1 |
;/************************************************************************** |
;/**************************************************************************** |
2 |
; * |
; * |
3 |
; * XVID MPEG-4 VIDEO CODEC |
; * XVID MPEG-4 VIDEO CODEC |
4 |
; * mmx 8bit<->16bit transfers |
; * - 8<->16 bit transfer functions - |
5 |
; * |
; * |
6 |
; * This program is an implementation of a part of one or more MPEG-4 |
; * Copyright (C) 2002 Jaan Kalda |
|
; * Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
|
|
; * to use this software module in hardware or software products are |
|
|
; * advised that its use may infringe existing patents or copyrights, and |
|
|
; * any such use would be at such party's own risk. The original |
|
|
; * developer of this software module and his/her company, and subsequent |
|
|
; * editors and their companies, will have no liability for use of this |
|
|
; * software or modifications or derivatives thereof. |
|
7 |
; * |
; * |
8 |
; * This program is free software; you can redistribute it and/or modify |
; * This program is free software; you can redistribute it and/or modify |
9 |
; * 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 |
17 |
; * |
; * |
18 |
; * You should have received a copy of the GNU General Public License |
; * You should have received a copy of the GNU General Public License |
19 |
; * along with this program; if not, write to the Free Software |
; * along with this program; if not, write to the Free Software |
20 |
; * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
; * |
; * |
22 |
; *************************************************************************/ |
; * $Id$ |
23 |
|
; * |
24 |
; these 3dne functions are compatible with iSSE, but are optimized specifically for |
; ***************************************************************************/ |
|
; K7 pipelines |
|
|
; |
|
|
;------------------------------------------------------------------------------ |
|
|
; 09.12.2002 Athlon optimizations contributed by Jaan Kalda |
|
|
;------------------------------------------------------------------------------ |
|
|
|
|
|
|
|
|
bits 32 |
|
|
%ifdef FORMAT_COFF |
|
|
section .data data |
|
|
%else |
|
|
section .data data align=16 |
|
|
%endif |
|
|
|
|
25 |
|
|
26 |
align 8 |
; these 3dne functions are compatible with iSSE, but are optimized specifically |
27 |
mm_zero: |
; for K7 pipelines |
|
dd 0,0 |
|
28 |
|
|
29 |
|
BITS 32 |
30 |
|
|
31 |
%macro cglobal 1 |
%macro cglobal 1 |
32 |
%ifdef PREFIX |
%ifdef PREFIX |
36 |
global %1 |
global %1 |
37 |
%endif |
%endif |
38 |
%endmacro |
%endmacro |
39 |
|
|
40 |
|
;============================================================================= |
41 |
|
; Read only data |
42 |
|
;============================================================================= |
43 |
|
|
44 |
|
%ifdef FORMAT_COFF |
45 |
|
SECTION .rodata |
46 |
|
%else |
47 |
|
SECTION .rodata align=16 |
48 |
|
%endif |
49 |
|
|
50 |
|
ALIGN 8 |
51 |
|
mm_zero: |
52 |
|
dd 0,0 |
53 |
|
;============================================================================= |
54 |
|
; Macros |
55 |
|
;============================================================================= |
56 |
|
|
57 |
%macro nop4 0 |
%macro nop4 0 |
58 |
DB 08Dh,074h,026h,0 |
db 08Dh, 074h, 026h, 0 |
59 |
%endmacro |
%endmacro |
60 |
|
|
61 |
section .text |
;============================================================================= |
62 |
|
; Code |
63 |
|
;============================================================================= |
64 |
|
|
65 |
|
SECTION .text |
66 |
|
|
67 |
cglobal transfer_8to16copy_3dne |
cglobal transfer_8to16copy_3dne |
68 |
cglobal transfer_16to8copy_3dne |
cglobal transfer_16to8copy_3dne |
72 |
cglobal transfer_16to8add_3dne |
cglobal transfer_16to8add_3dne |
73 |
cglobal transfer8x8_copy_3dne |
cglobal transfer8x8_copy_3dne |
74 |
|
|
75 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
76 |
; |
; |
77 |
; void transfer_8to16copy_3dne(int16_t * const dst, |
; void transfer_8to16copy_3dne(int16_t * const dst, |
78 |
; const uint8_t * const src, |
; const uint8_t * const src, |
79 |
; uint32_t stride); |
; uint32_t stride); |
80 |
; |
; |
81 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
82 |
|
|
83 |
align 16 |
ALIGN 16 |
84 |
transfer_8to16copy_3dne: |
transfer_8to16copy_3dne: |
85 |
|
|
86 |
mov eax, [esp+ 8] ; Src |
mov eax, [esp+ 8] ; Src |
142 |
|
|
143 |
|
|
144 |
|
|
145 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
146 |
; |
; |
147 |
; void transfer_16to8copy_3dne(uint8_t * const dst, |
; void transfer_16to8copy_3dne(uint8_t * const dst, |
148 |
; const int16_t * const src, |
; const int16_t * const src, |
149 |
; uint32_t stride); |
; uint32_t stride); |
150 |
; |
; |
151 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
152 |
|
|
153 |
align 16 |
ALIGN 16 |
154 |
transfer_16to8copy_3dne: |
transfer_16to8copy_3dne: |
155 |
|
|
156 |
mov eax, [esp+ 8] ; Src |
mov eax, [esp+ 8] ; Src |
186 |
movq [ecx+2*edx], mm6 |
movq [ecx+2*edx], mm6 |
187 |
ret |
ret |
188 |
|
|
189 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
190 |
; |
; |
191 |
; void transfer_8to16sub_3dne(int16_t * const dct, |
; void transfer_8to16sub_3dne(int16_t * const dct, |
192 |
; uint8_t * const cur, |
; uint8_t * const cur, |
193 |
; const uint8_t * const ref, |
; const uint8_t * const ref, |
194 |
; const uint32_t stride); |
; const uint32_t stride); |
195 |
; |
; |
196 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
|
;/************************************************************************** |
|
|
; * |
|
|
; * History: |
|
|
; * |
|
|
; * 27.12.2001 renamed from 'compensate' to 'transfer_8to16sub' |
|
|
; * 02.12.2001 loop unrolled, code runs 10% faster now (Isibaar) |
|
|
; * 30.11.2001 16 pixels are processed per iteration (Isibaar) |
|
|
; * 30.11.2001 .text missing |
|
|
; * 06.11.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
; * |
|
|
; *************************************************************************/ |
|
197 |
|
|
198 |
; when second argument == 1, reference (ebx) block is to current (eax) |
; when second argument == 1, reference (ebx) block is to current (eax) |
199 |
%macro COPY_8_TO_16_SUB 2 |
%macro COPY_8_TO_16_SUB 2 |
207 |
punpckhbw mm1, mm7 |
punpckhbw mm1, mm7 |
208 |
punpckhbw mm6, mm7 |
punpckhbw mm6, mm7 |
209 |
punpcklbw mm4, mm7 |
punpcklbw mm4, mm7 |
210 |
align 8 |
ALIGN 8 |
211 |
movq mm2, [byte eax+edx] |
movq mm2, [byte eax+edx] |
212 |
punpcklbw mm0, mm7 |
punpcklbw mm0, mm7 |
213 |
movq mm3, [byte eax+edx] |
movq mm3, [byte eax+edx] |
238 |
movq [edi+%1*32+24], mm3 |
movq [edi+%1*32+24], mm3 |
239 |
%endmacro |
%endmacro |
240 |
|
|
241 |
align 16 |
ALIGN 16 |
242 |
transfer_8to16sub_3dne: |
transfer_8to16sub_3dne: |
243 |
mov eax, [esp + 8] ; Cur |
mov eax, [esp + 8] ; Cur |
244 |
mov ecx, [esp +12] ; Ref |
mov ecx, [esp +12] ; Ref |
247 |
mov edi, [esp+4+ 4] ; Dst |
mov edi, [esp+4+ 4] ; Dst |
248 |
pxor mm7, mm7 |
pxor mm7, mm7 |
249 |
nop |
nop |
250 |
align 4 |
ALIGN 4 |
251 |
COPY_8_TO_16_SUB 0, 1 |
COPY_8_TO_16_SUB 0, 1 |
252 |
COPY_8_TO_16_SUB 1, 1 |
COPY_8_TO_16_SUB 1, 1 |
253 |
COPY_8_TO_16_SUB 2, 1 |
COPY_8_TO_16_SUB 2, 1 |
255 |
mov edi,ecx |
mov edi,ecx |
256 |
ret |
ret |
257 |
|
|
258 |
align 16 |
ALIGN 16 |
259 |
transfer_8to16subro_3dne: |
transfer_8to16subro_3dne: |
260 |
mov eax, [esp + 8] ; Cur |
mov eax, [esp + 8] ; Cur |
261 |
mov ecx, [esp +12] ; Ref |
mov ecx, [esp +12] ; Ref |
264 |
mov edi, [esp+4+ 4] ; Dst |
mov edi, [esp+4+ 4] ; Dst |
265 |
pxor mm7, mm7 |
pxor mm7, mm7 |
266 |
nop |
nop |
267 |
align 4 |
ALIGN 4 |
268 |
COPY_8_TO_16_SUB 0, 0 |
COPY_8_TO_16_SUB 0, 0 |
269 |
COPY_8_TO_16_SUB 1, 0 |
COPY_8_TO_16_SUB 1, 0 |
270 |
COPY_8_TO_16_SUB 2, 0 |
COPY_8_TO_16_SUB 2, 0 |
273 |
ret |
ret |
274 |
|
|
275 |
|
|
276 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
277 |
; |
; |
278 |
; void transfer_8to16sub2_3dne(int16_t * const dct, |
; void transfer_8to16sub2_3dne(int16_t * const dct, |
279 |
; uint8_t * const cur, |
; uint8_t * const cur, |
281 |
; const uint8_t * ref2, |
; const uint8_t * ref2, |
282 |
; const uint32_t stride) |
; const uint32_t stride) |
283 |
; |
; |
284 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
285 |
|
|
286 |
%macro COPY_8_TO_16_SUB2_SSE 1 |
%macro COPY_8_TO_16_SUB2_SSE 1 |
287 |
db 0Fh, 6Fh, 44h, 20h, 00 ;movq mm0, [byte eax] ; cur |
db 0Fh, 6Fh, 44h, 20h, 00 ;movq mm0, [byte eax] ; cur |
295 |
|
|
296 |
movq mm4, [byte ebx] ; ref1 |
movq mm4, [byte ebx] ; ref1 |
297 |
pavgb mm4, [byte esi] ; ref2 |
pavgb mm4, [byte esi] ; ref2 |
298 |
|
movq [eax], mm4 |
299 |
movq mm5, [ebx+edx] ; ref |
movq mm5, [ebx+edx] ; ref |
300 |
pavgb mm5, [esi+edx] ; ref2 |
pavgb mm5, [esi+edx] ; ref2 |
301 |
|
movq [eax+edx], mm5 |
302 |
movq mm6, mm4 |
movq mm6, mm4 |
303 |
punpcklbw mm4, mm7 |
punpcklbw mm4, mm7 |
304 |
punpckhbw mm6, mm7 |
punpckhbw mm6, mm7 |
324 |
movq [ecx+%1*32+24], mm3 |
movq [ecx+%1*32+24], mm3 |
325 |
%endmacro |
%endmacro |
326 |
|
|
327 |
align 16 |
ALIGN 16 |
328 |
transfer_8to16sub2_3dne: |
transfer_8to16sub2_3dne: |
329 |
mov edx, [esp +20] ; Stride |
mov edx, [esp +20] ; Stride |
330 |
mov ecx, [esp + 4] ; Dst |
mov ecx, [esp + 4] ; Dst |
344 |
ret |
ret |
345 |
|
|
346 |
|
|
347 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
348 |
; |
; |
349 |
; void transfer_16to8add_3dne(uint8_t * const dst, |
; void transfer_16to8add_3dne(uint8_t * const dst, |
350 |
; const int16_t * const src, |
; const int16_t * const src, |
351 |
; uint32_t stride); |
; uint32_t stride); |
352 |
; |
; |
353 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
354 |
|
|
355 |
%macro COPY_16_TO_8_ADD 1 |
%macro COPY_16_TO_8_ADD 1 |
356 |
db 0Fh, 6Fh, 44h, 21h, 00 ;movq mm0, [byte ecx] |
db 0Fh, 6Fh, 44h, 21h, 00 ;movq mm0, [byte ecx] |
373 |
%endmacro |
%endmacro |
374 |
|
|
375 |
|
|
376 |
align 16 |
ALIGN 16 |
377 |
transfer_16to8add_3dne: |
transfer_16to8add_3dne: |
378 |
mov ecx, [esp+ 4] ; Dst |
mov ecx, [esp+ 4] ; Dst |
379 |
mov edx, [esp+12] ; Stride |
mov edx, [esp+12] ; Stride |
390 |
COPY_16_TO_8_ADD 3 |
COPY_16_TO_8_ADD 3 |
391 |
ret |
ret |
392 |
|
|
393 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
394 |
; |
; |
395 |
; void transfer8x8_copy_3dne(uint8_t * const dst, |
; void transfer8x8_copy_3dne(uint8_t * const dst, |
396 |
; const uint8_t * const src, |
; const uint8_t * const src, |
397 |
; const uint32_t stride); |
; const uint32_t stride); |
398 |
; |
; |
399 |
; |
; |
400 |
;=========================================================================== |
;----------------------------------------------------------------------------- |
401 |
|
|
402 |
%macro COPY_8_TO_8 0 |
%macro COPY_8_TO_8 0 |
403 |
movq mm0, [byte eax] |
movq mm0, [byte eax] |
407 |
movq [ecx+edx], mm1 |
movq [ecx+edx], mm1 |
408 |
%endmacro |
%endmacro |
409 |
|
|
410 |
align 16 |
ALIGN 16 |
411 |
transfer8x8_copy_3dne: |
transfer8x8_copy_3dne: |
412 |
mov eax, [esp+ 8] ; Src |
mov eax, [esp+ 8] ; Src |
413 |
mov edx, [esp+12] ; Stride |
mov edx, [esp+12] ; Stride |