Mercurial > sdl-ios-xcode
annotate src/hermes/mmxp2_32.asm @ 1187:19d8949b4584
To: sdl@libsdl.org
From: Staffan Ulfberg <staffan@ulfberg.se>
Date: 19 Nov 2005 01:00:48 +0100
Subject: [SDL] New driver for OpenBSD/wscons
Hello,
I've written an SDL driver for OpenBSD/wscons (console mode, somewhat
resembling the functionality of the svga driver for Linux). I use it
for playing MAME on my Sharp Zaurus. The alternative is to play under
X, which is slower.
I asked how to submit the driver a few days ago, and posted a link to
the patch in a follow-up, so maybe it was missed?
Anyway, the patch is on the web at:
http://multivac.fatburen.org/SDL-wscons.patch
Comments?
Staffan
| author | Ryan C. Gordon <icculus@icculus.org> |
|---|---|
| date | Tue, 22 Nov 2005 15:19:50 +0000 |
| parents | da33b7e6d181 |
| children | 2d6dc7de1145 |
| rev | line source |
|---|---|
| 0 | 1 ; |
| 2 ; pII-optimised MMX format converters for HERMES | |
| 3 ; Copyright (c) 1998 Christian Nentwich (c.nentwich@cs.ucl.ac.uk) | |
| 4 ; and (c) 1999 Jonathan Matthew (jmatthew@uq.net.au) | |
| 5 ; This source code is licensed under the GNU LGPL | |
| 6 ; | |
| 7 ; Please refer to the file COPYING.LIB contained in the distribution for | |
| 8 ; licensing conditions | |
| 9 ; | |
| 10 ; COPYRIGHT NOTICE | |
| 11 ; | |
| 12 ; This file partly contains code that is (c) Intel Corporation, specifically | |
| 13 ; the mode detection routine, and the converter to 15 bit (8 pixel | |
| 14 ; conversion routine from the mmx programming tutorial pages). | |
| 15 ; | |
| 16 ; | |
| 17 ; These routines aren't exactly pII optimised - it's just that as they | |
| 18 ; are, they're terrible on p5 MMXs, but less so on pIIs. Someone needs to | |
| 19 ; optimise them for p5 MMXs.. | |
| 20 | |
| 21 BITS 32 | |
| 22 | |
| 23 | |
| 24 GLOBAL _ConvertMMXpII32_24RGB888 | |
| 25 GLOBAL _ConvertMMXpII32_16RGB565 | |
| 26 GLOBAL _ConvertMMXpII32_16BGR565 | |
| 27 GLOBAL _ConvertMMXpII32_16RGB555 | |
| 28 GLOBAL _ConvertMMXpII32_16BGR555 | |
| 29 | |
| 30 EXTERN _mmxreturn | |
| 31 | |
|
1166
da33b7e6d181
Date: Tue, 1 Nov 2005 20:25:10 +0100
Sam Lantinga <slouken@libsdl.org>
parents:
289
diff
changeset
|
32 SECTION .note.GNU-stack noalloc progbits noexec nowrite |
| 0 | 33 SECTION .data |
| 34 | |
| 35 ALIGN 8 | |
| 36 | |
| 37 ;; Constants for conversion routines | |
| 38 | |
| 39 mmx32_rgb888_mask dd 00ffffffh,00ffffffh | |
| 40 | |
| 41 mmx32_rgb565_b dd 000000f8h, 000000f8h | |
| 42 mmx32_rgb565_g dd 0000fc00h, 0000fc00h | |
| 43 mmx32_rgb565_r dd 00f80000h, 00f80000h | |
| 44 | |
| 45 mmx32_rgb555_rb dd 00f800f8h,00f800f8h | |
| 46 mmx32_rgb555_g dd 0000f800h,0000f800h | |
| 47 mmx32_rgb555_mul dd 20000008h,20000008h | |
| 48 mmx32_bgr555_mul dd 00082000h,00082000h | |
| 49 | |
| 50 | |
| 51 | |
| 52 SECTION .text | |
| 53 | |
| 54 _ConvertMMXpII32_24RGB888: | |
| 55 | |
| 56 ; set up mm6 as the mask, mm7 as zero | |
| 57 movq mm6, qword [mmx32_rgb888_mask] | |
| 58 pxor mm7, mm7 | |
| 59 | |
| 60 mov edx, ecx ; save ecx | |
| 61 and ecx, 0fffffffch ; clear lower two bits | |
| 62 jnz .L1 | |
| 63 jmp .L2 | |
| 64 | |
| 65 .L1: | |
| 66 | |
| 67 movq mm0, [esi] ; A R G B a r g b | |
| 68 pand mm0, mm6 ; 0 R G B 0 r g b | |
| 69 movq mm1, [esi+8] ; A R G B a r g b | |
| 70 pand mm1, mm6 ; 0 R G B 0 r g b | |
| 71 | |
| 72 movq mm2, mm0 ; 0 R G B 0 r g b | |
| 73 punpckhdq mm2, mm7 ; 0 0 0 0 0 R G B | |
| 74 punpckldq mm0, mm7 ; 0 0 0 0 0 r g b | |
| 75 psllq mm2, 24 ; 0 0 R G B 0 0 0 | |
| 76 por mm0, mm2 ; 0 0 R G B r g b | |
| 77 | |
| 78 movq mm3, mm1 ; 0 R G B 0 r g b | |
| 79 psllq mm3, 48 ; g b 0 0 0 0 0 0 | |
| 80 por mm0, mm3 ; g b R G B r g b | |
| 81 | |
| 82 movq mm4, mm1 ; 0 R G B 0 r g b | |
| 83 punpckhdq mm4, mm7 ; 0 0 0 0 0 R G B | |
| 84 punpckldq mm1, mm7 ; 0 0 0 0 0 r g b | |
| 85 psrlq mm1, 16 ; 0 0 0 R G B 0 r | |
| 86 psllq mm4, 8 ; 0 0 0 0 R G B 0 | |
| 87 por mm1, mm4 ; 0 0 0 0 R G B r | |
| 88 | |
| 89 movq [edi], mm0 | |
| 90 add esi, BYTE 16 | |
| 91 movd [edi+8], mm1 | |
| 92 add edi, BYTE 12 | |
| 93 sub ecx, BYTE 4 | |
| 94 jnz .L1 | |
| 95 | |
| 96 .L2: | |
| 97 mov ecx, edx | |
| 98 and ecx, BYTE 3 | |
| 99 jz .L4 | |
| 100 .L3: | |
| 101 mov al, [esi] | |
| 102 mov bl, [esi+1] | |
| 103 mov dl, [esi+2] | |
| 104 mov [edi], al | |
| 105 mov [edi+1], bl | |
| 106 mov [edi+2], dl | |
| 107 add esi, BYTE 4 | |
| 108 add edi, BYTE 3 | |
| 109 dec ecx | |
| 110 jnz .L3 | |
| 111 .L4: | |
| 112 jmp _mmxreturn | |
| 113 | |
| 114 | |
| 115 | |
| 116 _ConvertMMXpII32_16RGB565: | |
| 117 | |
| 118 ; set up masks | |
| 119 movq mm5, [mmx32_rgb565_b] | |
| 120 movq mm6, [mmx32_rgb565_g] | |
| 121 movq mm7, [mmx32_rgb565_r] | |
| 122 | |
| 123 mov edx, ecx | |
| 124 shr ecx, 2 | |
| 125 jnz .L1 | |
| 126 jmp .L2 ; not necessary at the moment, but doesn't hurt (much) | |
| 127 | |
| 128 .L1: | |
| 129 movq mm0, [esi] ; argb | |
| 130 movq mm1, mm0 ; argb | |
| 131 pand mm0, mm6 ; 00g0 | |
| 132 movq mm3, mm1 ; argb | |
| 133 pand mm1, mm5 ; 000b | |
| 134 pand mm3, mm7 ; 0r00 | |
| 135 pslld mm1, 2 ; 0 0 000000bb bbb00000 | |
| 136 por mm0, mm1 ; 0 0 ggggggbb bbb00000 | |
| 137 psrld mm0, 5 ; 0 0 00000ggg gggbbbbb | |
| 138 | |
| 139 movq mm4, [esi+8] ; argb | |
| 140 movq mm2, mm4 ; argb | |
| 141 pand mm4, mm6 ; 00g0 | |
| 142 movq mm1, mm2 ; argb | |
| 143 pand mm2, mm5 ; 000b | |
| 144 pand mm1, mm7 ; 0r00 | |
| 145 pslld mm2, 2 ; 0 0 000000bb bbb00000 | |
| 146 por mm4, mm2 ; 0 0 ggggggbb bbb00000 | |
| 147 psrld mm4, 5 ; 0 0 00000ggg gggbbbbb | |
| 148 | |
| 149 packuswb mm3, mm1 ; R 0 r 0 | |
| 150 packssdw mm0, mm4 ; as above.. ish | |
| 151 por mm0, mm3 ; done. | |
| 152 movq [edi], mm0 | |
| 153 | |
| 154 add esi, 16 | |
| 155 add edi, 8 | |
| 156 dec ecx | |
| 157 jnz .L1 | |
| 158 | |
| 159 .L2: | |
| 160 mov ecx, edx | |
| 161 and ecx, BYTE 3 | |
| 162 jz .L4 | |
| 163 .L3: | |
| 164 mov al, [esi] | |
| 165 mov bh, [esi+1] | |
| 166 mov ah, [esi+2] | |
| 167 shr al, 3 | |
| 168 and eax, 0F81Fh ; BYTE? | |
| 169 shr ebx, 5 | |
| 170 and ebx, 07E0h ; BYTE? | |
| 171 add eax, ebx | |
| 172 mov [edi], al | |
| 173 mov [edi+1], ah | |
| 174 add esi, BYTE 4 | |
| 175 add edi, BYTE 2 | |
| 176 dec ecx | |
| 177 jnz .L3 | |
| 178 | |
| 179 .L4: | |
| 180 jmp _mmxreturn | |
| 181 | |
| 182 | |
| 183 _ConvertMMXpII32_16BGR565: | |
| 184 | |
| 185 movq mm5, [mmx32_rgb565_r] | |
| 186 movq mm6, [mmx32_rgb565_g] | |
| 187 movq mm7, [mmx32_rgb565_b] | |
| 188 | |
| 189 mov edx, ecx | |
| 190 shr ecx, 2 | |
| 191 jnz .L1 | |
| 192 jmp .L2 | |
| 193 | |
| 194 .L1: | |
| 195 movq mm0, [esi] ; a r g b | |
| 196 movq mm1, mm0 ; a r g b | |
| 197 pand mm0, mm6 ; 0 0 g 0 | |
| 198 movq mm3, mm1 ; a r g b | |
| 199 pand mm1, mm5 ; 0 r 0 0 | |
| 200 pand mm3, mm7 ; 0 0 0 b | |
| 201 | |
| 202 psllq mm3, 16 ; 0 b 0 0 | |
| 203 psrld mm1, 14 ; 0 0 000000rr rrr00000 | |
| 204 por mm0, mm1 ; 0 0 ggggggrr rrr00000 | |
| 205 psrld mm0, 5 ; 0 0 00000ggg gggrrrrr | |
| 206 | |
| 207 movq mm4, [esi+8] ; a r g b | |
| 208 movq mm2, mm4 ; a r g b | |
| 209 pand mm4, mm6 ; 0 0 g 0 | |
| 210 movq mm1, mm2 ; a r g b | |
| 211 pand mm2, mm5 ; 0 r 0 0 | |
| 212 pand mm1, mm7 ; 0 0 0 b | |
| 213 | |
| 214 psllq mm1, 16 ; 0 b 0 0 | |
| 215 psrld mm2, 14 ; 0 0 000000rr rrr00000 | |
| 216 por mm4, mm2 ; 0 0 ggggggrr rrr00000 | |
| 217 psrld mm4, 5 ; 0 0 00000ggg gggrrrrr | |
| 218 | |
| 219 packuswb mm3, mm1 ; BBBBB000 00000000 bbbbb000 00000000 | |
| 220 packssdw mm0, mm4 ; 00000GGG GGGRRRRR 00000GGG GGGRRRRR | |
| 221 por mm0, mm3 ; BBBBBGGG GGGRRRRR bbbbbggg gggrrrrr | |
| 222 movq [edi], mm0 | |
| 223 | |
| 224 add esi, BYTE 16 | |
| 225 add edi, BYTE 8 | |
| 226 dec ecx | |
| 227 jnz .L1 | |
| 228 | |
| 229 .L2: | |
| 230 and edx, BYTE 3 | |
| 231 jz .L4 | |
| 232 .L3: | |
| 233 mov al, [esi+2] | |
| 234 mov bh, [esi+1] | |
| 235 mov ah, [esi] | |
| 236 shr al, 3 | |
| 237 and eax, 0F81Fh ; BYTE ? | |
| 238 shr ebx, 5 | |
| 239 and ebx, 07E0h ; BYTE ? | |
| 240 add eax, ebx | |
| 241 mov [edi], al | |
| 242 mov [edi+1], ah | |
| 243 add esi, BYTE 4 | |
| 244 add edi, BYTE 2 | |
| 245 dec edx | |
| 246 jnz .L3 | |
| 247 | |
| 248 .L4: | |
| 249 jmp _mmxreturn | |
| 250 | |
| 251 _ConvertMMXpII32_16BGR555: | |
| 252 | |
| 253 ; the 16BGR555 converter is identical to the RGB555 one, | |
| 254 ; except it uses a different multiplier for the pmaddwd | |
| 255 ; instruction. cool huh. | |
| 256 | |
| 257 movq mm7, qword [mmx32_bgr555_mul] | |
| 258 jmp _convert_bgr555_cheat | |
| 259 | |
| 260 ; This is the same as the Intel version.. they obviously went to | |
| 261 ; much more trouble to expand/coil the loop than I did, so theirs | |
| 262 ; would almost certainly be faster, even if only a little. | |
| 263 ; I did rename 'mmx32_rgb555_add' to 'mmx32_rgb555_mul', which is | |
| 264 ; (I think) a more accurate name.. | |
| 265 _ConvertMMXpII32_16RGB555: | |
| 266 | |
| 267 movq mm7,qword [mmx32_rgb555_mul] | |
| 268 _convert_bgr555_cheat: | |
| 269 movq mm6,qword [mmx32_rgb555_g] | |
| 270 | |
| 271 mov edx,ecx ; Save ecx | |
| 272 | |
| 273 and ecx,BYTE 0fffffff8h ; clear lower three bits | |
| 274 jnz .L_OK | |
|
289
77b6110c797d
Fixed "short jump out of range" error in MMX code
Sam Lantinga <slouken@libsdl.org>
parents:
0
diff
changeset
|
275 jmp near .L2 |
| 0 | 276 |
| 277 .L_OK: | |
| 278 | |
| 279 movq mm2,[esi+8] | |
| 280 | |
| 281 movq mm0,[esi] | |
| 282 movq mm3,mm2 | |
| 283 | |
| 284 pand mm3,qword [mmx32_rgb555_rb] | |
| 285 movq mm1,mm0 | |
| 286 | |
| 287 pand mm1,qword [mmx32_rgb555_rb] | |
| 288 pmaddwd mm3,mm7 | |
| 289 | |
| 290 pmaddwd mm1,mm7 | |
| 291 pand mm2,mm6 | |
| 292 | |
| 293 .L1: | |
| 294 movq mm4,[esi+24] | |
| 295 pand mm0,mm6 | |
| 296 | |
| 297 movq mm5,[esi+16] | |
| 298 por mm3,mm2 | |
| 299 | |
| 300 psrld mm3,6 | |
| 301 por mm1,mm0 | |
| 302 | |
| 303 movq mm0,mm4 | |
| 304 psrld mm1,6 | |
| 305 | |
| 306 pand mm0,qword [mmx32_rgb555_rb] | |
| 307 packssdw mm1,mm3 | |
| 308 | |
| 309 movq mm3,mm5 | |
| 310 pmaddwd mm0,mm7 | |
| 311 | |
| 312 pand mm3,qword [mmx32_rgb555_rb] | |
| 313 pand mm4,mm6 | |
| 314 | |
| 315 movq [edi],mm1 | |
| 316 pmaddwd mm3,mm7 | |
| 317 | |
| 318 add esi,BYTE 32 | |
| 319 por mm4,mm0 | |
| 320 | |
| 321 pand mm5,mm6 | |
| 322 psrld mm4,6 | |
| 323 | |
| 324 movq mm2,[esi+8] | |
| 325 por mm5,mm3 | |
| 326 | |
| 327 movq mm0,[esi] | |
| 328 psrld mm5,6 | |
| 329 | |
| 330 movq mm3,mm2 | |
| 331 movq mm1,mm0 | |
| 332 | |
| 333 pand mm3,qword [mmx32_rgb555_rb] | |
| 334 packssdw mm5,mm4 | |
| 335 | |
| 336 pand mm1,qword [mmx32_rgb555_rb] | |
| 337 pand mm2,mm6 | |
| 338 | |
| 339 movq [edi+8],mm5 | |
| 340 pmaddwd mm3,mm7 | |
| 341 | |
| 342 pmaddwd mm1,mm7 | |
| 343 add edi,BYTE 16 | |
| 344 | |
| 345 sub ecx,BYTE 8 | |
| 346 jz .L2 | |
| 347 jmp .L1 | |
| 348 | |
| 349 | |
| 350 .L2: | |
| 351 mov ecx,edx | |
| 352 | |
| 353 and ecx,BYTE 7 | |
| 354 jz .L4 | |
| 355 | |
| 356 .L3: | |
| 357 mov ebx,[esi] | |
| 358 add esi,BYTE 4 | |
| 359 | |
| 360 mov eax,ebx | |
| 361 mov edx,ebx | |
| 362 | |
| 363 shr eax,3 | |
| 364 shr edx,6 | |
| 365 | |
| 366 and eax,BYTE 0000000000011111b | |
| 367 and edx, 0000001111100000b | |
| 368 | |
| 369 shr ebx,9 | |
| 370 | |
| 371 or eax,edx | |
| 372 | |
| 373 and ebx, 0111110000000000b | |
| 374 | |
| 375 or eax,ebx | |
| 376 | |
| 377 mov [edi],ax | |
| 378 add edi,BYTE 2 | |
| 379 | |
| 380 dec ecx | |
| 381 jnz .L3 | |
| 382 | |
| 383 .L4: | |
| 384 jmp _mmxreturn | |
| 385 | |
| 386 | |
| 387 |