har0ke/FFmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
FFmpeg/libavcodec/x86/h264_idct_10bit.asm
James Darnley 13d71c28cc avcodec/h264: sse2 and avx 4:2:2 idct add8 10-bit functions 
Yorkfield:
 - sse2:
   - complex: 4.13x faster (1514 vs. 367 cycles)
   - simple:  4.38x faster (1836 vs. 419 cycles)

Skylake:
 - sse2:
   - complex: 3.61x faster ( 936 vs. 260 cycles)
   - simple:  3.97x faster (1126 vs. 284 cycles)
 - avx (versus sse2):
   - complex: 1.07x faster (260 vs. 244 cycles)
   - simple:  1.03x faster (284 vs. 274 cycles)
2016-11-30 22:58:28 +01:00

658 lines
16 KiB
NASM
Raw Blame History

;*****************************************************************************
;* MMX/SSE2/AVX-optimized 10-bit H.264 iDCT 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 .text
cextern pw_1023
%define pw_pixel_max pw_1023
cextern pd_32
;-----------------------------------------------------------------------------
; void ff_h264_idct_add_10(pixel *dst, int16_t *block, int stride)
;-----------------------------------------------------------------------------
%macro STORE_DIFFx2 6
psrad %1, 6
psrad %2, 6
packssdw %1, %2
movq %3, [%5]
movhps %3, [%5+%6]
paddsw %1, %3
CLIPW %1, %4, [pw_pixel_max]
movq [%5], %1
movhps [%5+%6], %1
%endmacro
%macro STORE_DIFF16 5
psrad %1, 6
psrad %2, 6
packssdw %1, %2
paddsw %1, [%5]
CLIPW %1, %3, %4
mova [%5], %1
%endmacro
;dst, in, stride
%macro IDCT4_ADD_10 3
mova m0, [%2+ 0]
mova m1, [%2+16]
mova m2, [%2+32]
mova m3, [%2+48]
IDCT4_1D d,0,1,2,3,4,5
TRANSPOSE4x4D 0,1,2,3,4
paddd m0, [pd_32]
IDCT4_1D d,0,1,2,3,4,5
pxor m5, m5
mova [%2+ 0], m5
mova [%2+16], m5
mova [%2+32], m5
mova [%2+48], m5
STORE_DIFFx2 m0, m1, m4, m5, %1, %3
lea %1, [%1+%3*2]
STORE_DIFFx2 m2, m3, m4, m5, %1, %3
%endmacro
%macro IDCT_ADD_10 0
cglobal h264_idct_add_10, 3,3
movsxdifnidn r2, r2d
IDCT4_ADD_10 r0, r1, r2
RET
%endmacro
INIT_XMM sse2
IDCT_ADD_10
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
IDCT_ADD_10
%endif
;-----------------------------------------------------------------------------
; void ff_h264_idct_add16_10(pixel *dst, const int *block_offset,
; int16_t *block, int stride,
; const uint8_t nnzc[6*8])
;-----------------------------------------------------------------------------
;;;;;;; NO FATE SAMPLES TRIGGER THIS
%macro ADD4x4IDCT 0
add4x4_idct %+ SUFFIX:
add r5, r0
mova m0, [r2+ 0]
mova m1, [r2+16]
mova m2, [r2+32]
mova m3, [r2+48]
IDCT4_1D d,0,1,2,3,4,5
TRANSPOSE4x4D 0,1,2,3,4
paddd m0, [pd_32]
IDCT4_1D d,0,1,2,3,4,5
pxor m5, m5
mova [r2+ 0], m5
mova [r2+16], m5
mova [r2+32], m5
mova [r2+48], m5
STORE_DIFFx2 m0, m1, m4, m5, r5, r3
lea r5, [r5+r3*2]
STORE_DIFFx2 m2, m3, m4, m5, r5, r3
ret
%endmacro
INIT_XMM sse2
ALIGN 16
ADD4x4IDCT
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
ALIGN 16
ADD4x4IDCT
%endif
%macro ADD16_OP 2
cmp byte [r4+%2], 0
jz .skipblock%1
mov r5d, [r1+%1*4]
call add4x4_idct %+ SUFFIX
.skipblock%1:
%if %1<15
add r2, 64
%endif
%endmacro
%macro IDCT_ADD16_10 0
cglobal h264_idct_add16_10, 5,6
movsxdifnidn r3, r3d
ADD16_OP 0, 4+1*8
ADD16_OP 1, 5+1*8
ADD16_OP 2, 4+2*8
ADD16_OP 3, 5+2*8
ADD16_OP 4, 6+1*8
ADD16_OP 5, 7+1*8
ADD16_OP 6, 6+2*8
ADD16_OP 7, 7+2*8
ADD16_OP 8, 4+3*8
ADD16_OP 9, 5+3*8
ADD16_OP 10, 4+4*8
ADD16_OP 11, 5+4*8
ADD16_OP 12, 6+3*8
ADD16_OP 13, 7+3*8
ADD16_OP 14, 6+4*8
ADD16_OP 15, 7+4*8
REP_RET
%endmacro
INIT_XMM sse2
IDCT_ADD16_10
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
IDCT_ADD16_10
%endif
;-----------------------------------------------------------------------------
; void ff_h264_idct_dc_add_10(pixel *dst, int16_t *block, int stride)
;-----------------------------------------------------------------------------
%macro IDCT_DC_ADD_OP_10 3
pxor m5, m5
%if avx_enabled
paddw m1, m0, [%1+0 ]
paddw m2, m0, [%1+%2 ]
paddw m3, m0, [%1+%2*2]
paddw m4, m0, [%1+%3 ]
%else
mova m1, [%1+0 ]
mova m2, [%1+%2 ]
mova m3, [%1+%2*2]
mova m4, [%1+%3 ]
paddw m1, m0
paddw m2, m0
paddw m3, m0
paddw m4, m0
%endif
CLIPW m1, m5, m6
CLIPW m2, m5, m6
CLIPW m3, m5, m6
CLIPW m4, m5, m6
mova [%1+0 ], m1
mova [%1+%2 ], m2
mova [%1+%2*2], m3
mova [%1+%3 ], m4
%endmacro
INIT_MMX mmxext
cglobal h264_idct_dc_add_10,3,3
movsxdifnidn r2, r2d
movd m0, [r1]
mov dword [r1], 0
paddd m0, [pd_32]
psrad m0, 6
lea r1, [r2*3]
pshufw m0, m0, 0
mova m6, [pw_pixel_max]
IDCT_DC_ADD_OP_10 r0, r2, r1
RET
;-----------------------------------------------------------------------------
; void ff_h264_idct8_dc_add_10(pixel *dst, int16_t *block, int stride)
;-----------------------------------------------------------------------------
%macro IDCT8_DC_ADD 0
cglobal h264_idct8_dc_add_10,3,4,7
movsxdifnidn r2, r2d
movd m0, [r1]
mov dword[r1], 0
paddd m0, [pd_32]
psrad m0, 6
lea r1, [r2*3]
SPLATW m0, m0, 0
mova m6, [pw_pixel_max]
IDCT_DC_ADD_OP_10 r0, r2, r1
lea r0, [r0+r2*4]
IDCT_DC_ADD_OP_10 r0, r2, r1
RET
%endmacro
INIT_XMM sse2
IDCT8_DC_ADD
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
IDCT8_DC_ADD
%endif
;-----------------------------------------------------------------------------
; void ff_h264_idct_add16intra_10(pixel *dst, const int *block_offset,
; int16_t *block, int stride,
; const uint8_t nnzc[6*8])
;-----------------------------------------------------------------------------
%macro AC 1
.ac%1:
mov r5d, [r1+(%1+0)*4]
call add4x4_idct %+ SUFFIX
mov r5d, [r1+(%1+1)*4]
add r2, 64
call add4x4_idct %+ SUFFIX
add r2, 64
jmp .skipadd%1
%endmacro
%assign last_block 16
%macro ADD16_OP_INTRA 2
cmp word [r4+%2], 0
jnz .ac%1
mov r5d, [r2+ 0]
or r5d, [r2+64]
jz .skipblock%1
mov r5d, [r1+(%1+0)*4]
call idct_dc_add %+ SUFFIX
.skipblock%1:
%if %1<last_block-2
add r2, 128
%endif
.skipadd%1:
%endmacro
%macro IDCT_ADD16INTRA_10 0
idct_dc_add %+ SUFFIX:
add r5, r0
movq m0, [r2+ 0]
movhps m0, [r2+64]
mov dword [r2+ 0], 0
mov dword [r2+64], 0
paddd m0, [pd_32]
psrad m0, 6
pshufhw m0, m0, 0
pshuflw m0, m0, 0
lea r6, [r3*3]
mova m6, [pw_pixel_max]
IDCT_DC_ADD_OP_10 r5, r3, r6
ret
cglobal h264_idct_add16intra_10,5,7,8
movsxdifnidn r3, r3d
ADD16_OP_INTRA 0, 4+1*8
ADD16_OP_INTRA 2, 4+2*8
ADD16_OP_INTRA 4, 6+1*8
ADD16_OP_INTRA 6, 6+2*8
ADD16_OP_INTRA 8, 4+3*8
ADD16_OP_INTRA 10, 4+4*8
ADD16_OP_INTRA 12, 6+3*8
ADD16_OP_INTRA 14, 6+4*8
REP_RET
AC 8
AC 10
AC 12
AC 14
AC 0
AC 2
AC 4
AC 6
%endmacro
INIT_XMM sse2
IDCT_ADD16INTRA_10
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
IDCT_ADD16INTRA_10
%endif
%assign last_block 36
;-----------------------------------------------------------------------------
; void ff_h264_idct_add8_10(pixel **dst, const int *block_offset,
; int16_t *block, int stride,
; const uint8_t nnzc[6*8])
;-----------------------------------------------------------------------------
%macro IDCT_ADD8 0
cglobal h264_idct_add8_10,5,8,7
movsxdifnidn r3, r3d
%if ARCH_X86_64
mov r7, r0
%endif
add r2, 1024
mov r0, [r0]
ADD16_OP_INTRA 16, 4+ 6*8
ADD16_OP_INTRA 18, 4+ 7*8
add r2, 1024-128*2
%if ARCH_X86_64
mov r0, [r7+gprsize]
%else
mov r0, r0m
mov r0, [r0+gprsize]
%endif
ADD16_OP_INTRA 32, 4+11*8
ADD16_OP_INTRA 34, 4+12*8
REP_RET
AC 16
AC 18
AC 32
AC 34
%endmacro ; IDCT_ADD8
INIT_XMM sse2
IDCT_ADD8
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
IDCT_ADD8
%endif
;-----------------------------------------------------------------------------
; void ff_h264_idct_add8_422_10(pixel **dst, const int *block_offset,
; int16_t *block, int stride,
; const uint8_t nnzc[6*8])
;-----------------------------------------------------------------------------
%assign last_block 44
%macro IDCT_ADD8_422 0
cglobal h264_idct_add8_422_10, 5, 8, 7
movsxdifnidn r3, r3d
%if ARCH_X86_64
mov r7, r0
%endif
add r2, 1024
mov r0, [r0]
ADD16_OP_INTRA 16, 4+ 6*8
ADD16_OP_INTRA 18, 4+ 7*8
ADD16_OP_INTRA 24, 4+ 8*8 ; i+4
ADD16_OP_INTRA 26, 4+ 9*8 ; i+4
add r2, 1024-128*4
%if ARCH_X86_64
mov r0, [r7+gprsize]
%else
mov r0, r0m
mov r0, [r0+gprsize]
%endif
ADD16_OP_INTRA 32, 4+11*8
ADD16_OP_INTRA 34, 4+12*8
ADD16_OP_INTRA 40, 4+13*8 ; i+4
ADD16_OP_INTRA 42, 4+14*8 ; i+4
REP_RET
AC 16
AC 18
AC 24 ; i+4
AC 26 ; i+4
AC 32
AC 34
AC 40 ; i+4
AC 42 ; i+4
%endmacro
INIT_XMM sse2
IDCT_ADD8_422
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
IDCT_ADD8_422
%endif
;-----------------------------------------------------------------------------
; void ff_h264_idct8_add_10(pixel *dst, int16_t *block, int stride)
;-----------------------------------------------------------------------------
%macro IDCT8_1D 2
SWAP 0, 1
psrad m4, m5, 1
psrad m1, m0, 1
paddd m4, m5
paddd m1, m0
paddd m4, m7
paddd m1, m5
psubd m4, m0
paddd m1, m3
psubd m0, m3
psubd m5, m3
paddd m0, m7
psubd m5, m7
psrad m3, 1
psrad m7, 1
psubd m0, m3
psubd m5, m7
SWAP 1, 7
psrad m1, m7, 2
psrad m3, m4, 2
paddd m3, m0
psrad m0, 2
paddd m1, m5
psrad m5, 2
psubd m0, m4
psubd m7, m5
SWAP 5, 6
psrad m4, m2, 1
psrad m6, m5, 1
psubd m4, m5
paddd m6, m2
mova m2, %1
mova m5, %2
SUMSUB_BA d, 5, 2
SUMSUB_BA d, 6, 5
SUMSUB_BA d, 4, 2
SUMSUB_BA d, 7, 6
SUMSUB_BA d, 0, 4
SUMSUB_BA d, 3, 2
SUMSUB_BA d, 1, 5
SWAP 7, 6, 4, 5, 2, 3, 1, 0 ; 70315246 -> 01234567
%endmacro
%macro IDCT8_1D_FULL 1
mova m7, [%1+112*2]
mova m6, [%1+ 96*2]
mova m5, [%1+ 80*2]
mova m3, [%1+ 48*2]
mova m2, [%1+ 32*2]
mova m1, [%1+ 16*2]
IDCT8_1D [%1], [%1+ 64*2]
%endmacro
; %1=int16_t *block, %2=int16_t *dstblock
%macro IDCT8_ADD_SSE_START 2
IDCT8_1D_FULL %1
%if ARCH_X86_64
TRANSPOSE4x4D 0,1,2,3,8
mova [%2 ], m0
TRANSPOSE4x4D 4,5,6,7,8
mova [%2+8*2], m4
%else
mova [%1], m7
TRANSPOSE4x4D 0,1,2,3,7
mova m7, [%1]
mova [%2 ], m0
mova [%2+16*2], m1
mova [%2+32*2], m2
mova [%2+48*2], m3
TRANSPOSE4x4D 4,5,6,7,3
mova [%2+ 8*2], m4
mova [%2+24*2], m5
mova [%2+40*2], m6
mova [%2+56*2], m7
%endif
%endmacro
; %1=uint8_t *dst, %2=int16_t *block, %3=int stride
%macro IDCT8_ADD_SSE_END 3
IDCT8_1D_FULL %2
mova [%2 ], m6
mova [%2+16*2], m7
pxor m7, m7
STORE_DIFFx2 m0, m1, m6, m7, %1, %3
lea %1, [%1+%3*2]
STORE_DIFFx2 m2, m3, m6, m7, %1, %3
mova m0, [%2 ]
mova m1, [%2+16*2]
lea %1, [%1+%3*2]
STORE_DIFFx2 m4, m5, m6, m7, %1, %3
lea %1, [%1+%3*2]
STORE_DIFFx2 m0, m1, m6, m7, %1, %3
%endmacro
%macro IDCT8_ADD 0
cglobal h264_idct8_add_10, 3,4,16
movsxdifnidn r2, r2d
%if UNIX64 == 0
%assign pad 16-gprsize-(stack_offset&15)
sub rsp, pad
call h264_idct8_add1_10 %+ SUFFIX
add rsp, pad
RET
%endif
ALIGN 16
; TODO: does not need to use stack
h264_idct8_add1_10 %+ SUFFIX:
%assign pad 256+16-gprsize
sub rsp, pad
add dword [r1], 32
%if ARCH_X86_64
IDCT8_ADD_SSE_START r1, rsp
SWAP 1, 9
SWAP 2, 10
SWAP 3, 11
SWAP 5, 13
SWAP 6, 14
SWAP 7, 15
IDCT8_ADD_SSE_START r1+16, rsp+128
PERMUTE 1,9, 2,10, 3,11, 5,1, 6,2, 7,3, 9,13, 10,14, 11,15, 13,5, 14,6, 15,7
IDCT8_1D [rsp], [rsp+128]
SWAP 0, 8
SWAP 1, 9
SWAP 2, 10
SWAP 3, 11
SWAP 4, 12
SWAP 5, 13
SWAP 6, 14
SWAP 7, 15
IDCT8_1D [rsp+16], [rsp+144]
psrad m8, 6
psrad m0, 6
packssdw m8, m0
paddsw m8, [r0]
pxor m0, m0
mova [r1+ 0], m0
mova [r1+ 16], m0
mova [r1+ 32], m0
mova [r1+ 48], m0
mova [r1+ 64], m0
mova [r1+ 80], m0
mova [r1+ 96], m0
mova [r1+112], m0
mova [r1+128], m0
mova [r1+144], m0
mova [r1+160], m0
mova [r1+176], m0
mova [r1+192], m0
mova [r1+208], m0
mova [r1+224], m0
mova [r1+240], m0
CLIPW m8, m0, [pw_pixel_max]
mova [r0], m8
mova m8, [pw_pixel_max]
STORE_DIFF16 m9, m1, m0, m8, r0+r2
lea r0, [r0+r2*2]
STORE_DIFF16 m10, m2, m0, m8, r0
STORE_DIFF16 m11, m3, m0, m8, r0+r2
lea r0, [r0+r2*2]
STORE_DIFF16 m12, m4, m0, m8, r0
STORE_DIFF16 m13, m5, m0, m8, r0+r2
lea r0, [r0+r2*2]
STORE_DIFF16 m14, m6, m0, m8, r0
STORE_DIFF16 m15, m7, m0, m8, r0+r2
%else
IDCT8_ADD_SSE_START r1, rsp
IDCT8_ADD_SSE_START r1+16, rsp+128
lea r3, [r0+8]
IDCT8_ADD_SSE_END r0, rsp, r2
IDCT8_ADD_SSE_END r3, rsp+16, r2
mova [r1+ 0], m7
mova [r1+ 16], m7
mova [r1+ 32], m7
mova [r1+ 48], m7
mova [r1+ 64], m7
mova [r1+ 80], m7
mova [r1+ 96], m7
mova [r1+112], m7
mova [r1+128], m7
mova [r1+144], m7
mova [r1+160], m7
mova [r1+176], m7
mova [r1+192], m7
mova [r1+208], m7
mova [r1+224], m7
mova [r1+240], m7
%endif ; ARCH_X86_64
add rsp, pad
ret
%endmacro
INIT_XMM sse2
IDCT8_ADD
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
IDCT8_ADD
%endif
;-----------------------------------------------------------------------------
; void ff_h264_idct8_add4_10(pixel **dst, const int *block_offset,
; int16_t *block, int stride,
; const uint8_t nnzc[6*8])
;-----------------------------------------------------------------------------
;;;;;;; NO FATE SAMPLES TRIGGER THIS
%macro IDCT8_ADD4_OP 2
cmp byte [r4+%2], 0
jz .skipblock%1
mov r0d, [r6+%1*4]
add r0, r5
call h264_idct8_add1_10 %+ SUFFIX
.skipblock%1:
%if %1<12
add r1, 256
%endif
%endmacro
%macro IDCT8_ADD4 0
cglobal h264_idct8_add4_10, 0,7,16
movsxdifnidn r3, r3d
%assign pad 16-gprsize-(stack_offset&15)
SUB rsp, pad
mov r5, r0mp
mov r6, r1mp
mov r1, r2mp
mov r2d, r3m
movifnidn r4, r4mp
IDCT8_ADD4_OP 0, 4+1*8
IDCT8_ADD4_OP 4, 6+1*8
IDCT8_ADD4_OP 8, 4+3*8
IDCT8_ADD4_OP 12, 6+3*8
ADD rsp, pad
RET
%endmacro ; IDCT8_ADD4
INIT_XMM sse2
IDCT8_ADD4
%if HAVE_AVX_EXTERNAL
INIT_XMM avx
IDCT8_ADD4
%endif
Reference in New Issue View Git Blame Copy Permalink