1 |
;/************************************************************************** |
;/***************************************************************************** |
2 |
; * |
; * |
3 |
; * XVID MPEG-4 VIDEO CODEC |
; * XVID MPEG-4 VIDEO CODEC |
4 |
; * mmx 8x8 block-based halfpel interpolation |
; * - mmx 8x8 block-based halfpel interpolation - |
5 |
|
; * |
6 |
|
; * Copyright(C) 2001 Peter Ross <pross@xvid.org> |
7 |
|
; * 2002 Michael Militzer <isibaar@xvid.org> |
8 |
; * |
; * |
9 |
; * This program is free software; you can redistribute it and/or modify |
; * This program is free software; you can redistribute it and/or modify |
10 |
; * it under the terms of the GNU General Public License as published by |
; * it under the terms of the GNU General Public License as published by |
18 |
; * |
; * |
19 |
; * You should have received a copy of the GNU General Public License |
; * You should have received a copy of the GNU General Public License |
20 |
; * along with this program; if not, write to the Free Software |
; * along with this program; if not, write to the Free Software |
21 |
; * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
|
; * |
|
|
; *************************************************************************/ |
|
|
|
|
|
;/************************************************************************** |
|
|
; * |
|
|
; * History: |
|
22 |
; * |
; * |
23 |
; * 05.10.2002 added some qpel mmx code - Isibaar |
; ****************************************************************************/ |
|
; * 06.07.2002 mmx cleanup - Isibaar |
|
|
; * 22.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
; * |
|
|
; *************************************************************************/ |
|
|
|
|
24 |
|
|
25 |
bits 32 |
BITS 32 |
26 |
|
|
27 |
%macro cglobal 1 |
%macro cglobal 1 |
28 |
%ifdef PREFIX |
%ifdef PREFIX |
33 |
%endif |
%endif |
34 |
%endmacro |
%endmacro |
35 |
|
|
36 |
section .data |
;============================================================================= |
37 |
|
; Read only data |
38 |
|
;============================================================================= |
39 |
|
|
40 |
align 16 |
%ifdef FORMAT_COFF |
41 |
|
SECTION .rodata data |
42 |
|
%else |
43 |
|
SECTION .rodata data align=16 |
44 |
|
%endif |
45 |
|
|
46 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
47 |
; (16 - r) rounding table |
; (16 - r) rounding table |
48 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
49 |
|
|
50 |
rounding_lowpass_mmx |
ALIGN 16 |
51 |
|
rounding_lowpass_mmx: |
52 |
times 4 dw 16 |
times 4 dw 16 |
53 |
times 4 dw 15 |
times 4 dw 15 |
54 |
|
|
55 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
56 |
; (1 - r) rounding table |
; (1 - r) rounding table |
57 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
58 |
|
|
59 |
rounding1_mmx |
rounding1_mmx: |
60 |
times 4 dw 1 |
times 4 dw 1 |
61 |
times 4 dw 0 |
times 4 dw 0 |
62 |
|
|
63 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
64 |
; (2 - r) rounding table |
; (2 - r) rounding table |
65 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
66 |
|
|
67 |
rounding2_mmx |
rounding2_mmx: |
68 |
times 4 dw 2 |
times 4 dw 2 |
69 |
times 4 dw 1 |
times 4 dw 1 |
70 |
|
|
71 |
mmx_one |
mmx_one: |
72 |
times 8 db 1 |
times 8 db 1 |
73 |
|
|
74 |
mmx_two |
mmx_two: |
75 |
times 8 db 2 |
times 8 db 2 |
76 |
|
|
77 |
mmx_three |
mmx_three: |
78 |
times 8 db 3 |
times 8 db 3 |
79 |
|
|
80 |
mmx_five |
mmx_five: |
81 |
times 4 dw 5 |
times 4 dw 5 |
82 |
|
|
83 |
mmx_mask |
mmx_mask: |
84 |
times 8 db 254 |
times 8 db 254 |
85 |
|
|
86 |
mmx_mask2 |
mmx_mask2: |
87 |
times 8 db 252 |
times 8 db 252 |
88 |
|
|
89 |
section .text |
;============================================================================= |
90 |
|
; Code |
91 |
|
;============================================================================= |
92 |
|
|
93 |
|
SECTION .text |
94 |
|
|
95 |
|
cglobal interpolate8x8_halfpel_h_mmx |
96 |
|
cglobal interpolate8x8_halfpel_v_mmx |
97 |
|
cglobal interpolate8x8_halfpel_hv_mmx |
98 |
|
cglobal interpolate8x8_avg4_mmx |
99 |
|
cglobal interpolate8x8_avg2_mmx |
100 |
|
cglobal interpolate8x8_6tap_lowpass_h_mmx |
101 |
|
cglobal interpolate8x8_6tap_lowpass_v_mmx |
102 |
|
|
103 |
%macro CALC_AVG 6 |
%macro CALC_AVG 6 |
104 |
punpcklbw %3, %6 |
punpcklbw %3, %6 |
111 |
|
|
112 |
psrlw %1, 1 ; mm01 >>= 1 |
psrlw %1, 1 ; mm01 >>= 1 |
113 |
psrlw %2, 1 |
psrlw %2, 1 |
|
|
|
114 |
%endmacro |
%endmacro |
115 |
|
|
116 |
|
|
117 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
118 |
; |
; |
119 |
; void interpolate8x8_halfpel_h_mmx(uint8_t * const dst, |
; void interpolate8x8_halfpel_h_mmx(uint8_t * const dst, |
120 |
; const uint8_t * const src, |
; const uint8_t * const src, |
121 |
; const uint32_t stride, |
; const uint32_t stride, |
122 |
; const uint32_t rounding); |
; const uint32_t rounding); |
123 |
; |
; |
124 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
125 |
|
|
126 |
%macro COPY_H_MMX 0 |
%macro COPY_H_MMX 0 |
127 |
movq mm0, [esi] |
movq mm0, [esi] |
141 |
add edi, edx ; dst += stride |
add edi, edx ; dst += stride |
142 |
%endmacro |
%endmacro |
143 |
|
|
144 |
align 16 |
ALIGN 16 |
145 |
cglobal interpolate8x8_halfpel_h_mmx |
interpolate8x8_halfpel_h_mmx: |
|
interpolate8x8_halfpel_h_mmx |
|
146 |
|
|
147 |
push esi |
push esi |
148 |
push edi |
push edi |
|
|
|
149 |
mov eax, [esp + 8 + 16] ; rounding |
mov eax, [esp + 8 + 16] ; rounding |
150 |
|
|
|
interpolate8x8_halfpel_h_mmx.start |
|
151 |
movq mm7, [rounding1_mmx + eax * 8] |
movq mm7, [rounding1_mmx + eax * 8] |
152 |
|
|
153 |
mov edi, [esp + 8 + 4] ; dst |
mov edi, [esp + 8 + 4] ; dst |
171 |
ret |
ret |
172 |
|
|
173 |
|
|
174 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
175 |
; |
; |
176 |
; void interpolate8x8_halfpel_v_mmx(uint8_t * const dst, |
; void interpolate8x8_halfpel_v_mmx(uint8_t * const dst, |
177 |
; const uint8_t * const src, |
; const uint8_t * const src, |
178 |
; const uint32_t stride, |
; const uint32_t stride, |
179 |
; const uint32_t rounding); |
; const uint32_t rounding); |
180 |
; |
; |
181 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
182 |
|
|
183 |
%macro COPY_V_MMX 0 |
%macro COPY_V_MMX 0 |
184 |
movq mm0, [esi] |
movq mm0, [esi] |
198 |
add edi, edx ; dst += stride |
add edi, edx ; dst += stride |
199 |
%endmacro |
%endmacro |
200 |
|
|
201 |
align 16 |
ALIGN 16 |
202 |
cglobal interpolate8x8_halfpel_v_mmx |
interpolate8x8_halfpel_v_mmx: |
|
interpolate8x8_halfpel_v_mmx |
|
203 |
|
|
204 |
push esi |
push esi |
205 |
push edi |
push edi |
206 |
|
|
207 |
mov eax, [esp + 8 + 16] ; rounding |
mov eax, [esp + 8 + 16] ; rounding |
208 |
|
|
|
interpolate8x8_halfpel_v_mmx.start |
|
209 |
movq mm7, [rounding1_mmx + eax * 8] |
movq mm7, [rounding1_mmx + eax * 8] |
210 |
|
|
211 |
mov edi, [esp + 8 + 4] ; dst |
mov edi, [esp + 8 + 4] ; dst |
230 |
ret |
ret |
231 |
|
|
232 |
|
|
233 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
234 |
; |
; |
235 |
; void interpolate8x8_halfpel_hv_mmx(uint8_t * const dst, |
; void interpolate8x8_halfpel_hv_mmx(uint8_t * const dst, |
236 |
; const uint8_t * const src, |
; const uint8_t * const src, |
238 |
; const uint32_t rounding); |
; const uint32_t rounding); |
239 |
; |
; |
240 |
; |
; |
241 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
242 |
|
|
243 |
%macro COPY_HV_MMX 0 |
%macro COPY_HV_MMX 0 |
244 |
; current row |
; current row |
|
|
|
245 |
movq mm0, [esi] |
movq mm0, [esi] |
246 |
movq mm2, [esi + 1] |
movq mm2, [esi + 1] |
247 |
|
|
257 |
paddusw mm1, mm3 |
paddusw mm1, mm3 |
258 |
|
|
259 |
; next row |
; next row |
|
|
|
260 |
movq mm4, [esi + edx] |
movq mm4, [esi + edx] |
261 |
movq mm2, [esi + edx + 1] |
movq mm2, [esi + edx + 1] |
262 |
|
|
272 |
paddusw mm5, mm3 |
paddusw mm5, mm3 |
273 |
|
|
274 |
; add current + next row |
; add current + next row |
|
|
|
275 |
paddusw mm0, mm4 ; mm01 += mm45 |
paddusw mm0, mm4 ; mm01 += mm45 |
276 |
paddusw mm1, mm5 |
paddusw mm1, mm5 |
277 |
paddusw mm0, mm7 ; mm01 += rounding2 |
paddusw mm0, mm7 ; mm01 += rounding2 |
287 |
add edi, edx ; dst += stride |
add edi, edx ; dst += stride |
288 |
%endmacro |
%endmacro |
289 |
|
|
290 |
align 16 |
ALIGN 16 |
291 |
cglobal interpolate8x8_halfpel_hv_mmx |
interpolate8x8_halfpel_hv_mmx: |
|
interpolate8x8_halfpel_hv_mmx |
|
292 |
|
|
293 |
push esi |
push esi |
294 |
push edi |
push edi |
295 |
|
|
296 |
mov eax, [esp + 8 + 16] ; rounding |
mov eax, [esp + 8 + 16] ; rounding |
|
interpolate8x8_halfpel_hv_mmx.start |
|
297 |
|
|
298 |
movq mm7, [rounding2_mmx + eax * 8] |
movq mm7, [rounding2_mmx + eax * 8] |
299 |
|
|
320 |
|
|
321 |
ret |
ret |
322 |
|
|
323 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
324 |
; |
; |
325 |
; void interpolate8x8_avg2_mmx(uint8_t const *dst, |
; void interpolate8x8_avg2_mmx(uint8_t const *dst, |
326 |
; const uint8_t * const src1, |
; const uint8_t * const src1, |
329 |
; const uint32_t rounding, |
; const uint32_t rounding, |
330 |
; const uint32_t height); |
; const uint32_t height); |
331 |
; |
; |
332 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
333 |
|
|
334 |
%macro AVG2_MMX_RND0 0 |
%macro AVG2_MMX_RND0 0 |
335 |
movq mm0, [eax] ; src1 -> mm0 |
movq mm0, [eax] ; src1 -> mm0 |
425 |
movq [ecx+edx], mm4 |
movq [ecx+edx], mm4 |
426 |
%endmacro |
%endmacro |
427 |
|
|
428 |
align 16 |
ALIGN 16 |
429 |
cglobal interpolate8x8_avg2_mmx |
interpolate8x8_avg2_mmx: |
|
interpolate8x8_avg2_mmx |
|
430 |
|
|
431 |
push ebx |
push ebx |
432 |
|
|
495 |
ret |
ret |
496 |
|
|
497 |
|
|
498 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
499 |
; |
; |
500 |
; void interpolate8x8_avg4_mmx(uint8_t const *dst, |
; void interpolate8x8_avg4_mmx(uint8_t const *dst, |
501 |
; const uint8_t * const src1, |
; const uint8_t * const src1, |
505 |
; const uint32_t stride, |
; const uint32_t stride, |
506 |
; const uint32_t rounding); |
; const uint32_t rounding); |
507 |
; |
; |
508 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
509 |
|
|
510 |
%macro AVG4_MMX_RND0 0 |
%macro AVG4_MMX_RND0 0 |
511 |
movq mm0, [eax] ; src1 -> mm0 |
movq mm0, [eax] ; src1 -> mm0 |
617 |
movq [ecx], mm0 ; (src1 + src2 + src3 + src4 + 2) / 4 -> dst |
movq [ecx], mm0 ; (src1 + src2 + src3 + src4 + 2) / 4 -> dst |
618 |
%endmacro |
%endmacro |
619 |
|
|
620 |
align 16 |
ALIGN 16 |
621 |
cglobal interpolate8x8_avg4_mmx |
interpolate8x8_avg4_mmx: |
|
interpolate8x8_avg4_mmx |
|
622 |
|
|
623 |
push ebx |
push ebx |
624 |
push edi |
push edi |
683 |
ret |
ret |
684 |
|
|
685 |
|
|
686 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
687 |
; |
; |
688 |
; void interpolate8x8_6tap_lowpass_h_mmx(uint8_t const *dst, |
; void interpolate8x8_6tap_lowpass_h_mmx(uint8_t const *dst, |
689 |
; const uint8_t * const src, |
; const uint8_t * const src, |
690 |
; const uint32_t stride, |
; const uint32_t stride, |
691 |
; const uint32_t rounding); |
; const uint32_t rounding); |
692 |
; |
; |
693 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
694 |
|
|
695 |
%macro LOWPASS_6TAP_H_MMX 0 |
%macro LOWPASS_6TAP_H_MMX 0 |
696 |
movq mm0, [eax] |
movq mm0, [eax] |
761 |
movq [ecx], mm0 |
movq [ecx], mm0 |
762 |
%endmacro |
%endmacro |
763 |
|
|
764 |
align 16 |
ALIGN 16 |
765 |
cglobal interpolate8x8_6tap_lowpass_h_mmx |
interpolate8x8_6tap_lowpass_h_mmx: |
|
interpolate8x8_6tap_lowpass_h_mmx |
|
766 |
|
|
767 |
mov eax, [esp + 16] ; rounding |
mov eax, [esp + 16] ; rounding |
768 |
|
|
792 |
|
|
793 |
ret |
ret |
794 |
|
|
795 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
796 |
; |
; |
797 |
; void interpolate8x8_6tap_lowpass_v_mmx(uint8_t const *dst, |
; void interpolate8x8_6tap_lowpass_v_mmx(uint8_t const *dst, |
798 |
; const uint8_t * const src, |
; const uint8_t * const src, |
799 |
; const uint32_t stride, |
; const uint32_t stride, |
800 |
; const uint32_t rounding); |
; const uint32_t rounding); |
801 |
; |
; |
802 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
803 |
|
|
804 |
%macro LOWPASS_6TAP_V_MMX 0 |
%macro LOWPASS_6TAP_V_MMX 0 |
805 |
movq mm0, [eax] |
movq mm0, [eax] |
871 |
movq [ecx], mm0 |
movq [ecx], mm0 |
872 |
%endmacro |
%endmacro |
873 |
|
|
874 |
align 16 |
ALIGN 16 |
875 |
cglobal interpolate8x8_6tap_lowpass_v_mmx |
interpolate8x8_6tap_lowpass_v_mmx: |
|
interpolate8x8_6tap_lowpass_v_mmx |
|
876 |
|
|
877 |
push ebx |
push ebx |
878 |
|
|