Hi,
the impact is relatively important (3 to 25 cycles, ie up to 2%) so I
also include bench.patch if anybody wants to confirm the timings.
Although I thought openhevc's MC code was faster, the benchmarked
functions are faster in ffmpeg. I didn't investigate if there are any
MC functions much slower in ffmpeg or if there's something else
occurring.
--
Christophe
From 36dbc9b67269579e23345ec225ffa270d472b94e Mon Sep 17 00:00:00 2001
From: Christophe Gisquet <christophe.gisq...@gmail.com>
Date: Thu, 24 Jul 2014 17:23:47 +0200
Subject: [PATCH 10/10] x86: hevc_mc: replace simple leas by adds
lea is detrimental for those simple cases. No impact overall to
the change though.
Before:
15017 decicycles in q, 1016152 runs, 32424 skips
15382 decicycles in q_bi, 1013673 runs, 34903 skips
3713 decicycles in e, 2074534 runs, 22618 skips
3901 decicycles in e_bi, 2065509 runs, 31643 skips
7852 decicycles in q_uni, 520165 runs, 4123 skips
2398 decicycles in e_uni, 1043339 runs, 5237 skips
After:
14898 decicycles in q, 1016295 runs, 32281 skips
15119 decicycles in q_bi, 1015392 runs, 33184 skips
3682 decicycles in e, 2073224 runs, 23928 skips
3720 decicycles in e_bi, 2065043 runs, 32109 skips
7643 decicycles in q_uni, 520280 runs, 4008 skips
2363 decicycles in e_uni, 1043780 runs, 4796 skips
---
libavcodec/x86/hevc_mc.asm | 120 ++++++++++++++++++++++-----------------------
1 file changed, 60 insertions(+), 60 deletions(-)
diff --git a/libavcodec/x86/hevc_mc.asm b/libavcodec/x86/hevc_mc.asm
index 5cf37d0..30edc52 100644
--- a/libavcodec/x86/hevc_mc.asm
+++ b/libavcodec/x86/hevc_mc.asm
@@ -520,8 +520,8 @@ cglobal hevc_put_hevc_uni_pel_pixels%1_%2, 5, 5, 3, dst, dststride, src, srcstri
.loop
SIMPLE_LOAD %1, %2, srcq, m0
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
RET
@@ -535,8 +535,8 @@ cglobal hevc_put_hevc_bi_pel_pixels%1_%2, 7, 7, 6, dst, dststride, src, srcstrid
MC_PIXEL_COMPUTE %1, %2
BI_COMPUTE %1, %2, m0, m1, m3, m4, m5
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
lea src2q, [src2q+2*src2strideq] ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
@@ -573,8 +573,8 @@ cglobal hevc_put_hevc_uni_epel_h%1_%2, 6, 7, 7, dst, dststride, src, srcstride,
EPEL_COMPUTE %2, %1, m4, m5
UNI_COMPUTE %1, %2, m0, m1, m6
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
RET
@@ -588,8 +588,8 @@ cglobal hevc_put_hevc_bi_epel_h%1_%2, 8, 9, 7, dst, dststride, src, srcstride, s
SIMPLE_BILOAD %1, src2q, m2, m3
BI_COMPUTE %1, %2, m0, m1, m2, m3, m6
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
lea src2q, [src2q+2*src2strideq] ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
@@ -623,8 +623,8 @@ cglobal hevc_put_hevc_uni_epel_v%1_%2, 7, 8, 7, dst, dststride, src, srcstride,
EPEL_COMPUTE %2, %1, m4, m5
UNI_COMPUTE %1, %2, m0, m1, m6
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
RET
@@ -641,8 +641,8 @@ cglobal hevc_put_hevc_bi_epel_v%1_%2, 9, 10, 7, dst, dststride, src, srcstride,
SIMPLE_BILOAD %1, src2q, m2, m3
BI_COMPUTE %1, %2, m0, m1, m2, m3, m6
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
lea src2q, [src2q+2*src2strideq] ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
@@ -664,15 +664,15 @@ cglobal hevc_put_hevc_epel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstride, h
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m4, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m5, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m6, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
.loop
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
@@ -698,15 +698,15 @@ cglobal hevc_put_hevc_uni_epel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstrid
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m4, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m5, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m6, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
.loop
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
@@ -723,8 +723,8 @@ cglobal hevc_put_hevc_uni_epel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstrid
movdqa m4, m5
movdqa m5, m6
movdqa m6, m7
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
RET
@@ -737,15 +737,15 @@ cglobal hevc_put_hevc_bi_epel_hv%1_%2, 9, 11, 16, dst, dststride, src, srcstride
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m4, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m5, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
SWAP m6, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
.loop
EPEL_LOAD %2, srcq-%%stride, %%stride, %1
EPEL_COMPUTE %2, %1, m14, m15
@@ -763,8 +763,8 @@ cglobal hevc_put_hevc_bi_epel_hv%1_%2, 9, 11, 16, dst, dststride, src, srcstride
movdqa m4, m5
movdqa m5, m6
movdqa m6, m7
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
lea src2q, [src2q+2*src2strideq] ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
@@ -801,8 +801,8 @@ cglobal hevc_put_hevc_uni_qpel_h%1_%2, 6, 7, 15 , dst, dststride, src, srcstride
%endif
UNI_COMPUTE %1, %2, m0, m1, m9
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
RET
@@ -819,8 +819,8 @@ cglobal hevc_put_hevc_bi_qpel_h%1_%2, 8, 9, 16 , dst, dststride, src, srcstride,
SIMPLE_BILOAD %1, src2q, m10, m11
BI_COMPUTE %1, %2, m0, m1, m10, m11, m9
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
lea src2q, [src2q+2*src2strideq] ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
@@ -858,8 +858,8 @@ cglobal hevc_put_hevc_uni_qpel_v%1_%2, 7, 14, 15 , dst, dststride, src, srcstrid
%endif
UNI_COMPUTE %1, %2, m0, m1, m9
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
RET
@@ -877,8 +877,8 @@ cglobal hevc_put_hevc_bi_qpel_v%1_%2, 9, 14, 16 , dst, dststride, src, srcstride
%endif
BI_COMPUTE %1, %2, m0, m1, m10, m11, m9
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
lea src2q, [src2q+2*src2strideq] ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
@@ -900,31 +900,31 @@ cglobal hevc_put_hevc_qpel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstride, h
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m8, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m9, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m10, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m11, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m12, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m13, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m14, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
.loop
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
@@ -969,31 +969,31 @@ cglobal hevc_put_hevc_uni_qpel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstrid
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m8, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m9, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m10, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m11, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m12, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m13, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m14, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
.loop
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
@@ -1029,8 +1029,8 @@ cglobal hevc_put_hevc_uni_qpel_hv%1_%2, 7, 9, 12 , dst, dststride, src, srcstrid
movdqa m13, m14
movdqa m14, m15
%endif
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
RET
@@ -1043,31 +1043,31 @@ cglobal hevc_put_hevc_bi_qpel_hv%1_%2, 9, 11, 16, dst, dststride, src, srcstride
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m8, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m9, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m10, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m11, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m12, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m13, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
SWAP m14, m0
- lea srcq, [srcq + srcstrideq]
+ add srcq, srcstrideq
.loop
QPEL_H_LOAD %2, srcq, %1, 15
QPEL_HV_COMPUTE %1, %2, mx, ackssdw
@@ -1104,8 +1104,8 @@ cglobal hevc_put_hevc_bi_qpel_hv%1_%2, 9, 11, 16, dst, dststride, src, srcstride
movdqa m13, m14
movdqa m14, m15
%endif
- lea dstq, [dstq+dststrideq] ; dst += dststride
- lea srcq, [srcq+srcstrideq] ; src += srcstride
+ add dstq, dststrideq ; dst += dststride
+ add srcq, srcstrideq ; src += srcstride
lea src2q, [src2q+2*src2strideq] ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
@@ -1158,7 +1158,7 @@ cglobal hevc_put_hevc_uni_w%1_%2, 6, 6, 7, dst, dststride, src, srcstride, heigh
pminsw m0, [max_pixels_%2]
%endif
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
+ add dstq, dststrideq ; dst += dststride
lea srcq, [srcq+2*srcstrideq] ; src += srcstride
dec heightd ; cmp height
jnz .loop ; height loop
@@ -1211,7 +1211,7 @@ cglobal hevc_put_hevc_bi_w%1_%2, 6, 7, 10, dst, dststride, src, srcstride, src2,
pminsw m0, [max_pixels_%2]
%endif
PEL_%2STORE%1 dstq, m0, m1
- lea dstq, [dstq+dststrideq] ; dst += dststride
+ add dstq, dststrideq ; dst += dststride
lea srcq, [srcq+2*srcstrideq] ; src += srcstride
lea src2q, [src2q+2*src2strideq] ; src2 += srcstride
dec r6d ; cmp height
--
1.9.2.msysgit.0
diff --git a/libavcodec/hevc.c b/libavcodec/hevc.c
index 913385a..881562c 100644
--- a/libavcodec/hevc.c
+++ b/libavcodec/hevc.c
@@ -1339,13 +1339,18 @@ static void luma_mc_uni(HEVCContext *s, uint8_t *dst, ptrdiff_t dststride,
srcstride = edge_emu_stride;
}
- if (!weight_flag)
+ if (!weight_flag) {
+ START_TIMER
s->hevcdsp.put_hevc_qpel_uni[idx][!!my][!!mx](dst, dststride, src, srcstride,
block_h, mx, my, block_w);
- else
+ STOP_TIMER("q_uni")
+ } else {
+ START_TIMER
s->hevcdsp.put_hevc_qpel_uni_w[idx][!!my][!!mx](dst, dststride, src, srcstride,
block_h, s->sh.luma_log2_weight_denom,
luma_weight, luma_offset, mx, my, block_w);
+ STOP_TIMER("q_uni_w")
+ }
}
/**
@@ -1423,12 +1428,20 @@ static void luma_mc_uni(HEVCContext *s, uint8_t *dst, ptrdiff_t dststride,
src1stride = edge_emu_stride;
}
+ {
+ START_TIMER
s->hevcdsp.put_hevc_qpel[idx][!!my0][!!mx0](tmp, MAX_PB_SIZE, src0, src0stride,
block_h, mx0, my0, block_w);
- if (!weight_flag)
+ STOP_TIMER("q")
+ }
+ if (!weight_flag) {
+ START_TIMER
s->hevcdsp.put_hevc_qpel_bi[idx][!!my1][!!mx1](dst, dststride, src1, src1stride, tmp, MAX_PB_SIZE,
block_h, mx1, my1, block_w);
- else
+ STOP_TIMER("q_bi")
+ }
+ else {
+ START_TIMER
s->hevcdsp.put_hevc_qpel_bi_w[idx][!!my1][!!mx1](dst, dststride, src1, src1stride, tmp, MAX_PB_SIZE,
block_h, s->sh.luma_log2_weight_denom,
s->sh.luma_weight_l0[current_mv->ref_idx[0]],
@@ -1436,7 +1449,8 @@ static void luma_mc_uni(HEVCContext *s, uint8_t *dst, ptrdiff_t dststride,
s->sh.luma_offset_l0[current_mv->ref_idx[0]],
s->sh.luma_offset_l1[current_mv->ref_idx[1]],
mx1, my1, block_w);
-
+ STOP_TIMER("q_bi_w")
+ }
}
/**
@@ -1495,13 +1509,19 @@ static void chroma_mc_uni(HEVCContext *s, uint8_t *dst0,
src0 = lc->edge_emu_buffer + buf_offset0;
srcstride = edge_emu_stride;
}
- if (!weight_flag)
+ if (!weight_flag) {
+ START_TIMER
s->hevcdsp.put_hevc_epel_uni[idx][!!my][!!mx](dst0, dststride, src0, srcstride,
block_h, _mx, _my, block_w);
- else
+ STOP_TIMER("e_uni")
+ }
+ else {
+ START_TIMER
s->hevcdsp.put_hevc_epel_uni_w[idx][!!my][!!mx](dst0, dststride, src0, srcstride,
block_h, s->sh.chroma_log2_weight_denom,
chroma_weight, chroma_offset, _mx, _my, block_w);
+ STOP_TIMER("e_uni_w")
+ }
}
/**
@@ -1595,13 +1615,21 @@ static void chroma_mc_bi(HEVCContext *s, uint8_t *dst0, ptrdiff_t dststride, AVF
src2stride = edge_emu_stride;
}
+ {
+ START_TIMER
s->hevcdsp.put_hevc_epel[idx][!!my0][!!mx0](tmp, tmpstride, src1, src1stride,
block_h, _mx0, _my0, block_w);
- if (!weight_flag)
+ STOP_TIMER("e")
+ }
+ if (!weight_flag) {
+ START_TIMER
s->hevcdsp.put_hevc_epel_bi[idx][!!my1][!!mx1](dst0, s->frame->linesize[cidx+1],
src2, src2stride, tmp, tmpstride,
block_h, _mx1, _my1, block_w);
- else
+ STOP_TIMER("e_bi")
+ }
+ else {
+ START_TIMER
s->hevcdsp.put_hevc_epel_bi_w[idx][!!my1][!!mx1](dst0, s->frame->linesize[cidx+1],
src2, src2stride, tmp, tmpstride,
block_h,
@@ -1611,6 +1639,8 @@ static void chroma_mc_bi(HEVCContext *s, uint8_t *dst0, ptrdiff_t dststride, AVF
s->sh.chroma_offset_l0[current_mv->ref_idx[0]][cidx],
s->sh.chroma_offset_l1[current_mv->ref_idx[1]][cidx],
_mx1, _my1, block_w);
+ STOP_TIMER("e_bi_w")
+ }
}
static void hevc_await_progress(HEVCContext *s, HEVCFrame *ref,
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
