/libavcodec/x86/vp8dsp.asm
diff options
context:
space:
mode:
authorMoonchild <mcwerewolf@gmail.com>2018-04-27 19:15:26 +0200
committerGitHub <noreply@github.com>2018-04-27 19:15:26 +0200
commitc75dae3ed21bfa5a8ae46cd83d18329af5bea05a (patch)
treeef657c38feb2368a3c86765778d1f068aa5eb614 /media/ffvpx/libavcodec/x86/vp8dsp.asm
parentc82c6d960a7f19d6595171f9705c43514f20c1ec (diff)
parent6ada4b14e4cfc91f5f1b2556623cab691f3ab813 (diff)
downloadUXP-c75dae3ed21bfa5a8ae46cd83d18329af5bea05a.tar
UXP-c75dae3ed21bfa5a8ae46cd83d18329af5bea05a.tar.gz
UXP-c75dae3ed21bfa5a8ae46cd83d18329af5bea05a.tar.lz
UXP-c75dae3ed21bfa5a8ae46cd83d18329af5bea05a.tar.xz
UXP-c75dae3ed21bfa5a8ae46cd83d18329af5bea05a.zip
Merge pull request #275 from trav90/ffvpx-resync
Resync ffvpx code with 3.4.2-release from upstream
Diffstat (limited to 'media/ffvpx/libavcodec/x86/vp8dsp.asm')
-rw-r--r--media/ffvpx/libavcodec/x86/vp8dsp.asm31
1 files changed, 24 insertions, 7 deletions
diff --git a/media/ffvpx/libavcodec/x86/vp8dsp.asm b/media/ffvpx/libavcodec/x86/vp8dsp.asm
index 538b3f4a9..e303b8029 100644
--- a/media/ffvpx/libavcodec/x86/vp8dsp.asm
+++ b/media/ffvpx/libavcodec/x86/vp8dsp.asm
@@ -156,8 +156,8 @@ SECTION .text
;-------------------------------------------------------------------------------
; subpel MC functions:
;
-; void ff_put_vp8_epel<size>_h<htap>v<vtap>_<opt>(uint8_t *dst, int deststride,
-; uint8_t *src, int srcstride,
+; void ff_put_vp8_epel<size>_h<htap>v<vtap>_<opt>(uint8_t *dst, ptrdiff_t deststride,
+; uint8_t *src, ptrdiff_t srcstride,
; int height, int mx, int my);
;-------------------------------------------------------------------------------
@@ -884,7 +884,7 @@ cglobal put_vp8_pixels16, 5, 5, 2, dst, dststride, src, srcstride, height
REP_RET
;-----------------------------------------------------------------------------
-; void ff_vp8_idct_dc_add_<opt>(uint8_t *dst, int16_t block[16], int stride);
+; void ff_vp8_idct_dc_add_<opt>(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
;-----------------------------------------------------------------------------
%macro ADD_DC 4
@@ -906,6 +906,7 @@ cglobal put_vp8_pixels16, 5, 5, 2, dst, dststride, src, srcstride, height
%4 [dst2q+strideq+%3], m5
%endmacro
+%if ARCH_X86_32
INIT_MMX mmx
cglobal vp8_idct_dc_add, 3, 3, 0, dst, block, stride
; load data
@@ -929,8 +930,9 @@ cglobal vp8_idct_dc_add, 3, 3, 0, dst, block, stride
lea dst2q, [dst1q+strideq*2]
ADD_DC m0, m1, 0, movh
RET
+%endif
-INIT_XMM sse4
+%macro VP8_IDCT_DC_ADD 0
cglobal vp8_idct_dc_add, 3, 3, 6, dst, block, stride
; load data
movd m0, [blockq]
@@ -956,13 +958,28 @@ cglobal vp8_idct_dc_add, 3, 3, 6, dst, block, stride
paddw m4, m0
packuswb m2, m4
movd [dst1q], m2
+%if cpuflag(sse4)
pextrd [dst1q+strideq], m2, 1
pextrd [dst2q], m2, 2
pextrd [dst2q+strideq], m2, 3
+%else
+ psrldq m2, 4
+ movd [dst1q+strideq], m2
+ psrldq m2, 4
+ movd [dst2q], m2
+ psrldq m2, 4
+ movd [dst2q+strideq], m2
+%endif
RET
+%endmacro
+
+INIT_XMM sse2
+VP8_IDCT_DC_ADD
+INIT_XMM sse4
+VP8_IDCT_DC_ADD
;-----------------------------------------------------------------------------
-; void ff_vp8_idct_dc_add4y_<opt>(uint8_t *dst, int16_t block[4][16], int stride);
+; void ff_vp8_idct_dc_add4y_<opt>(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride);
;-----------------------------------------------------------------------------
%if ARCH_X86_32
@@ -1035,7 +1052,7 @@ cglobal vp8_idct_dc_add4y, 3, 3, 6, dst, block, stride
RET
;-----------------------------------------------------------------------------
-; void ff_vp8_idct_dc_add4uv_<opt>(uint8_t *dst, int16_t block[4][16], int stride);
+; void ff_vp8_idct_dc_add4uv_<opt>(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride);
;-----------------------------------------------------------------------------
INIT_MMX mmx
@@ -1077,7 +1094,7 @@ cglobal vp8_idct_dc_add4uv, 3, 3, 0, dst, block, stride
RET
;-----------------------------------------------------------------------------
-; void ff_vp8_idct_add_<opt>(uint8_t *dst, int16_t block[16], int stride);
+; void ff_vp8_idct_add_<opt>(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
;-----------------------------------------------------------------------------
; calculate %1=mul_35468(%1)-mul_20091(%2); %2=mul_20091(%1)+mul_35468(%2)