2014-01-03 07:49:45 +01:00
|
|
|
;*!
|
|
|
|
|
;* \copy
|
|
|
|
|
;* Copyright (c) 2009-2013, Cisco Systems
|
|
|
|
|
;* All rights reserved.
|
|
|
|
|
;*
|
|
|
|
|
;* Redistribution and use in source and binary forms, with or without
|
|
|
|
|
;* modification, are permitted provided that the following conditions
|
|
|
|
|
;* are met:
|
|
|
|
|
;*
|
|
|
|
|
;* * Redistributions of source code must retain the above copyright
|
|
|
|
|
;* notice, this list of conditions and the following disclaimer.
|
|
|
|
|
;*
|
|
|
|
|
;* * Redistributions in binary form must reproduce the above copyright
|
|
|
|
|
;* notice, this list of conditions and the following disclaimer in
|
|
|
|
|
;* the documentation and/or other materials provided with the
|
|
|
|
|
;* distribution.
|
|
|
|
|
;*
|
|
|
|
|
;* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
|
|
|
|
;* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
|
|
|
|
;* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
|
|
|
|
|
;* FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
|
|
|
|
|
;* COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
|
|
|
|
|
;* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
|
|
|
|
|
;* BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
|
|
|
|
|
;* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
|
|
|
|
|
;* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
|
|
|
;* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
|
|
|
|
|
;* ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
|
|
|
;* POSSIBILITY OF SUCH DAMAGE.
|
|
|
|
|
;*
|
|
|
|
|
;*
|
|
|
|
|
;* expand_picture.asm
|
|
|
|
|
;*
|
|
|
|
|
;* Abstract
|
|
|
|
|
;* mmxext/sse for expand_frame
|
|
|
|
|
;*
|
|
|
|
|
;* History
|
|
|
|
|
;* 09/25/2009 Created
|
|
|
|
|
;*
|
|
|
|
|
;*
|
|
|
|
|
;*************************************************************************/
|
|
|
|
|
|
|
|
|
|
%include "asm_inc.asm"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
;***********************************************************************
|
|
|
|
|
; Macros and other preprocessor constants
|
|
|
|
|
;***********************************************************************
|
|
|
|
|
|
|
|
|
|
;***********************************************************************
|
|
|
|
|
; Code
|
|
|
|
|
;***********************************************************************
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
SECTION .text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
;;;;;;;expanding result;;;;;;;
|
|
|
|
|
|
|
|
|
|
;aaaa|attttttttttttttttb|bbbb
|
|
|
|
|
;aaaa|attttttttttttttttb|bbbb
|
|
|
|
|
;aaaa|attttttttttttttttb|bbbb
|
|
|
|
|
;aaaa|attttttttttttttttb|bbbb
|
|
|
|
|
;----------------------------
|
|
|
|
|
;aaaa|attttttttttttttttb|bbbb
|
|
|
|
|
;llll|l r|rrrr
|
|
|
|
|
;llll|l r|rrrr
|
|
|
|
|
;llll|l r|rrrr
|
|
|
|
|
;llll|l r|rrrr
|
|
|
|
|
;llll|l r|rrrr
|
|
|
|
|
;cccc|ceeeeeeeeeeeeeeeed|dddd
|
|
|
|
|
;----------------------------
|
|
|
|
|
;cccc|ceeeeeeeeeeeeeeeed|dddd
|
|
|
|
|
;cccc|ceeeeeeeeeeeeeeeed|dddd
|
|
|
|
|
;cccc|ceeeeeeeeeeeeeeeed|dddd
|
|
|
|
|
;cccc|ceeeeeeeeeeeeeeeed|dddd
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro mov_line_8x4_mmx 3 ; dst, stride, mm?
|
|
|
|
|
movq [%1], %3
|
|
|
|
|
movq [%1+%2], %3
|
|
|
|
|
lea %1, [%1+2*%2]
|
|
|
|
|
movq [%1], %3
|
|
|
|
|
movq [%1+%2], %3
|
|
|
|
|
lea %1, [%1+2*%2]
|
2014-01-03 07:49:45 +01:00
|
|
|
%endmacro
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro mov_line_end8x4_mmx 3 ; dst, stride, mm?
|
|
|
|
|
movq [%1], %3
|
|
|
|
|
movq [%1+%2], %3
|
|
|
|
|
lea %1, [%1+2*%2]
|
|
|
|
|
movq [%1], %3
|
|
|
|
|
movq [%1+%2], %3
|
|
|
|
|
lea %1, [%1+%2]
|
2014-01-03 07:49:45 +01:00
|
|
|
%endmacro
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro mov_line_16x4_sse2 4 ; dst, stride, xmm?, u/a
|
|
|
|
|
movdq%4 [%1], %3 ; top(bottom)_0
|
|
|
|
|
movdq%4 [%1+%2], %3 ; top(bottom)_1
|
|
|
|
|
lea %1, [%1+2*%2]
|
|
|
|
|
movdq%4 [%1], %3 ; top(bottom)_2
|
|
|
|
|
movdq%4 [%1+%2], %3 ; top(bottom)_3
|
|
|
|
|
lea %1, [%1+2*%2]
|
2014-01-03 07:49:45 +01:00
|
|
|
%endmacro
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro mov_line_end16x4_sse2 4 ; dst, stride, xmm?, u/a
|
|
|
|
|
movdq%4 [%1], %3 ; top(bottom)_0
|
|
|
|
|
movdq%4 [%1+%2], %3 ; top(bottom)_1
|
|
|
|
|
lea %1, [%1+2*%2]
|
|
|
|
|
movdq%4 [%1], %3 ; top(bottom)_2
|
|
|
|
|
movdq%4 [%1+%2], %3 ; top(bottom)_3
|
|
|
|
|
lea %1, [%1+%2]
|
2014-01-03 07:49:45 +01:00
|
|
|
%endmacro
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro mov_line_32x4_sse2 3 ; dst, stride, xmm?
|
|
|
|
|
movdqa [%1], %3 ; top(bottom)_0
|
|
|
|
|
movdqa [%1+16], %3 ; top(bottom)_0
|
|
|
|
|
movdqa [%1+%2], %3 ; top(bottom)_1
|
|
|
|
|
movdqa [%1+%2+16], %3 ; top(bottom)_1
|
|
|
|
|
lea %1, [%1+2*%2]
|
|
|
|
|
movdqa [%1], %3 ; top(bottom)_2
|
|
|
|
|
movdqa [%1+16], %3 ; top(bottom)_2
|
|
|
|
|
movdqa [%1+%2], %3 ; top(bottom)_3
|
|
|
|
|
movdqa [%1+%2+16], %3 ; top(bottom)_3
|
|
|
|
|
lea %1, [%1+2*%2]
|
2014-01-03 07:49:45 +01:00
|
|
|
%endmacro
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro mov_line_end32x4_sse2 3 ; dst, stride, xmm?
|
|
|
|
|
movdqa [%1], %3 ; top(bottom)_0
|
|
|
|
|
movdqa [%1+16], %3 ; top(bottom)_0
|
|
|
|
|
movdqa [%1+%2], %3 ; top(bottom)_1
|
|
|
|
|
movdqa [%1+%2+16], %3 ; top(bottom)_1
|
|
|
|
|
lea %1, [%1+2*%2]
|
|
|
|
|
movdqa [%1], %3 ; top(bottom)_2
|
|
|
|
|
movdqa [%1+16], %3 ; top(bottom)_2
|
|
|
|
|
movdqa [%1+%2], %3 ; top(bottom)_3
|
|
|
|
|
movdqa [%1+%2+16], %3 ; top(bottom)_3
|
|
|
|
|
lea %1, [%1+%2]
|
2014-01-03 07:49:45 +01:00
|
|
|
%endmacro
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro exp_top_bottom_sse2 1 ; iPaddingSize [luma(32)/chroma(16)]
|
2014-01-03 07:49:45 +01:00
|
|
|
;r2 [width/16(8)]
|
|
|
|
|
;r0 [pSrc +0], r5 [pSrc -width] r1[-stride], 32(16) ;top
|
|
|
|
|
;r3 [pSrc +(h-1)*stride], r4 [pSrc + (h+31)*stride],32(16); bottom
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%if %1 == 32 ; for luma
|
|
|
|
|
sar r2, 04h ; width / 16(8) pixels
|
2014-01-03 07:49:45 +01:00
|
|
|
.top_bottom_loops:
|
2014-05-31 13:13:34 +02:00
|
|
|
; top
|
|
|
|
|
movdqa xmm0, [r0] ; first line of picture pData
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
mov_line_end16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
|
|
|
|
|
; bottom
|
|
|
|
|
movdqa xmm1, [r3] ; last line of picture pData
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
mov_line_end16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
|
|
|
|
|
lea r0, [r0+16] ; top pSrc
|
|
|
|
|
lea r5, [r5+16] ; top dst
|
|
|
|
|
lea r3, [r3+16] ; bottom pSrc
|
|
|
|
|
lea r4, [r4+16] ; bottom dst
|
|
|
|
|
neg r1 ; positive/negative stride need for next loop?
|
|
|
|
|
|
|
|
|
|
dec r2
|
|
|
|
|
jnz near .top_bottom_loops
|
|
|
|
|
%elif %1 == 16 ; for chroma ??
|
|
|
|
|
mov r6, r2
|
|
|
|
|
sar r2, 04h ; (width / 16) pixels
|
2014-01-03 07:49:45 +01:00
|
|
|
.top_bottom_loops:
|
2014-05-31 13:13:34 +02:00
|
|
|
; top
|
|
|
|
|
movdqa xmm0, [r0] ; first line of picture pData
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
mov_line_end16x4_sse2 r5, r1, xmm0, a
|
|
|
|
|
|
|
|
|
|
; bottom
|
|
|
|
|
movdqa xmm1, [r3] ; last line of picture pData
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
mov_line_end16x4_sse2 r4, r1, xmm1, a
|
|
|
|
|
|
|
|
|
|
lea r0, [r0+16] ; top pSrc
|
|
|
|
|
lea r5, [r5+16] ; top dst
|
|
|
|
|
lea r3, [r3+16] ; bottom pSrc
|
|
|
|
|
lea r4, [r4+16] ; bottom dst
|
|
|
|
|
neg r1 ; positive/negative stride need for next loop?
|
|
|
|
|
|
|
|
|
|
dec r2
|
|
|
|
|
jnz near .top_bottom_loops
|
|
|
|
|
|
|
|
|
|
; for remaining 8 bytes
|
|
|
|
|
and r6, 0fh ; any 8 bytes left?
|
|
|
|
|
test r6, r6
|
|
|
|
|
jz near .to_be_continued ; no left to exit here
|
|
|
|
|
|
|
|
|
|
; top
|
|
|
|
|
movq mm0, [r0] ; remained 8 byte
|
|
|
|
|
mov_line_8x4_mmx r5, r1, mm0 ; dst, stride, mm?
|
|
|
|
|
mov_line_8x4_mmx r5, r1, mm0 ; dst, stride, mm?
|
|
|
|
|
mov_line_8x4_mmx r5, r1, mm0 ; dst, stride, mm?
|
|
|
|
|
mov_line_end8x4_mmx r5, r1, mm0 ; dst, stride, mm?
|
|
|
|
|
; bottom
|
|
|
|
|
movq mm1, [r3]
|
|
|
|
|
mov_line_8x4_mmx r4, r1, mm1 ; dst, stride, mm?
|
|
|
|
|
mov_line_8x4_mmx r4, r1, mm1 ; dst, stride, mm?
|
|
|
|
|
mov_line_8x4_mmx r4, r1, mm1 ; dst, stride, mm?
|
|
|
|
|
mov_line_end8x4_mmx r4, r1, mm1 ; dst, stride, mm?
|
|
|
|
|
WELSEMMS
|
2014-01-03 07:49:45 +01:00
|
|
|
|
|
|
|
|
.to_be_continued:
|
|
|
|
|
%endif
|
|
|
|
|
%endmacro
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro exp_left_right_sse2 2 ; iPaddingSize [luma(32)/chroma(16)], u/a
|
2014-01-03 07:49:45 +01:00
|
|
|
;r6 [height]
|
|
|
|
|
;r0 [pSrc+0] r5[pSrc-32] r1[stride]
|
2014-01-05 13:16:22 +01:00
|
|
|
;r3 [pSrc+(w-1)] r4[pSrc+w]
|
2014-01-03 07:49:45 +01:00
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%if %1 == 32 ; for luma
|
2014-01-03 07:49:45 +01:00
|
|
|
.left_right_loops:
|
2014-05-31 13:13:34 +02:00
|
|
|
; left
|
|
|
|
|
movzx r2d, byte [r0] ; pixel pData for left border
|
|
|
|
|
SSE2_Copy16Times xmm0, r2d ; dst, tmp, pSrc [generic register name: a/b/c/d]
|
|
|
|
|
movdqa [r5], xmm0
|
|
|
|
|
movdqa [r5+16], xmm0
|
|
|
|
|
|
|
|
|
|
; right
|
|
|
|
|
movzx r2d, byte [r3]
|
|
|
|
|
SSE2_Copy16Times xmm1, r2d ; dst, tmp, pSrc [generic register name: a/b/c/d]
|
|
|
|
|
movdqa [r4], xmm1
|
|
|
|
|
movdqa [r4+16], xmm1
|
|
|
|
|
|
|
|
|
|
lea r0, [r0+r1] ; left pSrc
|
|
|
|
|
lea r5, [r5+r1] ; left dst
|
|
|
|
|
lea r3, [r3+r1] ; right pSrc
|
|
|
|
|
lea r4, [r4+r1] ; right dst
|
|
|
|
|
|
|
|
|
|
dec r6
|
|
|
|
|
jnz near .left_right_loops
|
|
|
|
|
%elif %1 == 16 ; for chroma ??
|
2014-01-03 07:49:45 +01:00
|
|
|
.left_right_loops:
|
2014-05-31 13:13:34 +02:00
|
|
|
; left
|
|
|
|
|
movzx r2d, byte [r0] ; pixel pData for left border
|
|
|
|
|
SSE2_Copy16Times xmm0, r2d ; dst, tmp, pSrc [generic register name: a/b/c/d]
|
|
|
|
|
movdqa [r5], xmm0
|
|
|
|
|
|
|
|
|
|
; right
|
|
|
|
|
movzx r2d, byte [r3]
|
|
|
|
|
SSE2_Copy16Times xmm1, r2d ; dst, tmp, pSrc [generic register name: a/b/c/d]
|
|
|
|
|
movdq%2 [r4], xmm1 ; might not be aligned 16 bytes in case chroma planes
|
|
|
|
|
|
|
|
|
|
lea r0, [r0+r1] ; left pSrc
|
|
|
|
|
lea r5, [r5+r1] ; left dst
|
|
|
|
|
lea r3, [r3+r1] ; right pSrc
|
|
|
|
|
lea r4, [r4+r1] ; right dst
|
|
|
|
|
|
|
|
|
|
dec r6
|
|
|
|
|
jnz near .left_right_loops
|
2014-01-03 07:49:45 +01:00
|
|
|
%endif
|
|
|
|
|
%endmacro
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
%macro exp_cross_sse2 2 ; iPaddingSize [luma(32)/chroma(16)], u/a
|
|
|
|
|
; top-left: (x)mm3, top-right: (x)mm4, bottom-left: (x)mm5, bottom-right: (x)mm6
|
|
|
|
|
; edi: TL, ebp: TR, eax: BL, ebx: BR, ecx, -stride
|
2014-01-03 07:49:45 +01:00
|
|
|
;r3:TL ,r4:TR,r5:BL,r6:BR r1:-stride
|
2014-05-31 13:13:34 +02:00
|
|
|
%if %1 == 32 ; luma
|
|
|
|
|
; TL
|
|
|
|
|
mov_line_32x4_sse2 r3, r1, xmm3 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r3, r1, xmm3 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r3, r1, xmm3 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r3, r1, xmm3 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r3, r1, xmm3 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r3, r1, xmm3 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r3, r1, xmm3 ; dst, stride, xmm?
|
|
|
|
|
mov_line_end32x4_sse2 r3, r1, xmm3 ; dst, stride, xmm?
|
|
|
|
|
|
|
|
|
|
; TR
|
|
|
|
|
mov_line_32x4_sse2 r4, r1, xmm4 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r4, r1, xmm4 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r4, r1, xmm4 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r4, r1, xmm4 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r4, r1, xmm4 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r4, r1, xmm4 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r4, r1, xmm4 ; dst, stride, xmm?
|
|
|
|
|
mov_line_end32x4_sse2 r4, r1, xmm4 ; dst, stride, xmm?
|
|
|
|
|
|
|
|
|
|
; BL
|
|
|
|
|
mov_line_32x4_sse2 r5, r1, xmm5 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r5, r1, xmm5 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r5, r1, xmm5 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r5, r1, xmm5 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r5, r1, xmm5 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r5, r1, xmm5 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r5, r1, xmm5 ; dst, stride, xmm?
|
|
|
|
|
mov_line_end32x4_sse2 r5, r1, xmm5 ; dst, stride, xmm?
|
|
|
|
|
|
|
|
|
|
; BR
|
|
|
|
|
mov_line_32x4_sse2 r6, r1, xmm6 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r6, r1, xmm6 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r6, r1, xmm6 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r6, r1, xmm6 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r6, r1, xmm6 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r6, r1, xmm6 ; dst, stride, xmm?
|
|
|
|
|
mov_line_32x4_sse2 r6, r1, xmm6 ; dst, stride, xmm?
|
|
|
|
|
mov_line_end32x4_sse2 r6, r1, xmm6 ; dst, stride, xmm?
|
|
|
|
|
%elif %1 == 16 ; chroma
|
|
|
|
|
; TL
|
|
|
|
|
mov_line_16x4_sse2 r3, r1, xmm3, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r3, r1, xmm3, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r3, r1, xmm3, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_end16x4_sse2 r3, r1, xmm3, a ; dst, stride, xmm?
|
|
|
|
|
|
|
|
|
|
; TR
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm4, %2 ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm4, %2 ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r4, r1, xmm4, %2 ; dst, stride, xmm?
|
|
|
|
|
mov_line_end16x4_sse2 r4, r1, xmm4, %2 ; dst, stride, xmm?
|
|
|
|
|
|
|
|
|
|
; BL
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm5, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm5, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r5, r1, xmm5, a ; dst, stride, xmm?
|
|
|
|
|
mov_line_end16x4_sse2 r5, r1, xmm5, a ; dst, stride, xmm?
|
|
|
|
|
|
|
|
|
|
; BR
|
|
|
|
|
mov_line_16x4_sse2 r6, r1, xmm6, %2 ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r6, r1, xmm6, %2 ; dst, stride, xmm?
|
|
|
|
|
mov_line_16x4_sse2 r6, r1, xmm6, %2 ; dst, stride, xmm?
|
|
|
|
|
mov_line_end16x4_sse2 r6, r1, xmm6, %2 ; dst, stride, xmm?
|
2014-01-03 07:49:45 +01:00
|
|
|
%endif
|
|
|
|
|
%endmacro
|
|
|
|
|
|
|
|
|
|
;***********************************************************************----------------
|
2014-05-31 13:13:34 +02:00
|
|
|
; void ExpandPictureLuma_sse2( uint8_t *pDst,
|
|
|
|
|
; const int32_t iStride,
|
|
|
|
|
; const int32_t iWidth,
|
|
|
|
|
; const int32_t iHeight );
|
2014-01-03 07:49:45 +01:00
|
|
|
;***********************************************************************----------------
|
2014-03-16 12:23:24 +01:00
|
|
|
WELS_EXTERN ExpandPictureLuma_sse2
|
2014-01-03 07:49:45 +01:00
|
|
|
|
|
|
|
|
push r4
|
|
|
|
|
push r5
|
|
|
|
|
push r6
|
|
|
|
|
|
|
|
|
|
%assign push_num 3
|
|
|
|
|
LOAD_4_PARA
|
Backup/restore the xmm6-xmm15 SSE registers within asm functions on win64
According to the Win64 ABI, these registers need to be preserved,
and compilers are allowed to rely on their content to stay
available - not only for float usage but for any usage, anywhere,
in the calling C++ code.
This adds a macro which pushes the clobbered registers onto the
stack if targeting win64 (and a matching one which restores them).
The parameter to the macro is the number of xmm registers used
(e.g. if using xmm0 - xmm7, the parameter is 8), or in other
words, the number of the highest xmm register used plus one.
This is similar to how the same issue is handled for the NEON
registers q4-q7 with the vpush instruction, except that they needed
to be preserved on all platforms, not only on one particular platform.
This allows removing the XMMREG_PROTECT_* hacks, which can
easily fail if the compiler chooses to use the callee saved
xmm registers in an unexpected spot.
2014-03-14 09:29:53 +01:00
|
|
|
PUSH_XMM 7
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-03-14 09:13:18 +01:00
|
|
|
SIGN_EXTENSION r1, r1d
|
|
|
|
|
SIGN_EXTENSION r2, r2d
|
|
|
|
|
SIGN_EXTENSION r3, r3d
|
2014-01-03 07:49:45 +01:00
|
|
|
|
|
|
|
|
;also prepare for cross border pData top-left:xmm3
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
movzx r6d,byte[r0]
|
|
|
|
|
SSE2_Copy16Times xmm3,r6d ;xmm3: pSrc[0]
|
|
|
|
|
|
|
|
|
|
neg r1
|
|
|
|
|
lea r5,[r0+r1] ;last line of top border r5= dst top pSrc[-stride]
|
|
|
|
|
neg r1
|
|
|
|
|
|
|
|
|
|
push r3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
dec r3 ;h-1
|
|
|
|
|
imul r3,r1 ;(h-1)*stride
|
|
|
|
|
lea r3,[r0+r3] ;pSrc[(h-1)*stride] r3 = src bottom
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
mov r6,r1 ;r6 = stride
|
|
|
|
|
sal r6,05h ;r6 = 32*stride
|
|
|
|
|
lea r4,[r3+r6] ;r4 = dst bottom
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
;also prepare for cross border data: bottom-left with xmm5,bottom-right xmm6
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
movzx r6d,byte [r3] ;bottom-left
|
|
|
|
|
SSE2_Copy16Times xmm5,r6d
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r6,[r3+r2-1]
|
|
|
|
|
movzx r6d,byte [r6]
|
|
|
|
|
SSE2_Copy16Times xmm6,r6d ;bottom-right
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = -stride
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
push r0
|
|
|
|
|
push r1
|
|
|
|
|
push r2
|
|
|
|
|
|
|
|
|
|
exp_top_bottom_sse2 32
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
; for both left and right border
|
|
|
|
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
pop r2
|
|
|
|
|
pop r1
|
|
|
|
|
pop r0
|
|
|
|
|
|
|
|
|
|
lea r5,[r0-32] ;left border dst luma =32 chroma = -16
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r3,[r0+r2-1] ;right border src
|
|
|
|
|
lea r4,[r3+1] ;right border dst
|
|
|
|
|
|
|
|
|
|
;prepare for cross border data: top-rigth with xmm4
|
2014-05-31 13:13:34 +02:00
|
|
|
movzx r6d,byte [r3] ;top -rigth
|
|
|
|
|
SSE2_Copy16Times xmm4,r6d
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = stride
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pop r6 ; r6 = height
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
push r0
|
|
|
|
|
push r1
|
|
|
|
|
push r2
|
|
|
|
|
push r6
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
exp_left_right_sse2 32,a
|
|
|
|
|
|
|
|
|
|
pop r6
|
|
|
|
|
pop r2
|
|
|
|
|
pop r1
|
|
|
|
|
pop r0
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
; for cross border [top-left, top-right, bottom-left, bottom-right]
|
|
|
|
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|
2014-01-03 07:49:45 +01:00
|
|
|
; have done xmm3,..,xmm6 cross pData initialization above, perform pading as below, To be continued..
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = -stride
|
|
|
|
|
lea r3,[r0-32]
|
|
|
|
|
lea r3,[r3+r1] ;last line of top-left border
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r4,[r0+r2] ;psrc +width
|
|
|
|
|
lea r4,[r4+r1] ;psrc +width -stride
|
2014-01-05 13:16:22 +01:00
|
|
|
|
|
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = stride
|
|
|
|
|
add r6,32 ;height +32(16) ,luma = 32, chroma = 16
|
|
|
|
|
imul r6,r1
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r5,[r3+r6] ;last line of bottom-left border
|
|
|
|
|
lea r6,[r4+r6] ;last line of botoom-right border
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ; r1 = -stride
|
|
|
|
|
|
|
|
|
|
; for left & right border expanding
|
|
|
|
|
exp_cross_sse2 32,a
|
|
|
|
|
|
Backup/restore the xmm6-xmm15 SSE registers within asm functions on win64
According to the Win64 ABI, these registers need to be preserved,
and compilers are allowed to rely on their content to stay
available - not only for float usage but for any usage, anywhere,
in the calling C++ code.
This adds a macro which pushes the clobbered registers onto the
stack if targeting win64 (and a matching one which restores them).
The parameter to the macro is the number of xmm registers used
(e.g. if using xmm0 - xmm7, the parameter is 8), or in other
words, the number of the highest xmm register used plus one.
This is similar to how the same issue is handled for the NEON
registers q4-q7 with the vpush instruction, except that they needed
to be preserved on all platforms, not only on one particular platform.
This allows removing the XMMREG_PROTECT_* hacks, which can
easily fail if the compiler chooses to use the callee saved
xmm registers in an unexpected spot.
2014-03-14 09:29:53 +01:00
|
|
|
POP_XMM
|
2014-01-03 07:49:45 +01:00
|
|
|
LOAD_4_PARA_POP
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
pop r6
|
|
|
|
|
pop r5
|
|
|
|
|
pop r4
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
%assign push_num 0
|
|
|
|
|
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
ret
|
2014-01-03 07:49:45 +01:00
|
|
|
|
|
|
|
|
;***********************************************************************----------------
|
2014-05-31 13:13:34 +02:00
|
|
|
; void ExpandPictureChromaAlign_sse2( uint8_t *pDst,
|
|
|
|
|
; const int32_t iStride,
|
|
|
|
|
; const int32_t iWidth,
|
|
|
|
|
; const int32_t iHeight );
|
2014-01-03 07:49:45 +01:00
|
|
|
;***********************************************************************----------------
|
2014-03-16 12:23:24 +01:00
|
|
|
WELS_EXTERN ExpandPictureChromaAlign_sse2
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
push r4
|
|
|
|
|
push r5
|
|
|
|
|
push r6
|
|
|
|
|
|
|
|
|
|
%assign push_num 3
|
|
|
|
|
LOAD_4_PARA
|
Backup/restore the xmm6-xmm15 SSE registers within asm functions on win64
According to the Win64 ABI, these registers need to be preserved,
and compilers are allowed to rely on their content to stay
available - not only for float usage but for any usage, anywhere,
in the calling C++ code.
This adds a macro which pushes the clobbered registers onto the
stack if targeting win64 (and a matching one which restores them).
The parameter to the macro is the number of xmm registers used
(e.g. if using xmm0 - xmm7, the parameter is 8), or in other
words, the number of the highest xmm register used plus one.
This is similar to how the same issue is handled for the NEON
registers q4-q7 with the vpush instruction, except that they needed
to be preserved on all platforms, not only on one particular platform.
This allows removing the XMMREG_PROTECT_* hacks, which can
easily fail if the compiler chooses to use the callee saved
xmm registers in an unexpected spot.
2014-03-14 09:29:53 +01:00
|
|
|
PUSH_XMM 7
|
2014-01-03 07:49:45 +01:00
|
|
|
|
2014-03-14 09:13:18 +01:00
|
|
|
SIGN_EXTENSION r1,r1d
|
|
|
|
|
SIGN_EXTENSION r2,r2d
|
|
|
|
|
SIGN_EXTENSION r3,r3d
|
2014-01-03 07:49:45 +01:00
|
|
|
|
|
|
|
|
;also prepare for cross border pData top-left:xmm3
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
movzx r6d,byte [r0]
|
|
|
|
|
SSE2_Copy16Times xmm3,r6d ;xmm3: pSrc[0]
|
|
|
|
|
|
|
|
|
|
neg r1
|
|
|
|
|
lea r5,[r0+r1] ;last line of top border r5= dst top pSrc[-stride]
|
|
|
|
|
neg r1
|
|
|
|
|
|
|
|
|
|
push r3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
dec r3 ;h-1
|
|
|
|
|
imul r3,r1 ;(h-1)*stride
|
|
|
|
|
lea r3,[r0+r3] ;pSrc[(h-1)*stride] r3 = src bottom
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
mov r6,r1 ;r6 = stride
|
|
|
|
|
sal r6,04h ;r6 = 32*stride
|
2014-01-05 13:16:22 +01:00
|
|
|
lea r4,[r3+r6] ;r4 = dst bottom
|
|
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
;also prepare for cross border data: bottom-left with xmm5,bottom-right xmm6
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
movzx r6d,byte [r3] ;bottom-left
|
|
|
|
|
SSE2_Copy16Times xmm5,r6d
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r6,[r3+r2-1]
|
|
|
|
|
movzx r6d,byte [r6]
|
|
|
|
|
SSE2_Copy16Times xmm6,r6d ;bottom-right
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = -stride
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
push r0
|
2014-01-05 13:16:22 +01:00
|
|
|
push r1
|
2014-01-03 07:49:45 +01:00
|
|
|
push r2
|
|
|
|
|
|
|
|
|
|
exp_top_bottom_sse2 16
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
; for both left and right border
|
|
|
|
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
pop r2
|
|
|
|
|
pop r1
|
|
|
|
|
pop r0
|
|
|
|
|
|
|
|
|
|
lea r5,[r0-16] ;left border dst luma =32 chroma = -16
|
2014-01-05 13:16:22 +01:00
|
|
|
|
|
|
|
|
lea r3,[r0+r2-1] ;right border src
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r4,[r3+1] ;right border dst
|
|
|
|
|
|
|
|
|
|
;prepare for cross border data: top-rigth with xmm4
|
|
|
|
|
movzx r6d,byte [r3] ;top -rigth
|
|
|
|
|
SSE2_Copy16Times xmm4,r6d
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = stride
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pop r6 ; r6 = height
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
push r0
|
2014-01-05 13:16:22 +01:00
|
|
|
push r1
|
2014-01-03 07:49:45 +01:00
|
|
|
push r2
|
2014-05-31 13:13:34 +02:00
|
|
|
push r6
|
2014-01-03 07:49:45 +01:00
|
|
|
exp_left_right_sse2 16,a
|
|
|
|
|
|
|
|
|
|
pop r6
|
|
|
|
|
pop r2
|
|
|
|
|
pop r1
|
|
|
|
|
pop r0
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
; for cross border [top-left, top-right, bottom-left, bottom-right]
|
|
|
|
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|
2014-01-03 07:49:45 +01:00
|
|
|
; have done xmm3,..,xmm6 cross pData initialization above, perform pading as below, To be continued..
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = -stride
|
|
|
|
|
lea r3,[r0-16]
|
|
|
|
|
lea r3,[r3+r1] ;last line of top-left border
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r4,[r0+r2] ;psrc +width
|
2014-01-05 13:16:22 +01:00
|
|
|
lea r4,[r4+r1] ;psrc +width -stride
|
|
|
|
|
|
|
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = stride
|
|
|
|
|
add r6,16 ;height +32(16) ,luma = 32, chroma = 16
|
|
|
|
|
imul r6,r1
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r5,[r3+r6] ;last line of bottom-left border
|
|
|
|
|
lea r6,[r4+r6] ;last line of botoom-right border
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ; r1 = -stride
|
|
|
|
|
|
|
|
|
|
; for left & right border expanding
|
|
|
|
|
exp_cross_sse2 16,a
|
|
|
|
|
|
Backup/restore the xmm6-xmm15 SSE registers within asm functions on win64
According to the Win64 ABI, these registers need to be preserved,
and compilers are allowed to rely on their content to stay
available - not only for float usage but for any usage, anywhere,
in the calling C++ code.
This adds a macro which pushes the clobbered registers onto the
stack if targeting win64 (and a matching one which restores them).
The parameter to the macro is the number of xmm registers used
(e.g. if using xmm0 - xmm7, the parameter is 8), or in other
words, the number of the highest xmm register used plus one.
This is similar to how the same issue is handled for the NEON
registers q4-q7 with the vpush instruction, except that they needed
to be preserved on all platforms, not only on one particular platform.
This allows removing the XMMREG_PROTECT_* hacks, which can
easily fail if the compiler chooses to use the callee saved
xmm registers in an unexpected spot.
2014-03-14 09:29:53 +01:00
|
|
|
POP_XMM
|
2014-01-03 07:49:45 +01:00
|
|
|
LOAD_4_PARA_POP
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
pop r6
|
|
|
|
|
pop r5
|
|
|
|
|
pop r4
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
%assign push_num 0
|
|
|
|
|
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
ret
|
2014-01-03 07:49:45 +01:00
|
|
|
|
|
|
|
|
;***********************************************************************----------------
|
2014-05-31 13:13:34 +02:00
|
|
|
; void ExpandPictureChromaUnalign_sse2( uint8_t *pDst,
|
|
|
|
|
; const int32_t iStride,
|
|
|
|
|
; const int32_t iWidth,
|
|
|
|
|
; const int32_t iHeight );
|
2014-01-03 07:49:45 +01:00
|
|
|
;***********************************************************************----------------
|
2014-03-16 12:23:24 +01:00
|
|
|
WELS_EXTERN ExpandPictureChromaUnalign_sse2
|
2014-05-31 13:13:34 +02:00
|
|
|
push r4
|
2014-01-03 07:49:45 +01:00
|
|
|
push r5
|
|
|
|
|
push r6
|
|
|
|
|
|
|
|
|
|
%assign push_num 3
|
|
|
|
|
LOAD_4_PARA
|
Backup/restore the xmm6-xmm15 SSE registers within asm functions on win64
According to the Win64 ABI, these registers need to be preserved,
and compilers are allowed to rely on their content to stay
available - not only for float usage but for any usage, anywhere,
in the calling C++ code.
This adds a macro which pushes the clobbered registers onto the
stack if targeting win64 (and a matching one which restores them).
The parameter to the macro is the number of xmm registers used
(e.g. if using xmm0 - xmm7, the parameter is 8), or in other
words, the number of the highest xmm register used plus one.
This is similar to how the same issue is handled for the NEON
registers q4-q7 with the vpush instruction, except that they needed
to be preserved on all platforms, not only on one particular platform.
This allows removing the XMMREG_PROTECT_* hacks, which can
easily fail if the compiler chooses to use the callee saved
xmm registers in an unexpected spot.
2014-03-14 09:29:53 +01:00
|
|
|
PUSH_XMM 7
|
2014-01-03 07:49:45 +01:00
|
|
|
|
2014-03-14 09:13:18 +01:00
|
|
|
SIGN_EXTENSION r1,r1d
|
|
|
|
|
SIGN_EXTENSION r2,r2d
|
|
|
|
|
SIGN_EXTENSION r3,r3d
|
2014-01-03 07:49:45 +01:00
|
|
|
|
|
|
|
|
;also prepare for cross border pData top-left:xmm3
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
movzx r6d,byte [r0]
|
|
|
|
|
SSE2_Copy16Times xmm3,r6d ;xmm3: pSrc[0]
|
|
|
|
|
|
|
|
|
|
neg r1
|
|
|
|
|
lea r5,[r0+r1] ;last line of top border r5= dst top pSrc[-stride]
|
|
|
|
|
neg r1
|
|
|
|
|
|
|
|
|
|
push r3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
dec r3 ;h-1
|
|
|
|
|
imul r3,r1 ;(h-1)*stride
|
|
|
|
|
lea r3,[r0+r3] ;pSrc[(h-1)*stride] r3 = src bottom
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
mov r6,r1 ;r6 = stride
|
|
|
|
|
sal r6,04h ;r6 = 32*stride
|
2014-01-05 13:16:22 +01:00
|
|
|
lea r4,[r3+r6] ;r4 = dst bottom
|
|
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
;also prepare for cross border data: bottom-left with xmm5,bottom-right xmm6
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
movzx r6d,byte [r3] ;bottom-left
|
|
|
|
|
SSE2_Copy16Times xmm5,r6d
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r6,[r3+r2-1]
|
|
|
|
|
movzx r6d,byte [r6]
|
|
|
|
|
SSE2_Copy16Times xmm6,r6d ;bottom-right
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = -stride
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
push r0
|
2014-01-05 13:16:22 +01:00
|
|
|
push r1
|
2014-01-03 07:49:45 +01:00
|
|
|
push r2
|
|
|
|
|
|
|
|
|
|
exp_top_bottom_sse2 16
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
; for both left and right border
|
|
|
|
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
pop r2
|
|
|
|
|
pop r1
|
|
|
|
|
pop r0
|
|
|
|
|
|
|
|
|
|
lea r5,[r0-16] ;left border dst luma =32 chroma = -16
|
2014-01-05 13:16:22 +01:00
|
|
|
|
|
|
|
|
lea r3,[r0+r2-1] ;right border src
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r4,[r3+1] ;right border dst
|
|
|
|
|
|
|
|
|
|
;prepare for cross border data: top-rigth with xmm4
|
|
|
|
|
movzx r6d,byte [r3] ;top -rigth
|
|
|
|
|
SSE2_Copy16Times xmm4,r6d
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = stride
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pop r6 ; r6 = height
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
push r0
|
2014-01-05 13:16:22 +01:00
|
|
|
push r1
|
2014-01-03 07:49:45 +01:00
|
|
|
push r2
|
2014-05-31 13:13:34 +02:00
|
|
|
push r6
|
2014-01-03 07:49:45 +01:00
|
|
|
exp_left_right_sse2 16,u
|
|
|
|
|
|
|
|
|
|
pop r6
|
|
|
|
|
pop r2
|
|
|
|
|
pop r1
|
|
|
|
|
pop r0
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
; for cross border [top-left, top-right, bottom-left, bottom-right]
|
|
|
|
|
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
|
2014-01-03 07:49:45 +01:00
|
|
|
; have done xmm3,..,xmm6 cross pData initialization above, perform pading as below, To be continued..
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = -stride
|
|
|
|
|
lea r3,[r0-16]
|
|
|
|
|
lea r3,[r3+r1] ;last line of top-left border
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r4,[r0+r2] ;psrc +width
|
2014-01-05 13:16:22 +01:00
|
|
|
lea r4,[r4+r1] ;psrc +width -stride
|
|
|
|
|
|
|
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ;r1 = stride
|
|
|
|
|
add r6,16 ;height +32(16) ,luma = 32, chroma = 16
|
|
|
|
|
imul r6,r1
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
lea r5,[r3+r6] ;last line of bottom-left border
|
|
|
|
|
lea r6,[r4+r6] ;last line of botoom-right border
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
neg r1 ; r1 = -stride
|
|
|
|
|
|
|
|
|
|
; for left & right border expanding
|
|
|
|
|
exp_cross_sse2 16,u
|
|
|
|
|
|
Backup/restore the xmm6-xmm15 SSE registers within asm functions on win64
According to the Win64 ABI, these registers need to be preserved,
and compilers are allowed to rely on their content to stay
available - not only for float usage but for any usage, anywhere,
in the calling C++ code.
This adds a macro which pushes the clobbered registers onto the
stack if targeting win64 (and a matching one which restores them).
The parameter to the macro is the number of xmm registers used
(e.g. if using xmm0 - xmm7, the parameter is 8), or in other
words, the number of the highest xmm register used plus one.
This is similar to how the same issue is handled for the NEON
registers q4-q7 with the vpush instruction, except that they needed
to be preserved on all platforms, not only on one particular platform.
This allows removing the XMMREG_PROTECT_* hacks, which can
easily fail if the compiler chooses to use the callee saved
xmm registers in an unexpected spot.
2014-03-14 09:29:53 +01:00
|
|
|
POP_XMM
|
2014-01-03 07:49:45 +01:00
|
|
|
LOAD_4_PARA_POP
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
pop r6
|
|
|
|
|
pop r5
|
|
|
|
|
pop r4
|
2014-01-05 13:16:22 +01:00
|
|
|
|
2014-01-03 07:49:45 +01:00
|
|
|
%assign push_num 0
|
|
|
|
|
|
|
|
|
|
|
2014-05-31 13:13:34 +02:00
|
|
|
ret
|
