[FFmpeg-cvslog] r25119 - in trunk/libavcodec: h264.c x86/Makefile x86/h264_idct.asm x86/h264_idct_sse2.asm x86/h264dsp_mmx.c
rbultje
subversion
Tue Sep 14 15:36:26 CEST 2010
Author: rbultje
Date: Tue Sep 14 15:36:26 2010
New Revision: 25119
Log:
Rename h264_idct_sse2.asm to h264_idct.asm; move inline IDCT asm from
h264dsp_mmx.c to h264_idct.asm (as yasm code). Because the loops are now
coded in asm instead of C, this is (depending on the function) up to 50%
faster for cases where gcc didn't do a great job at looping.
Since h264_idct_add8() is now faster than the manual loop setup in h264.c,
in-asm idct calling can now be enabled for chroma as well (see r16207). For
MMX, this is 5% faster. For SSE2 (which isn't done for chroma if h264.c does
the looping), this makes it up to 50% faster. Speed gain overall is ~0.5-1.0%.
Added:
trunk/libavcodec/x86/h264_idct.asm
- copied, changed from r25092, trunk/libavcodec/x86/h264_idct_sse2.asm
Deleted:
trunk/libavcodec/x86/h264_idct_sse2.asm
Modified:
trunk/libavcodec/h264.c
trunk/libavcodec/x86/Makefile
trunk/libavcodec/x86/h264dsp_mmx.c
Modified: trunk/libavcodec/h264.c
==============================================================================
--- trunk/libavcodec/h264.c Tue Sep 14 15:21:13 2010 (r25118)
+++ trunk/libavcodec/h264.c Tue Sep 14 15:36:26 2010 (r25119)
@@ -1318,14 +1318,9 @@ static av_always_inline void hl_decode_m
chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp[0], h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]);
chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp[1], h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]);
if(is_h264){
- idct_add = h->h264dsp.h264_idct_add;
- idct_dc_add = h->h264dsp.h264_idct_dc_add;
- for(i=16; i<16+8; i++){
- if(h->non_zero_count_cache[ scan8[i] ])
- idct_add (dest[(i&4)>>2] + block_offset[i], h->mb + i*16, uvlinesize);
- else if(h->mb[i*16])
- idct_dc_add(dest[(i&4)>>2] + block_offset[i], h->mb + i*16, uvlinesize);
- }
+ h->h264dsp.h264_idct_add8(dest, block_offset,
+ h->mb, uvlinesize,
+ h->non_zero_count_cache);
}else{
for(i=16; i<16+8; i++){
if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
Modified: trunk/libavcodec/x86/Makefile
==============================================================================
--- trunk/libavcodec/x86/Makefile Tue Sep 14 15:21:13 2010 (r25118)
+++ trunk/libavcodec/x86/Makefile Tue Sep 14 15:36:26 2010 (r25119)
@@ -10,7 +10,7 @@ YASM-OBJS-$(CONFIG_FFT) +
MMX-OBJS-$(CONFIG_H264DSP) += x86/h264dsp_mmx.o
YASM-OBJS-$(CONFIG_H264DSP) += x86/h264_deblock.o \
x86/h264_weight.o \
- x86/h264_idct_sse2.o \
+ x86/h264_idct.o \
YASM-OBJS-$(CONFIG_H264PRED) += x86/h264_intrapred.o
MMX-OBJS-$(CONFIG_H264PRED) += x86/h264_intrapred_init.o
Copied and modified: trunk/libavcodec/x86/h264_idct.asm (from r25092, trunk/libavcodec/x86/h264_idct_sse2.asm)
==============================================================================
--- trunk/libavcodec/x86/h264_idct_sse2.asm Fri Sep 10 04:25:12 2010 (r25092, copy source)
+++ trunk/libavcodec/x86/h264_idct.asm Tue Sep 14 15:36:26 2010 (r25119)
@@ -1,6 +1,7 @@
;*****************************************************************************
-;* SSE2-optimized H.264 iDCT
+;* MMX/SSE2-optimized H.264 iDCT
;*****************************************************************************
+;* Copyright (C) 2004-2005 Michael Niedermayer, Loren Merritt
;* Copyright (C) 2003-2008 x264 project
;*
;* Authors: Laurent Aimar <fenrir at via.ecp.fr>
@@ -29,28 +30,836 @@
%include "x86util.asm"
SECTION_RODATA
-pw_32: times 8 dw 32
+
+; FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split
+scan8_mem: db 4+1*8, 5+1*8, 4+2*8, 5+2*8
+ db 6+1*8, 7+1*8, 6+2*8, 7+2*8
+ db 4+3*8, 5+3*8, 4+4*8, 5+4*8
+ db 6+3*8, 7+3*8, 6+4*8, 7+4*8
+ db 1+1*8, 2+1*8
+ db 1+2*8, 2+2*8
+ db 1+4*8, 2+4*8
+ db 1+5*8, 2+5*8
+%ifdef PIC
+%define scan8 r11
+%else
+%define scan8 scan8_mem
+%endif
+
+cextern pw_32
SECTION .text
+; %1=uint8_t *dst, %2=int16_t *block, %3=int stride
+%macro IDCT4_ADD 3
+ ; Load dct coeffs
+ movq m0, [%2]
+ movq m1, [%2+8]
+ movq m2, [%2+16]
+ movq m3, [%2+24]
+
+ IDCT4_1D 0, 1, 2, 3, 4, 5
+ mova m6, [pw_32]
+ TRANSPOSE4x4W 0, 1, 2, 3, 4
+ paddw m0, m6
+ IDCT4_1D 0, 1, 2, 3, 4, 5
+ pxor m7, m7
+
+ STORE_DIFFx2 m0, m1, m4, m5, m7, 6, %1, %3
+ lea %1, [%1+%3*2]
+ STORE_DIFFx2 m2, m3, m4, m5, m7, 6, %1, %3
+%endmacro
+
+INIT_MMX
+; ff_h264_idct_add_mmx(uint8_t *dst, int16_t *block, int stride)
+cglobal h264_idct_add_mmx, 3, 3, 0
+ IDCT4_ADD r0, r1, r2
+ RET
+
+%macro IDCT8_1D 2
+ mova m4, m5
+ mova m0, m1
+ psraw m4, 1
+ psraw m1, 1
+ paddw m4, m5
+ paddw m1, m0
+ paddw m4, m7
+ paddw m1, m5
+ psubw m4, m0
+ paddw m1, m3
+
+ psubw m0, m3
+ psubw m5, m3
+ paddw m0, m7
+ psubw m5, m7
+ psraw m3, 1
+ psraw m7, 1
+ psubw m0, m3
+ psubw m5, m7
+
+ mova m3, m4
+ mova m7, m1
+ psraw m1, 2
+ psraw m3, 2
+ paddw m3, m0
+ psraw m0, 2
+ paddw m1, m5
+ psraw m5, 2
+ psubw m0, m4
+ psubw m7, m5
+
+ mova m4, m2
+ mova m5, m6
+ psraw m4, 1
+ psraw m6, 1
+ psubw m4, m5
+ paddw m6, m2
+
+ mova m2, %1
+ mova m5, %2
+ SUMSUB_BA m5, m2
+ SUMSUB_BA m6, m5
+ SUMSUB_BA m4, m2
+ SUMSUB_BA m7, m6
+ SUMSUB_BA m0, m4
+ SUMSUB_BA m3, m2
+ SUMSUB_BA m1, m5
+ SWAP 7, 6, 4, 5, 2, 3, 1, 0 ; 70315246 -> 01234567
+%endmacro
+
+%macro IDCT8_1D_FULL 1
+ mova m7, [%1+112]
+ mova m6, [%1+ 96]
+ mova m5, [%1+ 80]
+ mova m3, [%1+ 48]
+ mova m2, [%1+ 32]
+ mova m1, [%1+ 16]
+ IDCT8_1D [%1], [%1+ 64]
+%endmacro
+
+; %1=int16_t *block, %2=int16_t *dstblock
+%macro IDCT8_ADD_MMX_START 2
+ IDCT8_1D_FULL %1
+ mova [%1], m7
+ TRANSPOSE4x4W 0, 1, 2, 3, 7
+ mova m7, [%1]
+ mova [%2 ], m0
+ mova [%2+16], m1
+ mova [%2+32], m2
+ mova [%2+48], m3
+ TRANSPOSE4x4W 4, 5, 6, 7, 3
+ mova [%2+ 8], m4
+ mova [%2+24], m5
+ mova [%2+40], m6
+ mova [%2+56], m7
+%endmacro
+
+; %1=uint8_t *dst, %2=int16_t *block, %3=int stride
+%macro IDCT8_ADD_MMX_END 3
+ IDCT8_1D_FULL %2
+ mova [%2 ], m5
+ mova [%2+16], m6
+ mova [%2+32], m7
+
+ pxor m7, m7
+ STORE_DIFFx2 m0, m1, m5, m6, m7, 6, %1, %3
+ lea %1, [%1+%3*2]
+ STORE_DIFFx2 m2, m3, m5, m6, m7, 6, %1, %3
+ mova m0, [%2 ]
+ mova m1, [%2+16]
+ mova m2, [%2+32]
+ lea %1, [%1+%3*2]
+ STORE_DIFFx2 m4, m0, m5, m6, m7, 6, %1, %3
+ lea %1, [%1+%3*2]
+ STORE_DIFFx2 m1, m2, m5, m6, m7, 6, %1, %3
+%endmacro
+
+INIT_MMX
+; ff_h264_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
+cglobal h264_idct8_add_mmx, 3, 4, 0
+ %assign pad 128+4-(stack_offset&7)
+ SUB rsp, pad
+
+ add word [r1], 32
+ IDCT8_ADD_MMX_START r1 , rsp
+ IDCT8_ADD_MMX_START r1+8, rsp+64
+ lea r3, [r0+4]
+ IDCT8_ADD_MMX_END r0 , rsp, r2
+ IDCT8_ADD_MMX_END r3 , rsp+8, r2
+
+ ADD rsp, pad
+ RET
+
+; %1=uint8_t *dst, %2=int16_t *block, %3=int stride
+%macro IDCT8_ADD_SSE 4
+ IDCT8_1D_FULL %2
+%ifdef ARCH_X86_64
+ TRANSPOSE8x8W 0, 1, 2, 3, 4, 5, 6, 7, 8
+%else
+ TRANSPOSE8x8W 0, 1, 2, 3, 4, 5, 6, 7, [%2], [%2+16]
+%endif
+ paddw m0, [pw_32]
+
+%ifndef ARCH_X86_64
+ mova [%2 ], m0
+ mova [%2+16], m4
+ IDCT8_1D [%2], [%2+ 16]
+ mova [%2 ], m6
+ mova [%2+16], m7
+%else
+ SWAP 0, 8
+ SWAP 4, 9
+ IDCT8_1D m8, m9
+ SWAP 6, 8
+ SWAP 7, 9
+%endif
+
+ pxor m7, m7
+ lea %4, [%3*3]
+ STORE_DIFF m0, m6, m7, [%1 ]
+ STORE_DIFF m1, m6, m7, [%1+%3 ]
+ STORE_DIFF m2, m6, m7, [%1+%3*2]
+ STORE_DIFF m3, m6, m7, [%1+%4 ]
+%ifndef ARCH_X86_64
+ mova m0, [%2 ]
+ mova m1, [%2+16]
+%else
+ SWAP 0, 8
+ SWAP 1, 9
+%endif
+ lea %1, [%1+%3*4]
+ STORE_DIFF m4, m6, m7, [%1 ]
+ STORE_DIFF m5, m6, m7, [%1+%3 ]
+ STORE_DIFF m0, m6, m7, [%1+%3*2]
+ STORE_DIFF m1, m6, m7, [%1+%4 ]
+%endmacro
+
INIT_XMM
-cglobal x264_add8x4_idct_sse2, 3,3,8
- movq m0, [r1+ 0]
- movq m1, [r1+ 8]
- movq m2, [r1+16]
- movq m3, [r1+24]
- movhps m0, [r1+32]
- movhps m1, [r1+40]
- movhps m2, [r1+48]
- movhps m3, [r1+56]
+; ff_h264_idct8_add_sse2(uint8_t *dst, int16_t *block, int stride)
+cglobal h264_idct8_add_sse2, 3, 4, 10
+ IDCT8_ADD_SSE r0, r1, r2, r3
+ RET
+
+%macro DC_ADD_MMX2_INIT 2-3
+%if %0 == 2
+ movsx %1, word [%1]
+ add %1, 32
+ sar %1, 6
+ movd m0, %1
+ lea %1, [%2*3]
+%else
+ add %3, 32
+ sar %3, 6
+ movd m0, %3
+ lea %3, [%2*3]
+%endif
+ pshufw m0, m0, 0
+ pxor m1, m1
+ psubw m1, m0
+ packuswb m0, m0
+ packuswb m1, m1
+%endmacro
+
+%macro DC_ADD_MMX2_OP 3-4
+ %1 m2, [%2 ]
+ %1 m3, [%2+%3 ]
+ %1 m4, [%2+%3*2]
+ %1 m5, [%2+%4 ]
+ paddusb m2, m0
+ paddusb m3, m0
+ paddusb m4, m0
+ paddusb m5, m0
+ psubusb m2, m1
+ psubusb m3, m1
+ psubusb m4, m1
+ psubusb m5, m1
+ %1 [%2 ], m2
+ %1 [%2+%3 ], m3
+ %1 [%2+%3*2], m4
+ %1 [%2+%4 ], m5
+%endmacro
+
+INIT_MMX
+; ff_h264_idct_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
+cglobal h264_idct_dc_add_mmx2, 3, 3, 0
+ DC_ADD_MMX2_INIT r1, r2
+ DC_ADD_MMX2_OP movh, r0, r2, r1
+ RET
+
+; ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
+cglobal h264_idct8_dc_add_mmx2, 3, 3, 0
+ DC_ADD_MMX2_INIT r1, r2
+ DC_ADD_MMX2_OP mova, r0, r2, r1
+ lea r0, [r0+r2*4]
+ DC_ADD_MMX2_OP mova, r0, r2, r1
+ RET
+
+; ff_h264_idct_add16_mmx(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add16_mmx, 5, 7, 0
+ xor r5, r5
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ test r6, r6
+ jz .skipblock
+ mov r6d, dword [r1+r5*4]
+ lea r6, [r0+r6]
+ IDCT4_ADD r6, r2, r3
+.skipblock
+ inc r5
+ add r2, 32
+ cmp r5, 16
+ jl .nextblock
+ REP_RET
+
+; ff_h264_idct8_add4_mmx(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct8_add4_mmx, 5, 7, 0
+ %assign pad 128+4-(stack_offset&7)
+ SUB rsp, pad
+
+ xor r5, r5
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ test r6, r6
+ jz .skipblock
+ mov r6d, dword [r1+r5*4]
+ lea r6, [r0+r6]
+ add word [r2], 32
+ IDCT8_ADD_MMX_START r2 , rsp
+ IDCT8_ADD_MMX_START r2+8, rsp+64
+ IDCT8_ADD_MMX_END r6 , rsp, r3
+ mov r6d, dword [r1+r5*4]
+ lea r6, [r0+r6+4]
+ IDCT8_ADD_MMX_END r6 , rsp+8, r3
+.skipblock
+ add r5, 4
+ add r2, 128
+ cmp r5, 16
+ jl .nextblock
+ ADD rsp, pad
+ RET
+
+; ff_h264_idct_add16_mmx2(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add16_mmx2, 5, 7, 0
+ xor r5, r5
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ test r6, r6
+ jz .skipblock
+ cmp r6, 1
+ jnz .no_dc
+ movsx r6, word [r2]
+ test r6, r6
+ jz .no_dc
+ DC_ADD_MMX2_INIT r2, r3, r6
+%ifdef ARCH_X86_64
+%define dst_reg r10
+%define dst_regd r10d
+%else
+%define dst_reg r1
+%define dst_regd r1d
+%endif
+ mov dst_regd, dword [r1+r5*4]
+ lea dst_reg, [r0+dst_reg]
+ DC_ADD_MMX2_OP movh, dst_reg, r3, r6
+%ifndef ARCH_X86_64
+ mov r1, r1m
+%endif
+ inc r5
+ add r2, 32
+ cmp r5, 16
+ jl .nextblock
+ REP_RET
+.no_dc
+ mov r6d, dword [r1+r5*4]
+ lea r6, [r0+r6]
+ IDCT4_ADD r6, r2, r3
+.skipblock
+ inc r5
+ add r2, 32
+ cmp r5, 16
+ jl .nextblock
+ REP_RET
+
+; ff_h264_idct_add16intra_mmx(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add16intra_mmx, 5, 7, 0
+ xor r5, r5
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ or r6w, word [r2]
+ test r6, r6
+ jz .skipblock
+ mov r6d, dword [r1+r5*4]
+ lea r6, [r0+r6]
+ IDCT4_ADD r6, r2, r3
+.skipblock
+ inc r5
+ add r2, 32
+ cmp r5, 16
+ jl .nextblock
+ REP_RET
+
+; ff_h264_idct_add16intra_mmx2(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add16intra_mmx2, 5, 7, 0
+ xor r5, r5
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ test r6, r6
+ jz .try_dc
+ mov r6d, dword [r1+r5*4]
+ lea r6, [r0+r6]
+ IDCT4_ADD r6, r2, r3
+ inc r5
+ add r2, 32
+ cmp r5, 16
+ jl .nextblock
+ REP_RET
+.try_dc
+ movsx r6, word [r2]
+ test r6, r6
+ jz .skipblock
+ DC_ADD_MMX2_INIT r2, r3, r6
+%ifdef ARCH_X86_64
+%define dst_reg r10
+%define dst_regd r10d
+%else
+%define dst_reg r1
+%define dst_regd r1d
+%endif
+ mov dst_regd, dword [r1+r5*4]
+ lea dst_reg, [r0+dst_reg]
+ DC_ADD_MMX2_OP movh, dst_reg, r3, r6
+%ifndef ARCH_X86_64
+ mov r1, r1m
+%endif
+.skipblock
+ inc r5
+ add r2, 32
+ cmp r5, 16
+ jl .nextblock
+ REP_RET
+
+; ff_h264_idct8_add4_mmx2(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct8_add4_mmx2, 5, 7, 0
+ %assign pad 128+4-(stack_offset&7)
+ SUB rsp, pad
+
+ xor r5, r5
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ test r6, r6
+ jz .skipblock
+ cmp r6, 1
+ jnz .no_dc
+ movsx r6, word [r2]
+ test r6, r6
+ jz .no_dc
+ DC_ADD_MMX2_INIT r2, r3, r6
+%ifdef ARCH_X86_64
+%define dst_reg r10
+%define dst_regd r10d
+%else
+%define dst_reg r1
+%define dst_regd r1d
+%endif
+ mov dst_regd, dword [r1+r5*4]
+ lea dst_reg, [r0+dst_reg]
+ DC_ADD_MMX2_OP mova, dst_reg, r3, r6
+ lea dst_reg, [dst_reg+r3*4]
+ DC_ADD_MMX2_OP mova, dst_reg, r3, r6
+%ifndef ARCH_X86_64
+ mov r1, r1m
+%endif
+ add r5, 4
+ add r2, 128
+ cmp r5, 16
+ jl .nextblock
+
+ ADD rsp, pad
+ RET
+.no_dc
+ mov r6d, dword [r1+r5*4]
+ lea r6, [r0+r6]
+ add word [r2], 32
+ IDCT8_ADD_MMX_START r2 , rsp
+ IDCT8_ADD_MMX_START r2+8, rsp+64
+ IDCT8_ADD_MMX_END r6 , rsp, r3
+ mov r6d, dword [r1+r5*4]
+ lea r6, [r0+r6+4]
+ IDCT8_ADD_MMX_END r6 , rsp+8, r3
+.skipblock
+ add r5, 4
+ add r2, 128
+ cmp r5, 16
+ jl .nextblock
+
+ ADD rsp, pad
+ RET
+
+INIT_XMM
+; ff_h264_idct8_add4_sse2(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct8_add4_sse2, 5, 7, 10
+ xor r5, r5
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ test r6, r6
+ jz .skipblock
+ cmp r6, 1
+ jnz .no_dc
+ movsx r6, word [r2]
+ test r6, r6
+ jz .no_dc
+INIT_MMX
+ DC_ADD_MMX2_INIT r2, r3, r6
+%ifdef ARCH_X86_64
+%define dst_reg r10
+%define dst_regd r10d
+%else
+%define dst_reg r1
+%define dst_regd r1d
+%endif
+ mov dst_regd, dword [r1+r5*4]
+ lea dst_reg, [r0+dst_reg]
+ DC_ADD_MMX2_OP mova, dst_reg, r3, r6
+ lea dst_reg, [dst_reg+r3*4]
+ DC_ADD_MMX2_OP mova, dst_reg, r3, r6
+%ifndef ARCH_X86_64
+ mov r1, r1m
+%endif
+ add r5, 4
+ add r2, 128
+ cmp r5, 16
+ jl .nextblock
+ REP_RET
+.no_dc
+INIT_XMM
+ mov dst_regd, dword [r1+r5*4]
+ lea dst_reg, [r0+dst_reg]
+ IDCT8_ADD_SSE dst_reg, r2, r3, r6
+%ifndef ARCH_X86_64
+ mov r1, r1m
+%endif
+.skipblock
+ add r5, 4
+ add r2, 128
+ cmp r5, 16
+ jl .nextblock
+ REP_RET
+
+INIT_MMX
+h264_idct_add8_mmx_plane:
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ or r6w, word [r2]
+ test r6, r6
+ jz .skipblock
+%ifdef ARCH_X86_64
+ mov r0d, dword [r1+r5*4]
+ add r0, [r10]
+%else
+ mov r0, r1m ; XXX r1m here is actually r0m of the calling func
+ mov r0, [r0]
+ add r0, dword [r1+r5*4]
+%endif
+ IDCT4_ADD r0, r2, r3
+.skipblock
+ inc r5
+ add r2, 32
+ test r5, 3
+ jnz .nextblock
+ rep ret
+
+; ff_h264_idct_add8_mmx(uint8_t **dest, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add8_mmx, 5, 7, 0
+ mov r5, 16
+ add r2, 512
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+%ifdef ARCH_X86_64
+ mov r10, r0
+%endif
+ call h264_idct_add8_mmx_plane
+%ifdef ARCH_X86_64
+ add r10, gprsize
+%else
+ add r0mp, gprsize
+%endif
+ call h264_idct_add8_mmx_plane
+ RET
+
+h264_idct_add8_mmx2_plane
+.nextblock
+ movzx r6, byte [scan8+r5]
+ movzx r6, byte [r4+r6]
+ test r6, r6
+ jz .try_dc
+%ifdef ARCH_X86_64
+ mov r0d, dword [r1+r5*4]
+ add r0, [r10]
+%else
+ mov r0, r1m ; XXX r1m here is actually r0m of the calling func
+ mov r0, [r0]
+ add r0, dword [r1+r5*4]
+%endif
+ IDCT4_ADD r0, r2, r3
+ inc r5
+ add r2, 32
+ test r5, 3
+ jnz .nextblock
+ rep ret
+.try_dc
+ movsx r6, word [r2]
+ test r6, r6
+ jz .skipblock
+ DC_ADD_MMX2_INIT r2, r3, r6
+%ifdef ARCH_X86_64
+ mov r0d, dword [r1+r5*4]
+ add r0, [r10]
+%else
+ mov r0, r1m ; XXX r1m here is actually r0m of the calling func
+ mov r0, [r0]
+ add r0, dword [r1+r5*4]
+%endif
+ DC_ADD_MMX2_OP movh, r0, r3, r6
+.skipblock
+ inc r5
+ add r2, 32
+ test r5, 3
+ jnz .nextblock
+ rep ret
+
+; ff_h264_idct_add8_mmx2(uint8_t **dest, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add8_mmx2, 5, 7, 0
+ mov r5, 16
+ add r2, 512
+%ifdef ARCH_X86_64
+ mov r10, r0
+%endif
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+ call h264_idct_add8_mmx2_plane
+%ifdef ARCH_X86_64
+ add r10, gprsize
+%else
+ add r0mp, gprsize
+%endif
+ call h264_idct_add8_mmx2_plane
+ RET
+
+INIT_MMX
+; r0 = uint8_t *dst, r2 = int16_t *block, r3 = int stride, r6=clobbered
+h264_idct_dc_add8_mmx2:
+ movd m0, [r2 ] ; 0 0 X D
+ punpcklwd m0, [r2+32] ; x X d D
+ paddsw m0, [pw_32]
+ psraw m0, 6
+ punpcklwd m0, m0 ; d d D D
+ pxor m1, m1 ; 0 0 0 0
+ psubw m1, m0 ; -d-d-D-D
+ packuswb m0, m1 ; -d-d-D-D d d D D
+ pshufw m1, m0, 0xFA ; -d-d-d-d-D-D-D-D
+ punpcklwd m0, m0 ; d d d d D D D D
+ lea r6, [r3*3]
+ DC_ADD_MMX2_OP movq, r0, r3, r6
+ ret
+
+ALIGN 16
+INIT_XMM
+; r0 = uint8_t *dst (clobbered), r2 = int16_t *block, r3 = int stride
+x264_add8x4_idct_sse2:
+ movq m0, [r2+ 0]
+ movq m1, [r2+ 8]
+ movq m2, [r2+16]
+ movq m3, [r2+24]
+ movhps m0, [r2+32]
+ movhps m1, [r2+40]
+ movhps m2, [r2+48]
+ movhps m3, [r2+56]
IDCT4_1D 0,1,2,3,4,5
TRANSPOSE2x4x4W 0,1,2,3,4
paddw m0, [pw_32]
IDCT4_1D 0,1,2,3,4,5
pxor m7, m7
- STORE_DIFF m0, m4, m7, [r0]
- STORE_DIFF m1, m4, m7, [r0+r2]
- lea r0, [r0+r2*2]
- STORE_DIFF m2, m4, m7, [r0]
- STORE_DIFF m3, m4, m7, [r0+r2]
+ STORE_DIFFx2 m0, m1, m4, m5, m7, 6, r0, r3
+ lea r0, [r0+r3*2]
+ STORE_DIFFx2 m2, m3, m4, m5, m7, 6, r0, r3
+ ret
+
+%macro add16_sse2_cycle 2
+ movzx r0, word [r4+%2]
+ test r0, r0
+ jz .cycle%1end
+ mov r0d, dword [r1+%1*8]
+%ifdef ARCH_X86_64
+ add r0, r10
+%else
+ add r0, r0m
+%endif
+ call x264_add8x4_idct_sse2
+.cycle%1end
+%if %1 < 7
+ add r2, 64
+%endif
+%endmacro
+
+; ff_h264_idct_add16_sse2(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add16_sse2, 5, 5, 8
+%ifdef ARCH_X86_64
+ mov r10, r0
+%endif
+ ; unrolling of the loop leads to an average performance gain of
+ ; 20-25%
+ add16_sse2_cycle 0, 0xc
+ add16_sse2_cycle 1, 0x14
+ add16_sse2_cycle 2, 0xe
+ add16_sse2_cycle 3, 0x16
+ add16_sse2_cycle 4, 0x1c
+ add16_sse2_cycle 5, 0x24
+ add16_sse2_cycle 6, 0x1e
+ add16_sse2_cycle 7, 0x26
+ RET
+
+; ff_h264_idct_add16intra_sse2(uint8_t *dst, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add16intra_sse2, 5, 7, 8
+ xor r5, r5
+%ifdef ARCH_X86_64
+ mov r10, r0
+%endif
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+.next2blocks
+ movzx r0, byte [scan8+r5]
+ movzx r0, word [r4+r0]
+ test r0, r0
+ jz .try_dc
+ mov r0d, dword [r1+r5*4]
+%ifdef ARCH_X86_64
+ add r0, r10
+%else
+ add r0, r0m
+%endif
+ call x264_add8x4_idct_sse2
+ add r5, 2
+ add r2, 64
+ cmp r5, 16
+ jl .next2blocks
+ REP_RET
+.try_dc
+ movsx r0, word [r2 ]
+ or r0w, word [r2+32]
+ jz .skip2blocks
+ mov r0d, dword [r1+r5*4]
+%ifdef ARCH_X86_64
+ add r0, r10
+%else
+ add r0, r0m
+%endif
+ call h264_idct_dc_add8_mmx2
+.skip2blocks
+ add r5, 2
+ add r2, 64
+ cmp r5, 16
+ jl .next2blocks
+ REP_RET
+
+h264_idct_add8_sse2_plane:
+.next2blocks
+ movzx r0, byte [scan8+r5]
+ movzx r0, word [r4+r0]
+ test r0, r0
+ jz .try_dc
+%ifdef ARCH_X86_64
+ mov r0d, dword [r1+r5*4]
+ add r0, [r10]
+%else
+ mov r0, r1m ; XXX r1m here is actually r0m of the calling func
+ mov r0, [r0]
+ add r0, dword [r1+r5*4]
+%endif
+ call x264_add8x4_idct_sse2
+ add r5, 2
+ add r2, 64
+ test r5, 3
+ jnz .next2blocks
+ rep ret
+.try_dc
+ movsx r0, word [r2 ]
+ or r0w, word [r2+32]
+ jz .skip2blocks
+%ifdef ARCH_X86_64
+ mov r0d, dword [r1+r5*4]
+ add r0, [r10]
+%else
+ mov r0, r1m ; XXX r1m here is actually r0m of the calling func
+ mov r0, [r0]
+ add r0, dword [r1+r5*4]
+%endif
+ call h264_idct_dc_add8_mmx2
+.skip2blocks
+ add r5, 2
+ add r2, 64
+ test r5, 3
+ jnz .next2blocks
+ rep ret
+
+; ff_h264_idct_add8_sse2(uint8_t **dest, const int *block_offset,
+; DCTELEM *block, int stride, const uint8_t nnzc[6*8])
+cglobal h264_idct_add8_sse2, 5, 7, 8
+ mov r5, 16
+ add r2, 512
+%ifdef PIC
+ lea r11, [scan8_mem]
+%endif
+%ifdef ARCH_X86_64
+ mov r10, r0
+%endif
+ call h264_idct_add8_sse2_plane
+%ifdef ARCH_X86_64
+ add r10, gprsize
+%else
+ add r0mp, gprsize
+%endif
+ call h264_idct_add8_sse2_plane
RET
Modified: trunk/libavcodec/x86/h264dsp_mmx.c
==============================================================================
--- trunk/libavcodec/x86/h264dsp_mmx.c Tue Sep 14 15:21:13 2010 (r25118)
+++ trunk/libavcodec/x86/h264dsp_mmx.c Tue Sep 14 15:36:26 2010 (r25119)
@@ -29,523 +29,37 @@ DECLARE_ALIGNED(8, static const uint64_t
/***********************************/
/* IDCT */
-#define SUMSUB_BADC( a, b, c, d ) \
- "paddw "#b", "#a" \n\t"\
- "paddw "#d", "#c" \n\t"\
- "paddw "#b", "#b" \n\t"\
- "paddw "#d", "#d" \n\t"\
- "psubw "#a", "#b" \n\t"\
- "psubw "#c", "#d" \n\t"
-
-#define SUMSUBD2_AB( a, b, t ) \
- "movq "#b", "#t" \n\t"\
- "psraw $1 , "#b" \n\t"\
- "paddw "#a", "#b" \n\t"\
- "psraw $1 , "#a" \n\t"\
- "psubw "#t", "#a" \n\t"
-
-#define IDCT4_1D( s02, s13, d02, d13, t ) \
- SUMSUB_BA ( s02, d02 )\
- SUMSUBD2_AB( s13, d13, t )\
- SUMSUB_BADC( d13, s02, s13, d02 )
-
-#define STORE_DIFF_4P( p, t, z ) \
- "psraw $6, "#p" \n\t"\
- "movd (%0), "#t" \n\t"\
- "punpcklbw "#z", "#t" \n\t"\
- "paddsw "#t", "#p" \n\t"\
- "packuswb "#z", "#p" \n\t"\
- "movd "#p", (%0) \n\t"
-
-static void ff_h264_idct_add_mmx(uint8_t *dst, int16_t *block, int stride)
-{
- /* Load dct coeffs */
- __asm__ volatile(
- "movq (%0), %%mm0 \n\t"
- "movq 8(%0), %%mm1 \n\t"
- "movq 16(%0), %%mm2 \n\t"
- "movq 24(%0), %%mm3 \n\t"
- :: "r"(block) );
-
- __asm__ volatile(
- /* mm1=s02+s13 mm2=s02-s13 mm4=d02+d13 mm0=d02-d13 */
- IDCT4_1D( %%mm2, %%mm1, %%mm0, %%mm3, %%mm4 )
-
- "movq %0, %%mm6 \n\t"
- /* in: 1,4,0,2 out: 1,2,3,0 */
- TRANSPOSE4( %%mm3, %%mm1, %%mm0, %%mm2, %%mm4 )
-
- "paddw %%mm6, %%mm3 \n\t"
-
- /* mm2=s02+s13 mm3=s02-s13 mm4=d02+d13 mm1=d02-d13 */
- IDCT4_1D( %%mm4, %%mm2, %%mm3, %%mm0, %%mm1 )
-
- "pxor %%mm7, %%mm7 \n\t"
- :: "m"(ff_pw_32));
-
- __asm__ volatile(
- STORE_DIFF_4P( %%mm0, %%mm1, %%mm7)
- "add %1, %0 \n\t"
- STORE_DIFF_4P( %%mm2, %%mm1, %%mm7)
- "add %1, %0 \n\t"
- STORE_DIFF_4P( %%mm3, %%mm1, %%mm7)
- "add %1, %0 \n\t"
- STORE_DIFF_4P( %%mm4, %%mm1, %%mm7)
- : "+r"(dst)
- : "r" ((x86_reg)stride)
- );
-}
-
-static inline void h264_idct8_1d(int16_t *block)
-{
- __asm__ volatile(
- "movq 112(%0), %%mm7 \n\t"
- "movq 80(%0), %%mm0 \n\t"
- "movq 48(%0), %%mm3 \n\t"
- "movq 16(%0), %%mm5 \n\t"
-
- "movq %%mm0, %%mm4 \n\t"
- "movq %%mm5, %%mm1 \n\t"
- "psraw $1, %%mm4 \n\t"
- "psraw $1, %%mm1 \n\t"
- "paddw %%mm0, %%mm4 \n\t"
- "paddw %%mm5, %%mm1 \n\t"
- "paddw %%mm7, %%mm4 \n\t"
- "paddw %%mm0, %%mm1 \n\t"
- "psubw %%mm5, %%mm4 \n\t"
- "paddw %%mm3, %%mm1 \n\t"
-
- "psubw %%mm3, %%mm5 \n\t"
- "psubw %%mm3, %%mm0 \n\t"
- "paddw %%mm7, %%mm5 \n\t"
- "psubw %%mm7, %%mm0 \n\t"
- "psraw $1, %%mm3 \n\t"
- "psraw $1, %%mm7 \n\t"
- "psubw %%mm3, %%mm5 \n\t"
- "psubw %%mm7, %%mm0 \n\t"
-
- "movq %%mm4, %%mm3 \n\t"
- "movq %%mm1, %%mm7 \n\t"
- "psraw $2, %%mm1 \n\t"
- "psraw $2, %%mm3 \n\t"
- "paddw %%mm5, %%mm3 \n\t"
- "psraw $2, %%mm5 \n\t"
- "paddw %%mm0, %%mm1 \n\t"
- "psraw $2, %%mm0 \n\t"
- "psubw %%mm4, %%mm5 \n\t"
- "psubw %%mm0, %%mm7 \n\t"
-
- "movq 32(%0), %%mm2 \n\t"
- "movq 96(%0), %%mm6 \n\t"
- "movq %%mm2, %%mm4 \n\t"
- "movq %%mm6, %%mm0 \n\t"
- "psraw $1, %%mm4 \n\t"
- "psraw $1, %%mm6 \n\t"
- "psubw %%mm0, %%mm4 \n\t"
- "paddw %%mm2, %%mm6 \n\t"
-
- "movq (%0), %%mm2 \n\t"
- "movq 64(%0), %%mm0 \n\t"
- SUMSUB_BA( %%mm0, %%mm2 )
- SUMSUB_BA( %%mm6, %%mm0 )
- SUMSUB_BA( %%mm4, %%mm2 )
- SUMSUB_BA( %%mm7, %%mm6 )
- SUMSUB_BA( %%mm5, %%mm4 )
- SUMSUB_BA( %%mm3, %%mm2 )
- SUMSUB_BA( %%mm1, %%mm0 )
- :: "r"(block)
- );
-}
-
-static void ff_h264_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
-{
- int i;
- DECLARE_ALIGNED(8, int16_t, b2)[64];
-
- block[0] += 32;
-
- for(i=0; i<2; i++){
- DECLARE_ALIGNED(8, uint64_t, tmp);
-
- h264_idct8_1d(block+4*i);
-
- __asm__ volatile(
- "movq %%mm7, %0 \n\t"
- TRANSPOSE4( %%mm0, %%mm2, %%mm4, %%mm6, %%mm7 )
- "movq %%mm0, 8(%1) \n\t"
- "movq %%mm6, 24(%1) \n\t"
- "movq %%mm7, 40(%1) \n\t"
- "movq %%mm4, 56(%1) \n\t"
- "movq %0, %%mm7 \n\t"
- TRANSPOSE4( %%mm7, %%mm5, %%mm3, %%mm1, %%mm0 )
- "movq %%mm7, (%1) \n\t"
- "movq %%mm1, 16(%1) \n\t"
- "movq %%mm0, 32(%1) \n\t"
- "movq %%mm3, 48(%1) \n\t"
- : "=m"(tmp)
- : "r"(b2+32*i)
- : "memory"
- );
- }
-
- for(i=0; i<2; i++){
- h264_idct8_1d(b2+4*i);
-
- __asm__ volatile(
- "psraw $6, %%mm7 \n\t"
- "psraw $6, %%mm6 \n\t"
- "psraw $6, %%mm5 \n\t"
- "psraw $6, %%mm4 \n\t"
- "psraw $6, %%mm3 \n\t"
- "psraw $6, %%mm2 \n\t"
- "psraw $6, %%mm1 \n\t"
- "psraw $6, %%mm0 \n\t"
-
- "movq %%mm7, (%0) \n\t"
- "movq %%mm5, 16(%0) \n\t"
- "movq %%mm3, 32(%0) \n\t"
- "movq %%mm1, 48(%0) \n\t"
- "movq %%mm0, 64(%0) \n\t"
- "movq %%mm2, 80(%0) \n\t"
- "movq %%mm4, 96(%0) \n\t"
- "movq %%mm6, 112(%0) \n\t"
- :: "r"(b2+4*i)
- : "memory"
- );
- }
-
- ff_add_pixels_clamped_mmx(b2, dst, stride);
-}
-
-#define STORE_DIFF_8P( p, d, t, z )\
- "movq "#d", "#t" \n"\
- "psraw $6, "#p" \n"\
- "punpcklbw "#z", "#t" \n"\
- "paddsw "#t", "#p" \n"\
- "packuswb "#p", "#p" \n"\
- "movq "#p", "#d" \n"
-
-#define H264_IDCT8_1D_SSE2(a,b,c,d,e,f,g,h)\
- "movdqa "#c", "#a" \n"\
- "movdqa "#g", "#e" \n"\
- "psraw $1, "#c" \n"\
- "psraw $1, "#g" \n"\
- "psubw "#e", "#c" \n"\
- "paddw "#a", "#g" \n"\
- "movdqa "#b", "#e" \n"\
- "psraw $1, "#e" \n"\
- "paddw "#b", "#e" \n"\
- "paddw "#d", "#e" \n"\
- "paddw "#f", "#e" \n"\
- "movdqa "#f", "#a" \n"\
- "psraw $1, "#a" \n"\
- "paddw "#f", "#a" \n"\
- "paddw "#h", "#a" \n"\
- "psubw "#b", "#a" \n"\
- "psubw "#d", "#b" \n"\
- "psubw "#d", "#f" \n"\
- "paddw "#h", "#b" \n"\
- "psubw "#h", "#f" \n"\
- "psraw $1, "#d" \n"\
- "psraw $1, "#h" \n"\
- "psubw "#d", "#b" \n"\
- "psubw "#h", "#f" \n"\
- "movdqa "#e", "#d" \n"\
- "movdqa "#a", "#h" \n"\
- "psraw $2, "#d" \n"\
- "psraw $2, "#h" \n"\
- "paddw "#f", "#d" \n"\
- "paddw "#b", "#h" \n"\
- "psraw $2, "#f" \n"\
- "psraw $2, "#b" \n"\
- "psubw "#f", "#e" \n"\
- "psubw "#a", "#b" \n"\
- "movdqa 0x00(%1), "#a" \n"\
- "movdqa 0x40(%1), "#f" \n"\
- SUMSUB_BA(f, a)\
- SUMSUB_BA(g, f)\
- SUMSUB_BA(c, a)\
- SUMSUB_BA(e, g)\
- SUMSUB_BA(b, c)\
- SUMSUB_BA(h, a)\
- SUMSUB_BA(d, f)
-
-static void ff_h264_idct8_add_sse2(uint8_t *dst, int16_t *block, int stride)
-{
- __asm__ volatile(
- "movdqa 0x10(%1), %%xmm1 \n"
- "movdqa 0x20(%1), %%xmm2 \n"
- "movdqa 0x30(%1), %%xmm3 \n"
- "movdqa 0x50(%1), %%xmm5 \n"
- "movdqa 0x60(%1), %%xmm6 \n"
- "movdqa 0x70(%1), %%xmm7 \n"
- H264_IDCT8_1D_SSE2(%%xmm0, %%xmm1, %%xmm2, %%xmm3, %%xmm4, %%xmm5, %%xmm6, %%xmm7)
- TRANSPOSE8(%%xmm4, %%xmm1, %%xmm7, %%xmm3, %%xmm5, %%xmm0, %%xmm2, %%xmm6, (%1))
- "paddw %4, %%xmm4 \n"
- "movdqa %%xmm4, 0x00(%1) \n"
- "movdqa %%xmm2, 0x40(%1) \n"
- H264_IDCT8_1D_SSE2(%%xmm4, %%xmm0, %%xmm6, %%xmm3, %%xmm2, %%xmm5, %%xmm7, %%xmm1)
- "movdqa %%xmm6, 0x60(%1) \n"
- "movdqa %%xmm7, 0x70(%1) \n"
- "pxor %%xmm7, %%xmm7 \n"
- STORE_DIFF_8P(%%xmm2, (%0), %%xmm6, %%xmm7)
- STORE_DIFF_8P(%%xmm0, (%0,%2), %%xmm6, %%xmm7)
- STORE_DIFF_8P(%%xmm1, (%0,%2,2), %%xmm6, %%xmm7)
- STORE_DIFF_8P(%%xmm3, (%0,%3), %%xmm6, %%xmm7)
- "lea (%0,%2,4), %0 \n"
- STORE_DIFF_8P(%%xmm5, (%0), %%xmm6, %%xmm7)
- STORE_DIFF_8P(%%xmm4, (%0,%2), %%xmm6, %%xmm7)
- "movdqa 0x60(%1), %%xmm0 \n"
- "movdqa 0x70(%1), %%xmm1 \n"
- STORE_DIFF_8P(%%xmm0, (%0,%2,2), %%xmm6, %%xmm7)
- STORE_DIFF_8P(%%xmm1, (%0,%3), %%xmm6, %%xmm7)
- :"+r"(dst)
- :"r"(block), "r"((x86_reg)stride), "r"((x86_reg)3L*stride), "m"(ff_pw_32)
- );
-}
-
-static void ff_h264_idct_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
-{
- int dc = (block[0] + 32) >> 6;
- __asm__ volatile(
- "movd %0, %%mm0 \n\t"
- "pshufw $0, %%mm0, %%mm0 \n\t"
- "pxor %%mm1, %%mm1 \n\t"
- "psubw %%mm0, %%mm1 \n\t"
- "packuswb %%mm0, %%mm0 \n\t"
- "packuswb %%mm1, %%mm1 \n\t"
- ::"r"(dc)
- );
- __asm__ volatile(
- "movd %0, %%mm2 \n\t"
- "movd %1, %%mm3 \n\t"
- "movd %2, %%mm4 \n\t"
- "movd %3, %%mm5 \n\t"
- "paddusb %%mm0, %%mm2 \n\t"
- "paddusb %%mm0, %%mm3 \n\t"
- "paddusb %%mm0, %%mm4 \n\t"
- "paddusb %%mm0, %%mm5 \n\t"
- "psubusb %%mm1, %%mm2 \n\t"
- "psubusb %%mm1, %%mm3 \n\t"
- "psubusb %%mm1, %%mm4 \n\t"
- "psubusb %%mm1, %%mm5 \n\t"
- "movd %%mm2, %0 \n\t"
- "movd %%mm3, %1 \n\t"
- "movd %%mm4, %2 \n\t"
- "movd %%mm5, %3 \n\t"
- :"+m"(*(uint32_t*)(dst+0*stride)),
- "+m"(*(uint32_t*)(dst+1*stride)),
- "+m"(*(uint32_t*)(dst+2*stride)),
- "+m"(*(uint32_t*)(dst+3*stride))
- );
-}
-
-static void ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride)
-{
- int dc = (block[0] + 32) >> 6;
- int y;
- __asm__ volatile(
- "movd %0, %%mm0 \n\t"
- "pshufw $0, %%mm0, %%mm0 \n\t"
- "pxor %%mm1, %%mm1 \n\t"
- "psubw %%mm0, %%mm1 \n\t"
- "packuswb %%mm0, %%mm0 \n\t"
- "packuswb %%mm1, %%mm1 \n\t"
- ::"r"(dc)
- );
- for(y=2; y--; dst += 4*stride){
- __asm__ volatile(
- "movq %0, %%mm2 \n\t"
- "movq %1, %%mm3 \n\t"
- "movq %2, %%mm4 \n\t"
- "movq %3, %%mm5 \n\t"
- "paddusb %%mm0, %%mm2 \n\t"
- "paddusb %%mm0, %%mm3 \n\t"
- "paddusb %%mm0, %%mm4 \n\t"
- "paddusb %%mm0, %%mm5 \n\t"
- "psubusb %%mm1, %%mm2 \n\t"
- "psubusb %%mm1, %%mm3 \n\t"
- "psubusb %%mm1, %%mm4 \n\t"
- "psubusb %%mm1, %%mm5 \n\t"
- "movq %%mm2, %0 \n\t"
- "movq %%mm3, %1 \n\t"
- "movq %%mm4, %2 \n\t"
- "movq %%mm5, %3 \n\t"
- :"+m"(*(uint64_t*)(dst+0*stride)),
- "+m"(*(uint64_t*)(dst+1*stride)),
- "+m"(*(uint64_t*)(dst+2*stride)),
- "+m"(*(uint64_t*)(dst+3*stride))
- );
- }
-}
-
-//FIXME this table is a duplicate from h264data.h, and will be removed once the tables from, h264 have been split
-static const uint8_t scan8[16 + 2*4]={
- 4+1*8, 5+1*8, 4+2*8, 5+2*8,
- 6+1*8, 7+1*8, 6+2*8, 7+2*8,
- 4+3*8, 5+3*8, 4+4*8, 5+4*8,
- 6+3*8, 7+3*8, 6+4*8, 7+4*8,
- 1+1*8, 2+1*8,
- 1+2*8, 2+2*8,
- 1+4*8, 2+4*8,
- 1+5*8, 2+5*8,
-};
-
-static void ff_h264_idct_add16_mmx(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i++){
- if(nnzc[ scan8[i] ])
- ff_h264_idct_add_mmx(dst + block_offset[i], block + i*16, stride);
- }
-}
-
-static void ff_h264_idct8_add4_mmx(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i+=4){
- if(nnzc[ scan8[i] ])
- ff_h264_idct8_add_mmx(dst + block_offset[i], block + i*16, stride);
- }
-}
-
-
-static void ff_h264_idct_add16_mmx2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i++){
- int nnz = nnzc[ scan8[i] ];
- if(nnz){
- if(nnz==1 && block[i*16]) ff_h264_idct_dc_add_mmx2(dst + block_offset[i], block + i*16, stride);
- else ff_h264_idct_add_mmx (dst + block_offset[i], block + i*16, stride);
- }
- }
-}
-
-static void ff_h264_idct_add16intra_mmx(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i++){
- if(nnzc[ scan8[i] ] || block[i*16])
- ff_h264_idct_add_mmx(dst + block_offset[i], block + i*16, stride);
- }
-}
-
-static void ff_h264_idct_add16intra_mmx2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i++){
- if(nnzc[ scan8[i] ]) ff_h264_idct_add_mmx (dst + block_offset[i], block + i*16, stride);
- else if(block[i*16]) ff_h264_idct_dc_add_mmx2(dst + block_offset[i], block + i*16, stride);
- }
-}
-
-static void ff_h264_idct8_add4_mmx2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i+=4){
- int nnz = nnzc[ scan8[i] ];
- if(nnz){
- if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_mmx2(dst + block_offset[i], block + i*16, stride);
- else ff_h264_idct8_add_mmx (dst + block_offset[i], block + i*16, stride);
- }
- }
-}
-
-static void ff_h264_idct8_add4_sse2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i+=4){
- int nnz = nnzc[ scan8[i] ];
- if(nnz){
- if(nnz==1 && block[i*16]) ff_h264_idct8_dc_add_mmx2(dst + block_offset[i], block + i*16, stride);
- else ff_h264_idct8_add_sse2 (dst + block_offset[i], block + i*16, stride);
- }
- }
-}
-
-static void ff_h264_idct_add8_mmx(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=16; i<16+8; i++){
- if(nnzc[ scan8[i] ] || block[i*16])
- ff_h264_idct_add_mmx (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
- }
-}
-
-static void ff_h264_idct_add8_mmx2(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=16; i<16+8; i++){
- if(nnzc[ scan8[i] ])
- ff_h264_idct_add_mmx (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
- else if(block[i*16])
- ff_h264_idct_dc_add_mmx2(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
- }
-}
-
-#if HAVE_YASM
-static void ff_h264_idct_dc_add8_mmx2(uint8_t *dst, int16_t *block, int stride)
-{
- __asm__ volatile(
- "movd %0, %%mm0 \n\t" // 0 0 X D
- "punpcklwd %1, %%mm0 \n\t" // x X d D
- "paddsw %2, %%mm0 \n\t"
- "psraw $6, %%mm0 \n\t"
- "punpcklwd %%mm0, %%mm0 \n\t" // d d D D
- "pxor %%mm1, %%mm1 \n\t" // 0 0 0 0
- "psubw %%mm0, %%mm1 \n\t" // -d-d-D-D
- "packuswb %%mm1, %%mm0 \n\t" // -d-d-D-D d d D D
- "pshufw $0xFA, %%mm0, %%mm1 \n\t" // -d-d-d-d-D-D-D-D
- "punpcklwd %%mm0, %%mm0 \n\t" // d d d d D D D D
- ::"m"(block[ 0]),
- "m"(block[16]),
- "m"(ff_pw_32)
- );
- __asm__ volatile(
- "movq %0, %%mm2 \n\t"
- "movq %1, %%mm3 \n\t"
- "movq %2, %%mm4 \n\t"
- "movq %3, %%mm5 \n\t"
- "paddusb %%mm0, %%mm2 \n\t"
- "paddusb %%mm0, %%mm3 \n\t"
- "paddusb %%mm0, %%mm4 \n\t"
- "paddusb %%mm0, %%mm5 \n\t"
- "psubusb %%mm1, %%mm2 \n\t"
- "psubusb %%mm1, %%mm3 \n\t"
- "psubusb %%mm1, %%mm4 \n\t"
- "psubusb %%mm1, %%mm5 \n\t"
- "movq %%mm2, %0 \n\t"
- "movq %%mm3, %1 \n\t"
- "movq %%mm4, %2 \n\t"
- "movq %%mm5, %3 \n\t"
- :"+m"(*(uint64_t*)(dst+0*stride)),
- "+m"(*(uint64_t*)(dst+1*stride)),
- "+m"(*(uint64_t*)(dst+2*stride)),
- "+m"(*(uint64_t*)(dst+3*stride))
- );
-}
-
-extern void ff_x264_add8x4_idct_sse2(uint8_t *dst, int16_t *block, int stride);
-
-static void ff_h264_idct_add16_sse2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i+=2)
- if(nnzc[ scan8[i+0] ]|nnzc[ scan8[i+1] ])
- ff_x264_add8x4_idct_sse2 (dst + block_offset[i], block + i*16, stride);
-}
+void ff_h264_idct_add_mmx (uint8_t *dst, int16_t *block, int stride);
+void ff_h264_idct8_add_mmx (uint8_t *dst, int16_t *block, int stride);
+void ff_h264_idct8_add_sse2 (uint8_t *dst, int16_t *block, int stride);
+void ff_h264_idct_dc_add_mmx2 (uint8_t *dst, int16_t *block, int stride);
+void ff_h264_idct8_dc_add_mmx2(uint8_t *dst, int16_t *block, int stride);
-static void ff_h264_idct_add16intra_sse2(uint8_t *dst, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=0; i<16; i+=2){
- if(nnzc[ scan8[i+0] ]|nnzc[ scan8[i+1] ])
- ff_x264_add8x4_idct_sse2 (dst + block_offset[i], block + i*16, stride);
- else if(block[i*16]|block[i*16+16])
- ff_h264_idct_dc_add8_mmx2(dst + block_offset[i], block + i*16, stride);
- }
-}
+void ff_h264_idct_add16_mmx (uint8_t *dst, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct8_add4_mmx (uint8_t *dst, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add16_mmx2 (uint8_t *dst, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add16intra_mmx (uint8_t *dst, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add16intra_mmx2(uint8_t *dst, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct8_add4_mmx2 (uint8_t *dst, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct8_add4_sse2 (uint8_t *dst, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add8_mmx (uint8_t **dest, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add8_mmx2 (uint8_t **dest, const int *block_offset,
+ DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
-static void ff_h264_idct_add8_sse2(uint8_t **dest, const int *block_offset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]){
- int i;
- for(i=16; i<16+8; i+=2){
- if(nnzc[ scan8[i+0] ]|nnzc[ scan8[i+1] ])
- ff_x264_add8x4_idct_sse2 (dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
- else if(block[i*16]|block[i*16+16])
- ff_h264_idct_dc_add8_mmx2(dest[(i&4)>>2] + block_offset[i], block + i*16, stride);
- }
-}
-#endif
+void ff_h264_idct_add16_sse2 (uint8_t *dst, const int *block_offset, DCTELEM *block,
+ int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add16intra_sse2(uint8_t *dst, const int *block_offset, DCTELEM *block,
+ int stride, const uint8_t nnzc[6*8]);
+void ff_h264_idct_add8_sse2 (uint8_t **dest, const int *block_offset, DCTELEM *block,
+ int stride, const uint8_t nnzc[6*8]);
/***********************************/
/* deblocking */
@@ -745,6 +259,10 @@ void ff_h264dsp_init_x86(H264DSPContext
{
int mm_flags = av_get_cpu_flags();
+ if (mm_flags & AV_CPU_FLAG_MMX2) {
+ c->h264_loop_filter_strength= h264_loop_filter_strength_mmx2;
+ }
+#if HAVE_YASM
if (mm_flags & AV_CPU_FLAG_MMX) {
c->h264_idct_dc_add=
c->h264_idct_add= ff_h264_idct_add_mmx;
@@ -764,15 +282,6 @@ void ff_h264dsp_init_x86(H264DSPContext
c->h264_idct_add8 = ff_h264_idct_add8_mmx2;
c->h264_idct_add16intra= ff_h264_idct_add16intra_mmx2;
- c->h264_loop_filter_strength= h264_loop_filter_strength_mmx2;
- }
- if(mm_flags & AV_CPU_FLAG_SSE2){
- c->h264_idct8_add = ff_h264_idct8_add_sse2;
- c->h264_idct8_add4= ff_h264_idct8_add4_sse2;
- }
-
-#if HAVE_YASM
- if (mm_flags & AV_CPU_FLAG_MMX2){
c->h264_v_loop_filter_chroma= ff_x264_deblock_v_chroma_mmxext;
c->h264_h_loop_filter_chroma= ff_x264_deblock_h_chroma_mmxext;
c->h264_v_loop_filter_chroma_intra= ff_x264_deblock_v_chroma_intra_mmxext;
@@ -802,6 +311,9 @@ void ff_h264dsp_init_x86(H264DSPContext
c->biweight_h264_pixels_tab[7]= ff_h264_biweight_4x2_mmx2;
if (mm_flags&AV_CPU_FLAG_SSE2) {
+ c->h264_idct8_add = ff_h264_idct8_add_sse2;
+ c->h264_idct8_add4= ff_h264_idct8_add4_sse2;
+
c->weight_h264_pixels_tab[0]= ff_h264_weight_16x16_sse2;
c->weight_h264_pixels_tab[1]= ff_h264_weight_16x8_sse2;
c->weight_h264_pixels_tab[2]= ff_h264_weight_8x16_sse2;
@@ -832,6 +344,6 @@ void ff_h264dsp_init_x86(H264DSPContext
c->biweight_h264_pixels_tab[4]= ff_h264_biweight_8x4_ssse3;
}
}
-#endif
}
+#endif
}
More information about the ffmpeg-cvslog
mailing list