patch to import Jon Smirl's work from Bitkeeper
[mesa.git] / src / mesa / drivers / dri / radeon / radeon_span.c
1 /* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_span.c,v 1.6 2002/10/30 12:51:56 alanh Exp $ */
2 /**************************************************************************
3
4 Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
5 VA Linux Systems Inc., Fremont, California.
6
7 All Rights Reserved.
8
9 Permission is hereby granted, free of charge, to any person obtaining
10 a copy of this software and associated documentation files (the
11 "Software"), to deal in the Software without restriction, including
12 without limitation the rights to use, copy, modify, merge, publish,
13 distribute, sublicense, and/or sell copies of the Software, and to
14 permit persons to whom the Software is furnished to do so, subject to
15 the following conditions:
16
17 The above copyright notice and this permission notice (including the
18 next paragraph) shall be included in all copies or substantial
19 portions of the Software.
20
21 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
22 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
23 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
24 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
25 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
26 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
27 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
28
29 **************************************************************************/
30
31 /*
32 * Authors:
33 * Kevin E. Martin <martin@valinux.com>
34 * Gareth Hughes <gareth@valinux.com>
35 * Keith Whitwell <keith@tungstengraphics.com>
36 *
37 */
38
39 #include "glheader.h"
40 #include "swrast/swrast.h"
41
42 #include "radeon_context.h"
43 #include "radeon_ioctl.h"
44 #include "radeon_state.h"
45 #include "radeon_span.h"
46 #include "radeon_tex.h"
47
48 #define DBG 0
49
50 #define LOCAL_VARS \
51 radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \
52 radeonScreenPtr radeonScreen = rmesa->radeonScreen; \
53 __DRIscreenPrivate *sPriv = rmesa->dri.screen; \
54 __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; \
55 GLuint pitch = radeonScreen->frontPitch * radeonScreen->cpp; \
56 GLuint height = dPriv->h; \
57 char *buf = (char *)(sPriv->pFB + \
58 rmesa->state.color.drawOffset + \
59 (dPriv->x * radeonScreen->cpp) + \
60 (dPriv->y * pitch)); \
61 char *read_buf = (char *)(sPriv->pFB + \
62 rmesa->state.pixel.readOffset + \
63 (dPriv->x * radeonScreen->cpp) + \
64 (dPriv->y * pitch)); \
65 GLuint p; \
66 (void) read_buf; (void) buf; (void) p
67
68 #define LOCAL_DEPTH_VARS \
69 radeonContextPtr rmesa = RADEON_CONTEXT(ctx); \
70 radeonScreenPtr radeonScreen = rmesa->radeonScreen; \
71 __DRIscreenPrivate *sPriv = rmesa->dri.screen; \
72 __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; \
73 GLuint height = dPriv->h; \
74 GLuint xo = dPriv->x; \
75 GLuint yo = dPriv->y; \
76 char *buf = (char *)(sPriv->pFB + radeonScreen->depthOffset); \
77 (void) buf
78
79 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
80
81
82 #define CLIPPIXEL( _x, _y ) \
83 ((_x >= minx) && (_x < maxx) && (_y >= miny) && (_y < maxy))
84
85
86 #define CLIPSPAN( _x, _y, _n, _x1, _n1, _i ) \
87 if ( _y < miny || _y >= maxy ) { \
88 _n1 = 0, _x1 = x; \
89 } else { \
90 _n1 = _n; \
91 _x1 = _x; \
92 if ( _x1 < minx ) _i += (minx-_x1), n1 -= (minx-_x1), _x1 = minx; \
93 if ( _x1 + _n1 >= maxx ) n1 -= (_x1 + n1 - maxx); \
94 }
95
96 #define Y_FLIP( _y ) (height - _y - 1)
97
98
99 #define HW_LOCK()
100
101 #define HW_CLIPLOOP() \
102 do { \
103 __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; \
104 int _nc = dPriv->numClipRects; \
105 \
106 while ( _nc-- ) { \
107 int minx = dPriv->pClipRects[_nc].x1 - dPriv->x; \
108 int miny = dPriv->pClipRects[_nc].y1 - dPriv->y; \
109 int maxx = dPriv->pClipRects[_nc].x2 - dPriv->x; \
110 int maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;
111
112 #define HW_ENDCLIPLOOP() \
113 } \
114 } while (0)
115
116 #define HW_UNLOCK()
117
118
119
120 /* ================================================================
121 * Color buffer
122 */
123
124 /* 16 bit, RGB565 color spanline and pixel functions
125 */
126 #define INIT_MONO_PIXEL(p, color) \
127 p = PACK_COLOR_565( color[0], color[1], color[2] )
128
129 #define WRITE_RGBA( _x, _y, r, g, b, a ) \
130 *(GLushort *)(buf + _x*2 + _y*pitch) = ((((int)r & 0xf8) << 8) | \
131 (((int)g & 0xfc) << 3) | \
132 (((int)b & 0xf8) >> 3))
133
134 #define WRITE_PIXEL( _x, _y, p ) \
135 *(GLushort *)(buf + _x*2 + _y*pitch) = p
136
137 #define READ_RGBA( rgba, _x, _y ) \
138 do { \
139 GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch); \
140 rgba[0] = ((p >> 8) & 0xf8) * 255 / 0xf8; \
141 rgba[1] = ((p >> 3) & 0xfc) * 255 / 0xfc; \
142 rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \
143 rgba[3] = 0xff; \
144 } while (0)
145
146 #define TAG(x) radeon##x##_RGB565
147 #include "spantmp.h"
148
149 /* 32 bit, ARGB8888 color spanline and pixel functions
150 */
151 #undef INIT_MONO_PIXEL
152 #define INIT_MONO_PIXEL(p, color) \
153 p = PACK_COLOR_8888( color[3], color[0], color[1], color[2] )
154
155 #define WRITE_RGBA( _x, _y, r, g, b, a ) \
156 do { \
157 *(GLuint *)(buf + _x*4 + _y*pitch) = ((b << 0) | \
158 (g << 8) | \
159 (r << 16) | \
160 (a << 24) ); \
161 } while (0)
162
163 #define WRITE_PIXEL( _x, _y, p ) \
164 do { \
165 *(GLuint *)(buf + _x*4 + _y*pitch) = p; \
166 } while (0)
167
168 #define READ_RGBA( rgba, _x, _y ) \
169 do { \
170 volatile GLuint *ptr = (volatile GLuint *)(read_buf + _x*4 + _y*pitch); \
171 GLuint p = *ptr; \
172 rgba[0] = (p >> 16) & 0xff; \
173 rgba[1] = (p >> 8) & 0xff; \
174 rgba[2] = (p >> 0) & 0xff; \
175 rgba[3] = (p >> 24) & 0xff; \
176 } while (0)
177
178 #define TAG(x) radeon##x##_ARGB8888
179 #include "spantmp.h"
180
181
182
183 /* ================================================================
184 * Depth buffer
185 */
186
187 /* The Radeon family has depth tiling on all the time, so we have to convert
188 * the x,y coordinates into the memory bus address (mba) in the same
189 * manner as the engine. In each case, the linear block address (ba)
190 * is calculated, and then wired with x and y to produce the final
191 * memory address.
192 */
193
194 static GLuint radeon_mba_z32( radeonContextPtr rmesa,
195 GLint x, GLint y )
196 {
197 GLuint pitch = rmesa->radeonScreen->frontPitch;
198 GLuint ba, address = 0; /* a[0..1] = 0 */
199
200 ba = (y / 16) * (pitch / 16) + (x / 16);
201
202 address |= (x & 0x7) << 2; /* a[2..4] = x[0..2] */
203 address |= (y & 0x3) << 5; /* a[5..6] = y[0..1] */
204 address |=
205 (((x & 0x10) >> 2) ^ (y & 0x4)) << 5; /* a[7] = x[4] ^ y[2] */
206 address |= (ba & 0x3) << 8; /* a[8..9] = ba[0..1] */
207
208 address |= (y & 0x8) << 7; /* a[10] = y[3] */
209 address |=
210 (((x & 0x8) << 1) ^ (y & 0x10)) << 7; /* a[11] = x[3] ^ y[4] */
211 address |= (ba & ~0x3) << 10; /* a[12..] = ba[2..] */
212
213 return address;
214 }
215
216 static __inline GLuint radeon_mba_z16( radeonContextPtr rmesa, GLint x, GLint y )
217 {
218 GLuint pitch = rmesa->radeonScreen->frontPitch;
219 GLuint ba, address = 0; /* a[0] = 0 */
220
221 ba = (y / 16) * (pitch / 32) + (x / 32);
222
223 address |= (x & 0x7) << 1; /* a[1..3] = x[0..2] */
224 address |= (y & 0x7) << 4; /* a[4..6] = y[0..2] */
225 address |= (x & 0x8) << 4; /* a[7] = x[3] */
226 address |= (ba & 0x3) << 8; /* a[8..9] = ba[0..1] */
227 address |= (y & 0x8) << 7; /* a[10] = y[3] */
228 address |= ((x & 0x10) ^ (y & 0x10)) << 7; /* a[11] = x[4] ^ y[4] */
229 address |= (ba & ~0x3) << 10; /* a[12..] = ba[2..] */
230
231 return address;
232 }
233
234
235 /* 16-bit depth buffer functions
236 */
237 #define WRITE_DEPTH( _x, _y, d ) \
238 *(GLushort *)(buf + radeon_mba_z16( rmesa, _x + xo, _y + yo )) = d;
239
240 #define READ_DEPTH( d, _x, _y ) \
241 d = *(GLushort *)(buf + radeon_mba_z16( rmesa, _x + xo, _y + yo ));
242
243 #define TAG(x) radeon##x##_16
244 #include "depthtmp.h"
245
246 /* 24 bit depth, 8 bit stencil depthbuffer functions
247 */
248 #define WRITE_DEPTH( _x, _y, d ) \
249 do { \
250 GLuint offset = radeon_mba_z32( rmesa, _x + xo, _y + yo ); \
251 GLuint tmp = *(GLuint *)(buf + offset); \
252 tmp &= 0xff000000; \
253 tmp |= ((d) & 0x00ffffff); \
254 *(GLuint *)(buf + offset) = tmp; \
255 } while (0)
256
257 #define READ_DEPTH( d, _x, _y ) \
258 d = *(GLuint *)(buf + radeon_mba_z32( rmesa, _x + xo, \
259 _y + yo )) & 0x00ffffff;
260
261 #define TAG(x) radeon##x##_24_8
262 #include "depthtmp.h"
263
264
265 /* ================================================================
266 * Stencil buffer
267 */
268
269 /* 24 bit depth, 8 bit stencil depthbuffer functions
270 */
271 #define WRITE_STENCIL( _x, _y, d ) \
272 do { \
273 GLuint offset = radeon_mba_z32( rmesa, _x + xo, _y + yo ); \
274 GLuint tmp = *(GLuint *)(buf + offset); \
275 tmp &= 0x00ffffff; \
276 tmp |= (((d) & 0xff) << 24); \
277 *(GLuint *)(buf + offset) = tmp; \
278 } while (0)
279
280 #define READ_STENCIL( d, _x, _y ) \
281 do { \
282 GLuint offset = radeon_mba_z32( rmesa, _x + xo, _y + yo ); \
283 GLuint tmp = *(GLuint *)(buf + offset); \
284 tmp &= 0xff000000; \
285 d = tmp >> 24; \
286 } while (0)
287
288 #define TAG(x) radeon##x##_24_8
289 #include "stenciltmp.h"
290
291
292 /*
293 * This function is called to specify which buffer to read and write
294 * for software rasterization (swrast) fallbacks. This doesn't necessarily
295 * correspond to glDrawBuffer() or glReadBuffer() calls.
296 */
297 static void radeonSetBuffer( GLcontext *ctx,
298 GLframebuffer *colorBuffer,
299 GLuint bufferBit )
300 {
301 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
302
303 switch ( bufferBit ) {
304 case FRONT_LEFT_BIT:
305 if ( rmesa->sarea->pfCurrentPage == 1 ) {
306 rmesa->state.pixel.readOffset = rmesa->radeonScreen->backOffset;
307 rmesa->state.pixel.readPitch = rmesa->radeonScreen->backPitch;
308 rmesa->state.color.drawOffset = rmesa->radeonScreen->backOffset;
309 rmesa->state.color.drawPitch = rmesa->radeonScreen->backPitch;
310 } else {
311 rmesa->state.pixel.readOffset = rmesa->radeonScreen->frontOffset;
312 rmesa->state.pixel.readPitch = rmesa->radeonScreen->frontPitch;
313 rmesa->state.color.drawOffset = rmesa->radeonScreen->frontOffset;
314 rmesa->state.color.drawPitch = rmesa->radeonScreen->frontPitch;
315 }
316 break;
317 case BACK_LEFT_BIT:
318 if ( rmesa->sarea->pfCurrentPage == 1 ) {
319 rmesa->state.pixel.readOffset = rmesa->radeonScreen->frontOffset;
320 rmesa->state.pixel.readPitch = rmesa->radeonScreen->frontPitch;
321 rmesa->state.color.drawOffset = rmesa->radeonScreen->frontOffset;
322 rmesa->state.color.drawPitch = rmesa->radeonScreen->frontPitch;
323 } else {
324 rmesa->state.pixel.readOffset = rmesa->radeonScreen->backOffset;
325 rmesa->state.pixel.readPitch = rmesa->radeonScreen->backPitch;
326 rmesa->state.color.drawOffset = rmesa->radeonScreen->backOffset;
327 rmesa->state.color.drawPitch = rmesa->radeonScreen->backPitch;
328 }
329 break;
330 default:
331 assert(0);
332 break;
333 }
334 }
335
336 /* Move locking out to get reasonable span performance (10x better
337 * than doing this in HW_LOCK above). WaitForIdle() is the main
338 * culprit.
339 */
340
341 static void radeonSpanRenderStart( GLcontext *ctx )
342 {
343 radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
344
345 RADEON_FIREVERTICES( rmesa );
346 LOCK_HARDWARE( rmesa );
347 radeonWaitForIdleLocked( rmesa );
348 }
349
350 static void radeonSpanRenderFinish( GLcontext *ctx )
351 {
352 radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
353 _swrast_flush( ctx );
354 UNLOCK_HARDWARE( rmesa );
355 }
356
357 void radeonInitSpanFuncs( GLcontext *ctx )
358 {
359 radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
360 struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
361
362 swdd->SetBuffer = radeonSetBuffer;
363
364 switch ( rmesa->radeonScreen->cpp ) {
365 case 2:
366 swdd->WriteRGBASpan = radeonWriteRGBASpan_RGB565;
367 swdd->WriteRGBSpan = radeonWriteRGBSpan_RGB565;
368 swdd->WriteMonoRGBASpan = radeonWriteMonoRGBASpan_RGB565;
369 swdd->WriteRGBAPixels = radeonWriteRGBAPixels_RGB565;
370 swdd->WriteMonoRGBAPixels = radeonWriteMonoRGBAPixels_RGB565;
371 swdd->ReadRGBASpan = radeonReadRGBASpan_RGB565;
372 swdd->ReadRGBAPixels = radeonReadRGBAPixels_RGB565;
373 break;
374
375 case 4:
376 swdd->WriteRGBASpan = radeonWriteRGBASpan_ARGB8888;
377 swdd->WriteRGBSpan = radeonWriteRGBSpan_ARGB8888;
378 swdd->WriteMonoRGBASpan = radeonWriteMonoRGBASpan_ARGB8888;
379 swdd->WriteRGBAPixels = radeonWriteRGBAPixels_ARGB8888;
380 swdd->WriteMonoRGBAPixels = radeonWriteMonoRGBAPixels_ARGB8888;
381 swdd->ReadRGBASpan = radeonReadRGBASpan_ARGB8888;
382 swdd->ReadRGBAPixels = radeonReadRGBAPixels_ARGB8888;
383 break;
384
385 default:
386 break;
387 }
388
389 switch ( rmesa->glCtx->Visual.depthBits ) {
390 case 16:
391 swdd->ReadDepthSpan = radeonReadDepthSpan_16;
392 swdd->WriteDepthSpan = radeonWriteDepthSpan_16;
393 swdd->ReadDepthPixels = radeonReadDepthPixels_16;
394 swdd->WriteDepthPixels = radeonWriteDepthPixels_16;
395 break;
396
397 case 24:
398 swdd->ReadDepthSpan = radeonReadDepthSpan_24_8;
399 swdd->WriteDepthSpan = radeonWriteDepthSpan_24_8;
400 swdd->ReadDepthPixels = radeonReadDepthPixels_24_8;
401 swdd->WriteDepthPixels = radeonWriteDepthPixels_24_8;
402
403 swdd->ReadStencilSpan = radeonReadStencilSpan_24_8;
404 swdd->WriteStencilSpan = radeonWriteStencilSpan_24_8;
405 swdd->ReadStencilPixels = radeonReadStencilPixels_24_8;
406 swdd->WriteStencilPixels = radeonWriteStencilPixels_24_8;
407 break;
408
409 default:
410 break;
411 }
412
413 swdd->SpanRenderStart = radeonSpanRenderStart;
414 swdd->SpanRenderFinish = radeonSpanRenderFinish;
415 }