1 /*****************************************************************************
2 * video_yuv_asm.h: YUV transformation assembly
3 *****************************************************************************
4 * Copyright (C) 1999, 2000 VideoLAN
7 * Olie Lho <ollie@sis.com.tw>
9 * Adapted to VideoLAN by:
10 * Gaël Hendryckx <jimmy@via.ecp.fr>
11 * Samuel Hocevar <sam@via.ecp.fr>
13 * This program is free software; you can redistribute it and/or modify
14 * it under the terms of the GNU General Public License as published by
15 * the Free Software Foundation; either version 2 of the License, or
16 * (at your option) any later version.
18 * This program is distributed in the hope that it will be useful,
19 * but WITHOUT ANY WARRANTY; without even the implied warranty of
20 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 * General Public License for more details.
23 * You should have received a copy of the GNU General Public
24 * License along with this program; if not, write to the
25 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
26 * Boston, MA 02111-1307, USA.
27 *****************************************************************************/
29 /* hope these constant values are cache line aligned */
30 static unsigned long long mmx_80w = 0x0080008000800080;
31 static unsigned long long mmx_10w = 0x1010101010101010;
32 static unsigned long long mmx_00ffw = 0x00ff00ff00ff00ff;
33 static unsigned long long mmx_Y_coeff = 0x253f253f253f253f;
35 static unsigned long long mmx_U_green = 0xf37df37df37df37d;
36 static unsigned long long mmx_U_blue = 0x4093409340934093;
37 static unsigned long long mmx_V_red = 0x3312331233123312;
38 static unsigned long long mmx_V_green = 0xe5fce5fce5fce5fc;
40 static unsigned long long mmx_redmask = 0xf8f8f8f8f8f8f8f8;
41 static unsigned long long mmx_grnmask = 0xfcfcfcfcfcfcfcfc;
42 static unsigned long long mmx_grnshift = 0x03;
43 static unsigned long long mmx_blueshift = 0x03;
45 #define MMX_INIT_16 " \n\
47 movd (%1), %%mm0 # Load 4 Cb 00 00 00 00 u3 u2 u1 u0 \n\
48 movd (%2), %%mm1 # Load 4 Cr 00 00 00 00 v3 v2 v1 v0 \n\
49 pxor %%mm4, %%mm4 # zero mm4 \n\
50 movq (%0), %%mm6 # Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 \n\
51 #movl $0, (%3) # cache preload for image \n\
54 #define MMX_INIT_32 " \n\
56 movd (%1), %%mm0 # Load 4 Cb 00 00 00 00 u3 u2 u1 u0 \n\
57 movl $0, (%3) # cache preload for image \n\
58 movd (%2), %%mm1 # Load 4 Cr 00 00 00 00 v3 v2 v1 v0 \n\
59 pxor %%mm4, %%mm4 # zero mm4 \n\
60 movq (%0), %%mm6 # Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 \n\
64 * Do the multiply part of the conversion for even and odd pixels,
66 * mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
67 * mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
68 * mm6 -> Y even, mm7 -> Y odd
71 #define MMX_YUV_MUL " \n\
73 # convert the chroma part \n\
74 punpcklbw %%mm4, %%mm0 # scatter 4 Cb 00 u3 00 u2 00 u1 00 u0 \n\
75 punpcklbw %%mm4, %%mm1 # scatter 4 Cr 00 v3 00 v2 00 v1 00 v0 \n\
76 psubsw mmx_80w, %%mm0 # Cb -= 128 \n\
77 psubsw mmx_80w, %%mm1 # Cr -= 128 \n\
78 psllw $3, %%mm0 # Promote precision \n\
79 psllw $3, %%mm1 # Promote precision \n\
80 movq %%mm0, %%mm2 # Copy 4 Cb 00 u3 00 u2 00 u1 00 u0 \n\
81 movq %%mm1, %%mm3 # Copy 4 Cr 00 v3 00 v2 00 v1 00 v0 \n\
82 pmulhw mmx_U_green, %%mm2 # Mul Cb with green coeff -> Cb green \n\
83 pmulhw mmx_V_green, %%mm3 # Mul Cr with green coeff -> Cr green \n\
84 pmulhw mmx_U_blue, %%mm0 # Mul Cb -> Cblue 00 b3 00 b2 00 b1 00 b0 \n\
85 pmulhw mmx_V_red, %%mm1 # Mul Cr -> Cred 00 r3 00 r2 00 r1 00 \n\
86 paddsw %%mm3, %%mm2 # Cb green + Cr green -> Cgreen \n\
88 # convert the luma part \n\
89 psubusb mmx_10w, %%mm6 # Y -= 16 \n\
90 movq %%mm6, %%mm7 # Copy 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 \n\
91 pand mmx_00ffw, %%mm6 # get Y even 00 Y6 00 Y4 00 Y2 00 Y0 \n\
92 psrlw $8, %%mm7 # get Y odd 00 Y7 00 Y5 00 Y3 00 Y1 \n\
93 psllw $3, %%mm6 # Promote precision \n\
94 psllw $3, %%mm7 # Promote precision \n\
95 pmulhw mmx_Y_coeff, %%mm6 # Mul 4 Y even 00 y6 00 y4 00 y2 00 y0 \n\
96 pmulhw mmx_Y_coeff, %%mm7 # Mul 4 Y odd 00 y7 00 y5 00 y3 00 y1 \n\
100 * Do the addition part of the conversion for even and odd pixels,
102 * mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
103 * mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
104 * mm6 -> Y even, mm7 -> Y odd
107 #define MMX_YUV_ADD " \n\
109 # Do horizontal and vertical scaling \n\
110 movq %%mm0, %%mm3 # Copy Cblue \n\
111 movq %%mm1, %%mm4 # Copy Cred \n\
112 movq %%mm2, %%mm5 # Copy Cgreen \n\
113 paddsw %%mm6, %%mm0 # Y even + Cblue 00 B6 00 B4 00 B2 00 B0 \n\
114 paddsw %%mm7, %%mm3 # Y odd + Cblue 00 B7 00 B5 00 B3 00 B1 \n\
115 paddsw %%mm6, %%mm1 # Y even + Cred 00 R6 00 R4 00 R2 00 R0 \n\
116 paddsw %%mm7, %%mm4 # Y odd + Cred 00 R7 00 R5 00 R3 00 R1 \n\
117 paddsw %%mm6, %%mm2 # Y even + Cgreen 00 G6 00 G4 00 G2 00 G0 \n\
118 paddsw %%mm7, %%mm5 # Y odd + Cgreen 00 G7 00 G5 00 G3 00 G1 \n\
120 # Limit RGB even to 0..255 \n\
121 packuswb %%mm0, %%mm0 # B6 B4 B2 B0 | B6 B4 B2 B0 \n\
122 packuswb %%mm1, %%mm1 # R6 R4 R2 R0 | R6 R4 R2 R0 \n\
123 packuswb %%mm2, %%mm2 # G6 G4 G2 G0 | G6 G4 G2 G0 \n\
125 # Limit RGB odd to 0..255 \n\
126 packuswb %%mm3, %%mm3 # B7 B5 B3 B1 | B7 B5 B3 B1 \n\
127 packuswb %%mm4, %%mm4 # R7 R5 R3 R1 | R7 R5 R3 R1 \n\
128 packuswb %%mm5, %%mm5 # G7 G5 G3 G1 | G7 G5 G3 G1 \n\
130 # Interleave RGB even and odd \n\
131 punpcklbw %%mm3, %%mm0 # B7 B6 B5 B4 B3 B2 B1 B0 \n\
132 punpcklbw %%mm4, %%mm1 # R7 R6 R5 R4 R3 R2 R1 R0 \n\
133 punpcklbw %%mm5, %%mm2 # G7 G6 G5 G4 G3 G2 G1 G0 \n\
136 #define MMX_UNPACK_16 " \n\
138 # mask unneeded bits off \n\
139 pand mmx_redmask, %%mm0 # b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 \n\
140 pand mmx_grnmask, %%mm2 # g7g6g5g4 g3g2_0_0 g7g6g5g4 g3g2_0_0 \n\
141 pand mmx_redmask, %%mm1 # r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 \n\
142 psrlw mmx_blueshift,%%mm0 # 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 \n\
143 pxor %%mm4, %%mm4 # zero mm4 \n\
144 movq %%mm0, %%mm5 # Copy B7-B0 \n\
145 movq %%mm2, %%mm7 # Copy G7-G0 \n\
147 # convert rgb24 plane to rgb16 pack for pixel 0-3 \n\
148 punpcklbw %%mm4, %%mm2 # 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0 \n\
149 punpcklbw %%mm1, %%mm0 # r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 \n\
150 psllw mmx_blueshift,%%mm2 # 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0 \n\
151 por %%mm2, %%mm0 # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 \n\
152 movq 8(%0), %%mm6 # Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 \n\
153 movq %%mm0, (%3) # store pixel 0-3 \n\
155 # convert rgb24 plane to rgb16 pack for pixel 0-3 \n\
156 punpckhbw %%mm4, %%mm7 # 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0 \n\
157 punpckhbw %%mm1, %%mm5 # r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 \n\
158 psllw mmx_blueshift,%%mm7 # 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0 \n\
159 movd 4(%1), %%mm0 # Load 4 Cb 00 00 00 00 u3 u2 u1 u0 \n\
160 por %%mm7, %%mm5 # r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 \n\
161 movd 4(%2), %%mm1 # Load 4 Cr 00 00 00 00 v3 v2 v1 v0 \n\
162 movq %%mm5, 8(%3) # store pixel 4-7 \n\
166 * convert RGB plane to RGB packed format,
167 * mm0 -> B, mm1 -> R, mm2 -> G, mm3 -> 0,
168 * mm4 -> GB, mm5 -> AR pixel 4-7,
169 * mm6 -> GB, mm7 -> AR pixel 0-3
172 #define MMX_UNPACK_32 " \n\
174 pxor %%mm3, %%mm3 # zero mm3 \n\
175 movq %%mm0, %%mm6 # B7 B6 B5 B4 B3 B2 B1 B0 \n\
176 movq %%mm1, %%mm7 # R7 R6 R5 R4 R3 R2 R1 R0 \n\
177 movq %%mm0, %%mm4 # B7 B6 B5 B4 B3 B2 B1 B0 \n\
178 movq %%mm1, %%mm5 # R7 R6 R5 R4 R3 R2 R1 R0 \n\
179 punpcklbw %%mm2, %%mm6 # G3 B3 G2 B2 G1 B1 G0 B0 \n\
180 punpcklbw %%mm3, %%mm7 # 00 R3 00 R2 00 R1 00 R0 \n\
181 punpcklwd %%mm7, %%mm6 # 00 R1 B1 G1 00 R0 B0 G0 \n\
182 movq %%mm6, (%3) # Store ARGB1 ARGB0 \n\
183 movq %%mm0, %%mm6 # B7 B6 B5 B4 B3 B2 B1 B0 \n\
184 punpcklbw %%mm2, %%mm6 # G3 B3 G2 B2 G1 B1 G0 B0 \n\
185 punpckhwd %%mm7, %%mm6 # 00 R3 G3 B3 00 R2 B3 G2 \n\
186 movq %%mm6, 8(%3) # Store ARGB3 ARGB2 \n\
187 punpckhbw %%mm2, %%mm4 # G7 B7 G6 B6 G5 B5 G4 B4 \n\
188 punpckhbw %%mm3, %%mm5 # 00 R7 00 R6 00 R5 00 R4 \n\
189 punpcklwd %%mm5, %%mm4 # 00 R5 B5 G5 00 R4 B4 G4 \n\
190 movq %%mm4, 16(%3) # Store ARGB5 ARGB4 \n\
191 movq %%mm0, %%mm4 # B7 B6 B5 B4 B3 B2 B1 B0 \n\
192 punpckhbw %%mm2, %%mm4 # G7 B7 G6 B6 G5 B5 G4 B4 \n\
193 punpckhwd %%mm5, %%mm4 # 00 R7 G7 B7 00 R6 B6 G6 \n\
194 movq %%mm4, 24(%3) # Store ARGB7 ARGB6 \n\
195 movd 4(%1), %%mm0 # Load 4 Cb 00 00 00 00 u3 u2 u1 u0 \n\
196 movd 4(%2), %%mm1 # Load 4 Cr 00 00 00 00 v3 v2 v1 v0 \n\
197 pxor %%mm4, %%mm4 # zero mm4 \n\
198 movq 8(%0), %%mm6 # Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 \n\