ffmpeg/FFmpeg
1
0
Fork 0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2024-10-20 15:13:26 +00:00
Code Issues Packages Projects Releases Wiki Activity
2f90d83981
FFmpeg/libavcodec/x86/h264_chromamc_10bit.asm
Lynne bbe95f7353
x86: replace explicit REP_RETs with RETs 
From x86inc:
> On AMD cpus <=K10, an ordinary ret is slow if it immediately follows either
> a branch or a branch target. So switch to a 2-byte form of ret in that case.
> We can automatically detect "follows a branch", but not a branch target.
> (SSSE3 is a sufficient condition to know that your cpu doesn't have this problem.)

x86inc can automatically determine whether to use REP_RET rather than
REP in most of these cases, so impact is minimal. Additionally, a few
REP_RETs were used unnecessary, despite the return being nowhere near a
branch.

The only CPUs affected were AMD K10s, made between 2007 and 2011, 16
years ago and 12 years ago, respectively.

In the future, everyone involved with x86inc should consider dropping
REP_RETs altogether.
2023-02-01 04:23:55 +01:00

270 lines
7.7 KiB
NASM
Raw Blame History

;*****************************************************************************
;* MMX/SSE2/AVX-optimized 10-bit H.264 chroma MC code
;*****************************************************************************
;* Copyright (C) 2005-2011 x264 project
;*
;* Authors: Daniel Kang <daniel.d.kang@gmail.com>
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or
;* modify it under the terms of the GNU Lesser General Public
;* License as published by the Free Software Foundation; either
;* version 2.1 of the License, or (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
;* Lesser General Public License for more details.
;*
;* You should have received a copy of the GNU Lesser General Public
;* License along with FFmpeg; if not, write to the Free Software
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
;******************************************************************************
%include "libavutil/x86/x86util.asm"
SECTION_RODATA
cextern pw_4
cextern pw_8
cextern pw_32
cextern pw_64
SECTION .text
%macro MV0_PIXELS_MC8 0
lea r4, [r2*3 ]
lea r5, [r2*4 ]
.next4rows:
movu m0, [r1 ]
movu m1, [r1+r2 ]
CHROMAMC_AVG m0, [r0 ]
CHROMAMC_AVG m1, [r0+r2 ]
mova [r0 ], m0
mova [r0+r2 ], m1
movu m0, [r1+r2*2]
movu m1, [r1+r4 ]
CHROMAMC_AVG m0, [r0+r2*2]
CHROMAMC_AVG m1, [r0+r4 ]
mova [r0+r2*2], m0
mova [r0+r4 ], m1
add r1, r5
add r0, r5
sub r3d, 4
jne .next4rows
%endmacro
;-----------------------------------------------------------------------------
; void ff_put/avg_h264_chroma_mc8(pixel *dst, const pixel *src, ptrdiff_t stride,
; int h, int mx, int my)
;-----------------------------------------------------------------------------
%macro CHROMA_MC8 1
cglobal %1_h264_chroma_mc8_10, 6,7,8
mov r6d, r5d
or r6d, r4d
jne .at_least_one_non_zero
; mx == 0 AND my == 0 - no filter needed
MV0_PIXELS_MC8
RET
.at_least_one_non_zero:
mov r6d, 2
test r5d, r5d
je .x_interpolation
mov r6, r2 ; dxy = x ? 1 : stride
test r4d, r4d
jne .xy_interpolation
.x_interpolation:
; mx == 0 XOR my == 0 - 1 dimensional filter only
or r4d, r5d ; x + y
movd m5, r4d
mova m4, [pw_8]
mova m6, [pw_4] ; mm6 = rnd >> 3
SPLATW m5, m5 ; mm5 = B = x
psubw m4, m5 ; mm4 = A = 8-x
.next1drow:
movu m0, [r1 ] ; mm0 = src[0..7]
movu m2, [r1+r6] ; mm2 = src[1..8]
pmullw m0, m4 ; mm0 = A * src[0..7]
pmullw m2, m5 ; mm2 = B * src[1..8]
paddw m0, m6
paddw m0, m2
psrlw m0, 3
CHROMAMC_AVG m0, [r0]
mova [r0], m0 ; dst[0..7] = (A * src[0..7] + B * src[1..8] + (rnd >> 3)) >> 3
add r0, r2
add r1, r2
dec r3d
jne .next1drow
RET
.xy_interpolation: ; general case, bilinear
movd m4, r4m ; x
movd m6, r5m ; y
SPLATW m4, m4 ; mm4 = x words
SPLATW m6, m6 ; mm6 = y words
psllw m5, m4, 3 ; mm5 = 8x
pmullw m4, m6 ; mm4 = x * y
psllw m6, 3 ; mm6 = 8y
paddw m1, m5, m6 ; mm7 = 8x+8y
mova m7, m4 ; DD = x * y
psubw m5, m4 ; mm5 = B = 8x - xy
psubw m6, m4 ; mm6 = C = 8y - xy
paddw m4, [pw_64]
psubw m4, m1 ; mm4 = A = xy - (8x+8y) + 64
movu m0, [r1 ] ; mm0 = src[0..7]
movu m1, [r1+2] ; mm1 = src[1..8]
.next2drow:
add r1, r2
pmullw m2, m0, m4
pmullw m1, m5
paddw m2, m1 ; mm2 = A * src[0..7] + B * src[1..8]
movu m0, [r1]
movu m1, [r1+2]
pmullw m3, m0, m6
paddw m2, m3 ; mm2 += C * src[0..7+strde]
pmullw m3, m1, m7
paddw m2, m3 ; mm2 += D * src[1..8+strde]
paddw m2, [pw_32]
psrlw m2, 6
CHROMAMC_AVG m2, [r0]
mova [r0], m2 ; dst[0..7] = (mm2 + 32) >> 6
add r0, r2
dec r3d
jne .next2drow
RET
%endmacro
;-----------------------------------------------------------------------------
; void ff_put/avg_h264_chroma_mc4(pixel *dst, pixel *src, ptrdiff_t stride,
; int h, int mx, int my)
;-----------------------------------------------------------------------------
;TODO: xmm mc4
%macro MC4_OP 2
movq %1, [r1 ]
movq m1, [r1+2]
add r1, r2
pmullw %1, m4
pmullw m1, m2
paddw m1, %1
mova %1, m1
pmullw %2, m5
pmullw m1, m3
paddw %2, [pw_32]
paddw m1, %2
psrlw m1, 6
CHROMAMC_AVG m1, %2, [r0]
movq [r0], m1
add r0, r2
%endmacro
%macro CHROMA_MC4 1
cglobal %1_h264_chroma_mc4_10, 6,6,7
movd m2, r4m ; x
movd m3, r5m ; y
mova m4, [pw_8]
mova m5, m4
SPLATW m2, m2
SPLATW m3, m3
psubw m4, m2
psubw m5, m3
movq m0, [r1 ]
movq m6, [r1+2]
add r1, r2
pmullw m0, m4
pmullw m6, m2
paddw m6, m0
.next2rows:
MC4_OP m0, m6
MC4_OP m6, m0
sub r3d, 2
jnz .next2rows
RET
%endmacro
;-----------------------------------------------------------------------------
; void ff_put/avg_h264_chroma_mc2(pixel *dst, const pixel *src, ptrdiff_t stride,
; int h, int mx, int my)
;-----------------------------------------------------------------------------
%macro CHROMA_MC2 1
cglobal %1_h264_chroma_mc2_10, 6,7
mov r6d, r4d
shl r4d, 16
sub r4d, r6d
add r4d, 8
imul r5d, r4d ; x*y<<16 | y*(8-x)
shl r4d, 3
sub r4d, r5d ; x*(8-y)<<16 | (8-x)*(8-y)
movd m5, r4d
movd m6, r5d
punpckldq m5, m5 ; mm5 = {A,B,A,B}
punpckldq m6, m6 ; mm6 = {C,D,C,D}
pxor m7, m7
pshufw m2, [r1], 0x94 ; mm0 = src[0,1,1,2]
.nextrow:
add r1, r2
movq m1, m2
pmaddwd m1, m5 ; mm1 = A * src[0,1] + B * src[1,2]
pshufw m0, [r1], 0x94 ; mm0 = src[0,1,1,2]
movq m2, m0
pmaddwd m0, m6
paddw m1, [pw_32]
paddw m1, m0 ; mm1 += C * src[0,1] + D * src[1,2]
psrlw m1, 6
packssdw m1, m7
CHROMAMC_AVG m1, m3, [r0]
movd [r0], m1
add r0, r2
dec r3d
jnz .nextrow
RET
%endmacro
%macro NOTHING 2-3
%endmacro
%macro AVG 2-3
%if %0==3
movq %2, %3
%endif
pavgw %1, %2
%endmacro
%define CHROMAMC_AVG NOTHING
INIT_XMM sse2
CHROMA_MC8 put
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
CHROMA_MC8 put
%endif
INIT_MMX mmxext
CHROMA_MC4 put
CHROMA_MC2 put
%define CHROMAMC_AVG AVG
INIT_XMM sse2
CHROMA_MC8 avg
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
CHROMA_MC8 avg
%endif
INIT_MMX mmxext
CHROMA_MC4 avg
CHROMA_MC2 avg
Reference in New Issue View Git Blame Copy Permalink