This is not a clean equivalent of the openhevc patch, as we don't have
the same history on hevc_mc.
--
Christophe
From 0eedfb7d5902cc388d018111c670cc31ad22db60 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Micka=C3=ABl=20Raulet?= <mrau...@insa-rennes.fr>
Date: Thu, 31 Jul 2014 19:26:57 +0200
Subject: [PATCH] x86: hevc_mc: remove non necessary moves
Signed-off-by: Christophe Gisquet <christophe.gisq...@gmail.com>
---
libavcodec/x86/hevc_mc.asm | 48 +++++++++++++++++-----------------------------
1 file changed, 18 insertions(+), 30 deletions(-)
diff --git a/libavcodec/x86/hevc_mc.asm b/libavcodec/x86/hevc_mc.asm
index 42fe65a..99f8c0a 100644
--- a/libavcodec/x86/hevc_mc.asm
+++ b/libavcodec/x86/hevc_mc.asm
@@ -81,20 +81,6 @@ QPEL_TABLE 12, 4, w, sse4
%if ARCH_X86_64
-%macro SIMPLE_BILOAD 4 ;width, tab, r1, r2
-%if %1 <= 4
- movq %3, [%2] ; load data from source2
-%elif %1 <= 8
- movdqa %3, [%2] ; load data from source2
-%elif %1 <= 12
- movdqa %3, [%2] ; load data from source2
- movq %4, [%2+16] ; load data from source2
-%else
- movdqa %3, [%2] ; load data from source2
- movdqa %4, [%2+16] ; load data from source2
-%endif
-%endmacro
-
%macro SIMPLE_LOAD 4 ;width, bitd, tab, r1
%if %1 == 2 || (%2 == 8 && %1 <= 4)
movd %4, [%3] ; load data from source
@@ -505,11 +491,20 @@ QPEL_TABLE 12, 4, w, sse4
%endif
%endmacro
-%macro BI_COMPUTE 7 ; width, bitd, src1l, src1h, scr2l, scr2h, pw
+%macro SIMPLE_BILOAD 5 ;width, tab, r1, r2
+%if %1 <= 4
+ movq %5, [%2] ; load data from source2
paddsw %3, %5
+%else
+ paddsw %3, [%2]
%if %1 > 8
- paddsw %4, %6
+ paddsw %4, [%2+16] ; load data from source2
+%endif
%endif
+%endmacro
+
+%macro BI_COMPUTE 7 ; width, bitd, src1l, src1h, scr2l, scr2h, pw
+ SIMPLE_BILOAD %1, %6, %3, %4, %5
UNI_COMPUTE %1, %2, %3, %4, %7
%endmacro
@@ -562,9 +557,8 @@ cglobal hevc_put_hevc_bi_pel_pixels%1_%2, 6, 6, 6, dst, dststride, src, srcstrid
movdqa m5, [pw_bi_%2]
.loop
SIMPLE_LOAD %1, %2, srcq, m0
- SIMPLE_BILOAD %1, src2q, m3, m4
MC_PIXEL_COMPUTE %1, %2
- BI_COMPUTE %1, %2, m0, m1, m3, m4, m5
+ BI_COMPUTE %1, %2, m0, m1, m3, src2q, m5
PEL_%2STORE%1 dstq, m0, m1
add dstq, dststrideq ; dst += dststride
add srcq, srcstrideq ; src += srcstride
@@ -616,8 +610,7 @@ cglobal hevc_put_hevc_bi_epel_h%1_%2, 7, 8, 11, dst, dststride, src, srcstride,
.loop
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m4, m5
- SIMPLE_BILOAD %1, src2q, m2, m3
- BI_COMPUTE %1, %2, m0, m1, m2, m3, m6
+ BI_COMPUTE %1, %2, m0, m1, m2, src2q, m6
PEL_%2STORE%1 dstq, m0, m1
add dstq, dststrideq ; dst += dststride
add srcq, srcstrideq ; src += srcstride
@@ -669,8 +662,7 @@ cglobal hevc_put_hevc_bi_epel_v%1_%2, 8, 9, 11, dst, dststride, src, srcstride,
.loop
EPEL_LOAD %2, srcq, srcstride, %1
EPEL_COMPUTE %2, %1, m4, m5
- SIMPLE_BILOAD %1, src2q, m2, m3
- BI_COMPUTE %1, %2, m0, m1, m2, m3, m6
+ BI_COMPUTE %1, %2, m0, m1, m2, src2q, m6
PEL_%2STORE%1 dstq, m0, m1
add dstq, dststrideq ; dst += dststride
add srcq, srcstrideq ; src += srcstride
@@ -788,8 +780,7 @@ cglobal hevc_put_hevc_bi_epel_hv%1_%2, 8, 10, 16, dst, dststride, src, srcstride
punpckhwd m3, m6, m7
%endif
EPEL_COMPUTE 14, %1, m12, m13
- SIMPLE_BILOAD %1, src2q, m8, m9
- BI_COMPUTE %1, %2, m0, m1, m8, m9, [pw_bi_%2]
+ BI_COMPUTE %1, %2, m0, m1, m8, src2q, [pw_bi_%2]
PEL_%2STORE%1 dstq, m0, m1
movdqa m4, m5
movdqa m5, m6
@@ -847,8 +838,7 @@ cglobal hevc_put_hevc_bi_qpel_h%1_%2, 7, 8, 16 , dst, dststride, src, srcstride,
%if %2 > 8
packssdw m0, m1
%endif
- SIMPLE_BILOAD %1, src2q, m10, m11
- BI_COMPUTE %1, %2, m0, m1, m10, m11, m9
+ BI_COMPUTE %1, %2, m0, m1, m10, src2q, m9
PEL_%2STORE%1 dstq, m0, m1
add dstq, dststrideq ; dst += dststride
add srcq, srcstrideq ; src += srcstride
@@ -900,13 +890,12 @@ cglobal hevc_put_hevc_bi_qpel_v%1_%2, 8, 10, 16, dst, dststride, src, srcstride,
lea r3srcq, [srcstrideq*3]
QPEL_FILTER %2, my
.loop
- SIMPLE_BILOAD %1, src2q, m10, m11
QPEL_V_LOAD %2, srcq, srcstride, %1, r9
QPEL_COMPUTE %1, %2
%if %2 > 8
packssdw m0, m1
%endif
- BI_COMPUTE %1, %2, m0, m1, m10, m11, m9
+ BI_COMPUTE %1, %2, m0, m1, m10, src2q, m9
PEL_%2STORE%1 dstq, m0, m1
add dstq, dststrideq ; dst += dststride
add srcq, srcstrideq ; src += srcstride
@@ -1120,8 +1109,7 @@ cglobal hevc_put_hevc_bi_qpel_hv%1_%2, 8, 10, 16, dst, dststride, src, srcstride
punpckhwd m7, m14, m15
%endif
QPEL_HV_COMPUTE %1, 14, my, ackssdw
- SIMPLE_BILOAD %1, src2q, m8, m9 ;m9 not used in this case
- BI_COMPUTE %1, %2, m0, m1, m8, m9, [pw_bi_%2]
+ BI_COMPUTE %1, %2, m0, m1, m8, src2q, [pw_bi_%2]
PEL_%2STORE%1 dstq, m0, m1
%if %1 <= 4
--
1.9.2.msysgit.0
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
