generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
1743f9486b
vpx/vp8/encoder/x86/dct_mmx.asm
Yaowu Xu d0dd01b8ce Redo the forward 4x4 dct 
The new fdct lowers the round trip sum squared error for a
4x4 block ~0.12. or ~0.008/pixel. For reference, the old
matrix multiply version has average round trip error 1.46
for a 4x4 block.

Thanks to "derf" for his suggestions and references.

Change-Id: I5559d1e81d333b319404ab16b336b739f87afc79
2010-06-24 13:17:58 -07:00

522 lines
15 KiB
NASM
Raw Blame History

;
; Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
;
; Use of this source code is governed by a BSD-style license
; that can be found in the LICENSE file in the root of the source
; tree. An additional intellectual property rights grant can be found
; in the file PATENTS. All contributing project authors may
; be found in the AUTHORS file in the root of the source tree.
;
%include "vpx_ports/x86_abi_support.asm"
section .text
global sym(vp8_short_fdct4x4_mmx)
global sym(vp8_short_fdct8x4_wmt)
%define DCTCONSTANTSBITS (16)
%define DCTROUNDINGVALUE (1<< (DCTCONSTANTSBITS-1))
%define x_c1 (60547) ; cos(pi /8) * (1<<15)
%define x_c2 (46341) ; cos(pi*2/8) * (1<<15)
%define x_c3 (25080) ; cos(pi*3/8) * (1<<15)
;void vp8_short_fdct4x4_mmx(short *input, short *output, int pitch)
sym(vp8_short_fdct4x4_mmx):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 3
GET_GOT rbx
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;input
mov rdi, arg(1) ;output
lea rdx, [dct_const_mmx GLOBAL]
movsxd rax, dword ptr arg(2) ;pitch
lea rcx, [rsi + rax*2]
; read the input data
movq mm0, [rsi]
movq mm1, [rsi + rax ]
movq mm2, [rcx]
movq mm3, [rcx + rax]
; get the constants
;shift to left by 1 for prescision
psllw mm0, 3
psllw mm1, 3
psllw mm2, 3
psllw mm3, 3
; transpose for the second stage
movq mm4, mm0 ; 00 01 02 03
movq mm5, mm2 ; 10 11 12 03
punpcklwd mm0, mm1 ; 00 10 01 11
punpckhwd mm4, mm1 ; 02 12 03 13
punpcklwd mm2, mm3 ; 20 30 21 31
punpckhwd mm5, mm3 ; 22 32 23 33
movq mm1, mm0 ; 00 10 01 11
punpckldq mm0, mm2 ; 00 10 20 30
punpckhdq mm1, mm2 ; 01 11 21 31
movq mm2, mm4 ; 02 12 03 13
punpckldq mm2, mm5 ; 02 12 22 32
punpckhdq mm4, mm5 ; 03 13 23 33
movq mm3, mm4
; first stage
movq mm5, mm0
movq mm4, mm1
paddw mm0, mm3 ; a = 0 + 3
paddw mm1, mm2 ; b = 1 + 2
psubw mm4, mm2 ; c = 1 - 2
psubw mm5, mm3 ; d = 0 - 3
; output 0 and 2
movq mm6, [rdx + 16] ; c2
movq mm2, mm0 ; a
paddw mm0, mm1 ; a + b
psubw mm2, mm1 ; a - b
movq mm1, mm0 ; a + b
pmulhw mm0, mm6 ; 00 01 02 03
paddw mm0, mm1 ; output 00 01 02 03
pmulhw mm6, mm2 ; 20 21 22 23
paddw mm2, mm6 ; output 20 21 22 23
; output 1 and 3
movq mm6, [rdx + 8] ; c1
movq mm7, [rdx + 24] ; c3
movq mm1, mm4 ; c
movq mm3, mm5 ; d
pmulhw mm1, mm7 ; c * c3
pmulhw mm3, mm6 ; d * c1
paddw mm3, mm5 ; d * c1 rounded
paddw mm1, mm3 ; output 10 11 12 13
movq mm3, mm4 ; c
pmulhw mm5, mm7 ; d * c3
pmulhw mm4, mm6 ; c * c1
paddw mm3, mm4 ; round c* c1
psubw mm5, mm3 ; output 30 31 32 33
movq mm3, mm5
; done with vertical
; transpose for the second stage
movq mm4, mm0 ; 00 01 02 03
movq mm5, mm2 ; 10 11 12 03
punpcklwd mm0, mm1 ; 00 10 01 11
punpckhwd mm4, mm1 ; 02 12 03 13
punpcklwd mm2, mm3 ; 20 30 21 31
punpckhwd mm5, mm3 ; 22 32 23 33
movq mm1, mm0 ; 00 10 01 11
punpckldq mm0, mm2 ; 00 10 20 30
punpckhdq mm1, mm2 ; 01 11 21 31
movq mm2, mm4 ; 02 12 03 13
punpckldq mm2, mm5 ; 02 12 22 32
punpckhdq mm4, mm5 ; 03 13 23 33
movq mm3, mm4
; first stage
movq mm5, mm0
movq mm4, mm1
paddw mm0, mm3 ; a = 0 + 3
paddw mm1, mm2 ; b = 1 + 2
psubw mm4, mm2 ; c = 1 - 2
psubw mm5, mm3 ; d = 0 - 3
; output 0 and 2
movq mm6, [rdx + 16] ; c2
movq mm2, mm0 ; a
paddw mm0, mm1 ; a + b
psubw mm2, mm1 ; a - b
movq mm1, mm0 ; a + b
pmulhw mm0, mm6 ; 00 01 02 03
paddw mm0, mm1 ; output 00 01 02 03
pmulhw mm6, mm2 ; 20 21 22 23
paddw mm2, mm6 ; output 20 21 22 23
; output 1 and 3
movq mm6, [rdx + 8] ; c1
movq mm7, [rdx + 24] ; c3
movq mm1, mm4 ; c
movq mm3, mm5 ; d
pmulhw mm1, mm7 ; c * c3
pmulhw mm3, mm6 ; d * c1
paddw mm3, mm5 ; d * c1 rounded
paddw mm1, mm3 ; output 10 11 12 13
movq mm3, mm4 ; c
pmulhw mm5, mm7 ; d * c3
pmulhw mm4, mm6 ; c * c1
paddw mm3, mm4 ; round c* c1
psubw mm5, mm3 ; output 30 31 32 33
movq mm3, mm5
; done with vertical
pcmpeqw mm4, mm4
pcmpeqw mm5, mm5
psrlw mm4, 15
psrlw mm5, 15
psllw mm4, 2
psllw mm5, 2
paddw mm0, mm4
paddw mm1, mm5
paddw mm2, mm4
paddw mm3, mm5
psraw mm0, 3
psraw mm1, 3
psraw mm2, 3
psraw mm3, 3
movq [rdi ], mm0
movq [rdi+ 8], mm1
movq [rdi+16], mm2
movq [rdi+24], mm3
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
;void vp8_short_fdct8x4_wmt(short *input, short *output, int pitch)
sym(vp8_short_fdct8x4_wmt):
push rbp
mov rbp, rsp
SHADOW_ARGS_TO_STACK 3
GET_GOT rbx
push rsi
push rdi
; end prolog
mov rsi, arg(0) ;input
mov rdi, arg(1) ;output
lea rdx, [dct_const_xmm GLOBAL]
movsxd rax, dword ptr arg(2) ;pitch
lea rcx, [rsi + rax*2]
; read the input data
movdqa xmm0, [rsi]
movdqa xmm2, [rsi + rax]
movdqa xmm4, [rcx]
movdqa xmm3, [rcx + rax]
; get the constants
;shift to left by 1 for prescision
psllw xmm0, 3
psllw xmm2, 3
psllw xmm4, 3
psllw xmm3, 3
; transpose for the second stage
movdqa xmm1, xmm0 ; 00 01 02 03 04 05 06 07
movdqa xmm5, xmm4 ; 20 21 22 23 24 25 26 27
punpcklwd xmm0, xmm2 ; 00 10 01 11 02 12 03 13
punpckhwd xmm1, xmm2 ; 04 14 05 15 06 16 07 17
punpcklwd xmm4, xmm3 ; 20 30 21 31 22 32 23 33
punpckhwd xmm5, xmm3 ; 24 34 25 35 26 36 27 37
movdqa xmm2, xmm0 ; 00 10 01 11 02 12 03 13
punpckldq xmm0, xmm4 ; 00 10 20 30 01 11 21 31
punpckhdq xmm2, xmm4 ; 02 12 22 32 03 13 23 33
movdqa xmm4, xmm1 ; 04 14 05 15 06 16 07 17
punpckldq xmm4, xmm5 ; 04 14 24 34 05 15 25 35
punpckhdq xmm1, xmm5 ; 06 16 26 36 07 17 27 37
movdqa xmm3, xmm2 ; 02 12 22 32 03 13 23 33
punpckhqdq xmm3, xmm1 ; 03 13 23 33 07 17 27 37
punpcklqdq xmm2, xmm1 ; 02 12 22 32 06 16 26 36
movdqa xmm1, xmm0 ; 00 10 20 30 01 11 21 31
punpcklqdq xmm0, xmm4 ; 00 10 20 30 04 14 24 34
punpckhqdq xmm1, xmm4 ; 01 11 21 32 05 15 25 35
; xmm0 0
; xmm1 1
; xmm2 2
; xmm3 3
; first stage
movdqa xmm5, xmm0
movdqa xmm4, xmm1
paddw xmm0, xmm3 ; a = 0 + 3
paddw xmm1, xmm2 ; b = 1 + 2
psubw xmm4, xmm2 ; c = 1 - 2
psubw xmm5, xmm3 ; d = 0 - 3
; output 0 and 2
movdqa xmm6, [rdx + 32] ; c2
movdqa xmm2, xmm0 ; a
paddw xmm0, xmm1 ; a + b
psubw xmm2, xmm1 ; a - b
movdqa xmm1, xmm0 ; a + b
pmulhw xmm0, xmm6 ; 00 01 02 03
paddw xmm0, xmm1 ; output 00 01 02 03
pmulhw xmm6, xmm2 ; 20 21 22 23
paddw xmm2, xmm6 ; output 20 21 22 23
; output 1 and 3
movdqa xmm6, [rdx + 16] ; c1
movdqa xmm7, [rdx + 48] ; c3
movdqa xmm1, xmm4 ; c
movdqa xmm3, xmm5 ; d
pmulhw xmm1, xmm7 ; c * c3
pmulhw xmm3, xmm6 ; d * c1
paddw xmm3, xmm5 ; d * c1 rounded
paddw xmm1, xmm3 ; output 10 11 12 13
movdqa xmm3, xmm4 ; c
pmulhw xmm5, xmm7 ; d * c3
pmulhw xmm4, xmm6 ; c * c1
paddw xmm3, xmm4 ; round c* c1
psubw xmm5, xmm3 ; output 30 31 32 33
movdqa xmm3, xmm5
; done with vertical
; transpose for the second stage
movdqa xmm4, xmm2 ; 02 12 22 32 06 16 26 36
movdqa xmm2, xmm1 ; 01 11 21 31 05 15 25 35
movdqa xmm1, xmm0 ; 00 10 20 30 04 14 24 34
movdqa xmm5, xmm4 ; 02 12 22 32 06 16 26 36
punpcklwd xmm0, xmm2 ; 00 01 10 11 20 21 30 31
punpckhwd xmm1, xmm2 ; 04 05 14 15 24 25 34 35
punpcklwd xmm4, xmm3 ; 02 03 12 13 22 23 32 33
punpckhwd xmm5, xmm3 ; 06 07 16 17 26 27 36 37
movdqa xmm2, xmm0 ; 00 01 10 11 20 21 30 31
punpckldq xmm0, xmm4 ; 00 01 02 03 10 11 12 13
punpckhdq xmm2, xmm4 ; 20 21 22 23 30 31 32 33
movdqa xmm4, xmm1 ; 04 05 14 15 24 25 34 35
punpckldq xmm4, xmm5 ; 04 05 06 07 14 15 16 17
punpckhdq xmm1, xmm5 ; 24 25 26 27 34 35 36 37
movdqa xmm3, xmm2 ; 20 21 22 23 30 31 32 33
punpckhqdq xmm3, xmm1 ; 30 31 32 33 34 35 36 37
punpcklqdq xmm2, xmm1 ; 20 21 22 23 24 25 26 27
movdqa xmm1, xmm0 ; 00 01 02 03 10 11 12 13
punpcklqdq xmm0, xmm4 ; 00 01 02 03 04 05 06 07
punpckhqdq xmm1, xmm4 ; 10 11 12 13 14 15 16 17
; first stage
movdqa xmm5, xmm0
movdqa xmm4, xmm1
paddw xmm0, xmm3 ; a = 0 + 3
paddw xmm1, xmm2 ; b = 1 + 2
psubw xmm4, xmm2 ; c = 1 - 2
psubw xmm5, xmm3 ; d = 0 - 3
; output 0 and 2
movdqa xmm6, [rdx + 32] ; c2
movdqa xmm2, xmm0 ; a
paddw xmm0, xmm1 ; a + b
psubw xmm2, xmm1 ; a - b
movdqa xmm1, xmm0 ; a + b
pmulhw xmm0, xmm6 ; 00 01 02 03
paddw xmm0, xmm1 ; output 00 01 02 03
pmulhw xmm6, xmm2 ; 20 21 22 23
paddw xmm2, xmm6 ; output 20 21 22 23
; output 1 and 3
movdqa xmm6, [rdx + 16] ; c1
movdqa xmm7, [rdx + 48] ; c3
movdqa xmm1, xmm4 ; c
movdqa xmm3, xmm5 ; d
pmulhw xmm1, xmm7 ; c * c3
pmulhw xmm3, xmm6 ; d * c1
paddw xmm3, xmm5 ; d * c1 rounded
paddw xmm1, xmm3 ; output 10 11 12 13
movdqa xmm3, xmm4 ; c
pmulhw xmm5, xmm7 ; d * c3
pmulhw xmm4, xmm6 ; c * c1
paddw xmm3, xmm4 ; round c* c1
psubw xmm5, xmm3 ; output 30 31 32 33
movdqa xmm3, xmm5
; done with vertical
pcmpeqw xmm4, xmm4
pcmpeqw xmm5, xmm5;
psrlw xmm4, 15
psrlw xmm5, 15
psllw xmm4, 2
psllw xmm5, 2
paddw xmm0, xmm4
paddw xmm1, xmm5
paddw xmm2, xmm4
paddw xmm3, xmm5
psraw xmm0, 3
psraw xmm1, 3
psraw xmm2, 3
psraw xmm3, 3
movq QWORD PTR[rdi ], xmm0
movq QWORD PTR[rdi+ 8], xmm1
movq QWORD PTR[rdi+16], xmm2
movq QWORD PTR[rdi+24], xmm3
psrldq xmm0, 8
psrldq xmm1, 8
psrldq xmm2, 8
psrldq xmm3, 8
movq QWORD PTR[rdi+32], xmm0
movq QWORD PTR[rdi+40], xmm1
movq QWORD PTR[rdi+48], xmm2
movq QWORD PTR[rdi+56], xmm3
; begin epilog
pop rdi
pop rsi
RESTORE_GOT
UNSHADOW_ARGS
pop rbp
ret
SECTION_RODATA
;static const unsigned int dct1st_stage_rounding_mmx[2] =
align 16
dct1st_stage_rounding_mmx:
times 2 dd 8192
;static const unsigned int dct2nd_stage_rounding_mmx[2] =
align 16
dct2nd_stage_rounding_mmx:
times 2 dd 32768
;static const short dct_matrix[4][4]=
align 16
dct_matrix:
times 4 dw 23170
dw 30274
dw 12540
dw -12540
dw -30274
dw 23170
times 2 dw -23170
dw 23170
dw 12540
dw -30274
dw 30274
dw -12540
;static const unsigned short dct_const_mmx[4 * 4]=
align 16
dct_const_mmx:
times 4 dw 0
times 4 dw 60547
times 4 dw 46341
times 4 dw 25080
;static const unsigned short dct_const_xmm[8 * 4]=
align 16
dct_const_xmm:
times 8 dw 0
times 8 dw 60547
times 8 dw 46341
times 8 dw 25080
Reference in New Issue View Git Blame Copy Permalink