2 * Copyright 2000-2001 VA Linux Systems, Inc.
3 * (C) Copyright IBM Corporation 2004
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * on the rights to use, copy, modify, merge, publish, distribute, sub
10 * license, and/or sell copies of the Software, and to permit persons to whom
11 * the Software is furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice (including the next
14 * paragraph) shall be included in all copies or substantial portions of the
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
20 * VA LINUX SYSTEM, IBM AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
21 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
22 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
23 * USE OR OTHER DEALINGS IN THE SOFTWARE.
29 * Template file of span read / write functions.
31 * \author Keith Whitwell <keithw@tungstengraphics.com>
32 * \author Gareth Hughes <gareth@nvidia.com>
33 * \author Ian Romanick <idr@us.ibm.com>
43 #define HW_WRITE_LOCK() HW_LOCK()
46 #ifndef HW_WRITE_UNLOCK
47 #define HW_WRITE_UNLOCK() HW_UNLOCK()
51 #define HW_READ_LOCK() HW_LOCK()
54 #ifndef HW_READ_UNLOCK
55 #define HW_READ_UNLOCK() HW_UNLOCK()
58 #ifndef HW_READ_CLIPLOOP
59 #define HW_READ_CLIPLOOP() HW_CLIPLOOP()
62 #ifndef HW_WRITE_CLIPLOOP
63 #define HW_WRITE_CLIPLOOP() HW_CLIPLOOP()
66 #if (SPANTMP_PIXEL_FMT == GL_RGB) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_5_6_5)
68 #define INIT_MONO_PIXEL(p, color) \
69 p = PACK_COLOR_565( color[0], color[1], color[2] )
71 #define WRITE_RGBA( _x, _y, r, g, b, a ) \
73 GLshort * _p = (GLshort *) GET_DST_PTR(_x, _y); \
74 _p[0] = ((((int)r & 0xf8) << 8) | (((int)g & 0xfc) << 3) | \
75 (((int)b & 0xf8) >> 3)); \
78 #define WRITE_PIXEL( _x, _y, p ) \
80 GLushort * _p = (GLushort *) GET_DST_PTR(_x, _y); \
84 #define READ_RGBA( rgba, _x, _y ) \
86 GLushort p = *(volatile GLshort *) GET_SRC_PTR(_x, _y); \
87 rgba[0] = ((p >> 8) & 0xf8) * 255 / 0xf8; \
88 rgba[1] = ((p >> 3) & 0xfc) * 255 / 0xfc; \
89 rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \
93 #elif (SPANTMP_PIXEL_FMT == GL_BGRA) && (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV)
95 # define INIT_MONO_PIXEL(p, color) \
96 p = PACK_COLOR_8888(color[3], color[0], color[1], color[2])
98 # define WRITE_RGBA(_x, _y, r, g, b, a) \
100 GLuint * _p = (GLuint *) GET_DST_PTR(_x, _y); \
101 _p[0] = ((r << 16) | (g << 8) | (b << 0) | (a << 24)); \
104 #define WRITE_PIXEL(_x, _y, p) \
106 GLuint * _p = (GLuint *) GET_DST_PTR(_x, _y); \
110 # if defined( USE_X86_ASM )
111 # define READ_RGBA(rgba, _x, _y) \
113 GLuint p = *(volatile GLuint *) GET_SRC_PTR(_x, _y); \
114 __asm__ __volatile__( "bswap %0; rorl $8, %0" \
115 : "=r" (p) : "r" (p) ); \
116 ((GLuint *)rgba)[0] = p; \
118 # elif defined( MESA_BIG_ENDIAN )
119 /* On PowerPC with GCC 3.4.2 the shift madness below becomes a single
120 * rotlwi instruction. It also produces good code on SPARC.
122 # define READ_RGBA( rgba, _x, _y ) \
124 GLuint p = *(volatile GLuint *) GET_SRC_PTR(_x, _y); \
126 *((uint32_t *) rgba) = (t >> 24) | (p << 8); \
129 # define READ_RGBA( rgba, _x, _y ) \
131 GLuint p = *(volatile GLuint *) GET_SRC_PTR(_x, _y); \
132 rgba[0] = (p >> 16) & 0xff; \
133 rgba[1] = (p >> 8) & 0xff; \
134 rgba[2] = (p >> 0) & 0xff; \
135 rgba[3] = (p >> 24) & 0xff; \
140 #error SPANTMP_PIXEL_FMT must be set to a valid value!
143 #if defined( USE_MMX_ASM ) || defined( USE_SSE_ASM )
144 #include "x86/read_rgba_span_x86.h"
145 #include "x86/common_x86_asm.h"
148 static void TAG(WriteRGBASpan
)( const GLcontext
*ctx
,
149 GLuint n
, GLint x
, GLint y
,
150 const GLubyte rgba
[][4],
151 const GLubyte mask
[] )
164 CLIPSPAN(x
,y
,n
,x1
,n1
,i
);
166 if (DBG
) fprintf(stderr
, "WriteRGBASpan %d..%d (x1 %d)\n",
167 (int)i
, (int)n1
, (int)x1
);
171 for (;n1
>0;i
++,x1
++,n1
--)
174 rgba
[i
][0], rgba
[i
][1],
175 rgba
[i
][2], rgba
[i
][3] );
179 for (;n1
>0;i
++,x1
++,n1
--)
181 rgba
[i
][0], rgba
[i
][1],
182 rgba
[i
][2], rgba
[i
][3] );
190 static void TAG(WriteRGBSpan
)( const GLcontext
*ctx
,
191 GLuint n
, GLint x
, GLint y
,
192 const GLubyte rgb
[][3],
193 const GLubyte mask
[] )
206 CLIPSPAN(x
,y
,n
,x1
,n1
,i
);
208 if (DBG
) fprintf(stderr
, "WriteRGBSpan %d..%d (x1 %d)\n",
209 (int)i
, (int)n1
, (int)x1
);
213 for (;n1
>0;i
++,x1
++,n1
--)
215 WRITE_RGBA( x1
, y
, rgb
[i
][0], rgb
[i
][1], rgb
[i
][2], 255 );
219 for (;n1
>0;i
++,x1
++,n1
--)
220 WRITE_RGBA( x1
, y
, rgb
[i
][0], rgb
[i
][1], rgb
[i
][2], 255 );
228 static void TAG(WriteRGBAPixels
)( const GLcontext
*ctx
,
232 const GLubyte rgba
[][4],
233 const GLubyte mask
[] )
240 if (DBG
) fprintf(stderr
, "WriteRGBAPixels\n");
249 const int fy
= Y_FLIP(y
[i
]);
250 if (CLIPPIXEL(x
[i
],fy
))
251 WRITE_RGBA( x
[i
], fy
,
252 rgba
[i
][0], rgba
[i
][1],
253 rgba
[i
][2], rgba
[i
][3] );
261 const int fy
= Y_FLIP(y
[i
]);
262 if (CLIPPIXEL(x
[i
],fy
))
263 WRITE_RGBA( x
[i
], fy
,
264 rgba
[i
][0], rgba
[i
][1],
265 rgba
[i
][2], rgba
[i
][3] );
275 static void TAG(WriteMonoRGBASpan
)( const GLcontext
*ctx
,
276 GLuint n
, GLint x
, GLint y
,
277 const GLchan color
[4],
278 const GLubyte mask
[] )
285 INIT_MONO_PIXEL(p
, color
);
289 if (DBG
) fprintf(stderr
, "WriteMonoRGBASpan\n");
294 CLIPSPAN(x
,y
,n
,x1
,n1
,i
);
297 for (;n1
>0;i
++,x1
++,n1
--)
299 WRITE_PIXEL( x1
, y
, p
);
303 for (;n1
>0;i
++,x1
++,n1
--)
304 WRITE_PIXEL( x1
, y
, p
);
313 static void TAG(WriteMonoRGBAPixels
)( const GLcontext
*ctx
,
315 const GLint x
[], const GLint y
[],
316 const GLchan color
[],
317 const GLubyte mask
[] )
323 INIT_MONO_PIXEL(p
, color
);
325 if (DBG
) fprintf(stderr
, "WriteMonoRGBAPixels\n");
333 int fy
= Y_FLIP(y
[i
]);
334 if (CLIPPIXEL( x
[i
], fy
))
335 WRITE_PIXEL( x
[i
], fy
, p
);
341 int fy
= Y_FLIP(y
[i
]);
342 if (CLIPPIXEL( x
[i
], fy
))
343 WRITE_PIXEL( x
[i
], fy
, p
);
353 static void TAG(ReadRGBASpan
)( const GLcontext
*ctx
,
354 GLuint n
, GLint x
, GLint y
,
364 if (DBG
) fprintf(stderr
, "ReadRGBASpan\n");
369 CLIPSPAN(x
,y
,n
,x1
,n1
,i
);
370 for (;n1
>0;i
++,x1
++,n1
--)
371 READ_RGBA( rgba
[i
], x1
, y
);
379 #if defined(USE_MMX_ASM) && \
380 (((SPANTMP_PIXEL_FMT == GL_BGRA) && \
381 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV)) || \
382 ((SPANTMP_PIXEL_FMT == GL_RGB) && \
383 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_5_6_5)))
384 static void TAG2(ReadRGBASpan
,_MMX
)( const GLcontext
*ctx
,
385 GLuint n
, GLint x
, GLint y
,
388 #ifndef USE_INNER_EMMS
389 /* The EMMS instruction is directly in-lined here because using GCC's
390 * built-in _mm_empty function was found to utterly destroy performance.
392 __asm__
__volatile__( "emms" );
402 if (DBG
) fprintf(stderr
, "ReadRGBASpan\n");
407 CLIPSPAN(x
,y
,n
,x1
,n1
,i
);
410 const char * src
= GET_SRC_PTR( x1
, y
);
411 #if (SPANTMP_PIXEL_FMT == GL_RGB) && \
412 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_5_6_5)
413 _generic_read_RGBA_span_RGB565_MMX( src
, rgba
[i
], n1
);
415 _generic_read_RGBA_span_BGRA8888_REV_MMX( src
, rgba
[i
], n1
);
422 #ifndef USE_INNER_EMMS
423 __asm__
__volatile__( "emms" );
429 #if defined(USE_SSE_ASM) && \
430 (SPANTMP_PIXEL_FMT == GL_BGRA) && \
431 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV)
432 static void TAG2(ReadRGBASpan
,_SSE2
)( const GLcontext
*ctx
,
433 GLuint n
, GLint x
, GLint y
,
443 if (DBG
) fprintf(stderr
, "ReadRGBASpan\n");
448 CLIPSPAN(x
,y
,n
,x1
,n1
,i
);
451 const char * src
= GET_SRC_PTR( x1
, y
);
452 _generic_read_RGBA_span_BGRA8888_REV_SSE2( src
, rgba
[i
], n1
);
461 #if defined(USE_SSE_ASM) && \
462 (SPANTMP_PIXEL_FMT == GL_BGRA) && \
463 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV)
464 static void TAG2(ReadRGBASpan
,_SSE
)( const GLcontext
*ctx
,
465 GLuint n
, GLint x
, GLint y
,
468 #ifndef USE_INNER_EMMS
469 /* The EMMS instruction is directly in-lined here because using GCC's
470 * built-in _mm_empty function was found to utterly destroy performance.
472 __asm__
__volatile__( "emms" );
482 if (DBG
) fprintf(stderr
, "ReadRGBASpan\n");
487 CLIPSPAN(x
,y
,n
,x1
,n1
,i
);
490 const char * src
= GET_SRC_PTR( x1
, y
);
491 _generic_read_RGBA_span_BGRA8888_REV_SSE( src
, rgba
[i
], n1
);
497 #ifndef USE_INNER_EMMS
498 __asm__
__volatile__( "emms" );
504 static void TAG(ReadRGBAPixels
)( const GLcontext
*ctx
,
505 GLuint n
, const GLint x
[], const GLint y
[],
506 GLubyte rgba
[][4], const GLubyte mask
[] )
513 if (DBG
) fprintf(stderr
, "ReadRGBAPixels\n");
521 int fy
= Y_FLIP( y
[i
] );
522 if (CLIPPIXEL( x
[i
], fy
))
523 READ_RGBA( rgba
[i
], x
[i
], fy
);
529 int fy
= Y_FLIP( y
[i
] );
530 if (CLIPPIXEL( x
[i
], fy
))
531 READ_RGBA( rgba
[i
], x
[i
], fy
);
540 static void TAG(InitPointers
)(struct swrast_device_driver
*swdd
)
542 swdd
->WriteRGBASpan
= TAG(WriteRGBASpan
);
543 swdd
->WriteRGBSpan
= TAG(WriteRGBSpan
);
544 swdd
->WriteMonoRGBASpan
= TAG(WriteMonoRGBASpan
);
545 swdd
->WriteRGBAPixels
= TAG(WriteRGBAPixels
);
546 swdd
->WriteMonoRGBAPixels
= TAG(WriteMonoRGBAPixels
);
547 swdd
->ReadRGBAPixels
= TAG(ReadRGBAPixels
);
549 #if defined(USE_SSE_ASM) && \
550 (SPANTMP_PIXEL_FMT == GL_BGRA) && \
551 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV)
552 if ( cpu_has_xmm2
) {
553 if (DBG
) fprintf( stderr
, "Using %s version of ReadRGBASpan\n", "SSE2" );
554 swdd
->ReadRGBASpan
= TAG2(ReadRGBASpan
, _SSE2
);
558 #if defined(USE_SSE_ASM) && \
559 (SPANTMP_PIXEL_FMT == GL_BGRA) && \
560 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV)
562 if (DBG
) fprintf( stderr
, "Using %s version of ReadRGBASpan\n", "SSE" );
563 swdd
->ReadRGBASpan
= TAG2(ReadRGBASpan
, _SSE
);
567 #if defined(USE_MMX_ASM) && \
568 (((SPANTMP_PIXEL_FMT == GL_BGRA) && \
569 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_INT_8_8_8_8_REV)) || \
570 ((SPANTMP_PIXEL_FMT == GL_RGB) && \
571 (SPANTMP_PIXEL_TYPE == GL_UNSIGNED_SHORT_5_6_5)))
573 if (DBG
) fprintf( stderr
, "Using %s version of ReadRGBASpan\n", "MMX" );
574 swdd
->ReadRGBASpan
= TAG2(ReadRGBASpan
, _MMX
);
579 if (DBG
) fprintf( stderr
, "Using %s version of ReadRGBASpan\n", "C" );
580 swdd
->ReadRGBASpan
= TAG(ReadRGBASpan
);
586 #undef INIT_MONO_PIXEL
594 #undef SPANTMP_PIXEL_FMT
595 #undef SPANTMP_PIXEL_TYPE