generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

x86: hevc_mc: replace simple leas by adds

Browse Source 
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

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Christophe Gisquet 2014-07-24 17:23:47 +02:00 committed by Michael Niedermayer
parent 7295ee7f51
commit 036f11bdb5
1 changed files with 60 additions and 60 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

120
libavcodec/x86/hevc_mc.asm
View File

@ -548,8 +548,8 @@ 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
@ -563,8 +563,8 @@ 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
@ -601,8 +601,8 @@ 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
@ -616,8 +616,8 @@ 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
@ -651,8 +651,8 @@ 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
@ -669,8 +669,8 @@ 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
@ -692,15 +692,15 @@ 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
@ -726,15 +726,15 @@ 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
@ -751,8 +751,8 @@ 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
@ -765,15 +765,15 @@ 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
@ -791,8 +791,8 @@ 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
@ -829,8 +829,8 @@ 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
@ -847,8 +847,8 @@ 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
@ -886,8 +886,8 @@ 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
@ -905,8 +905,8 @@ 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
@ -928,31 +928,31 @@ 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
@ -997,31 +997,31 @@ 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
@ -1057,8 +1057,8 @@ 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
@ -1071,31 +1071,31 @@ 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
@ -1132,8 +1132,8 @@ 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
@ -1186,7 +1186,7 @@ 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
@ -1239,7 +1239,7 @@ 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