1 |
;/************************************************************************** |
;/************************************************************************** |
2 |
; * |
; * |
3 |
; * XVID MPEG-4 VIDEO CODEC |
; * XVID MPEG-4 VIDEO CODEC |
4 |
; * - mmx quantization/dequantization - |
; * - 3dne Quantization/Dequantization - |
5 |
; * |
; * |
6 |
; * Copyright(C) 2001-2003 XviD Team <xvid-devel@xvid.org> |
; * Copyright(C) 2002-2003 Jaan Kalda |
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 |
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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
; * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
; * |
; * |
22 |
; * $Id: quantize_h263_3dne.asm,v 1.1.2.1 2003-10-07 13:02:35 edgomez Exp $ |
; * $Id: quantize_h263_3dne.asm,v 1.1.2.2 2003-10-09 18:50:22 edgomez Exp $ |
23 |
; * |
; * |
24 |
; *************************************************************************/ |
; *************************************************************************/ |
25 |
|
; |
26 |
; these 3dne functions are compatible with iSSE, but are optimized specifically for |
; these 3dne functions are compatible with iSSE, but are optimized specifically for |
27 |
; K7 pipelines |
; K7 pipelines |
|
; |
|
|
;------------------------------------------------------------------------------ |
|
|
; 09.12.2002 Athlon optimizations contributed by Jaan Kalda |
|
|
;------------------------------------------------------------------------------ |
|
28 |
|
|
29 |
; enable dequant saturate [-2048,2047], test purposes only. |
; enable dequant saturate [-2048,2047], test purposes only. |
30 |
%define SATURATE |
%define SATURATE |
34 |
|
|
35 |
bits 32 |
bits 32 |
36 |
|
|
|
%ifdef FORMAT_COFF |
|
|
section .data data |
|
|
%else |
|
|
section .data data align=16 |
|
|
%endif |
|
|
|
|
37 |
%macro cglobal 1 |
%macro cglobal 1 |
38 |
%ifdef PREFIX |
%ifdef PREFIX |
39 |
global _%1 |
global _%1 |
42 |
global %1 |
global %1 |
43 |
%endif |
%endif |
44 |
%endmacro |
%endmacro |
45 |
|
|
46 |
|
;*************************************************************************** |
47 |
|
; Local data |
48 |
|
;*************************************************************************** |
49 |
|
|
50 |
|
%ifdef FORMAT_COFF |
51 |
|
section .data data |
52 |
|
%else |
53 |
|
section .data data align=16 |
54 |
|
%endif |
55 |
|
|
56 |
align 4 |
align 4 |
57 |
int_div |
int_div: |
58 |
dd 0 |
dd 0 |
59 |
%assign i 1 |
%assign i 1 |
60 |
%rep 255 |
%rep 255 |
63 |
%endrep |
%endrep |
64 |
|
|
65 |
align 16 |
align 16 |
66 |
|
plus_one: |
67 |
plus_one times 8 dw 1 |
times 8 dw 1 |
68 |
|
|
69 |
;=========================================================================== |
;=========================================================================== |
70 |
; |
; |
72 |
; |
; |
73 |
;=========================================================================== |
;=========================================================================== |
74 |
|
|
|
%macro MMX_SUB 1 |
|
|
times 4 dw %1 / 2 |
|
|
%endmacro |
|
|
|
|
|
|
|
75 |
align 16 |
align 16 |
76 |
mmx_sub |
mmx_sub: |
77 |
%assign i 1 |
%assign i 1 |
78 |
%rep 31 |
%rep 31 |
79 |
times 4 dw i / 2 |
times 4 dw i / 2 |
92 |
;=========================================================================== |
;=========================================================================== |
93 |
|
|
94 |
align 16 |
align 16 |
95 |
mmx_div |
mmx_div: |
|
|
|
96 |
%assign i 1 |
%assign i 1 |
97 |
%rep 31 |
%rep 31 |
98 |
times 4 dw (1 << 16) / (i * 2) + 1 |
times 4 dw (1 << 16) / (i * 2) + 1 |
105 |
; |
; |
106 |
;=========================================================================== |
;=========================================================================== |
107 |
|
|
|
%macro MMX_ADD 1 |
|
|
%if %1 % 2 != 0 |
|
|
times 4 dw %1 |
|
|
%else |
|
|
times 4 dw %1 - 1 |
|
|
%endif |
|
|
%endmacro |
|
|
|
|
108 |
align 16 |
align 16 |
109 |
mmx_add |
mmx_add: |
|
|
|
110 |
%assign i 1 |
%assign i 1 |
111 |
%rep 31 |
%rep 31 |
112 |
MMX_ADD i |
%if i % 2 != 0 |
113 |
|
times 4 dw i |
114 |
|
%else |
115 |
|
times 4 dw i - 1 |
116 |
|
%endif |
117 |
%assign i i+1 |
%assign i i+1 |
118 |
%endrep |
%endrep |
119 |
|
|
123 |
; |
; |
124 |
;=========================================================================== |
;=========================================================================== |
125 |
|
|
|
%macro MMX_MUL 1 |
|
|
times 4 dw %1 * 2 |
|
|
%endmacro |
|
|
|
|
126 |
align 16 |
align 16 |
127 |
mmx_mul |
mmx_mul: |
|
|
|
128 |
%assign i 1 |
%assign i 1 |
129 |
%rep 31 |
%rep 31 |
130 |
times 4 dw i * 2 |
times 4 dw i * 2 |
138 |
;=========================================================================== |
;=========================================================================== |
139 |
|
|
140 |
align 8 |
align 8 |
141 |
mmx_32768_minus_2048 times 4 dw (32768-2048) |
mmx_32768_minus_2048: |
142 |
mmx_32767_minus_2047 times 4 dw (32767-2047) |
times 4 dw (32768-2048) |
143 |
|
mmx_32767_minus_2047: |
144 |
|
times 4 dw (32767-2047) |
145 |
|
|
146 |
align 16 |
align 16 |
147 |
mmx_2047 times 4 dw 2047 |
mmx_2047: |
148 |
|
times 4 dw 2047 |
149 |
|
|
150 |
align 8 |
align 8 |
151 |
mmzero dd 0, 0 |
mmzero: |
152 |
int2047 dd 2047 |
dd 0, 0 |
153 |
int_2048 dd -2048 |
int2047: |
154 |
|
dd 2047 |
155 |
|
int_2048: |
156 |
|
dd -2048 |
157 |
|
|
158 |
|
;*************************************************************************** |
159 |
|
; Code |
160 |
|
;*************************************************************************** |
161 |
|
|
162 |
section .text |
section .text |
163 |
|
|
164 |
|
|
165 |
;=========================================================================== |
;=========================================================================== |
166 |
; |
; |
167 |
; void quant_intra_3dne(int16_t * coeff, |
; uint32_t quant_h263_intra_3dne(int16_t * coeff, |
168 |
; const int16_t const * data, |
; const int16_t const * data, |
169 |
; const uint32_t quant, |
; const uint32_t quant, |
170 |
; const uint32_t dcscalar); |
; const uint32_t dcscalar); |
171 |
; |
; |
172 |
;=========================================================================== |
;=========================================================================== |
173 |
;This is Athlon-optimized code (ca 70 clk per call) |
;This is Athlon-optimized code (ca 70 clk per call) |
|
;Optimized by Jaan, 30 Nov 2002 |
|
174 |
|
|
175 |
%macro quant_intra1 1 |
%macro quant_intra1 1 |
176 |
|
|
374 |
movq [edx + 3 * 32 + 16], mm5 ;C9 |
movq [edx + 3 * 32 + 16], mm5 ;C9 |
375 |
movq [edx + 3 * 32 + 24], mm7 ;D9 |
movq [edx + 3 * 32 + 24], mm7 ;D9 |
376 |
|
|
377 |
|
xor eax, eax |
378 |
ret |
ret |
379 |
|
|
380 |
align 16 |
align 16 |
410 |
movq [edx + 3 * 32 + 16], mm5 ;C9 |
movq [edx + 3 * 32 + 16], mm5 ;C9 |
411 |
movq [edx + 3 * 32 + 24], mm7 ;D9 |
movq [edx + 3 * 32 + 24], mm7 ;D9 |
412 |
|
|
413 |
|
xor eax, eax |
414 |
ret |
ret |
415 |
|
|
416 |
|
|
418 |
|
|
419 |
;=========================================================================== |
;=========================================================================== |
420 |
; |
; |
421 |
; uint32_t quant_inter_3dne(int16_t * coeff, |
; uint32_t quant_h263_inter_3dne(int16_t * coeff, |
422 |
; const int16_t const * data, |
; const int16_t const * data, |
423 |
; const uint32_t quant); |
; const uint32_t quant); |
424 |
; |
; |
570 |
|
|
571 |
;=========================================================================== |
;=========================================================================== |
572 |
; |
; |
573 |
; void dequant_intra_3dne(int16_t *data, |
; uint32_t dequant_h263_intra_3dne(int16_t *data, |
574 |
; const int16_t const *coeff, |
; const int16_t const *coeff, |
575 |
; const uint32_t quant, |
; const uint32_t quant, |
576 |
; const uint32_t dcscalar); |
; const uint32_t dcscalar); |
714 |
pxor mm3, mm2 ;B13 (6th+) |
pxor mm3, mm2 ;B13 (6th+) |
715 |
movq [edx+4*24+8], mm3 ;B14 (7th) |
movq [edx+4*24+8], mm3 ;B14 (7th) |
716 |
mov [edx], ax |
mov [edx], ax |
717 |
|
|
718 |
|
xor eax, eax |
719 |
ret |
ret |
720 |
|
|
721 |
;=========================================================================== |
;=========================================================================== |
722 |
; |
; |
723 |
; void dequant_inter_3dne(int16_t * data, |
; uint32_t dequant_h263_inter_3dne(int16_t * data, |
724 |
; const int16_t * const coeff, |
; const int16_t * const coeff, |
725 |
; const uint32_t quant); |
; const uint32_t quant); |
726 |
; |
; |
729 |
; this is the same as dequant_inter_3dne, |
; this is the same as dequant_inter_3dne, |
730 |
; except that we're saturating using 'pminsw' (saves 2 cycles/loop) |
; except that we're saturating using 'pminsw' (saves 2 cycles/loop) |
731 |
;This is Athlon-optimized code (ca 100 clk per call) |
;This is Athlon-optimized code (ca 100 clk per call) |
|
;Optimized by Jaan, 30 Nov 2002 |
|
732 |
|
|
733 |
align ALIGN |
align ALIGN |
734 |
cglobal dequant_h263_inter_3dne |
cglobal dequant_h263_inter_3dne |
776 |
pxor mm3, mm2 ;B13 (6th+) |
pxor mm3, mm2 ;B13 (6th+) |
777 |
movq [edx+4*24+8], mm3 ;B14 (7th) |
movq [edx+4*24+8], mm3 ;B14 (7th) |
778 |
|
|
779 |
|
xor eax, eax |
780 |
ret |
ret |