mozilla
/
aom
зеркало из
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность

nasm: address labels 'rel label' vice 'wrt rip' 

nasm does not support `label wrt rip', it requires `rel label'. It is
still fully compatible with yasm.

Provide nasm compatibility. No binary change by this patch with yasm on
{x86_64,i686}-fedora13-linux-gnu. Few longer opcodes with nasm on
{x86_64,i686}-fedora13-linux-gnu have been checked as safe.

Change-Id: I488773a4e930a56e43b0cc72d867ee5291215f50
This commit is contained in:
Jan Kratochvil 2010-10-04 23:18:58 +02:00 коммит произвёл John Koleszar
Родитель e114f699f6
Коммит 5cdc3a4c29
15 изменённых файлов: 369 добавлений и 369 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

24
vp8/common/x86/idctllm_mmx.asm
Просмотреть файл

@ -58,11 +58,11 @@ sym(vp8_short_idct4x4llm_mmx):
movq mm5, mm1
paddw mm2, mm0 ; a1 =0+2
pmulhw mm5, [x_s1sqr2 GLOBAL] ;
pmulhw mm5, [GLOBAL(x_s1sqr2)] ;
paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
movq mm7, mm3 ;
pmulhw mm7, [x_c1sqr2less1 GLOBAL] ;
pmulhw mm7, [GLOBAL(x_c1sqr2less1)] ;
paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw mm7, mm5 ; c1
@ -70,10 +70,10 @@ sym(vp8_short_idct4x4llm_mmx):
movq mm5, mm1
movq mm4, mm3
pmulhw mm5, [x_c1sqr2less1 GLOBAL]
pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
paddw mm5, mm1
pmulhw mm3, [x_s1sqr2 GLOBAL]
pmulhw mm3, [GLOBAL(x_s1sqr2)]
paddw mm3, mm4
paddw mm3, mm5 ; d1
@ -113,11 +113,11 @@ sym(vp8_short_idct4x4llm_mmx):
movq mm5, mm1
paddw mm2, mm0 ; a1 =0+2
pmulhw mm5, [x_s1sqr2 GLOBAL] ;
pmulhw mm5, [GLOBAL(x_s1sqr2)] ;
paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
movq mm7, mm3 ;
pmulhw mm7, [x_c1sqr2less1 GLOBAL] ;
pmulhw mm7, [GLOBAL(x_c1sqr2less1)] ;
paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw mm7, mm5 ; c1
@ -125,16 +125,16 @@ sym(vp8_short_idct4x4llm_mmx):
movq mm5, mm1
movq mm4, mm3
pmulhw mm5, [x_c1sqr2less1 GLOBAL]
pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
paddw mm5, mm1
pmulhw mm3, [x_s1sqr2 GLOBAL]
pmulhw mm3, [GLOBAL(x_s1sqr2)]
paddw mm3, mm4
paddw mm3, mm5 ; d1
paddw mm0, [fours GLOBAL]
paddw mm0, [GLOBAL(fours)]
paddw mm2, [fours GLOBAL]
paddw mm2, [GLOBAL(fours)]
movq mm6, mm2 ; a1
movq mm4, mm0 ; b1
@ -196,7 +196,7 @@ sym(vp8_short_idct4x4llm_1_mmx):
mov rax, arg(0) ;input
movd mm0, [rax]
paddw mm0, [fours GLOBAL]
paddw mm0, [GLOBAL(fours)]
mov rdx, arg(1) ;output
psraw mm0, 3
@ -239,7 +239,7 @@ sym(vp8_dc_only_idct_add_mmx):
movd mm5, arg(0) ;input_dc
paddw mm5, [fours GLOBAL]
paddw mm5, [GLOBAL(fours)]
psraw mm5, 3

44
vp8/common/x86/idctllm_sse2.asm
Просмотреть файл

@ -51,7 +51,7 @@ sym(idct_dequant_0_2x_sse2):
pshufhw xmm4, xmm4, 00000000b
mov rax, arg(2) ; pre
paddw xmm4, [fours GLOBAL]
paddw xmm4, [GLOBAL(fours)]
movsxd rcx, dword ptr arg(5) ; blk_stride
psraw xmm4, 3
@ -160,11 +160,11 @@ sym(idct_dequant_full_2x_sse2):
movdqa xmm5, xmm1
paddw xmm2, xmm0 ; a1 = 0+2
pmulhw xmm5, [x_s1sqr2 GLOBAL]
pmulhw xmm5, [GLOBAL(x_s1sqr2)]
paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
movdqa xmm7, xmm3
pmulhw xmm7, [x_c1sqr2less1 GLOBAL]
pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw xmm7, xmm5 ; c1
@ -172,10 +172,10 @@ sym(idct_dequant_full_2x_sse2):
movdqa xmm5, xmm1
movdqa xmm4, xmm3
pmulhw xmm5, [x_c1sqr2less1 GLOBAL]
pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
paddw xmm5, xmm1
pmulhw xmm3, [x_s1sqr2 GLOBAL]
pmulhw xmm3, [GLOBAL(x_s1sqr2)]
paddw xmm3, xmm4
paddw xmm3, xmm5 ; d1
@ -229,11 +229,11 @@ sym(idct_dequant_full_2x_sse2):
movdqa xmm5, xmm1
paddw xmm2, xmm0 ; a1 = 0+2
pmulhw xmm5, [x_s1sqr2 GLOBAL]
pmulhw xmm5, [GLOBAL(x_s1sqr2)]
paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
movdqa xmm7, xmm3
pmulhw xmm7, [x_c1sqr2less1 GLOBAL]
pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw xmm7, xmm5 ; c1
@ -241,16 +241,16 @@ sym(idct_dequant_full_2x_sse2):
movdqa xmm5, xmm1
movdqa xmm4, xmm3
pmulhw xmm5, [x_c1sqr2less1 GLOBAL]
pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
paddw xmm5, xmm1
pmulhw xmm3, [x_s1sqr2 GLOBAL]
pmulhw xmm3, [GLOBAL(x_s1sqr2)]
paddw xmm3, xmm4
paddw xmm3, xmm5 ; d1
paddw xmm0, [fours GLOBAL]
paddw xmm0, [GLOBAL(fours)]
paddw xmm2, [fours GLOBAL]
paddw xmm2, [GLOBAL(fours)]
movdqa xmm6, xmm2 ; a1
movdqa xmm4, xmm0 ; b1
@ -394,7 +394,7 @@ sym(idct_dequant_dc_0_2x_sse2):
punpckldq xmm4, xmm4
; Rounding to dequant and downshift
paddw xmm4, [fours GLOBAL]
paddw xmm4, [GLOBAL(fours)]
psraw xmm4, 3
; Predict buffer needs to be expanded from bytes to words
@ -505,11 +505,11 @@ sym(idct_dequant_dc_full_2x_sse2):
movdqa xmm5, xmm1
paddw xmm2, xmm0 ; a1 = 0+2
pmulhw xmm5, [x_s1sqr2 GLOBAL]
pmulhw xmm5, [GLOBAL(x_s1sqr2)]
paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
movdqa xmm7, xmm3
pmulhw xmm7, [x_c1sqr2less1 GLOBAL]
pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw xmm7, xmm5 ; c1
@ -517,10 +517,10 @@ sym(idct_dequant_dc_full_2x_sse2):
movdqa xmm5, xmm1
movdqa xmm4, xmm3
pmulhw xmm5, [x_c1sqr2less1 GLOBAL]
pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
paddw xmm5, xmm1
pmulhw xmm3, [x_s1sqr2 GLOBAL]
pmulhw xmm3, [GLOBAL(x_s1sqr2)]
paddw xmm3, xmm4
paddw xmm3, xmm5 ; d1
@ -574,11 +574,11 @@ sym(idct_dequant_dc_full_2x_sse2):
movdqa xmm5, xmm1
paddw xmm2, xmm0 ; a1 = 0+2
pmulhw xmm5, [x_s1sqr2 GLOBAL]
pmulhw xmm5, [GLOBAL(x_s1sqr2)]
paddw xmm5, xmm1 ; ip1 * sin(pi/8) * sqrt(2)
movdqa xmm7, xmm3
pmulhw xmm7, [x_c1sqr2less1 GLOBAL]
pmulhw xmm7, [GLOBAL(x_c1sqr2less1)]
paddw xmm7, xmm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw xmm7, xmm5 ; c1
@ -586,16 +586,16 @@ sym(idct_dequant_dc_full_2x_sse2):
movdqa xmm5, xmm1
movdqa xmm4, xmm3
pmulhw xmm5, [x_c1sqr2less1 GLOBAL]
pmulhw xmm5, [GLOBAL(x_c1sqr2less1)]
paddw xmm5, xmm1
pmulhw xmm3, [x_s1sqr2 GLOBAL]
pmulhw xmm3, [GLOBAL(x_s1sqr2)]
paddw xmm3, xmm4
paddw xmm3, xmm5 ; d1
paddw xmm0, [fours GLOBAL]
paddw xmm0, [GLOBAL(fours)]
paddw xmm2, [fours GLOBAL]
paddw xmm2, [GLOBAL(fours)]
movdqa xmm6, xmm2 ; a1
movdqa xmm4, xmm0 ; b1

208
vp8/common/x86/loopfilter_mmx.asm
Просмотреть файл

@ -111,7 +111,7 @@ next8_h:
psubusb mm3, mm2 ; q1-=p1
psubusb mm2, mm4 ; p1-=q1
por mm2, mm3 ; abs(p1-q1)
pand mm2, [tfe GLOBAL] ; set lsb of each byte to zero
pand mm2, [GLOBAL(tfe)] ; set lsb of each byte to zero
psrlw mm2, 1 ; abs(p1-q1)/2
movq mm6, mm5 ; p0
@ -150,12 +150,12 @@ next8_h:
; start work on filters
movq mm2, [rsi+2*rax] ; p1
movq mm7, [rdi] ; q1
pxor mm2, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor mm7, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor mm2, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor mm7, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb mm2, mm7 ; p1 - q1
pand mm2, mm4 ; high var mask (hvm)(p1 - q1)
pxor mm6, [t80 GLOBAL] ; offset to convert to signed values
pxor mm0, [t80 GLOBAL] ; offset to convert to signed values
pxor mm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor mm0, [GLOBAL(t80)] ; offset to convert to signed values
movq mm3, mm0 ; q0
psubsb mm0, mm6 ; q0 - p0
paddsb mm2, mm0 ; 1 * (q0 - p0) + hvm(p1 - q1)
@ -163,8 +163,8 @@ next8_h:
paddsb mm2, mm0 ; 3 * (q0 - p0) + hvm(p1 - q1)
pand mm1, mm2 ; mask filter values we don't care about
movq mm2, mm1
paddsb mm1, [t4 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 4
paddsb mm2, [t3 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 3
paddsb mm1, [GLOBAL(t4)] ; 3* (q0 - p0) + hvm(p1 - q1) + 4
paddsb mm2, [GLOBAL(t3)] ; 3* (q0 - p0) + hvm(p1 - q1) + 3
pxor mm0, mm0 ;
pxor mm5, mm5
@ -185,29 +185,29 @@ next8_h:
movq mm5, mm0 ; save results
packsswb mm0, mm1 ; (3* (q0 - p0) + hvm(p1 - q1) + 4) >>3
paddsw mm5, [ones GLOBAL]
paddsw mm1, [ones GLOBAL]
paddsw mm5, [GLOBAL(ones)]
paddsw mm1, [GLOBAL(ones)]
psraw mm5, 1 ; partial shifted one more time for 2nd tap
psraw mm1, 1 ; partial shifted one more time for 2nd tap
packsswb mm5, mm1 ; (3* (q0 - p0) + hvm(p1 - q1) + 4) >>4
pandn mm4, mm5 ; high edge variance additive
paddsb mm6, mm2 ; p0+= p0 add
pxor mm6, [t80 GLOBAL] ; unoffset
pxor mm6, [GLOBAL(t80)] ; unoffset
movq [rsi+rax], mm6 ; write back
movq mm6, [rsi+2*rax] ; p1
pxor mm6, [t80 GLOBAL] ; reoffset
pxor mm6, [GLOBAL(t80)] ; reoffset
paddsb mm6, mm4 ; p1+= p1 add
pxor mm6, [t80 GLOBAL] ; unoffset
pxor mm6, [GLOBAL(t80)] ; unoffset
movq [rsi+2*rax], mm6 ; write back
psubsb mm3, mm0 ; q0-= q0 add
pxor mm3, [t80 GLOBAL] ; unoffset
pxor mm3, [GLOBAL(t80)] ; unoffset
movq [rsi], mm3 ; write back
psubsb mm7, mm4 ; q1-= q1 add
pxor mm7, [t80 GLOBAL] ; unoffset
pxor mm7, [GLOBAL(t80)] ; unoffset
movq [rdi], mm7 ; write back
add rsi,8
@ -403,7 +403,7 @@ next8_v:
psubusb mm5, mm1 ; q1-=p1
psubusb mm1, mm2 ; p1-=q1
por mm5, mm1 ; abs(p1-q1)
pand mm5, [tfe GLOBAL] ; set lsb of each byte to zero
pand mm5, [GLOBAL(tfe)] ; set lsb of each byte to zero
psrlw mm5, 1 ; abs(p1-q1)/2
mov rdx, arg(2) ;flimit ;
@ -455,14 +455,14 @@ next8_v:
movq mm6, [rdx+8] ; p0
movq mm0, [rdx+16] ; q0
pxor mm2, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor mm7, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor mm2, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor mm7, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb mm2, mm7 ; p1 - q1
pand mm2, mm4 ; high var mask (hvm)(p1 - q1)
pxor mm6, [t80 GLOBAL] ; offset to convert to signed values
pxor mm0, [t80 GLOBAL] ; offset to convert to signed values
pxor mm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor mm0, [GLOBAL(t80)] ; offset to convert to signed values
movq mm3, mm0 ; q0
psubsb mm0, mm6 ; q0 - p0
@ -474,9 +474,9 @@ next8_v:
pand mm1, mm2 ; mask filter values we don't care about
movq mm2, mm1
paddsb mm1, [t4 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 4
paddsb mm1, [GLOBAL(t4)] ; 3* (q0 - p0) + hvm(p1 - q1) + 4
paddsb mm2, [t3 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 3
paddsb mm2, [GLOBAL(t3)] ; 3* (q0 - p0) + hvm(p1 - q1) + 3
pxor mm0, mm0 ;
pxor mm5, mm5
@ -503,9 +503,9 @@ next8_v:
movq mm5, mm0 ; save results
packsswb mm0, mm1 ; (3* (q0 - p0) + hvm(p1 - q1) + 4) >>3
paddsw mm5, [ones GLOBAL]
paddsw mm5, [GLOBAL(ones)]
paddsw mm1, [ones GLOBAL]
paddsw mm1, [GLOBAL(ones)]
psraw mm5, 1 ; partial shifted one more time for 2nd tap
psraw mm1, 1 ; partial shifted one more time for 2nd tap
@ -514,22 +514,22 @@ next8_v:
pandn mm4, mm5 ; high edge variance additive
paddsb mm6, mm2 ; p0+= p0 add
pxor mm6, [t80 GLOBAL] ; unoffset
pxor mm6, [GLOBAL(t80)] ; unoffset
; mm6=p0 ;
movq mm1, [rdx] ; p1
pxor mm1, [t80 GLOBAL] ; reoffset
pxor mm1, [GLOBAL(t80)] ; reoffset
paddsb mm1, mm4 ; p1+= p1 add
pxor mm1, [t80 GLOBAL] ; unoffset
pxor mm1, [GLOBAL(t80)] ; unoffset
; mm6 = p0 mm1 = p1
psubsb mm3, mm0 ; q0-= q0 add
pxor mm3, [t80 GLOBAL] ; unoffset
pxor mm3, [GLOBAL(t80)] ; unoffset
; mm3 = q0
psubsb mm7, mm4 ; q1-= q1 add
pxor mm7, [t80 GLOBAL] ; unoffset
pxor mm7, [GLOBAL(t80)] ; unoffset
; mm7 = q1
; tranpose and write back
@ -708,7 +708,7 @@ next8_mbh:
psubusb mm3, mm2 ; q1-=p1
psubusb mm2, mm4 ; p1-=q1
por mm2, mm3 ; abs(p1-q1)
pand mm2, [tfe GLOBAL] ; set lsb of each byte to zero
pand mm2, [GLOBAL(tfe)] ; set lsb of each byte to zero
psrlw mm2, 1 ; abs(p1-q1)/2
movq mm6, mm5 ; p0
@ -753,12 +753,12 @@ next8_mbh:
; start work on filters
movq mm2, [rsi+2*rax] ; p1
movq mm7, [rdi] ; q1
pxor mm2, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor mm7, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor mm2, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor mm7, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb mm2, mm7 ; p1 - q1
pxor mm6, [t80 GLOBAL] ; offset to convert to signed values
pxor mm0, [t80 GLOBAL] ; offset to convert to signed values
pxor mm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor mm0, [GLOBAL(t80)] ; offset to convert to signed values
movq mm3, mm0 ; q0
psubsb mm0, mm6 ; q0 - p0
paddsb mm2, mm0 ; 1 * (q0 - p0) + (p1 - q1)
@ -772,7 +772,7 @@ next8_mbh:
pand mm2, mm4; ; Filter2 = vp8_filter & hev
movq mm5, mm2 ;
paddsb mm5, [t3 GLOBAL];
paddsb mm5, [GLOBAL(t3)];
pxor mm0, mm0 ; 0
pxor mm7, mm7 ; 0
@ -785,7 +785,7 @@ next8_mbh:
movq mm5, mm0 ; Filter2
paddsb mm2, [t4 GLOBAL] ; vp8_signed_char_clamp(Filter2 + 4)
paddsb mm2, [GLOBAL(t4)] ; vp8_signed_char_clamp(Filter2 + 4)
pxor mm0, mm0 ; 0
pxor mm7, mm7 ; 0
@ -818,10 +818,10 @@ next8_mbh:
pxor mm2, mm2
punpcklbw mm1, mm4
punpckhbw mm2, mm4
pmulhw mm1, [s27 GLOBAL]
pmulhw mm2, [s27 GLOBAL]
paddw mm1, [s63 GLOBAL]
paddw mm2, [s63 GLOBAL]
pmulhw mm1, [GLOBAL(s27)]
pmulhw mm2, [GLOBAL(s27)]
paddw mm1, [GLOBAL(s63)]
paddw mm2, [GLOBAL(s63)]
psraw mm1, 7
psraw mm2, 7
packsswb mm1, mm2
@ -829,8 +829,8 @@ next8_mbh:
psubsb mm3, mm1
paddsb mm6, mm1
pxor mm3, [t80 GLOBAL]
pxor mm6, [t80 GLOBAL]
pxor mm3, [GLOBAL(t80)]
pxor mm6, [GLOBAL(t80)]
movq [rsi+rax], mm6
movq [rsi], mm3
@ -844,10 +844,10 @@ next8_mbh:
pxor mm2, mm2
punpcklbw mm1, mm4
punpckhbw mm2, mm4
pmulhw mm1, [s18 GLOBAL]
pmulhw mm2, [s18 GLOBAL]
paddw mm1, [s63 GLOBAL]
paddw mm2, [s63 GLOBAL]
pmulhw mm1, [GLOBAL(s18)]
pmulhw mm2, [GLOBAL(s18)]
paddw mm1, [GLOBAL(s63)]
paddw mm2, [GLOBAL(s63)]
psraw mm1, 7
psraw mm2, 7
packsswb mm1, mm2
@ -855,14 +855,14 @@ next8_mbh:
movq mm3, [rdi]
movq mm6, [rsi+rax*2] ; p1
pxor mm3, [t80 GLOBAL]
pxor mm6, [t80 GLOBAL]
pxor mm3, [GLOBAL(t80)]
pxor mm6, [GLOBAL(t80)]
paddsb mm6, mm1
psubsb mm3, mm1
pxor mm6, [t80 GLOBAL]
pxor mm3, [t80 GLOBAL]
pxor mm6, [GLOBAL(t80)]
pxor mm3, [GLOBAL(t80)]
movq [rdi], mm3
movq [rsi+rax*2], mm6
@ -876,10 +876,10 @@ next8_mbh:
pxor mm2, mm2
punpcklbw mm1, mm4
punpckhbw mm2, mm4
pmulhw mm1, [s9 GLOBAL]
pmulhw mm2, [s9 GLOBAL]
paddw mm1, [s63 GLOBAL]
paddw mm2, [s63 GLOBAL]
pmulhw mm1, [GLOBAL(s9)]
pmulhw mm2, [GLOBAL(s9)]
paddw mm1, [GLOBAL(s63)]
paddw mm2, [GLOBAL(s63)]
psraw mm1, 7
psraw mm2, 7
packsswb mm1, mm2
@ -889,14 +889,14 @@ next8_mbh:
neg rax
movq mm3, [rdi+rax ]
pxor mm6, [t80 GLOBAL]
pxor mm3, [t80 GLOBAL]
pxor mm6, [GLOBAL(t80)]
pxor mm3, [GLOBAL(t80)]
paddsb mm6, mm1
psubsb mm3, mm1
pxor mm6, [t80 GLOBAL]
pxor mm3, [t80 GLOBAL]
pxor mm6, [GLOBAL(t80)]
pxor mm3, [GLOBAL(t80)]
movq [rdi+rax ], mm3
neg rax
movq [rdi+rax*4], mm6
@ -1105,7 +1105,7 @@ next8_mbv:
psubusb mm5, mm1 ; q1-=p1
psubusb mm1, mm2 ; p1-=q1
por mm5, mm1 ; abs(p1-q1)
pand mm5, [tfe GLOBAL] ; set lsb of each byte to zero
pand mm5, [GLOBAL(tfe)] ; set lsb of each byte to zero
psrlw mm5, 1 ; abs(p1-q1)/2
mov rdx, arg(2) ;flimit ;
@ -1155,14 +1155,14 @@ next8_mbv:
; start work on filters
movq mm2, [rdx+16] ; p1
movq mm7, [rdx+40] ; q1
pxor mm2, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor mm7, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor mm2, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor mm7, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb mm2, mm7 ; p1 - q1
movq mm6, [rdx+24] ; p0
movq mm0, [rdx+32] ; q0
pxor mm6, [t80 GLOBAL] ; offset to convert to signed values
pxor mm0, [t80 GLOBAL] ; offset to convert to signed values
pxor mm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor mm0, [GLOBAL(t80)] ; offset to convert to signed values
movq mm3, mm0 ; q0
psubsb mm0, mm6 ; q0 - p0
@ -1176,7 +1176,7 @@ next8_mbv:
pand mm2, mm4; ; Filter2 = vp8_filter & hev
movq mm5, mm2 ;
paddsb mm5, [t3 GLOBAL];
paddsb mm5, [GLOBAL(t3)];
pxor mm0, mm0 ; 0
pxor mm7, mm7 ; 0
@ -1189,7 +1189,7 @@ next8_mbv:
movq mm5, mm0 ; Filter2
paddsb mm2, [t4 GLOBAL] ; vp8_signed_char_clamp(Filter2 + 4)
paddsb mm2, [GLOBAL(t4)] ; vp8_signed_char_clamp(Filter2 + 4)
pxor mm0, mm0 ; 0
pxor mm7, mm7 ; 0
@ -1222,10 +1222,10 @@ next8_mbv:
pxor mm2, mm2
punpcklbw mm1, mm4
punpckhbw mm2, mm4
pmulhw mm1, [s27 GLOBAL]
pmulhw mm2, [s27 GLOBAL]
paddw mm1, [s63 GLOBAL]
paddw mm2, [s63 GLOBAL]
pmulhw mm1, [GLOBAL(s27)]
pmulhw mm2, [GLOBAL(s27)]
paddw mm1, [GLOBAL(s63)]
paddw mm2, [GLOBAL(s63)]
psraw mm1, 7
psraw mm2, 7
packsswb mm1, mm2
@ -1233,8 +1233,8 @@ next8_mbv:
psubsb mm3, mm1
paddsb mm6, mm1
pxor mm3, [t80 GLOBAL]
pxor mm6, [t80 GLOBAL]
pxor mm3, [GLOBAL(t80)]
pxor mm6, [GLOBAL(t80)]
movq [rdx+24], mm6
movq [rdx+32], mm3
@ -1248,24 +1248,24 @@ next8_mbv:
pxor mm2, mm2
punpcklbw mm1, mm4
punpckhbw mm2, mm4
pmulhw mm1, [s18 GLOBAL]
pmulhw mm2, [s18 GLOBAL]
paddw mm1, [s63 GLOBAL]
paddw mm2, [s63 GLOBAL]
pmulhw mm1, [GLOBAL(s18)]
pmulhw mm2, [GLOBAL(s18)]
paddw mm1, [GLOBAL(s63)]
paddw mm2, [GLOBAL(s63)]
psraw mm1, 7
psraw mm2, 7
packsswb mm1, mm2
movq mm3, [rdx + 40]
movq mm6, [rdx + 16] ; p1
pxor mm3, [t80 GLOBAL]
pxor mm6, [t80 GLOBAL]
pxor mm3, [GLOBAL(t80)]
pxor mm6, [GLOBAL(t80)]
paddsb mm6, mm1
psubsb mm3, mm1
pxor mm6, [t80 GLOBAL]
pxor mm3, [t80 GLOBAL]
pxor mm6, [GLOBAL(t80)]
pxor mm3, [GLOBAL(t80)]
movq [rdx + 40], mm3
movq [rdx + 16], mm6
@ -1279,10 +1279,10 @@ next8_mbv:
pxor mm2, mm2
punpcklbw mm1, mm4
punpckhbw mm2, mm4
pmulhw mm1, [s9 GLOBAL]
pmulhw mm2, [s9 GLOBAL]
paddw mm1, [s63 GLOBAL]
paddw mm2, [s63 GLOBAL]
pmulhw mm1, [GLOBAL(s9)]
pmulhw mm2, [GLOBAL(s9)]
paddw mm1, [GLOBAL(s63)]
paddw mm2, [GLOBAL(s63)]
psraw mm1, 7
psraw mm2, 7
packsswb mm1, mm2
@ -1290,14 +1290,14 @@ next8_mbv:
movq mm6, [rdx+ 8]
movq mm3, [rdx+48]
pxor mm6, [t80 GLOBAL]
pxor mm3, [t80 GLOBAL]
pxor mm6, [GLOBAL(t80)]
pxor mm3, [GLOBAL(t80)]
paddsb mm6, mm1
psubsb mm3, mm1
pxor mm6, [t80 GLOBAL] ; mm6 = 71 61 51 41 31 21 11 01
pxor mm3, [t80 GLOBAL] ; mm3 = 76 66 56 46 36 26 15 06
pxor mm6, [GLOBAL(t80)] ; mm6 = 71 61 51 41 31 21 11 01
pxor mm3, [GLOBAL(t80)] ; mm3 = 76 66 56 46 36 26 15 06
; tranpose and write back
movq mm0, [rdx] ; mm0 = 70 60 50 40 30 20 10 00
@ -1432,7 +1432,7 @@ nexts8_h:
psubusb mm0, mm1 ; q1-=p1
psubusb mm1, mm4 ; p1-=q1
por mm1, mm0 ; abs(p1-q1)
pand mm1, [tfe GLOBAL] ; set lsb of each byte to zero
pand mm1, [GLOBAL(tfe)] ; set lsb of each byte to zero
psrlw mm1, 1 ; abs(p1-q1)/2
movq mm5, [rsi+rax] ; p0
@ -1450,12 +1450,12 @@ nexts8_h:
pcmpeqb mm5, mm3
; start work on filters
pxor mm2, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor mm7, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor mm2, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor mm7, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb mm2, mm7 ; p1 - q1
pxor mm6, [t80 GLOBAL] ; offset to convert to signed values
pxor mm0, [t80 GLOBAL] ; offset to convert to signed values
pxor mm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor mm0, [GLOBAL(t80)] ; offset to convert to signed values
movq mm3, mm0 ; q0
psubsb mm0, mm6 ; q0 - p0
paddsb mm2, mm0 ; p1 - q1 + 1 * (q0 - p0)
@ -1464,7 +1464,7 @@ nexts8_h:
pand mm5, mm2 ; mask filter values we don't care about
; do + 4 side
paddsb mm5, [t4 GLOBAL] ; 3* (q0 - p0) + (p1 - q1) + 4
paddsb mm5, [GLOBAL(t4)] ; 3* (q0 - p0) + (p1 - q1) + 4
movq mm0, mm5 ; get a copy of filters
psllw mm0, 8 ; shift left 8
@ -1477,12 +1477,12 @@ nexts8_h:
por mm0, mm1 ; put the two together to get result
psubsb mm3, mm0 ; q0-= q0 add
pxor mm3, [t80 GLOBAL] ; unoffset
pxor mm3, [GLOBAL(t80)] ; unoffset
movq [rsi], mm3 ; write back
; now do +3 side
psubsb mm5, [t1s GLOBAL] ; +3 instead of +4
psubsb mm5, [GLOBAL(t1s)] ; +3 instead of +4
movq mm0, mm5 ; get a copy of filters
psllw mm0, 8 ; shift left 8
@ -1494,7 +1494,7 @@ nexts8_h:
paddsb mm6, mm0 ; p0+= p0 add
pxor mm6, [t80 GLOBAL] ; unoffset
pxor mm6, [GLOBAL(t80)] ; unoffset
movq [rsi+rax], mm6 ; write back
add rsi,8
@ -1589,7 +1589,7 @@ nexts8_v:
psubusb mm7, mm6 ; q1-=p1
psubusb mm6, mm3 ; p1-=q1
por mm6, mm7 ; abs(p1-q1)
pand mm6, [tfe GLOBAL] ; set lsb of each byte to zero
pand mm6, [GLOBAL(tfe)] ; set lsb of each byte to zero
psrlw mm6, 1 ; abs(p1-q1)/2
movq mm5, mm1 ; p0
@ -1617,16 +1617,16 @@ nexts8_v:
movq t0, mm0
movq t1, mm3
pxor mm0, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor mm3, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor mm0, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor mm3, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb mm0, mm3 ; p1 - q1
movq mm6, mm1 ; p0
movq mm7, mm2 ; q0
pxor mm6, [t80 GLOBAL] ; offset to convert to signed values
pxor mm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor mm7, [t80 GLOBAL] ; offset to convert to signed values
pxor mm7, [GLOBAL(t80)] ; offset to convert to signed values
movq mm3, mm7 ; offseted ; q0
psubsb mm7, mm6 ; q0 - p0
@ -1637,7 +1637,7 @@ nexts8_v:
pand mm5, mm0 ; mask filter values we don't care about
paddsb mm5, [t4 GLOBAL] ; 3* (q0 - p0) + (p1 - q1) + 4
paddsb mm5, [GLOBAL(t4)] ; 3* (q0 - p0) + (p1 - q1) + 4
movq mm0, mm5 ; get a copy of filters
psllw mm0, 8 ; shift left 8
@ -1651,10 +1651,10 @@ nexts8_v:
por mm0, mm7 ; put the two together to get result
psubsb mm3, mm0 ; q0-= q0sz add
pxor mm3, [t80 GLOBAL] ; unoffset
pxor mm3, [GLOBAL(t80)] ; unoffset
; now do +3 side
psubsb mm5, [t1s GLOBAL] ; +3 instead of +4
psubsb mm5, [GLOBAL(t1s)] ; +3 instead of +4
movq mm0, mm5 ; get a copy of filters
psllw mm0, 8 ; shift left 8
@ -1666,7 +1666,7 @@ nexts8_v:
por mm0, mm5 ; put the two together to get result
paddsb mm6, mm0 ; p0+= p0 add
pxor mm6, [t80 GLOBAL] ; unoffset
pxor mm6, [GLOBAL(t80)] ; unoffset
movq mm0, t0

106
vp8/common/x86/loopfilter_sse2.asm
Просмотреть файл

@ -126,7 +126,7 @@
movdqa xmm4, XMMWORD PTR [rdx] ; flimit
movdqa xmm3, xmm0 ; q0
pand xmm2, [tfe GLOBAL] ; set lsb of each byte to zero
pand xmm2, [GLOBAL(tfe)] ; set lsb of each byte to zero
mov rdx, arg(4) ; hev get thresh
@ -182,14 +182,14 @@
movdqa xmm0, [rdx+32] ; q0
%endif
pxor xmm2, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor xmm7, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor xmm2, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor xmm7, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb xmm2, xmm7 ; p1 - q1
pxor xmm6, [t80 GLOBAL] ; offset to convert to signed values
pxor xmm6, [GLOBAL(t80)] ; offset to convert to signed values
pand xmm2, xmm4 ; high var mask (hvm)(p1 - q1)
pxor xmm0, [t80 GLOBAL] ; offset to convert to signed values
pxor xmm0, [GLOBAL(t80)] ; offset to convert to signed values
movdqa xmm3, xmm0 ; q0
psubsb xmm0, xmm6 ; q0 - p0
@ -204,8 +204,8 @@
movdqa xmm2, xmm1
paddsb xmm1, [t4 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 4
paddsb xmm2, [t3 GLOBAL] ; 3* (q0 - p0) + hvm(p1 - q1) + 3
paddsb xmm1, [GLOBAL(t4)] ; 3* (q0 - p0) + hvm(p1 - q1) + 4
paddsb xmm2, [GLOBAL(t3)] ; 3* (q0 - p0) + hvm(p1 - q1) + 3
punpckhbw xmm5, xmm2 ; axbxcxdx
punpcklbw xmm2, xmm2 ; exfxgxhx
@ -223,9 +223,9 @@
movdqa xmm5, xmm0 ; save results
packsswb xmm0, xmm1 ; (3* (q0 - p0) + hvm(p1 - q1) + 4) >>3
paddsw xmm5, [ones GLOBAL]
paddsw xmm5, [GLOBAL(ones)]
paddsw xmm1, [ones GLOBAL]
paddsw xmm1, [GLOBAL(ones)]
psraw xmm5, 1 ; partial shifted one more time for 2nd tap
psraw xmm1, 1 ; partial shifted one more time for 2nd tap
@ -241,18 +241,18 @@
movdqa xmm1, [rdx] ; p1
%endif
pandn xmm4, xmm5 ; high edge variance additive
pxor xmm6, [t80 GLOBAL] ; unoffset
pxor xmm6, [GLOBAL(t80)] ; unoffset
pxor xmm1, [t80 GLOBAL] ; reoffset
pxor xmm1, [GLOBAL(t80)] ; reoffset
psubsb xmm3, xmm0 ; q0-= q0 add
paddsb xmm1, xmm4 ; p1+= p1 add
pxor xmm3, [t80 GLOBAL] ; unoffset
pxor xmm3, [GLOBAL(t80)] ; unoffset
pxor xmm1, [t80 GLOBAL] ; unoffset
pxor xmm1, [GLOBAL(t80)] ; unoffset
psubsb xmm7, xmm4 ; q1-= q1 add
pxor xmm7, [t80 GLOBAL] ; unoffset
pxor xmm7, [GLOBAL(t80)] ; unoffset
%if %1 == 0
lea rsi, [rsi + rcx*2]
lea rdi, [rdi + rcx*2]
@ -401,10 +401,10 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
movdqa xmm0, [rdx+64] ; q0
%endif
pxor xmm2, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor xmm7, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor xmm6, [t80 GLOBAL] ; offset to convert to signed values
pxor xmm0, [t80 GLOBAL] ; offset to convert to signed values
pxor xmm2, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor xmm7, [GLOBAL(t80)] ; q1 offset to convert to signed values
pxor xmm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor xmm0, [GLOBAL(t80)] ; offset to convert to signed values
psubsb xmm2, xmm7 ; p1 - q1
movdqa xmm3, xmm0 ; q0
@ -431,14 +431,14 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
movdqa xmm5, xmm2
punpckhbw xmm1, xmm4 ; Filter 2 (lo)
paddsb xmm5, [t3 GLOBAL] ; vp8_signed_char_clamp(Filter2 + 3)
paddsb xmm5, [GLOBAL(t3)] ; vp8_signed_char_clamp(Filter2 + 3)
pmulhw xmm1, [s9 GLOBAL] ; Filter 2 (lo) * 9
pmulhw xmm1, [GLOBAL(s9)] ; Filter 2 (lo) * 9
pmulhw xmm0, [s9 GLOBAL] ; Filter 2 (hi) * 9
pmulhw xmm0, [GLOBAL(s9)] ; Filter 2 (hi) * 9
punpckhbw xmm7, xmm5 ; axbxcxdx
paddsb xmm2, [t4 GLOBAL] ; vp8_signed_char_clamp(Filter2 + 4)
paddsb xmm2, [GLOBAL(t4)] ; vp8_signed_char_clamp(Filter2 + 4)
punpcklbw xmm5, xmm5 ; exfxgxhx
psraw xmm7, 11 ; sign extended shift right by 3
@ -462,9 +462,9 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
movdqa xmm5, xmm0
movdqa xmm2, xmm5
paddw xmm0, [s63 GLOBAL] ; Filter 2 (hi) * 9 + 63
paddw xmm0, [GLOBAL(s63)] ; Filter 2 (hi) * 9 + 63
paddw xmm1, [s63 GLOBAL] ; Filter 2 (lo) * 9 + 63
paddw xmm1, [GLOBAL(s63)] ; Filter 2 (lo) * 9 + 63
paddw xmm5, xmm5 ; Filter 2 (hi) * 18
paddw xmm7, xmm7 ; Filter 2 (lo) * 18
@ -510,26 +510,26 @@ sym(vp8_loop_filter_horizontal_edge_uv_sse2):
movdqa xmm7, XMMWORD PTR [rdx+16] ; p2
%endif
pxor xmm3, [t80 GLOBAL] ; *oq0 = sq^0x80
pxor xmm6, [t80 GLOBAL] ; *oq0 = sp^0x80
pxor xmm3, [GLOBAL(t80)] ; *oq0 = sq^0x80
pxor xmm6, [GLOBAL(t80)] ; *oq0 = sp^0x80
pxor xmm1, [t80 GLOBAL]
pxor xmm4, [t80 GLOBAL]
pxor xmm1, [GLOBAL(t80)]
pxor xmm4, [GLOBAL(t80)]
psubsb xmm1, xmm2 ; sq = vp8_signed_char_clamp(qs1 - u2)
paddsb xmm4, xmm2 ; sp = vp8_signed_char_clamp(ps1 - u2)
pxor xmm1, [t80 GLOBAL] ; *oq1 = sq^0x80;
pxor xmm4, [t80 GLOBAL] ; *op1 = sp^0x80;
pxor xmm1, [GLOBAL(t80)] ; *oq1 = sq^0x80;
pxor xmm4, [GLOBAL(t80)] ; *op1 = sp^0x80;
pxor xmm7, [t80 GLOBAL]
pxor xmm5, [t80 GLOBAL]
pxor xmm7, [GLOBAL(t80)]
pxor xmm5, [GLOBAL(t80)]
paddsb xmm7, xmm0 ; sp = vp8_signed_char_clamp(ps2 - u)
psubsb xmm5, xmm0 ; sq = vp8_signed_char_clamp(qs2 - u)
pxor xmm7, [t80 GLOBAL] ; *op2 = sp^0x80;
pxor xmm5, [t80 GLOBAL] ; *oq2 = sq^0x80;
pxor xmm7, [GLOBAL(t80)] ; *op2 = sp^0x80;
pxor xmm5, [GLOBAL(t80)] ; *oq2 = sq^0x80;
%if %1 == 0
lea rsi, [rsi+rcx*2]
@ -915,7 +915,7 @@ sym(vp8_mbloop_filter_horizontal_edge_uv_sse2):
por xmm5, xmm1 ; abs(p1-q1)
movdqa xmm1, xmm3 ; p0
pand xmm5, [tfe GLOBAL] ; set lsb of each byte to zero
pand xmm5, [GLOBAL(tfe)] ; set lsb of each byte to zero
psubusb xmm1, xmm6 ; p0-q0
psrlw xmm5, 1 ; abs(p1-q1)/2
@ -1415,7 +1415,7 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
psubusb xmm0, xmm1 ; q1-=p1
psubusb xmm1, xmm4 ; p1-=q1
por xmm1, xmm0 ; abs(p1-q1)
pand xmm1, [tfe GLOBAL] ; set lsb of each byte to zero
pand xmm1, [GLOBAL(tfe)] ; set lsb of each byte to zero
psrlw xmm1, 1 ; abs(p1-q1)/2
movdqu xmm5, [rsi+rax] ; p0
@ -1433,12 +1433,12 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
pcmpeqb xmm5, xmm3
; start work on filters
pxor xmm2, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor xmm7, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor xmm2, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor xmm7, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb xmm2, xmm7 ; p1 - q1
pxor xmm6, [t80 GLOBAL] ; offset to convert to signed values
pxor xmm0, [t80 GLOBAL] ; offset to convert to signed values
pxor xmm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor xmm0, [GLOBAL(t80)] ; offset to convert to signed values
movdqa xmm3, xmm0 ; q0
psubsb xmm0, xmm6 ; q0 - p0
paddsb xmm2, xmm0 ; p1 - q1 + 1 * (q0 - p0)
@ -1447,7 +1447,7 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
pand xmm5, xmm2 ; mask filter values we don't care about
; do + 4 side
paddsb xmm5, [t4 GLOBAL] ; 3* (q0 - p0) + (p1 - q1) + 4
paddsb xmm5, [GLOBAL(t4)] ; 3* (q0 - p0) + (p1 - q1) + 4
movdqa xmm0, xmm5 ; get a copy of filters
psllw xmm0, 8 ; shift left 8
@ -1460,11 +1460,11 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
por xmm0, xmm1 ; put the two together to get result
psubsb xmm3, xmm0 ; q0-= q0 add
pxor xmm3, [t80 GLOBAL] ; unoffset
pxor xmm3, [GLOBAL(t80)] ; unoffset
movdqu [rsi], xmm3 ; write back
; now do +3 side
psubsb xmm5, [t1s GLOBAL] ; +3 instead of +4
psubsb xmm5, [GLOBAL(t1s)] ; +3 instead of +4
movdqa xmm0, xmm5 ; get a copy of filters
psllw xmm0, 8 ; shift left 8
@ -1476,7 +1476,7 @@ sym(vp8_loop_filter_simple_horizontal_edge_sse2):
paddsb xmm6, xmm0 ; p0+= p0 add
pxor xmm6, [t80 GLOBAL] ; unoffset
pxor xmm6, [GLOBAL(t80)] ; unoffset
movdqu [rsi+rax], xmm6 ; write back
; begin epilog
@ -1596,7 +1596,7 @@ sym(vp8_loop_filter_simple_vertical_edge_sse2):
psubusb xmm7, xmm0 ; q1-=p1
psubusb xmm6, xmm3 ; p1-=q1
por xmm6, xmm7 ; abs(p1-q1)
pand xmm6, [tfe GLOBAL] ; set lsb of each byte to zero
pand xmm6, [GLOBAL(tfe)] ; set lsb of each byte to zero
psrlw xmm6, 1 ; abs(p1-q1)/2
movdqa xmm5, xmm1 ; p0
@ -1622,16 +1622,16 @@ sym(vp8_loop_filter_simple_vertical_edge_sse2):
movdqa t0, xmm0
movdqa t1, xmm3
pxor xmm0, [t80 GLOBAL] ; p1 offset to convert to signed values
pxor xmm3, [t80 GLOBAL] ; q1 offset to convert to signed values
pxor xmm0, [GLOBAL(t80)] ; p1 offset to convert to signed values
pxor xmm3, [GLOBAL(t80)] ; q1 offset to convert to signed values
psubsb xmm0, xmm3 ; p1 - q1
movdqa xmm6, xmm1 ; p0
movdqa xmm7, xmm2 ; q0
pxor xmm6, [t80 GLOBAL] ; offset to convert to signed values
pxor xmm6, [GLOBAL(t80)] ; offset to convert to signed values
pxor xmm7, [t80 GLOBAL] ; offset to convert to signed values
pxor xmm7, [GLOBAL(t80)] ; offset to convert to signed values
movdqa xmm3, xmm7 ; offseted ; q0
psubsb xmm7, xmm6 ; q0 - p0
@ -1643,7 +1643,7 @@ sym(vp8_loop_filter_simple_vertical_edge_sse2):
pand xmm5, xmm0 ; mask filter values we don't care about
paddsb xmm5, [t4 GLOBAL] ; 3* (q0 - p0) + (p1 - q1) + 4
paddsb xmm5, [GLOBAL(t4)] ; 3* (q0 - p0) + (p1 - q1) + 4
movdqa xmm0, xmm5 ; get a copy of filters
psllw xmm0, 8 ; shift left 8
@ -1658,10 +1658,10 @@ sym(vp8_loop_filter_simple_vertical_edge_sse2):
por xmm0, xmm7 ; put the two together to get result
psubsb xmm3, xmm0 ; q0-= q0sz add
pxor xmm3, [t80 GLOBAL] ; unoffset q0
pxor xmm3, [GLOBAL(t80)] ; unoffset q0
; now do +3 side
psubsb xmm5, [t1s GLOBAL] ; +3 instead of +4
psubsb xmm5, [GLOBAL(t1s)] ; +3 instead of +4
movdqa xmm0, xmm5 ; get a copy of filters
psllw xmm0, 8 ; shift left 8
@ -1674,7 +1674,7 @@ sym(vp8_loop_filter_simple_vertical_edge_sse2):
por xmm0, xmm5 ; put the two together to get result
paddsb xmm6, xmm0 ; p0+= p0 add
pxor xmm6, [t80 GLOBAL] ; unoffset p0
pxor xmm6, [GLOBAL(t80)] ; unoffset p0
movdqa xmm0, t0 ; p1
movdqa xmm4, t1 ; q1

10
vp8/common/x86/postproc_mmx.asm
Просмотреть файл

@ -37,16 +37,16 @@ sym(vp8_post_proc_down_and_across_mmx):
%if ABI_IS_32BIT=1 && CONFIG_PIC=1
; move the global rd onto the stack, since we don't have enough registers
; to do PIC addressing
movq mm0, [rd GLOBAL]
movq mm0, [GLOBAL(rd)]
sub rsp, 8
movq [rsp], mm0
%define RD [rsp]
%else
%define RD [rd GLOBAL]
%define RD [GLOBAL(rd)]
%endif
push rbx
lea rbx, [Blur GLOBAL]
lea rbx, [GLOBAL(Blur)]
movd mm2, dword ptr arg(6) ;flimit
punpcklwd mm2, mm2
punpckldq mm2, mm2
@ -286,7 +286,7 @@ sym(vp8_mbpost_proc_down_mmx):
%define flimit2 [rsp+128]
%if ABI_IS_32BIT=0
lea r8, [sym(vp8_rv) GLOBAL]
lea r8, [GLOBAL(sym(vp8_rv))]
%endif
;rows +=8;
@ -404,7 +404,7 @@ loop_row:
and rcx, 127
%if ABI_IS_32BIT=1 && CONFIG_PIC=1
push rax
lea rax, [sym(vp8_rv) GLOBAL]
lea rax, [GLOBAL(sym(vp8_rv))]
movq mm4, [rax + rcx*2] ;vp8_rv[rcx*2]
pop rax
%elif ABI_IS_32BIT=0

10
vp8/common/x86/postproc_sse2.asm
Просмотреть файл

@ -36,12 +36,12 @@ sym(vp8_post_proc_down_and_across_xmm):
ALIGN_STACK 16, rax
; move the global rd onto the stack, since we don't have enough registers
; to do PIC addressing
movdqa xmm0, [rd42 GLOBAL]
movdqa xmm0, [GLOBAL(rd42)]
sub rsp, 16
movdqa [rsp], xmm0
%define RD42 [rsp]
%else
%define RD42 [rd42 GLOBAL]
%define RD42 [GLOBAL(rd42)]
%endif
@ -275,7 +275,7 @@ sym(vp8_mbpost_proc_down_xmm):
%define flimit4 [rsp+128]
%if ABI_IS_32BIT=0
lea r8, [sym(vp8_rv) GLOBAL]
lea r8, [GLOBAL(sym(vp8_rv))]
%endif
;rows +=8;
@ -393,7 +393,7 @@ loop_row:
and rcx, 127
%if ABI_IS_32BIT=1 && CONFIG_PIC=1
push rax
lea rax, [sym(vp8_rv) GLOBAL]
lea rax, [GLOBAL(sym(vp8_rv))]
movdqu xmm4, [rax + rcx*2] ;vp8_rv[rcx*2]
pop rax
%elif ABI_IS_32BIT=0
@ -579,7 +579,7 @@ nextcol4:
punpcklwd xmm1, xmm0
paddd xmm1, xmm6
paddd xmm1, [four8s GLOBAL]
paddd xmm1, [GLOBAL(four8s)]
psrad xmm1, 4
packssdw xmm1, xmm0

42
vp8/common/x86/subpixel_mmx.asm
Просмотреть файл

@ -84,7 +84,7 @@ nextrow:
pmullw mm5, [rdx] ; mm5 *= kernel 5 modifiers
paddsw mm3, mm5 ; mm3 += mm5
paddsw mm3, [rd GLOBAL] ; mm3 += round value
paddsw mm3, [GLOBAL(rd)] ; mm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; mm3 /= 128
packuswb mm3, mm0 ; pack and unpack to saturate
punpcklbw mm3, mm0 ;
@ -136,7 +136,7 @@ sym(vp8_filter_block1d_v6_mmx):
push rdi
; end prolog
movq mm5, [rd GLOBAL]
movq mm5, [GLOBAL(rd)]
push rbx
mov rbx, arg(6) ;vp8_filter
movq mm1, [rbx + 16] ; do both the negative taps first!!!
@ -225,7 +225,7 @@ sym(vp8_filter_block1dc_v6_mmx):
push rdi
; end prolog
movq mm5, [rd GLOBAL]
movq mm5, [GLOBAL(rd)]
push rbx
mov rbx, arg(7) ;vp8_filter
movq mm1, [rbx + 16] ; do both the negative taps first!!!
@ -320,7 +320,7 @@ sym(vp8_bilinear_predict8x8_mmx):
mov rdi, arg(4) ;dst_ptr ;
shl rax, 5 ; offset * 32
lea rcx, [sym(vp8_bilinear_filters_mmx) GLOBAL]
lea rcx, [GLOBAL(sym(vp8_bilinear_filters_mmx))]
add rax, rcx ; HFilter
mov rsi, arg(0) ;src_ptr ;
@ -363,10 +363,10 @@ sym(vp8_bilinear_predict8x8_mmx):
paddw mm3, mm5 ;
paddw mm4, mm6 ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [rd GLOBAL] ;
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP8_FILTER_SHIFT ;
movq mm7, mm3 ;
@ -404,10 +404,10 @@ next_row_8x8:
pmullw mm5, [rax] ;
pmullw mm6, [rax] ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [rd GLOBAL] ;
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP8_FILTER_SHIFT ;
movq mm7, mm3 ;
@ -421,10 +421,10 @@ next_row_8x8:
paddw mm4, mm6 ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [rd GLOBAL] ;
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP8_FILTER_SHIFT ;
packuswb mm3, mm4
@ -476,7 +476,7 @@ sym(vp8_bilinear_predict8x4_mmx):
movsxd rax, dword ptr arg(2) ;xoffset
mov rdi, arg(4) ;dst_ptr ;
lea rcx, [sym(vp8_bilinear_filters_mmx) GLOBAL]
lea rcx, [GLOBAL(sym(vp8_bilinear_filters_mmx))]
shl rax, 5
mov rsi, arg(0) ;src_ptr ;
@ -518,10 +518,10 @@ sym(vp8_bilinear_predict8x4_mmx):
paddw mm3, mm5 ;
paddw mm4, mm6 ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [rd GLOBAL] ;
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP8_FILTER_SHIFT ;
movq mm7, mm3 ;
@ -559,10 +559,10 @@ next_row_8x4:
pmullw mm5, [rax] ;
pmullw mm6, [rax] ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [rd GLOBAL] ;
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP8_FILTER_SHIFT ;
movq mm7, mm3 ;
@ -576,10 +576,10 @@ next_row_8x4:
paddw mm4, mm6 ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw mm4, [rd GLOBAL] ;
paddw mm4, [GLOBAL(rd)] ;
psraw mm4, VP8_FILTER_SHIFT ;
packuswb mm3, mm4
@ -631,7 +631,7 @@ sym(vp8_bilinear_predict4x4_mmx):
movsxd rax, dword ptr arg(2) ;xoffset
mov rdi, arg(4) ;dst_ptr ;
lea rcx, [sym(vp8_bilinear_filters_mmx) GLOBAL]
lea rcx, [GLOBAL(sym(vp8_bilinear_filters_mmx))]
shl rax, 5
add rax, rcx ; HFilter
@ -662,7 +662,7 @@ sym(vp8_bilinear_predict4x4_mmx):
pmullw mm5, mm2 ;
paddw mm3, mm5 ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
@ -686,7 +686,7 @@ next_row_4x4:
punpcklbw mm5, mm0 ;
pmullw mm5, [rax] ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
movq mm7, mm3 ;
@ -697,7 +697,7 @@ next_row_4x4:
paddw mm3, mm5 ;
paddw mm3, [rd GLOBAL] ; xmm3 += round value
paddw mm3, [GLOBAL(rd)] ; xmm3 += round value
psraw mm3, VP8_FILTER_SHIFT ; xmm3 /= 128
packuswb mm3, mm0

48
vp8/common/x86/subpixel_sse2.asm
Просмотреть файл

@ -107,7 +107,7 @@ filter_block1d8_h6_rowloop:
paddsw xmm4, xmm6
paddsw xmm4, xmm1
paddsw xmm4, [rd GLOBAL]
paddsw xmm4, [GLOBAL(rd)]
psraw xmm4, 7
@ -231,7 +231,7 @@ filter_block1d16_h6_sse2_rowloop:
paddsw xmm4, xmm6
paddsw xmm4, xmm1
paddsw xmm4, [rd GLOBAL]
paddsw xmm4, [GLOBAL(rd)]
psraw xmm4, 7
@ -284,7 +284,7 @@ filter_block1d16_h6_sse2_rowloop:
paddsw xmm4, xmm6
paddsw xmm4, xmm2
paddsw xmm4, [rd GLOBAL]
paddsw xmm4, [GLOBAL(rd)]
psraw xmm4, 7
@ -351,7 +351,7 @@ sym(vp8_filter_block1d8_v6_sse2):
movsxd rcx, DWORD PTR arg(5) ;[output_height]
pxor xmm0, xmm0 ; clear xmm0
movdqa xmm7, XMMWORD PTR [rd GLOBAL]
movdqa xmm7, XMMWORD PTR [GLOBAL(rd)]
%if ABI_IS_32BIT=0
movsxd r8, dword ptr arg(2) ; dst_ptich
%endif
@ -489,7 +489,7 @@ vp8_filter_block1d16_v6_sse2_loop:
pmullw xmm5, [rax + 80]
pmullw xmm6, [rax + 80]
movdqa xmm7, XMMWORD PTR [rd GLOBAL]
movdqa xmm7, XMMWORD PTR [GLOBAL(rd)]
pxor xmm0, xmm0 ; clear xmm0
paddsw xmm1, xmm3
@ -608,7 +608,7 @@ filter_block1d8_h6_only_rowloop:
paddsw xmm4, xmm6
paddsw xmm4, xmm1
paddsw xmm4, [rd GLOBAL]
paddsw xmm4, [GLOBAL(rd)]
psraw xmm4, 7
@ -723,7 +723,7 @@ filter_block1d16_h6_only_sse2_rowloop:
paddsw xmm4, xmm6
paddsw xmm4, xmm1
paddsw xmm4, [rd GLOBAL]
paddsw xmm4, [GLOBAL(rd)]
psraw xmm4, 7
@ -773,7 +773,7 @@ filter_block1d16_h6_only_sse2_rowloop:
paddsw xmm4, xmm6
paddsw xmm4, xmm2
paddsw xmm4, [rd GLOBAL]
paddsw xmm4, [GLOBAL(rd)]
psraw xmm4, 7
@ -832,7 +832,7 @@ sym(vp8_filter_block1d8_v6_only_sse2):
pxor xmm0, xmm0 ; clear xmm0
movdqa xmm7, XMMWORD PTR [rd GLOBAL]
movdqa xmm7, XMMWORD PTR [GLOBAL(rd)]
%if ABI_IS_32BIT=0
movsxd r8, dword ptr arg(3) ; dst_ptich
%endif
@ -978,7 +978,7 @@ sym(vp8_bilinear_predict16x16_sse2):
;const short *HFilter = bilinear_filters_mmx[xoffset]
;const short *VFilter = bilinear_filters_mmx[yoffset]
lea rcx, [sym(vp8_bilinear_filters_mmx) GLOBAL]
lea rcx, [GLOBAL(sym(vp8_bilinear_filters_mmx))]
movsxd rax, dword ptr arg(2) ;xoffset
cmp rax, 0 ;skip first_pass filter if xoffset=0
@ -1033,10 +1033,10 @@ sym(vp8_bilinear_predict16x16_sse2):
paddw xmm3, xmm5
paddw xmm4, xmm6
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw xmm4, [rd GLOBAL]
paddw xmm4, [GLOBAL(rd)]
psraw xmm4, VP8_FILTER_SHIFT
movdqa xmm7, xmm3
@ -1074,10 +1074,10 @@ next_row:
pmullw xmm5, [rax]
pmullw xmm6, [rax]
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw xmm4, [rd GLOBAL]
paddw xmm4, [GLOBAL(rd)]
psraw xmm4, VP8_FILTER_SHIFT
movdqa xmm7, xmm3
@ -1089,10 +1089,10 @@ next_row:
paddw xmm3, xmm5
paddw xmm4, xmm6
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw xmm4, [rd GLOBAL]
paddw xmm4, [GLOBAL(rd)]
psraw xmm4, VP8_FILTER_SHIFT
packuswb xmm3, xmm4
@ -1154,10 +1154,10 @@ next_row_spo:
paddw xmm3, xmm5
paddw xmm4, xmm6
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw xmm4, [rd GLOBAL]
paddw xmm4, [GLOBAL(rd)]
psraw xmm4, VP8_FILTER_SHIFT
packuswb xmm3, xmm4
@ -1198,10 +1198,10 @@ next_row_fpo:
paddw xmm3, xmm5
paddw xmm4, xmm6
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw xmm4, [rd GLOBAL]
paddw xmm4, [GLOBAL(rd)]
psraw xmm4, VP8_FILTER_SHIFT
packuswb xmm3, xmm4
@ -1249,7 +1249,7 @@ sym(vp8_bilinear_predict8x8_sse2):
;const short *HFilter = bilinear_filters_mmx[xoffset]
;const short *VFilter = bilinear_filters_mmx[yoffset]
lea rcx, [sym(vp8_bilinear_filters_mmx) GLOBAL]
lea rcx, [GLOBAL(sym(vp8_bilinear_filters_mmx))]
mov rsi, arg(0) ;src_ptr
movsxd rdx, dword ptr arg(1) ;src_pixels_per_line
@ -1315,7 +1315,7 @@ sym(vp8_bilinear_predict8x8_sse2):
paddw xmm3, xmm4
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
movdqa xmm7, xmm3
@ -1334,7 +1334,7 @@ next_row8x8:
paddw xmm3, xmm4
pmullw xmm7, xmm5
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
movdqa xmm4, xmm3
@ -1344,7 +1344,7 @@ next_row8x8:
movdqa xmm7, xmm4
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
packuswb xmm3, xmm0

140
vp8/common/x86/subpixel_ssse3.asm
Просмотреть файл

@ -48,9 +48,9 @@ sym(vp8_filter_block1d8_h6_ssse3):
xor rsi, rsi
shl rdx, 4
movdqa xmm7, [rd GLOBAL]
movdqa xmm7, [GLOBAL(rd)]
lea rax, [k0_k5 GLOBAL]
lea rax, [GLOBAL(k0_k5)]
add rax, rdx
mov rdi, arg(2) ;output_ptr
@ -80,9 +80,9 @@ filter_block1d8_h6_rowloop_ssse3:
pmaddubsw xmm0, xmm4
movdqa xmm2, xmm1
pshufb xmm1, [shuf2bfrom1 GLOBAL]
pshufb xmm1, [GLOBAL(shuf2bfrom1)]
pshufb xmm2, [shuf3bfrom1 GLOBAL]
pshufb xmm2, [GLOBAL(shuf3bfrom1)]
pmaddubsw xmm1, xmm5
lea rdi, [rdi + rdx]
@ -115,8 +115,8 @@ vp8_filter_block1d8_h4_ssse3:
movdqa xmm5, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm6, XMMWORD PTR [rax+128] ;k1_k3
movdqa xmm3, XMMWORD PTR [shuf2bfrom1 GLOBAL]
movdqa xmm4, XMMWORD PTR [shuf3bfrom1 GLOBAL]
movdqa xmm3, XMMWORD PTR [GLOBAL(shuf2bfrom1)]
movdqa xmm4, XMMWORD PTR [GLOBAL(shuf3bfrom1)]
mov rsi, arg(0) ;src_ptr
@ -189,7 +189,7 @@ sym(vp8_filter_block1d16_h6_ssse3):
xor rsi, rsi
shl rdx, 4 ;
lea rax, [k0_k5 GLOBAL]
lea rax, [GLOBAL(k0_k5)]
add rax, rdx
mov rdi, arg(2) ;output_ptr
@ -219,9 +219,9 @@ filter_block1d16_h6_rowloop_ssse3:
pmaddubsw xmm0, xmm4
movdqa xmm2, xmm1
pshufb xmm1, [shuf2bfrom1 GLOBAL]
pshufb xmm1, [GLOBAL(shuf2bfrom1)]
pshufb xmm2, [shuf3bfrom1 GLOBAL]
pshufb xmm2, [GLOBAL(shuf3bfrom1)]
movq xmm3, MMWORD PTR [rsi + 6]
pmaddubsw xmm1, xmm5
@ -237,10 +237,10 @@ filter_block1d16_h6_rowloop_ssse3:
paddsw xmm0, xmm2
movdqa xmm2, xmm1
paddsw xmm0, [rd GLOBAL]
paddsw xmm0, [GLOBAL(rd)]
pshufb xmm1, [shuf2bfrom1 GLOBAL]
pshufb xmm2, [shuf3bfrom1 GLOBAL]
pshufb xmm1, [GLOBAL(shuf2bfrom1)]
pshufb xmm2, [GLOBAL(shuf3bfrom1)]
psraw xmm0, 7
pmaddubsw xmm1, xmm5
@ -253,7 +253,7 @@ filter_block1d16_h6_rowloop_ssse3:
paddsw xmm3, xmm2
paddsw xmm3, [rd GLOBAL]
paddsw xmm3, [GLOBAL(rd)]
psraw xmm3, 7
@ -288,18 +288,18 @@ filter_block1d16_h4_rowloop_ssse3:
movdqu xmm1, XMMWORD PTR [rsi - 2]
movdqa xmm2, xmm1
pshufb xmm1, [shuf2b GLOBAL]
pshufb xmm2, [shuf3b GLOBAL]
pshufb xmm1, [GLOBAL(shuf2b)]
pshufb xmm2, [GLOBAL(shuf3b)]
pmaddubsw xmm1, xmm5
movdqu xmm3, XMMWORD PTR [rsi + 6]
pmaddubsw xmm2, xmm6
movdqa xmm0, xmm3
pshufb xmm3, [shuf3b GLOBAL]
pshufb xmm0, [shuf2b GLOBAL]
pshufb xmm3, [GLOBAL(shuf3b)]
pshufb xmm0, [GLOBAL(shuf2b)]
paddsw xmm1, [rd GLOBAL]
paddsw xmm1, [GLOBAL(rd)]
paddsw xmm1, xmm2
pmaddubsw xmm0, xmm5
@ -309,7 +309,7 @@ filter_block1d16_h4_rowloop_ssse3:
packuswb xmm1, xmm1
lea rsi, [rsi + rax]
paddsw xmm3, xmm0
paddsw xmm3, [rd GLOBAL]
paddsw xmm3, [GLOBAL(rd)]
psraw xmm3, 7
packuswb xmm3, xmm3
@ -353,9 +353,9 @@ sym(vp8_filter_block1d4_h6_ssse3):
xor rsi, rsi
shl rdx, 4 ;
lea rax, [k0_k5 GLOBAL]
lea rax, [GLOBAL(k0_k5)]
add rax, rdx
movdqa xmm7, [rd GLOBAL]
movdqa xmm7, [GLOBAL(rd)]
cmp esi, DWORD PTR [rax]
je vp8_filter_block1d4_h4_ssse3
@ -376,12 +376,12 @@ filter_block1d4_h6_rowloop_ssse3:
movdqu xmm0, XMMWORD PTR [rsi - 2]
movdqa xmm1, xmm0
pshufb xmm0, [shuf1b GLOBAL]
pshufb xmm0, [GLOBAL(shuf1b)]
movdqa xmm2, xmm1
pshufb xmm1, [shuf2b GLOBAL]
pshufb xmm1, [GLOBAL(shuf2b)]
pmaddubsw xmm0, xmm4
pshufb xmm2, [shuf3b GLOBAL]
pshufb xmm2, [GLOBAL(shuf3b)]
pmaddubsw xmm1, xmm5
;--
@ -413,8 +413,8 @@ filter_block1d4_h6_rowloop_ssse3:
vp8_filter_block1d4_h4_ssse3:
movdqa xmm5, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm6, XMMWORD PTR [rax+128] ;k1_k3
movdqa xmm0, XMMWORD PTR [shuf2b GLOBAL]
movdqa xmm3, XMMWORD PTR [shuf3b GLOBAL]
movdqa xmm0, XMMWORD PTR [GLOBAL(shuf2b)]
movdqa xmm3, XMMWORD PTR [GLOBAL(shuf3b)]
mov rsi, arg(0) ;src_ptr
mov rdi, arg(2) ;output_ptr
@ -427,8 +427,8 @@ filter_block1d4_h4_rowloop_ssse3:
movdqu xmm1, XMMWORD PTR [rsi - 2]
movdqa xmm2, xmm1
pshufb xmm1, xmm0 ;;[shuf2b GLOBAL]
pshufb xmm2, xmm3 ;;[shuf3b GLOBAL]
pshufb xmm1, xmm0 ;;[GLOBAL(shuf2b)]
pshufb xmm2, xmm3 ;;[GLOBAL(shuf3b)]
pmaddubsw xmm1, xmm5
;--
@ -480,7 +480,7 @@ sym(vp8_filter_block1d16_v6_ssse3):
xor rsi, rsi
shl rdx, 4 ;
lea rax, [k0_k5 GLOBAL]
lea rax, [GLOBAL(k0_k5)]
add rax, rdx
cmp esi, DWORD PTR [rax]
@ -521,7 +521,7 @@ vp8_filter_block1d16_v6_ssse3_loop:
paddsw xmm2, xmm3
paddsw xmm2, xmm1
paddsw xmm2, [rd GLOBAL]
paddsw xmm2, [GLOBAL(rd)]
psraw xmm2, 7
packuswb xmm2, xmm2
@ -548,7 +548,7 @@ vp8_filter_block1d16_v6_ssse3_loop:
;--
paddsw xmm2, xmm3
paddsw xmm2, xmm1
paddsw xmm2, [rd GLOBAL]
paddsw xmm2, [GLOBAL(rd)]
psraw xmm2, 7
packuswb xmm2, xmm2
@ -601,7 +601,7 @@ vp8_filter_block1d16_v4_ssse3_loop:
movq xmm4, MMWORD PTR [rax + rdx * 2 + 8] ;D
movq xmm0, MMWORD PTR [rsi + rdx * 4 + 8] ;E
paddsw xmm2, [rd GLOBAL]
paddsw xmm2, [GLOBAL(rd)]
paddsw xmm2, xmm3
psraw xmm2, 7
packuswb xmm2, xmm2
@ -612,7 +612,7 @@ vp8_filter_block1d16_v4_ssse3_loop:
pmaddubsw xmm1, xmm6
pmaddubsw xmm5, xmm7
movdqa xmm4, [rd GLOBAL]
movdqa xmm4, [GLOBAL(rd)]
add rsi, rdx
add rax, rdx
;--
@ -665,7 +665,7 @@ sym(vp8_filter_block1d8_v6_ssse3):
xor rsi, rsi
shl rdx, 4 ;
lea rax, [k0_k5 GLOBAL]
lea rax, [GLOBAL(k0_k5)]
add rax, rdx
movsxd rdx, DWORD PTR arg(1) ;pixels_per_line
@ -698,7 +698,7 @@ vp8_filter_block1d8_v6_ssse3_loop:
punpcklbw xmm3, xmm0 ;C E
movq xmm0, MMWORD PTR [rax + rdx * 4] ;F
movdqa xmm4, [rd GLOBAL]
movdqa xmm4, [GLOBAL(rd)]
pmaddubsw xmm3, xmm6
punpcklbw xmm1, xmm0 ;A F
@ -735,7 +735,7 @@ vp8_filter_block1d8_v6_ssse3_loop:
vp8_filter_block1d8_v4_ssse3:
movdqa xmm6, XMMWORD PTR [rax+256] ;k2_k4
movdqa xmm7, XMMWORD PTR [rax+128] ;k1_k3
movdqa xmm5, [rd GLOBAL]
movdqa xmm5, [GLOBAL(rd)]
mov rsi, arg(0) ;src_ptr
@ -802,7 +802,7 @@ sym(vp8_filter_block1d4_v6_ssse3):
xor rsi, rsi
shl rdx, 4 ;
lea rax, [k0_k5 GLOBAL]
lea rax, [GLOBAL(k0_k5)]
add rax, rdx
movsxd rdx, DWORD PTR arg(1) ;pixels_per_line
@ -836,7 +836,7 @@ vp8_filter_block1d4_v6_ssse3_loop:
movd mm0, DWORD PTR [rax + rdx * 4] ;F
movq mm4, [rd GLOBAL]
movq mm4, [GLOBAL(rd)]
pmaddubsw mm3, mm6
punpcklbw mm1, mm0 ;A F
@ -873,7 +873,7 @@ vp8_filter_block1d4_v6_ssse3_loop:
vp8_filter_block1d4_v4_ssse3:
movq mm6, MMWORD PTR [rax+256] ;k2_k4
movq mm7, MMWORD PTR [rax+128] ;k1_k3
movq mm5, MMWORD PTR [rd GLOBAL]
movq mm5, MMWORD PTR [GLOBAL(rd)]
mov rsi, arg(0) ;src_ptr
@ -938,7 +938,7 @@ sym(vp8_bilinear_predict16x16_ssse3):
push rdi
; end prolog
lea rcx, [vp8_bilinear_filters_ssse3 GLOBAL]
lea rcx, [GLOBAL(vp8_bilinear_filters_ssse3)]
movsxd rax, dword ptr arg(2) ; xoffset
cmp rax, 0 ; skip first_pass filter if xoffset=0
@ -985,10 +985,10 @@ sym(vp8_bilinear_predict16x16_ssse3):
punpcklbw xmm4, xmm5 ; 08 09 09 10 10 11 11 12 12 13 13 14 14 15 15 16
pmaddubsw xmm4, xmm1 ; 01 03 05 07 09 11 13 15
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
paddw xmm4, [rd GLOBAL] ; xmm4 += round value
paddw xmm4, [GLOBAL(rd)] ; xmm4 += round value
psraw xmm4, VP8_FILTER_SHIFT ; xmm4 /= 128
movdqa xmm7, xmm3
@ -1009,10 +1009,10 @@ sym(vp8_bilinear_predict16x16_ssse3):
punpcklbw xmm4, xmm5
pmaddubsw xmm4, xmm1
paddw xmm6, [rd GLOBAL] ; xmm6 += round value
paddw xmm6, [GLOBAL(rd)] ; xmm6 += round value
psraw xmm6, VP8_FILTER_SHIFT ; xmm6 /= 128
paddw xmm4, [rd GLOBAL] ; xmm4 += round value
paddw xmm4, [GLOBAL(rd)] ; xmm4 += round value
psraw xmm4, VP8_FILTER_SHIFT ; xmm4 /= 128
packuswb xmm6, xmm4
@ -1024,10 +1024,10 @@ sym(vp8_bilinear_predict16x16_ssse3):
punpckhbw xmm7, xmm6
pmaddubsw xmm7, xmm2
paddw xmm5, [rd GLOBAL] ; xmm5 += round value
paddw xmm5, [GLOBAL(rd)] ; xmm5 += round value
psraw xmm5, VP8_FILTER_SHIFT ; xmm5 /= 128
paddw xmm7, [rd GLOBAL] ; xmm7 += round value
paddw xmm7, [GLOBAL(rd)] ; xmm7 += round value
psraw xmm7, VP8_FILTER_SHIFT ; xmm7 /= 128
packuswb xmm5, xmm7
@ -1082,19 +1082,19 @@ b16x16_sp_only:
punpcklbw xmm5, xmm6
pmaddubsw xmm3, xmm1
paddw xmm4, [rd GLOBAL]
paddw xmm4, [GLOBAL(rd)]
pmaddubsw xmm5, xmm1
paddw xmm2, [rd GLOBAL]
paddw xmm2, [GLOBAL(rd)]
psraw xmm4, VP8_FILTER_SHIFT
psraw xmm2, VP8_FILTER_SHIFT
packuswb xmm4, xmm2
paddw xmm3, [rd GLOBAL]
paddw xmm3, [GLOBAL(rd)]
movdqa [rdi], xmm4 ; store row 0
paddw xmm5, [rd GLOBAL]
paddw xmm5, [GLOBAL(rd)]
psraw xmm3, VP8_FILTER_SHIFT
psraw xmm5, VP8_FILTER_SHIFT
@ -1134,7 +1134,7 @@ b16x16_fp_only:
pmaddubsw xmm3, xmm1
movq xmm5, [rsi]
paddw xmm2, [rd GLOBAL]
paddw xmm2, [GLOBAL(rd)]
movq xmm7, [rsi+1]
movq xmm6, [rsi+8]
@ -1143,7 +1143,7 @@ b16x16_fp_only:
punpcklbw xmm5, xmm7
movq xmm7, [rsi+9]
paddw xmm3, [rd GLOBAL]
paddw xmm3, [GLOBAL(rd)]
pmaddubsw xmm5, xmm1
psraw xmm3, VP8_FILTER_SHIFT
@ -1153,12 +1153,12 @@ b16x16_fp_only:
pmaddubsw xmm6, xmm1
movdqa [rdi], xmm2 ; store the results in the destination
paddw xmm5, [rd GLOBAL]
paddw xmm5, [GLOBAL(rd)]
lea rdi, [rdi + rdx] ; dst_pitch
psraw xmm5, VP8_FILTER_SHIFT
paddw xmm6, [rd GLOBAL]
paddw xmm6, [GLOBAL(rd)]
psraw xmm6, VP8_FILTER_SHIFT
packuswb xmm5, xmm6
@ -1204,7 +1204,7 @@ sym(vp8_bilinear_predict8x8_ssse3):
ALIGN_STACK 16, rax
sub rsp, 144 ; reserve 144 bytes
lea rcx, [vp8_bilinear_filters_ssse3 GLOBAL]
lea rcx, [GLOBAL(vp8_bilinear_filters_ssse3)]
mov rsi, arg(0) ;src_ptr
movsxd rdx, dword ptr arg(1) ;src_pixels_per_line
@ -1269,7 +1269,7 @@ sym(vp8_bilinear_predict8x8_ssse3):
punpcklbw xmm3, xmm5 ; 00 01 01 02 02 03 03 04 04 05 05 06 06 07 07 08
pmaddubsw xmm3, xmm0 ; 00 02 04 06 08 10 12 14
paddw xmm3, [rd GLOBAL] ; xmm3 += round value
paddw xmm3, [GLOBAL(rd)] ; xmm3 += round value
psraw xmm3, VP8_FILTER_SHIFT ; xmm3 /= 128
movdqa xmm7, xmm3
@ -1286,7 +1286,7 @@ sym(vp8_bilinear_predict8x8_ssse3):
punpcklbw xmm6, xmm5
pmaddubsw xmm6, xmm0
paddw xmm6, [rd GLOBAL] ; xmm6 += round value
paddw xmm6, [GLOBAL(rd)] ; xmm6 += round value
psraw xmm6, VP8_FILTER_SHIFT ; xmm6 /= 128
packuswb xmm6, xmm6
@ -1294,7 +1294,7 @@ sym(vp8_bilinear_predict8x8_ssse3):
punpcklbw xmm7, xmm6
pmaddubsw xmm7, xmm1
paddw xmm7, [rd GLOBAL] ; xmm7 += round value
paddw xmm7, [GLOBAL(rd)] ; xmm7 += round value
psraw xmm7, VP8_FILTER_SHIFT ; xmm7 /= 128
packuswb xmm7, xmm7
@ -1347,21 +1347,21 @@ b8x8_sp_only:
punpcklbw xmm6, xmm7
pmaddubsw xmm6, xmm0
paddw xmm1, [rd GLOBAL]
paddw xmm1, [GLOBAL(rd)]
paddw xmm2, [rd GLOBAL]
paddw xmm2, [GLOBAL(rd)]
psraw xmm1, VP8_FILTER_SHIFT
paddw xmm3, [rd GLOBAL]
paddw xmm3, [GLOBAL(rd)]
psraw xmm2, VP8_FILTER_SHIFT
paddw xmm4, [rd GLOBAL]
paddw xmm4, [GLOBAL(rd)]
psraw xmm3, VP8_FILTER_SHIFT
paddw xmm5, [rd GLOBAL]
paddw xmm5, [GLOBAL(rd)]
psraw xmm4, VP8_FILTER_SHIFT
paddw xmm6, [rd GLOBAL]
paddw xmm6, [GLOBAL(rd)]
psraw xmm5, VP8_FILTER_SHIFT
psraw xmm6, VP8_FILTER_SHIFT
@ -1395,10 +1395,10 @@ b8x8_sp_only:
punpcklbw xmm1, xmm2
pmaddubsw xmm1, xmm0
paddw xmm7, [rd GLOBAL]
paddw xmm7, [GLOBAL(rd)]
psraw xmm7, VP8_FILTER_SHIFT
paddw xmm1, [rd GLOBAL]
paddw xmm1, [GLOBAL(rd)]
psraw xmm1, VP8_FILTER_SHIFT
packuswb xmm7, xmm7
@ -1447,16 +1447,16 @@ b8x8_fp_only:
punpcklbw xmm7, xmm2
pmaddubsw xmm7, xmm0
paddw xmm1, [rd GLOBAL]
paddw xmm1, [GLOBAL(rd)]
psraw xmm1, VP8_FILTER_SHIFT
paddw xmm3, [rd GLOBAL]
paddw xmm3, [GLOBAL(rd)]
psraw xmm3, VP8_FILTER_SHIFT
paddw xmm5, [rd GLOBAL]
paddw xmm5, [GLOBAL(rd)]
psraw xmm5, VP8_FILTER_SHIFT
paddw xmm7, [rd GLOBAL]
paddw xmm7, [GLOBAL(rd)]
psraw xmm7, VP8_FILTER_SHIFT
packuswb xmm1, xmm1

40
vp8/decoder/x86/dequantize_mmx.asm
Просмотреть файл

@ -98,11 +98,11 @@ sym(vp8_dequant_idct_add_mmx):
movq mm5, mm1
paddw mm2, mm0 ; a1 =0+2
pmulhw mm5, [x_s1sqr2 GLOBAL];
pmulhw mm5, [GLOBAL(x_s1sqr2)];
paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
movq mm7, mm3 ;
pmulhw mm7, [x_c1sqr2less1 GLOBAL];
pmulhw mm7, [GLOBAL(x_c1sqr2less1)];
paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw mm7, mm5 ; c1
@ -110,10 +110,10 @@ sym(vp8_dequant_idct_add_mmx):
movq mm5, mm1
movq mm4, mm3
pmulhw mm5, [x_c1sqr2less1 GLOBAL]
pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
paddw mm5, mm1
pmulhw mm3, [x_s1sqr2 GLOBAL]
pmulhw mm3, [GLOBAL(x_s1sqr2)]
paddw mm3, mm4
paddw mm3, mm5 ; d1
@ -153,11 +153,11 @@ sym(vp8_dequant_idct_add_mmx):
movq mm5, mm1
paddw mm2, mm0 ; a1 =0+2
pmulhw mm5, [x_s1sqr2 GLOBAL];
pmulhw mm5, [GLOBAL(x_s1sqr2)];
paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
movq mm7, mm3 ;
pmulhw mm7, [x_c1sqr2less1 GLOBAL];
pmulhw mm7, [GLOBAL(x_c1sqr2less1)];
paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw mm7, mm5 ; c1
@ -165,16 +165,16 @@ sym(vp8_dequant_idct_add_mmx):
movq mm5, mm1
movq mm4, mm3
pmulhw mm5, [x_c1sqr2less1 GLOBAL]
pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
paddw mm5, mm1
pmulhw mm3, [x_s1sqr2 GLOBAL]
pmulhw mm3, [GLOBAL(x_s1sqr2)]
paddw mm3, mm4
paddw mm3, mm5 ; d1
paddw mm0, [fours GLOBAL]
paddw mm0, [GLOBAL(fours)]
paddw mm2, [fours GLOBAL]
paddw mm2, [GLOBAL(fours)]
movq mm6, mm2 ; a1
movq mm4, mm0 ; b1
@ -300,11 +300,11 @@ sym(vp8_dequant_dc_idct_add_mmx):
movq mm5, mm1
paddw mm2, mm0 ; a1 =0+2
pmulhw mm5, [x_s1sqr2 GLOBAL];
pmulhw mm5, [GLOBAL(x_s1sqr2)];
paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
movq mm7, mm3 ;
pmulhw mm7, [x_c1sqr2less1 GLOBAL];
pmulhw mm7, [GLOBAL(x_c1sqr2less1)];
paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw mm7, mm5 ; c1
@ -312,10 +312,10 @@ sym(vp8_dequant_dc_idct_add_mmx):
movq mm5, mm1
movq mm4, mm3
pmulhw mm5, [x_c1sqr2less1 GLOBAL]
pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
paddw mm5, mm1
pmulhw mm3, [x_s1sqr2 GLOBAL]
pmulhw mm3, [GLOBAL(x_s1sqr2)]
paddw mm3, mm4
paddw mm3, mm5 ; d1
@ -355,11 +355,11 @@ sym(vp8_dequant_dc_idct_add_mmx):
movq mm5, mm1
paddw mm2, mm0 ; a1 =0+2
pmulhw mm5, [x_s1sqr2 GLOBAL];
pmulhw mm5, [GLOBAL(x_s1sqr2)];
paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
movq mm7, mm3 ;
pmulhw mm7, [x_c1sqr2less1 GLOBAL];
pmulhw mm7, [GLOBAL(x_c1sqr2less1)];
paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
psubw mm7, mm5 ; c1
@ -367,16 +367,16 @@ sym(vp8_dequant_dc_idct_add_mmx):
movq mm5, mm1
movq mm4, mm3
pmulhw mm5, [x_c1sqr2less1 GLOBAL]
pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
paddw mm5, mm1
pmulhw mm3, [x_s1sqr2 GLOBAL]
pmulhw mm3, [GLOBAL(x_s1sqr2)]
paddw mm3, mm4
paddw mm3, mm5 ; d1
paddw mm0, [fours GLOBAL]
paddw mm0, [GLOBAL(fours)]
paddw mm2, [fours GLOBAL]
paddw mm2, [GLOBAL(fours)]
movq mm6, mm2 ; a1
movq mm4, mm0 ; b1

4
vp8/encoder/x86/dct_mmx.asm
Просмотреть файл

@ -35,7 +35,7 @@ sym(vp8_short_fdct4x4_mmx):
mov rsi, arg(0) ;input
mov rdi, arg(1) ;output
lea rdx, [dct_const_mmx GLOBAL]
lea rdx, [GLOBAL(dct_const_mmx)]
movsxd rax, dword ptr arg(2) ;pitch
lea rcx, [rsi + rax*2]
@ -243,7 +243,7 @@ sym(vp8_short_fdct8x4_wmt):
mov rsi, arg(0) ;input
mov rdi, arg(1) ;output
lea rdx, [dct_const_xmm GLOBAL]
lea rdx, [GLOBAL(dct_const_xmm)]
movsxd rax, dword ptr arg(2) ;pitch
lea rcx, [rsi + rax*2]

30
vp8/encoder/x86/dct_sse2.asm
Просмотреть файл

@ -52,14 +52,14 @@ sym(vp8_short_fdct4x4_sse2):
psllw xmm0, 3 ;b1 <<= 3 a1 <<= 3
psllw xmm3, 3 ;c1 <<= 3 d1 <<= 3
movdqa xmm1, xmm0
pmaddwd xmm0, XMMWORD PTR[_mult_add GLOBAL] ;a1 + b1
pmaddwd xmm1, XMMWORD PTR[_mult_sub GLOBAL] ;a1 - b1
pmaddwd xmm0, XMMWORD PTR[GLOBAL(_mult_add)] ;a1 + b1
pmaddwd xmm1, XMMWORD PTR[GLOBAL(_mult_sub)] ;a1 - b1
movdqa xmm4, xmm3
pmaddwd xmm3, XMMWORD PTR[_5352_2217 GLOBAL] ;c1*2217 + d1*5352
pmaddwd xmm4, XMMWORD PTR[_2217_neg5352 GLOBAL] ;d1*2217 - c1*5352
pmaddwd xmm3, XMMWORD PTR[GLOBAL(_5352_2217)] ;c1*2217 + d1*5352
pmaddwd xmm4, XMMWORD PTR[GLOBAL(_2217_neg5352)];d1*2217 - c1*5352
paddd xmm3, XMMWORD PTR[_14500 GLOBAL]
paddd xmm4, XMMWORD PTR[_7500 GLOBAL]
paddd xmm3, XMMWORD PTR[GLOBAL(_14500)]
paddd xmm4, XMMWORD PTR[GLOBAL(_7500)]
psrad xmm3, 12 ;(c1 * 2217 + d1 * 5352 + 14500)>>12
psrad xmm4, 12 ;(d1 * 2217 - c1 * 5352 + 7500)>>12
@ -80,7 +80,7 @@ sym(vp8_short_fdct4x4_sse2):
punpcklwd xmm0, xmm3 ;13 12 11 10 03 02 01 00
punpckhwd xmm2, xmm3 ;33 32 31 30 23 22 21 20
movdqa xmm5, XMMWORD PTR[_7 GLOBAL]
movdqa xmm5, XMMWORD PTR[GLOBAL(_7)]
pshufd xmm2, xmm2, 04eh
movdqa xmm3, xmm0
paddw xmm0, xmm2 ;b1 b1 b1 b1 a1 a1 a1 a1
@ -94,8 +94,8 @@ sym(vp8_short_fdct4x4_sse2):
pshufhw xmm0, xmm0, 0d8h ;b1 a1 b1 a1 b1 a1 b1 a1
pshufhw xmm3, xmm3, 0d8h ;c1 d1 c1 d1 c1 d1 c1 d1
movdqa xmm1, xmm0
pmaddwd xmm0, XMMWORD PTR[_mult_add GLOBAL] ;a1 + b1
pmaddwd xmm1, XMMWORD PTR[_mult_sub GLOBAL] ;a1 - b1
pmaddwd xmm0, XMMWORD PTR[GLOBAL(_mult_add)] ;a1 + b1
pmaddwd xmm1, XMMWORD PTR[GLOBAL(_mult_sub)] ;a1 - b1
pxor xmm4, xmm4 ;zero out for compare
paddd xmm0, xmm5
@ -103,14 +103,14 @@ sym(vp8_short_fdct4x4_sse2):
pcmpeqw xmm2, xmm4
psrad xmm0, 4 ;(a1 + b1 + 7)>>4
psrad xmm1, 4 ;(a1 - b1 + 7)>>4
pandn xmm2, XMMWORD PTR[_cmp_mask GLOBAL] ;clear upper,
;and keep bit 0 of lower
pandn xmm2, XMMWORD PTR[GLOBAL(_cmp_mask)] ;clear upper,
;and keep bit 0 of lower
movdqa xmm4, xmm3
pmaddwd xmm3, XMMWORD PTR[_5352_2217 GLOBAL] ;c1*2217 + d1*5352
pmaddwd xmm4, XMMWORD PTR[_2217_neg5352 GLOBAL] ;d1*2217 - c1*5352
paddd xmm3, XMMWORD PTR[_12000 GLOBAL]
paddd xmm4, XMMWORD PTR[_51000 GLOBAL]
pmaddwd xmm3, XMMWORD PTR[GLOBAL(_5352_2217)] ;c1*2217 + d1*5352
pmaddwd xmm4, XMMWORD PTR[GLOBAL(_2217_neg5352)] ;d1*2217 - c1*5352
paddd xmm3, XMMWORD PTR[GLOBAL(_12000)]
paddd xmm4, XMMWORD PTR[GLOBAL(_51000)]
packssdw xmm0, xmm1 ;op[8] op[0]
psrad xmm3, 16 ;(c1 * 2217 + d1 * 5352 + 12000)>>16
psrad xmm4, 16 ;(d1 * 2217 - c1 * 5352 + 51000)>>16

18
vp8/encoder/x86/variance_impl_mmx.asm
Просмотреть файл

@ -556,7 +556,7 @@ sym(vp8_filter_block2d_bil4x4_var_mmx):
pmullw mm3, [rax+8] ;
paddw mm1, mm3 ;
paddw mm1, [mmx_bi_rd GLOBAL] ;
paddw mm1, [GLOBAL(mmx_bi_rd)] ;
psraw mm1, mmx_filter_shift ;
movq mm5, mm1
@ -580,7 +580,7 @@ filter_block2d_bil4x4_var_mmx_loop:
pmullw mm3, [rax+8] ;
paddw mm1, mm3 ;
paddw mm1, [mmx_bi_rd GLOBAL] ;
paddw mm1, [GLOBAL(mmx_bi_rd)] ;
psraw mm1, mmx_filter_shift ;
movq mm3, mm5 ;
@ -592,7 +592,7 @@ filter_block2d_bil4x4_var_mmx_loop:
paddw mm1, mm3 ;
paddw mm1, [mmx_bi_rd GLOBAL] ;
paddw mm1, [GLOBAL(mmx_bi_rd)] ;
psraw mm1, mmx_filter_shift ;
movd mm3, [rdi] ;
@ -710,10 +710,10 @@ sym(vp8_filter_block2d_bil_var_mmx):
paddw mm1, mm3 ;
paddw mm2, mm4 ;
paddw mm1, [mmx_bi_rd GLOBAL] ;
paddw mm1, [GLOBAL(mmx_bi_rd)] ;
psraw mm1, mmx_filter_shift ;
paddw mm2, [mmx_bi_rd GLOBAL] ;
paddw mm2, [GLOBAL(mmx_bi_rd)] ;
psraw mm2, mmx_filter_shift ;
movq mm5, mm1
@ -749,10 +749,10 @@ filter_block2d_bil_var_mmx_loop:
paddw mm1, mm3 ;
paddw mm2, mm4 ;
paddw mm1, [mmx_bi_rd GLOBAL] ;
paddw mm1, [GLOBAL(mmx_bi_rd)] ;
psraw mm1, mmx_filter_shift ;
paddw mm2, [mmx_bi_rd GLOBAL] ;
paddw mm2, [GLOBAL(mmx_bi_rd)] ;
psraw mm2, mmx_filter_shift ;
movq mm3, mm5 ;
@ -773,8 +773,8 @@ filter_block2d_bil_var_mmx_loop:
paddw mm1, mm3 ;
paddw mm2, mm4 ;
paddw mm1, [mmx_bi_rd GLOBAL] ;
paddw mm2, [mmx_bi_rd GLOBAL] ;
paddw mm1, [GLOBAL(mmx_bi_rd)] ;
paddw mm2, [GLOBAL(mmx_bi_rd)] ;
psraw mm1, mmx_filter_shift ;
psraw mm2, mmx_filter_shift ;

6
vp8/encoder/x86/variance_impl_sse2.asm
Просмотреть файл

@ -532,7 +532,7 @@ sym(vp8_filter_block2d_bil_var_sse2):
pmullw xmm3, [rax+16] ;
paddw xmm1, xmm3 ;
paddw xmm1, [xmm_bi_rd GLOBAL] ;
paddw xmm1, [GLOBAL(xmm_bi_rd)] ;
psraw xmm1, xmm_filter_shift ;
movdqa xmm5, xmm1
@ -554,7 +554,7 @@ filter_block2d_bil_var_sse2_loop:
pmullw xmm3, [rax+16] ;
paddw xmm1, xmm3 ;
paddw xmm1, [xmm_bi_rd GLOBAL] ;
paddw xmm1, [GLOBAL(xmm_bi_rd)] ;
psraw xmm1, xmm_filter_shift ;
movdqa xmm3, xmm5 ;
@ -565,7 +565,7 @@ filter_block2d_bil_var_sse2_loop:
pmullw xmm1, [rdx+16] ;
paddw xmm1, xmm3 ;
paddw xmm1, [xmm_bi_rd GLOBAL] ;
paddw xmm1, [GLOBAL(xmm_bi_rd)] ;
psraw xmm1, xmm_filter_shift ;
movq xmm3, QWORD PTR [rdi] ;

8
vpx_ports/x86_abi_support.asm
Просмотреть файл

@ -160,7 +160,7 @@
ret
%%exitGG:
%undef GLOBAL
%define GLOBAL + %1 wrt ..gotoff
%define GLOBAL(x) x + %1 wrt ..gotoff
%undef RESTORE_GOT
%define RESTORE_GOT pop %1
%endmacro
@ -176,7 +176,7 @@
ret
%%exitGG:
%undef GLOBAL
%define GLOBAL + %1 - fake_got
%define GLOBAL(x) x + %1 - fake_got
%undef RESTORE_GOT
%define RESTORE_GOT pop %1
%endmacro
@ -186,7 +186,7 @@
%else
%macro GET_GOT 1
%endmacro
%define GLOBAL wrt rip
%define GLOBAL(x) rel x
%ifidn __OUTPUT_FORMAT__,elf64
%define WRT_PLT wrt ..plt
%define HIDDEN_DATA(x) x:data hidden
@ -197,7 +197,7 @@
%ifnmacro GET_GOT
%macro GET_GOT 1
%endmacro
%define GLOBAL
%define GLOBAL(x) x
%endif
%ifndef RESTORE_GOT
%define RESTORE_GOT