Mercurial > sdl-ios-xcode

comparison src/video/SDL_yuv_mmx.c @ 4046:3a9e60224efe SDL-1.2

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Cleaned up tabs.
author Ryan C. Gordon <icculus@icculus.org>
date Wed, 11 Jul 2007 07:21:39 +0000
parents f420bba13676
children 810c6f4ab7aa
comparison
equal deleted inserted replaced
4045:f420bba13676 4046:3a9e60224efe
93 row1 = (Uint32 *)out; // 32 bit target 93 row1 = (Uint32 *)out; // 32 bit target
94 row2 = (Uint32 *)out+cols+mod; // start of second row 94 row2 = (Uint32 *)out+cols+mod; // start of second row
95 mod = (mod+cols+mod)*4; // increment for row1 in byte 95 mod = (mod+cols+mod)*4; // increment for row1 in byte
96 96
97 __asm__ __volatile__ ( 97 __asm__ __volatile__ (
98 // tap dance to workaround the inability to use %%ebx at will... 98 // tap dance to workaround the inability to use %%ebx at will...
99 // move one thing to the stack... 99 // move one thing to the stack...
100 "pushl $0\n" // save a slot on the stack. 100 "pushl $0\n" // save a slot on the stack.
101 "pushl %%ebx\n" // save %%ebx. 101 "pushl %%ebx\n" // save %%ebx.
102 "movl %0, %%ebx\n" // put the thing in ebx. 102 "movl %0, %%ebx\n" // put the thing in ebx.
103 "movl %%ebx, 4(%%esp)\n" // put the thing in the stack slot. 103 "movl %%ebx,4(%%esp)\n" // put the thing in the stack slot.
104 "popl %%ebx\n" // get back %%ebx (the PIC register). 104 "popl %%ebx\n" // get back %%ebx (the PIC register).
105 105
106 ".align 8\n" 106 ".align 8\n"
107 "1:\n" 107 "1:\n"
108 108
109 // create Cr (result in mm1) 109 // create Cr (result in mm1)
110 "pushl %%ebx\n" 110 "pushl %%ebx\n"
111 "movl 4(%%esp), %%ebx\n" 111 "movl 4(%%esp),%%ebx\n"
112 "movd (%%ebx), %%mm1\n" // 0 0 0 0 v3 v2 v1 v0 112 "movd (%%ebx),%%mm1\n" // 0 0 0 0 v3 v2 v1 v0
113 "popl %%ebx\n" 113 "popl %%ebx\n"
114 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00 114 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00
115 "movd (%2), %%mm2\n" // 0 0 0 0 l3 l2 l1 l0 115 "movd (%2), %%mm2\n" // 0 0 0 0 l3 l2 l1 l0
116 "punpcklbw %%mm7,%%mm1\n" // 0 v3 0 v2 00 v1 00 v0 116 "punpcklbw %%mm7,%%mm1\n" // 0 v3 0 v2 00 v1 00 v0
117 "punpckldq %%mm1,%%mm1\n" // 00 v1 00 v0 00 v1 00 v0 117 "punpckldq %%mm1,%%mm1\n" // 00 v1 00 v0 00 v1 00 v0
118 "psubw %9,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0 118 "psubw %9,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0
119 119
120 // create Cr_g (result in mm0) 120 // create Cr_g (result in mm0)
121 "movq %%mm1,%%mm0\n" // r1 r1 r0 r0 r1 r1 r0 r0 121 "movq %%mm1,%%mm0\n" // r1 r1 r0 r0 r1 r1 r0 r0
122 "pmullw %10,%%mm0\n" // red*-46dec=0.7136*64 122 "pmullw %10,%%mm0\n" // red*-46dec=0.7136*64
123 "pmullw %11,%%mm1\n" // red*89dec=1.4013*64 123 "pmullw %11,%%mm1\n" // red*89dec=1.4013*64
124 "psraw $6, %%mm0\n" // red=red/64 124 "psraw $6, %%mm0\n" // red=red/64
125 "psraw $6, %%mm1\n" // red=red/64 125 "psraw $6, %%mm1\n" // red=red/64
126
127 // create L1 L2 (result in mm2,mm4)
128 // L2=lum+cols
129 "movq (%2,%4),%%mm3\n" // 0 0 0 0 L3 L2 L1 L0
130 "punpckldq %%mm3,%%mm2\n" // L3 L2 L1 L0 l3 l2 l1 l0
131 "movq %%mm2,%%mm4\n" // L3 L2 L1 L0 l3 l2 l1 l0
132 "pand %12,%%mm2\n" // L3 0 L1 0 l3 0 l1 0
133 "pand %13,%%mm4\n" // 0 L2 0 L0 0 l2 0 l0
134 "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1
135
136 // create R (result in mm6)
137 "movq %%mm2,%%mm5\n" // 0 L3 0 L1 0 l3 0 l1
138 "movq %%mm4,%%mm6\n" // 0 L2 0 L0 0 l2 0 l0
139 "paddsw %%mm1, %%mm5\n" // lum1+red:x R3 x R1 x r3 x r1
140 "paddsw %%mm1, %%mm6\n" // lum1+red:x R2 x R0 x r2 x r0
141 "packuswb %%mm5,%%mm5\n" // R3 R1 r3 r1 R3 R1 r3 r1
142 "packuswb %%mm6,%%mm6\n" // R2 R0 r2 r0 R2 R0 r2 r0
143 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00
144 "punpcklbw %%mm5,%%mm6\n" // R3 R2 R1 R0 r3 r2 r1 r0
145
146 // create Cb (result in mm1)
147 "movd (%1), %%mm1\n" // 0 0 0 0 u3 u2 u1 u0
148 "punpcklbw %%mm7,%%mm1\n" // 0 u3 0 u2 00 u1 00 u0
149 "punpckldq %%mm1,%%mm1\n" // 00 u1 00 u0 00 u1 00 u0
150 "psubw %9,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0
151 // create Cb_g (result in mm5)
152 "movq %%mm1,%%mm5\n" // u1 u1 u0 u0 u1 u1 u0 u0
153 "pmullw %14,%%mm5\n" // blue*-109dec=1.7129*64
154 "pmullw %15,%%mm1\n" // blue*114dec=1.78125*64
155 "psraw $6, %%mm5\n" // blue=red/64
156 "psraw $6, %%mm1\n" // blue=blue/64
157
158 // create G (result in mm7)
159 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1
160 "movq %%mm4,%%mm7\n" // 0 L2 0 L0 0 l2 0 l1
161 "paddsw %%mm5, %%mm3\n" // lum1+Cb_g:x G3t x G1t x g3t x g1t
162 "paddsw %%mm5, %%mm7\n" // lum1+Cb_g:x G2t x G0t x g2t x g0t
163 "paddsw %%mm0, %%mm3\n" // lum1+Cr_g:x G3 x G1 x g3 x g1
164 "paddsw %%mm0, %%mm7\n" // lum1+blue:x G2 x G0 x g2 x g0
165 "packuswb %%mm3,%%mm3\n" // G3 G1 g3 g1 G3 G1 g3 g1
166 "packuswb %%mm7,%%mm7\n" // G2 G0 g2 g0 G2 G0 g2 g0
167 "punpcklbw %%mm3,%%mm7\n" // G3 G2 G1 G0 g3 g2 g1 g0
168
169 // create B (result in mm5)
170 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1
171 "movq %%mm4,%%mm5\n" // 0 L2 0 L0 0 l2 0 l1
172 "paddsw %%mm1, %%mm3\n" // lum1+blue:x B3 x B1 x b3 x b1
173 "paddsw %%mm1, %%mm5\n" // lum1+blue:x B2 x B0 x b2 x b0
174 "packuswb %%mm3,%%mm3\n" // B3 B1 b3 b1 B3 B1 b3 b1
175 "packuswb %%mm5,%%mm5\n" // B2 B0 b2 b0 B2 B0 b2 b0
176 "punpcklbw %%mm3,%%mm5\n" // B3 B2 B1 B0 b3 b2 b1 b0
177
178 // fill destination row1 (needed are mm6=Rr,mm7=Gg,mm5=Bb)
179
180 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
181 "pxor %%mm4,%%mm4\n" // 0 0 0 0 0 0 0 0
182 "movq %%mm6,%%mm1\n" // R3 R2 R1 R0 r3 r2 r1 r0
183 "movq %%mm5,%%mm3\n" // B3 B2 B1 B0 b3 b2 b1 b0
184 // process lower lum
185 "punpcklbw %%mm4,%%mm1\n" // 0 r3 0 r2 0 r1 0 r0
186 "punpcklbw %%mm4,%%mm3\n" // 0 b3 0 b2 0 b1 0 b0
187 "movq %%mm1,%%mm2\n" // 0 r3 0 r2 0 r1 0 r0
188 "movq %%mm3,%%mm0\n" // 0 b3 0 b2 0 b1 0 b0
189 "punpcklwd %%mm1,%%mm3\n" // 0 r1 0 b1 0 r0 0 b0
190 "punpckhwd %%mm2,%%mm0\n" // 0 r3 0 b3 0 r2 0 b2
191
192 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
193 "movq %%mm7,%%mm1\n" // G3 G2 G1 G0 g3 g2 g1 g0
194 "punpcklbw %%mm1,%%mm2\n" // g3 0 g2 0 g1 0 g0 0
195 "punpcklwd %%mm4,%%mm2\n" // 0 0 g1 0 0 0 g0 0
196 "por %%mm3, %%mm2\n" // 0 r1 g1 b1 0 r0 g0 b0
197 "movq %%mm2,(%3)\n" // wrote out ! row1
198
199 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
200 "punpcklbw %%mm1,%%mm4\n" // g3 0 g2 0 g1 0 g0 0
201 "punpckhwd %%mm2,%%mm4\n" // 0 0 g3 0 0 0 g2 0
202 "por %%mm0, %%mm4\n" // 0 r3 g3 b3 0 r2 g2 b2
203 "movq %%mm4,8(%3)\n" // wrote out ! row1
126 204
127 // create L1 L2 (result in mm2,mm4) 205 // fill destination row2 (needed are mm6=Rr,mm7=Gg,mm5=Bb)
128 // L2=lum+cols 206 // this can be done "destructive"
129 "movq (%2,%4),%%mm3\n" // 0 0 0 0 L3 L2 L1 L0 207 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
130 "punpckldq %%mm3,%%mm2\n" // L3 L2 L1 L0 l3 l2 l1 l0 208 "punpckhbw %%mm2,%%mm6\n" // 0 R3 0 R2 0 R1 0 R0
131 "movq %%mm2,%%mm4\n" // L3 L2 L1 L0 l3 l2 l1 l0 209 "punpckhbw %%mm1,%%mm5\n" // G3 B3 G2 B2 G1 B1 G0 B0
132 "pand %12,%%mm2\n" // L3 0 L1 0 l3 0 l1 0 210 "movq %%mm5,%%mm1\n" // G3 B3 G2 B2 G1 B1 G0 B0
133 "pand %13,%%mm4\n" // 0 L2 0 L0 0 l2 0 l0 211 "punpcklwd %%mm6,%%mm1\n" // 0 R1 G1 B1 0 R0 G0 B0
134 "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1 212 "movq %%mm1,(%5)\n" // wrote out ! row2
135 213 "punpckhwd %%mm6,%%mm5\n" // 0 R3 G3 B3 0 R2 G2 B2
136 // create R (result in mm6) 214 "movq %%mm5,8(%5)\n" // wrote out ! row2
137 "movq %%mm2,%%mm5\n" // 0 L3 0 L1 0 l3 0 l1
138 "movq %%mm4,%%mm6\n" // 0 L2 0 L0 0 l2 0 l0
139 "paddsw %%mm1, %%mm5\n" // lum1+red:x R3 x R1 x r3 x r1
140 "paddsw %%mm1, %%mm6\n" // lum1+red:x R2 x R0 x r2 x r0
141 "packuswb %%mm5,%%mm5\n" // R3 R1 r3 r1 R3 R1 r3 r1
142 "packuswb %%mm6,%%mm6\n" // R2 R0 r2 r0 R2 R0 r2 r0
143 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00
144 "punpcklbw %%mm5,%%mm6\n" // R3 R2 R1 R0 r3 r2 r1 r0
145
146 // create Cb (result in mm1)
147 "movd (%1), %%mm1\n" // 0 0 0 0 u3 u2 u1 u0
148 "punpcklbw %%mm7,%%mm1\n" // 0 u3 0 u2 00 u1 00 u0
149 "punpckldq %%mm1,%%mm1\n" // 00 u1 00 u0 00 u1 00 u0
150 "psubw %9,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0
151 // create Cb_g (result in mm5)
152 "movq %%mm1,%%mm5\n" // u1 u1 u0 u0 u1 u1 u0 u0
153 "pmullw %14,%%mm5\n" // blue*-109dec=1.7129*64
154 "pmullw %15,%%mm1\n" // blue*114dec=1.78125*64
155 "psraw $6, %%mm5\n" // blue=red/64
156 "psraw $6, %%mm1\n" // blue=blue/64
157
158 // create G (result in mm7)
159 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1
160 "movq %%mm4,%%mm7\n" // 0 L2 0 L0 0 l2 0 l1
161 "paddsw %%mm5, %%mm3\n" // lum1+Cb_g:x G3t x G1t x g3t x g1t
162 "paddsw %%mm5, %%mm7\n" // lum1+Cb_g:x G2t x G0t x g2t x g0t
163 "paddsw %%mm0, %%mm3\n" // lum1+Cr_g:x G3 x G1 x g3 x g1
164 "paddsw %%mm0, %%mm7\n" // lum1+blue:x G2 x G0 x g2 x g0
165 "packuswb %%mm3,%%mm3\n" // G3 G1 g3 g1 G3 G1 g3 g1
166 "packuswb %%mm7,%%mm7\n" // G2 G0 g2 g0 G2 G0 g2 g0
167 "punpcklbw %%mm3,%%mm7\n" // G3 G2 G1 G0 g3 g2 g1 g0
168 215
169 // create B (result in mm5) 216 "addl $4,%2\n" // lum+4
170 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1 217 "leal 16(%3),%3\n" // row1+16
171 "movq %%mm4,%%mm5\n" // 0 L2 0 L0 0 l2 0 l1 218 "leal 16(%5),%5\n" // row2+16
172 "paddsw %%mm1, %%mm3\n" // lum1+blue:x B3 x B1 x b3 x b1 219 "addl $2,(%%esp)\n" // cr+2
173 "paddsw %%mm1, %%mm5\n" // lum1+blue:x B2 x B0 x b2 x b0 220 "addl $2,%1\n" // cb+2
174 "packuswb %%mm3,%%mm3\n" // B3 B1 b3 b1 B3 B1 b3 b1 221
175 "packuswb %%mm5,%%mm5\n" // B2 B0 b2 b0 B2 B0 b2 b0 222 "addl $4,%6\n" // x+4
176 "punpcklbw %%mm3,%%mm5\n" // B3 B2 B1 B0 b3 b2 b1 b0 223 "cmpl %4,%6\n"
177 224
178 // fill destination row1 (needed are mm6=Rr,mm7=Gg,mm5=Bb) 225 "jl 1b\n"
179 226 "addl %4,%2\n" // lum += cols
180 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0 227 "addl %8,%3\n" // row1+= mod
181 "pxor %%mm4,%%mm4\n" // 0 0 0 0 0 0 0 0 228 "addl %8,%5\n" // row2+= mod
182 "movq %%mm6,%%mm1\n" // R3 R2 R1 R0 r3 r2 r1 r0 229 "movl $0,%6\n" // x=0
183 "movq %%mm5,%%mm3\n" // B3 B2 B1 B0 b3 b2 b1 b0 230 "cmpl %7,%2\n"
184 // process lower lum 231 "jl 1b\n"
185 "punpcklbw %%mm4,%%mm1\n" // 0 r3 0 r2 0 r1 0 r0 232
186 "punpcklbw %%mm4,%%mm3\n" // 0 b3 0 b2 0 b1 0 b0 233 "addl $4,%%esp\n" // get rid of the stack slot we reserved.
187 "movq %%mm1,%%mm2\n" // 0 r3 0 r2 0 r1 0 r0 234 "emms\n" // reset MMX registers.
188 "movq %%mm3,%%mm0\n" // 0 b3 0 b2 0 b1 0 b0 235 :
189 "punpcklwd %%mm1,%%mm3\n" // 0 r1 0 b1 0 r0 0 b0 236 : "m" (cr), "r"(cb),"r"(lum),
190 "punpckhwd %%mm2,%%mm0\n" // 0 r3 0 b3 0 r2 0 b2 237 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod),
191 238 "m"(MMX_0080w),"m"(MMX_VgrnRGB),"m"(MMX_VredRGB),
192 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0 239 "m"(MMX_FF00w),"m"(MMX_00FFw),"m"(MMX_UgrnRGB),
193 "movq %%mm7,%%mm1\n" // G3 G2 G1 G0 g3 g2 g1 g0 240 "m"(MMX_UbluRGB)
194 "punpcklbw %%mm1,%%mm2\n" // g3 0 g2 0 g1 0 g0 0
195 "punpcklwd %%mm4,%%mm2\n" // 0 0 g1 0 0 0 g0 0
196 "por %%mm3, %%mm2\n" // 0 r1 g1 b1 0 r0 g0 b0
197 "movq %%mm2,(%3)\n" // wrote out ! row1
198
199 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
200 "punpcklbw %%mm1,%%mm4\n" // g3 0 g2 0 g1 0 g0 0
201 "punpckhwd %%mm2,%%mm4\n" // 0 0 g3 0 0 0 g2 0
202 "por %%mm0, %%mm4\n" // 0 r3 g3 b3 0 r2 g2 b2
203 "movq %%mm4,8(%3)\n" // wrote out ! row1
204
205 // fill destination row2 (needed are mm6=Rr,mm7=Gg,mm5=Bb)
206 // this can be done "destructive"
207 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
208 "punpckhbw %%mm2,%%mm6\n" // 0 R3 0 R2 0 R1 0 R0
209 "punpckhbw %%mm1,%%mm5\n" // G3 B3 G2 B2 G1 B1 G0 B0
210 "movq %%mm5,%%mm1\n" // G3 B3 G2 B2 G1 B1 G0 B0
211 "punpcklwd %%mm6,%%mm1\n" // 0 R1 G1 B1 0 R0 G0 B0
212 "movq %%mm1,(%5)\n" // wrote out ! row2
213 "punpckhwd %%mm6,%%mm5\n" // 0 R3 G3 B3 0 R2 G2 B2
214 "movq %%mm5,8(%5)\n" // wrote out ! row2
215
216 "addl $4,%2\n" // lum+4
217 "leal 16(%3),%3\n" // row1+16
218 "leal 16(%5),%5\n" // row2+16
219 "addl $2, (%%esp)\n" // cr+2
220 "addl $2, %1\n" // cb+2
221
222 "addl $4,%6\n" // x+4
223 "cmpl %4,%6\n"
224
225 "jl 1b\n"
226 "addl %4, %2\n" // lum += cols
227 "addl %8, %3\n" // row1+= mod
228 "addl %8, %5\n" // row2+= mod
229 "movl $0, %6\n" // x=0
230 "cmpl %7, %2\n"
231 "jl 1b\n"
232
233 "addl $4, %%esp\n" // get rid of the stack slot we reserved.
234 "emms\n" // reset MMX registers.
235 :
236 : "m" (cr), "r"(cb),"r"(lum),
237 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod),
238 "m"(MMX_0080w),"m"(MMX_VgrnRGB),"m"(MMX_VredRGB),
239 "m"(MMX_FF00w),"m"(MMX_00FFw),"m"(MMX_UgrnRGB),
240 "m"(MMX_UbluRGB)
241 ); 241 );
242 } 242 }
243 243
244 void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix, 244 void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
245 unsigned char *lum, unsigned char *cr, 245 unsigned char *lum, unsigned char *cr,
246 unsigned char *cb, unsigned char *out, 246 unsigned char *cb, unsigned char *out,
247 int rows, int cols, int mod ) 247 int rows, int cols, int mod )
248 { 248 {
249 Uint16 *row1; 249 Uint16 *row1;
250 Uint16 *row2; 250 Uint16 *row2;
251 251
252 unsigned char* y = lum +cols*rows; /* Pointer to the end */ 252 unsigned char* y = lum +cols*rows; /* Pointer to the end */
253 int x=0; 253 int x = 0;
254 row1 = (Uint16 *)out; /* 16 bit target */ 254 row1 = (Uint16 *)out; /* 16 bit target */
255 row2 = (Uint16 *)out+cols+mod; /* start of second row */ 255 row2 = (Uint16 *)out+cols+mod; /* start of second row */
256 mod = (mod+cols+mod)*2; /* increment for row1 in byte */ 256 mod = (mod+cols+mod)*2; /* increment for row1 in byte */
257 257
258 258 __asm__ __volatile__(
259 __asm__ __volatile__( 259 // tap dance to workaround the inability to use %%ebx at will...
260 // tap dance to workaround the inability to use %%ebx at will... 260 // move one thing to the stack...
261 // move one thing to the stack... 261 "pushl $0\n" // save a slot on the stack.
262 "pushl $0\n" // save a slot on the stack. 262 "pushl %%ebx\n" // save %%ebx.
263 "pushl %%ebx\n" // save %%ebx. 263 "movl %0, %%ebx\n" // put the thing in ebx.
264 "movl %0, %%ebx\n" // put the thing in ebx. 264 "movl %%ebx, 4(%%esp)\n" // put the thing in the stack slot.
265 "movl %%ebx, 4(%%esp)\n" // put the thing in the stack slot. 265 "popl %%ebx\n" // get back %%ebx (the PIC register).
266 "popl %%ebx\n" // get back %%ebx (the PIC register). 266
267 267 ".align 8\n"
268 ".align 8\n" 268 "1:\n"
269 "1:\n" 269 "movd (%1), %%mm0\n" // 4 Cb 0 0 0 0 u3 u2 u1 u0
270 "movd (%1), %%mm0\n" // 4 Cb 0 0 0 0 u3 u2 u1 u0 270 "pxor %%mm7, %%mm7\n"
271 "pxor %%mm7, %%mm7\n" 271 "pushl %%ebx\n"
272 "pushl %%ebx\n" 272 "movl 4(%%esp), %%ebx\n"
273 "movl 4(%%esp), %%ebx\n" 273 "movd (%%ebx), %%mm1\n" // 4 Cr 0 0 0 0 v3 v2 v1 v0
274 "movd (%%ebx), %%mm1\n" // 4 Cr 0 0 0 0 v3 v2 v1 v0 274 "popl %%ebx\n"
275 "popl %%ebx\n" 275
276 276 "punpcklbw %%mm7, %%mm0\n" // 4 W cb 0 u3 0 u2 0 u1 0 u0
277 "punpcklbw %%mm7, %%mm0\n" // 4 W cb 0 u3 0 u2 0 u1 0 u0 277 "punpcklbw %%mm7, %%mm1\n" // 4 W cr 0 v3 0 v2 0 v1 0 v0
278 "punpcklbw %%mm7, %%mm1\n" // 4 W cr 0 v3 0 v2 0 v1 0 v0 278 "psubw %9, %%mm0\n"
279 "psubw %9, %%mm0\n" 279 "psubw %9, %%mm1\n"
280 "psubw %9, %%mm1\n" 280 "movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0
281 "movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0 281 "movq %%mm1, %%mm3\n" // Cr
282 "movq %%mm1, %%mm3\n" // Cr 282 "pmullw %10, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0
283 "pmullw %10, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0 283 "movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0
284 "movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0 284 "pmullw %11, %%mm0\n" // Cb2blue
285 "pmullw %11, %%mm0\n" // Cb2blue 285 "pand %12, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0
286 "pand %12, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0 286 "pmullw %13, %%mm3\n" // Cr2green
287 "pmullw %13, %%mm3\n" // Cr2green 287 "movq (%2), %%mm7\n" // L2
288 "movq (%2), %%mm7\n" // L2 288 "pmullw %14, %%mm1\n" // Cr2red
289 "pmullw %14, %%mm1\n" // Cr2red 289 "psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1
290 "psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1 290 "pmullw %15, %%mm6\n" // lum1
291 "pmullw %15, %%mm6\n" // lum1 291 "paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green
292 "paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green 292 "pmullw %15, %%mm7\n" // lum2
293 "pmullw %15, %%mm7\n" // lum2 293
294 294 "movq %%mm6, %%mm4\n" // lum1
295 "movq %%mm6, %%mm4\n" // lum1 295 "paddw %%mm0, %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0
296 "paddw %%mm0, %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0 296 "movq %%mm4, %%mm5\n" // lum1
297 "movq %%mm4, %%mm5\n" // lum1 297 "paddw %%mm1, %%mm4\n" // lum1 +red 00 R6 00 R4 00 R2 00 R0
298 "paddw %%mm1, %%mm4\n" // lum1 +red 00 R6 00 R4 00 R2 00 R0 298 "paddw %%mm2, %%mm5\n" // lum1 +green 00 G6 00 G4 00 G2 00 G0
299 "paddw %%mm2, %%mm5\n" // lum1 +green 00 G6 00 G4 00 G2 00 G0 299 "psraw $6, %%mm4\n" // R1 0 .. 64
300 "psraw $6, %%mm4\n" // R1 0 .. 64 300 "movq %%mm7, %%mm3\n" // lum2 00 L7 00 L5 00 L3 00 L1
301 "movq %%mm7, %%mm3\n" // lum2 00 L7 00 L5 00 L3 00 L1 301 "psraw $6, %%mm5\n" // G1 - .. +
302 "psraw $6, %%mm5\n" // G1 - .. + 302 "paddw %%mm0, %%mm7\n" // Lum2 +blue 00 B7 00 B5 00 B3 00 B1
303 "paddw %%mm0, %%mm7\n" // Lum2 +blue 00 B7 00 B5 00 B3 00 B1 303 "psraw $6, %%mm6\n" // B1 0 .. 64
304 "psraw $6, %%mm6\n" // B1 0 .. 64 304 "packuswb %%mm4, %%mm4\n" // R1 R1
305 "packuswb %%mm4, %%mm4\n" // R1 R1 305 "packuswb %%mm5, %%mm5\n" // G1 G1
306 "packuswb %%mm5, %%mm5\n" // G1 G1 306 "packuswb %%mm6, %%mm6\n" // B1 B1
307 "packuswb %%mm6, %%mm6\n" // B1 B1 307 "punpcklbw %%mm4, %%mm4\n"
308 "punpcklbw %%mm4, %%mm4\n" 308 "punpcklbw %%mm5, %%mm5\n"
309 "punpcklbw %%mm5, %%mm5\n" 309
310 310 "pand %16, %%mm4\n"
311 "pand %16, %%mm4\n" 311 "psllw $3, %%mm5\n" // GREEN 1
312 "psllw $3, %%mm5\n" // GREEN 1 312 "punpcklbw %%mm6, %%mm6\n"
313 "punpcklbw %%mm6, %%mm6\n" 313 "pand %17, %%mm5\n"
314 "pand %17, %%mm5\n" 314 "pand %16, %%mm6\n"
315 "pand %16, %%mm6\n" 315 "por %%mm5, %%mm4\n" //
316 "por %%mm5, %%mm4\n" // 316 "psrlw $11, %%mm6\n" // BLUE 1
317 "psrlw $11, %%mm6\n" // BLUE 1 317 "movq %%mm3, %%mm5\n" // lum2
318 "movq %%mm3, %%mm5\n" // lum2 318 "paddw %%mm1, %%mm3\n" // lum2 +red 00 R7 00 R5 00 R3 00 R1
319 "paddw %%mm1, %%mm3\n" // lum2 +red 00 R7 00 R5 00 R3 00 R1 319 "paddw %%mm2, %%mm5\n" // lum2 +green 00 G7 00 G5 00 G3 00 G1
320 "paddw %%mm2, %%mm5\n" // lum2 +green 00 G7 00 G5 00 G3 00 G1 320 "psraw $6, %%mm3\n" // R2
321 "psraw $6, %%mm3\n" // R2 321 "por %%mm6, %%mm4\n" // MM4
322 "por %%mm6, %%mm4\n" // MM4 322 "psraw $6, %%mm5\n" // G2
323 "psraw $6, %%mm5\n" // G2 323 "movq (%2, %4), %%mm6\n" // L3 load lum2
324 "movq (%2, %4), %%mm6\n" // L3 load lum2 324 "psraw $6, %%mm7\n"
325 "psraw $6, %%mm7\n" 325 "packuswb %%mm3, %%mm3\n"
326 "packuswb %%mm3, %%mm3\n" 326 "packuswb %%mm5, %%mm5\n"
327 "packuswb %%mm5, %%mm5\n" 327 "packuswb %%mm7, %%mm7\n"
328 "packuswb %%mm7, %%mm7\n" 328 "pand %12, %%mm6\n" // L3
329 "pand %12, %%mm6\n" // L3 329 "punpcklbw %%mm3, %%mm3\n"
330 "punpcklbw %%mm3, %%mm3\n" 330 "punpcklbw %%mm5, %%mm5\n"
331 "punpcklbw %%mm5, %%mm5\n" 331 "pmullw %15, %%mm6\n" // lum3
332 "pmullw %15, %%mm6\n" // lum3 332 "punpcklbw %%mm7, %%mm7\n"
333 "punpcklbw %%mm7, %%mm7\n" 333 "psllw $3, %%mm5\n" // GREEN 2
334 "psllw $3, %%mm5\n" // GREEN 2 334 "pand %16, %%mm7\n"
335 "pand %16, %%mm7\n" 335 "pand %16, %%mm3\n"
336 "pand %16, %%mm3\n" 336 "psrlw $11, %%mm7\n" // BLUE 2
337 "psrlw $11, %%mm7\n" // BLUE 2 337 "pand %17, %%mm5\n"
338 "pand %17, %%mm5\n" 338 "por %%mm7, %%mm3\n"
339 "por %%mm7, %%mm3\n" 339 "movq (%2,%4), %%mm7\n" // L4 load lum2
340 "movq (%2,%4), %%mm7\n" // L4 load lum2 340 "por %%mm5, %%mm3\n" //
341 "por %%mm5, %%mm3\n" // 341 "psrlw $8, %%mm7\n" // L4
342 "psrlw $8, %%mm7\n" // L4 342 "movq %%mm4, %%mm5\n"
343 "movq %%mm4, %%mm5\n" 343 "punpcklwd %%mm3, %%mm4\n"
344 "punpcklwd %%mm3, %%mm4\n" 344 "pmullw %15, %%mm7\n" // lum4
345 "pmullw %15, %%mm7\n" // lum4 345 "punpckhwd %%mm3, %%mm5\n"
346 "punpckhwd %%mm3, %%mm5\n" 346
347 347 "movq %%mm4, (%3)\n" // write row1
348 "movq %%mm4, (%3)\n" // write row1 348 "movq %%mm5, 8(%3)\n" // write row1
349 "movq %%mm5, 8(%3)\n" // write row1 349
350 350 "movq %%mm6, %%mm4\n" // Lum3
351 "movq %%mm6, %%mm4\n" // Lum3 351 "paddw %%mm0, %%mm6\n" // Lum3 +blue
352 "paddw %%mm0, %%mm6\n" // Lum3 +blue 352
353 353 "movq %%mm4, %%mm5\n" // Lum3
354 "movq %%mm4, %%mm5\n" // Lum3 354 "paddw %%mm1, %%mm4\n" // Lum3 +red
355 "paddw %%mm1, %%mm4\n" // Lum3 +red 355 "paddw %%mm2, %%mm5\n" // Lum3 +green
356 "paddw %%mm2, %%mm5\n" // Lum3 +green 356 "psraw $6, %%mm4\n"
357 "psraw $6, %%mm4\n" 357 "movq %%mm7, %%mm3\n" // Lum4
358 "movq %%mm7, %%mm3\n" // Lum4 358 "psraw $6, %%mm5\n"
359 "psraw $6, %%mm5\n" 359 "paddw %%mm0, %%mm7\n" // Lum4 +blue
360 "paddw %%mm0, %%mm7\n" // Lum4 +blue 360 "psraw $6, %%mm6\n" // Lum3 +blue
361 "psraw $6, %%mm6\n" // Lum3 +blue 361 "movq %%mm3, %%mm0\n" // Lum4
362 "movq %%mm3, %%mm0\n" // Lum4 362 "packuswb %%mm4, %%mm4\n"
363 "packuswb %%mm4, %%mm4\n" 363 "paddw %%mm1, %%mm3\n" // Lum4 +red
364 "paddw %%mm1, %%mm3\n" // Lum4 +red 364 "packuswb %%mm5, %%mm5\n"
365 "packuswb %%mm5, %%mm5\n" 365 "paddw %%mm2, %%mm0\n" // Lum4 +green
366 "paddw %%mm2, %%mm0\n" // Lum4 +green 366 "packuswb %%mm6, %%mm6\n"
367 "packuswb %%mm6, %%mm6\n" 367 "punpcklbw %%mm4, %%mm4\n"
368 "punpcklbw %%mm4, %%mm4\n" 368 "punpcklbw %%mm5, %%mm5\n"
369 "punpcklbw %%mm5, %%mm5\n" 369 "punpcklbw %%mm6, %%mm6\n"
370 "punpcklbw %%mm6, %%mm6\n" 370 "psllw $3, %%mm5\n" // GREEN 3
371 "psllw $3, %%mm5\n" // GREEN 3 371 "pand %16, %%mm4\n"
372 "pand %16, %%mm4\n" 372 "psraw $6, %%mm3\n" // psr 6
373 "psraw $6, %%mm3\n" // psr 6 373 "psraw $6, %%mm0\n"
374 "psraw $6, %%mm0\n" 374 "pand %16, %%mm6\n" // BLUE
375 "pand %16, %%mm6\n" // BLUE 375 "pand %17, %%mm5\n"
376 "pand %17, %%mm5\n" 376 "psrlw $11, %%mm6\n" // BLUE 3
377 "psrlw $11, %%mm6\n" // BLUE 3 377 "por %%mm5, %%mm4\n"
378 "por %%mm5, %%mm4\n" 378 "psraw $6, %%mm7\n"
379 "psraw $6, %%mm7\n" 379 "por %%mm6, %%mm4\n"
380 "por %%mm6, %%mm4\n" 380 "packuswb %%mm3, %%mm3\n"
381 "packuswb %%mm3, %%mm3\n" 381 "packuswb %%mm0, %%mm0\n"
382 "packuswb %%mm0, %%mm0\n" 382 "packuswb %%mm7, %%mm7\n"
383 "packuswb %%mm7, %%mm7\n" 383 "punpcklbw %%mm3, %%mm3\n"
384 "punpcklbw %%mm3, %%mm3\n" 384 "punpcklbw %%mm0, %%mm0\n"
385 "punpcklbw %%mm0, %%mm0\n" 385 "punpcklbw %%mm7, %%mm7\n"
386 "punpcklbw %%mm7, %%mm7\n" 386 "pand %16, %%mm3\n"
387 "pand %16, %%mm3\n" 387 "pand %16, %%mm7\n" // BLUE
388 "pand %16, %%mm7\n" // BLUE 388 "psllw $3, %%mm0\n" // GREEN 4
389 "psllw $3, %%mm0\n" // GREEN 4 389 "psrlw $11, %%mm7\n"
390 "psrlw $11, %%mm7\n" 390 "pand %17, %%mm0\n"
391 "pand %17, %%mm0\n" 391 "por %%mm7, %%mm3\n"
392 "por %%mm7, %%mm3\n" 392 "por %%mm0, %%mm3\n"
393 "por %%mm0, %%mm3\n" 393
394 394 "movq %%mm4, %%mm5\n"
395 "movq %%mm4, %%mm5\n" 395
396 396 "punpcklwd %%mm3, %%mm4\n"
397 "punpcklwd %%mm3, %%mm4\n" 397 "punpckhwd %%mm3, %%mm5\n"
398 "punpckhwd %%mm3, %%mm5\n" 398
399 399 "movq %%mm4, (%5)\n"
400 "movq %%mm4, (%5)\n" 400 "movq %%mm5, 8(%5)\n"
401 "movq %%mm5, 8(%5)\n" 401
402 402 "addl $8, %6\n"
403 "addl $8, %6\n" 403 "addl $8, %2\n"
404 "addl $8, %2\n" 404 "addl $4, (%%esp)\n"
405 "addl $4, (%%esp)\n" 405 "addl $4, %1\n"
406 "addl $4, %1\n" 406 "cmpl %4, %6\n"
407 "cmpl %4, %6\n" 407 "leal 16(%3), %3\n"
408 "leal 16(%3), %3\n" 408 "leal 16(%5),%5\n" // row2+16
409 "leal 16(%5),%5\n" // row2+16 409
410 410 "jl 1b\n"
411 411 "addl %4, %2\n" // lum += cols
412 "jl 1b\n" 412 "addl %8, %3\n" // row1+= mod
413 "addl %4, %2\n" // lum += cols 413 "addl %8, %5\n" // row2+= mod
414 "addl %8, %3\n" // row1+= mod 414 "movl $0, %6\n" // x=0
415 "addl %8, %5\n" // row2+= mod 415 "cmpl %7, %2\n"
416 "movl $0, %6\n" // x=0 416 "jl 1b\n"
417 "cmpl %7, %2\n" 417 "addl $4, %%esp\n" // get rid of the stack slot we reserved.
418 "jl 1b\n" 418 "emms\n"
419 "addl $4, %%esp\n" // get rid of the stack slot we reserved. 419 :
420 "emms\n" 420 : "m" (cr), "r"(cb),"r"(lum),
421 : 421 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod),
422 : "m" (cr), "r"(cb),"r"(lum), 422 "m"(MMX_0080w),"m"(MMX_Ugrn565),"m"(MMX_Ublu5x5),
423 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod), 423 "m"(MMX_00FFw),"m"(MMX_Vgrn565),"m"(MMX_Vred5x5),
424 "m"(MMX_0080w),"m"(MMX_Ugrn565),"m"(MMX_Ublu5x5), 424 "m"(MMX_Ycoeff),"m"(MMX_red565),"m"(MMX_grn565)
425 "m"(MMX_00FFw),"m"(MMX_Vgrn565),"m"(MMX_Vred5x5), 425 );
426 "m"(MMX_Ycoeff),"m"(MMX_red565),"m"(MMX_grn565));
427 } 426 }
428 427
429 #endif /* GCC3 i386 inline assembly */ 428 #endif /* GCC3 i386 inline assembly */
430 429