2 Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved.
4 The Weather Channel (TM) funded Tungsten Graphics to develop the
5 initial release of the Radeon 8500 driver under the XFree86 license.
6 This notice must be preserved.
8 Permission is hereby granted, free of charge, to any person obtaining
9 a copy of this software and associated documentation files (the
10 "Software"), to deal in the Software without restriction, including
11 without limitation the rights to use, copy, modify, merge, publish,
12 distribute, sublicense, and/or sell copies of the Software, and to
13 permit persons to whom the Software is furnished to do so, subject to
14 the following conditions:
16 The above copyright notice and this permission notice (including the
17 next paragraph) shall be included in all copies or substantial
18 portions of the Software.
20 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
21 EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
22 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
23 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
24 LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
25 OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
26 WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
28 **************************************************************************/
32 * Keith Whitwell <keith@tungstengraphics.com>
37 #include "swrast/swrast.h"
40 #include "r200_context.h"
41 #include "radeon_ioctl.h"
42 #include "r300_ioctl.h"
43 #include "radeon_span.h"
48 radeonContextPtr radeon = RADEON_CONTEXT(ctx); \
49 radeonScreenPtr radeonScreen = radeon->radeonScreen; \
50 __DRIscreenPrivate *sPriv = radeon->dri.screen; \
51 __DRIdrawablePrivate *dPriv = radeon->dri.drawable; \
52 GLuint pitch = radeonScreen->frontPitch * radeonScreen->cpp; \
53 GLuint height = dPriv->h; \
54 char *buf = (char *)(sPriv->pFB + \
55 radeon->state.color.drawOffset + \
56 (dPriv->x * radeonScreen->cpp) + \
57 (dPriv->y * pitch)); \
58 char *read_buf = (char *)(sPriv->pFB + \
59 radeon->state.pixel.readOffset + \
60 (dPriv->x * radeonScreen->cpp) + \
61 (dPriv->y * pitch)); \
63 (void) read_buf; (void) buf; (void) p
65 #define LOCAL_DEPTH_VARS \
66 radeonContextPtr radeon = RADEON_CONTEXT(ctx); \
67 radeonScreenPtr radeonScreen = radeon->radeonScreen; \
68 __DRIscreenPrivate *sPriv = radeon->dri.screen; \
69 __DRIdrawablePrivate *dPriv = radeon->dri.drawable; \
70 GLuint height = dPriv->h; \
71 GLuint xo = dPriv->x; \
72 GLuint yo = dPriv->y; \
73 char *buf = (char *)(sPriv->pFB + radeonScreen->depthOffset); \
74 GLuint pitch = radeonScreen->depthPitch; \
75 (void) buf; (void) pitch
77 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
79 #define CLIPPIXEL( _x, _y ) \
80 ((_x >= minx) && (_x < maxx) && (_y >= miny) && (_y < maxy))
82 #define CLIPSPAN( _x, _y, _n, _x1, _n1, _i ) \
83 if ( _y < miny || _y >= maxy ) { \
88 if ( _x1 < minx ) _i += (minx-_x1), n1 -= (minx-_x1), _x1 = minx; \
89 if ( _x1 + _n1 >= maxx ) n1 -= (_x1 + n1 - maxx); \
92 #define Y_FLIP( _y ) (height - _y - 1)
96 #define HW_CLIPLOOP() \
98 __DRIdrawablePrivate *dPriv = radeon->dri.drawable; \
99 int _nc = dPriv->numClipRects; \
102 int minx = dPriv->pClipRects[_nc].x1 - dPriv->x; \
103 int miny = dPriv->pClipRects[_nc].y1 - dPriv->y; \
104 int maxx = dPriv->pClipRects[_nc].x2 - dPriv->x; \
105 int maxy = dPriv->pClipRects[_nc].y2 - dPriv->y;
107 #define HW_ENDCLIPLOOP() \
113 /* ================================================================
117 /* 16 bit, RGB565 color spanline and pixel functions
119 #define INIT_MONO_PIXEL(p, color) \
120 p = PACK_COLOR_565( color[0], color[1], color[2] )
122 #define WRITE_RGBA( _x, _y, r, g, b, a ) \
123 *(GLushort *)(buf + _x*2 + _y*pitch) = ((((int)r & 0xf8) << 8) | \
124 (((int)g & 0xfc) << 3) | \
125 (((int)b & 0xf8) >> 3))
127 #define WRITE_PIXEL( _x, _y, p ) \
128 *(GLushort *)(buf + _x*2 + _y*pitch) = p
130 #define READ_RGBA( rgba, _x, _y ) \
132 GLushort p = *(GLushort *)(read_buf + _x*2 + _y*pitch); \
133 rgba[0] = ((p >> 8) & 0xf8) * 255 / 0xf8; \
134 rgba[1] = ((p >> 3) & 0xfc) * 255 / 0xfc; \
135 rgba[2] = ((p << 3) & 0xf8) * 255 / 0xf8; \
139 #define TAG(x) radeon##x##_RGB565
142 /* 32 bit, ARGB8888 color spanline and pixel functions
144 #undef INIT_MONO_PIXEL
145 #define INIT_MONO_PIXEL(p, color) \
146 p = PACK_COLOR_8888( color[3], color[0], color[1], color[2] )
148 #define WRITE_RGBA( _x, _y, r, g, b, a ) \
150 *(GLuint *)(buf + _x*4 + _y*pitch) = ((b << 0) | \
156 #define WRITE_PIXEL( _x, _y, p ) \
158 *(GLuint *)(buf + _x*4 + _y*pitch) = p; \
161 #define READ_RGBA( rgba, _x, _y ) \
163 volatile GLuint *ptr = (volatile GLuint *)(read_buf + _x*4 + _y*pitch); \
165 rgba[0] = (p >> 16) & 0xff; \
166 rgba[1] = (p >> 8) & 0xff; \
167 rgba[2] = (p >> 0) & 0xff; \
168 rgba[3] = (p >> 24) & 0xff; \
171 #define TAG(x) radeon##x##_ARGB8888
174 /* ================================================================
178 /* The Radeon family has depth tiling on all the time, so we have to convert
179 * the x,y coordinates into the memory bus address (mba) in the same
180 * manner as the engine. In each case, the linear block address (ba)
181 * is calculated, and then wired with x and y to produce the final
185 #define BIT(x,b) ((x & (1<<b))>>b)
186 static GLuint
radeon_mba_z32(radeonContextPtr radeon
, GLint x
, GLint y
)
188 GLuint pitch
= radeon
->radeonScreen
->depthPitch
;
190 ((y
& 0x3FF) >> 4) * ((pitch
& 0xFFF) >> 5) + ((x
& 0x3FF) >> 5);
192 (BIT(x
, 0) << 2) | (BIT(y
, 0) << 3) | (BIT(x
, 1) << 4) | (BIT(y
, 1)
194 (BIT(x
, 3) << 6) | (BIT(x
, 4) << 7) | (BIT(x
, 2) << 8) | (BIT(y
, 2)
197 (((pitch
& 0x20) ? (b
& 0x01) : ((b
& 0x01) ^ (BIT(y
, 4)))) << 11) |
202 static GLuint
radeon_mba_z16(radeonContextPtr radeon
, GLint x
, GLint y
)
204 GLuint pitch
= radeon
->radeonScreen
->depthPitch
;
206 ((y
& 0x3FF) >> 4) * ((pitch
& 0xFFF) >> 6) + ((x
& 0x3FF) >> 6);
208 (BIT(x
, 0) << 1) | (BIT(y
, 0) << 2) | (BIT(x
, 1) << 3) | (BIT(y
, 1)
210 (BIT(x
, 2) << 5) | (BIT(x
, 4) << 6) | (BIT(x
, 5) << 7) | (BIT(x
, 3)
212 (BIT(y
, 2) << 9) | (BIT(y
, 3) << 10) |
213 (((pitch
& 0x40) ? (b
& 0x01) : ((b
& 0x01) ^ (BIT(y
, 4)))) << 11) |
219 /* 16-bit depth buffer functions
221 #define WRITE_DEPTH( _x, _y, d ) \
222 *(GLushort *)(buf + radeon_mba_z16( radeon, _x + xo, _y + yo )) = d;
224 #define READ_DEPTH( d, _x, _y ) \
225 d = *(GLushort *)(buf + radeon_mba_z16( radeon, _x + xo, _y + yo ));
227 #define TAG(x) radeon##x##_16_TILE
228 #include "depthtmp.h"
230 /* 24 bit depth, 8 bit stencil depthbuffer functions
232 #define WRITE_DEPTH( _x, _y, d ) \
234 GLuint offset = radeon_mba_z32( radeon, _x + xo, _y + yo ); \
235 GLuint tmp = *(GLuint *)(buf + offset); \
237 tmp |= ((d) & 0x00ffffff); \
238 *(GLuint *)(buf + offset) = tmp; \
241 #define READ_DEPTH( d, _x, _y ) \
242 d = *(GLuint *)(buf + radeon_mba_z32( radeon, _x + xo, \
243 _y + yo )) & 0x00ffffff;
245 #define TAG(x) radeon##x##_24_8_TILE
246 #include "depthtmp.h"
248 /* 16-bit depth buffer functions
250 #define WRITE_DEPTH( _x, _y, d ) \
251 *(GLushort *)(buf + (_x + xo + (_y + yo)*pitch)*2 ) = d;
253 #define READ_DEPTH( d, _x, _y ) \
254 d = *(GLushort *)(buf + (_x + xo + (_y + yo)*pitch)*2 );
256 #define TAG(x) radeon##x##_16_LINEAR
257 #include "depthtmp.h"
259 /* 24 bit depth, 8 bit stencil depthbuffer functions
261 * Careful: It looks like the R300 uses ZZZS byte order while the R200
262 * uses SZZZ for 24 bit depth, 8 bit stencil mode.
264 #define WRITE_DEPTH( _x, _y, d ) \
266 GLuint offset = (_x + xo + (_y + yo)*pitch)*4; \
267 GLuint tmp = *(GLuint *)(buf + offset); \
269 tmp |= ((d << 8) & 0xffffff00); \
270 *(GLuint *)(buf + offset) = tmp; \
273 #define READ_DEPTH( d, _x, _y ) \
274 d = (*(GLuint *)(buf + (_x + xo + (_y + yo)*pitch)*4) & 0xffffff00) >> 8;
276 #define TAG(x) radeon##x##_24_8_LINEAR
277 #include "depthtmp.h"
279 /* ================================================================
283 /* 24 bit depth, 8 bit stencil depthbuffer functions
285 #define WRITE_STENCIL( _x, _y, d ) \
287 GLuint offset = radeon_mba_z32( radeon, _x + xo, _y + yo ); \
288 GLuint tmp = *(GLuint *)(buf + offset); \
290 tmp |= (((d) & 0xff) << 24); \
291 *(GLuint *)(buf + offset) = tmp; \
294 #define READ_STENCIL( d, _x, _y ) \
296 GLuint offset = radeon_mba_z32( radeon, _x + xo, _y + yo ); \
297 GLuint tmp = *(GLuint *)(buf + offset); \
302 #define TAG(x) radeon##x##_24_8_TILE
303 #include "stenciltmp.h"
305 /* 24 bit depth, 8 bit stencil depthbuffer functions
307 #define WRITE_STENCIL( _x, _y, d ) \
309 GLuint offset = (_x + xo)*4 + (_y + yo)*pitch; \
310 GLuint tmp = *(GLuint *)(buf + offset); \
312 tmp |= (((d) & 0xff) << 24); \
313 *(GLuint *)(buf + offset) = tmp; \
316 #define READ_STENCIL( d, _x, _y ) \
318 GLuint offset = (_x + xo)*4 + (_y + yo)*pitch; \
319 GLuint tmp = *(GLuint *)(buf + offset); \
324 #define TAG(x) radeon##x##_24_8_LINEAR
325 #include "stenciltmp.h"
328 * This function is called to specify which buffer to read and write
329 * for software rasterization (swrast) fallbacks. This doesn't necessarily
330 * correspond to glDrawBuffer() or glReadBuffer() calls.
332 static void radeonSetBuffer(GLcontext
* ctx
,
333 GLframebuffer
* colorBuffer
, GLuint bufferBit
)
335 radeonContextPtr radeon
= RADEON_CONTEXT(ctx
);
339 case BUFFER_BIT_FRONT_LEFT
:
343 case BUFFER_BIT_BACK_LEFT
:
348 _mesa_problem(ctx
, "Bad bufferBit in %s", __FUNCTION__
);
352 if (radeon
->doPageFlip
&& radeon
->sarea
->pfCurrentPage
== 1)
356 fprintf(stderr
, "%s: using %s buffer\n", __FUNCTION__
,
357 buffer
? "back" : "front");
361 radeon
->state
.pixel
.readOffset
=
362 radeon
->radeonScreen
->backOffset
;
363 radeon
->state
.pixel
.readPitch
=
364 radeon
->radeonScreen
->backPitch
;
365 radeon
->state
.color
.drawOffset
=
366 radeon
->radeonScreen
->backOffset
;
367 radeon
->state
.color
.drawPitch
=
368 radeon
->radeonScreen
->backPitch
;
370 radeon
->state
.pixel
.readOffset
=
371 radeon
->radeonScreen
->frontOffset
;
372 radeon
->state
.pixel
.readPitch
=
373 radeon
->radeonScreen
->frontPitch
;
374 radeon
->state
.color
.drawOffset
=
375 radeon
->radeonScreen
->frontOffset
;
376 radeon
->state
.color
.drawPitch
=
377 radeon
->radeonScreen
->frontPitch
;
381 /* Move locking out to get reasonable span performance (10x better
382 * than doing this in HW_LOCK above). WaitForIdle() is the main
386 static void radeonSpanRenderStart(GLcontext
* ctx
)
388 radeonContextPtr radeon
= RADEON_CONTEXT(ctx
);
390 if (IS_FAMILY_R200(radeon
))
391 R200_FIREVERTICES((r200ContextPtr
)radeon
);
395 LOCK_HARDWARE(radeon
);
396 radeonWaitForIdleLocked(radeon
);
398 /* Read & rewrite the first pixel in the frame buffer. This should
399 * be a noop, right? In fact without this conform fails as reading
400 * from the framebuffer sometimes produces old results -- the
401 * on-card read cache gets mixed up and doesn't notice that the
402 * framebuffer has been updated.
404 * In the worst case this is buggy too as p might get the wrong
405 * value first time, so really need a hidden pixel somewhere for this.
409 volatile int *read_buf
=
410 (volatile int *)(radeon
->dri
.screen
->pFB
+
411 radeon
->state
.pixel
.readOffset
);
417 static void radeonSpanRenderFinish(GLcontext
* ctx
)
419 radeonContextPtr radeon
= RADEON_CONTEXT(ctx
);
422 UNLOCK_HARDWARE(radeon
);
425 void radeonInitSpanFuncs(GLcontext
* ctx
)
427 radeonContextPtr radeon
= RADEON_CONTEXT(ctx
);
428 struct swrast_device_driver
*swdd
=
429 _swrast_GetDeviceDriverReference(ctx
);
431 swdd
->SetBuffer
= radeonSetBuffer
;
433 switch (radeon
->radeonScreen
->cpp
) {
436 swdd
->WriteRGBASpan
= radeonWriteRGBASpan_RGB565
;
437 swdd
->WriteRGBSpan
= radeonWriteRGBSpan_RGB565
;
438 swdd
->WriteMonoRGBASpan
= radeonWriteMonoRGBASpan_RGB565
;
439 swdd
->WriteRGBAPixels
= radeonWriteRGBAPixels_RGB565
;
440 swdd
->WriteMonoRGBAPixels
= radeonWriteMonoRGBAPixels_RGB565
;
441 swdd
->ReadRGBASpan
= radeonReadRGBASpan_RGB565
;
442 swdd
->ReadRGBAPixels
= radeonReadRGBAPixels_RGB565
;
448 swdd
->WriteRGBASpan
= radeonWriteRGBASpan_ARGB8888
;
449 swdd
->WriteRGBSpan
= radeonWriteRGBSpan_ARGB8888
;
450 swdd
->WriteMonoRGBASpan
= radeonWriteMonoRGBASpan_ARGB8888
;
451 swdd
->WriteRGBAPixels
= radeonWriteRGBAPixels_ARGB8888
;
452 swdd
->WriteMonoRGBAPixels
= radeonWriteMonoRGBAPixels_ARGB8888
;
453 swdd
->ReadRGBASpan
= radeonReadRGBASpan_ARGB8888
;
454 swdd
->ReadRGBAPixels
= radeonReadRGBAPixels_ARGB8888
;
462 if (IS_FAMILY_R300(radeon
))
464 switch (radeon
->glCtx
->Visual
.depthBits
) {
467 swdd
->ReadDepthSpan
= radeonReadDepthSpan_16_LINEAR
;
468 swdd
->WriteDepthSpan
= radeonWriteDepthSpan_16_LINEAR
;
469 swdd
->WriteMonoDepthSpan
= radeonWriteMonoDepthSpan_16_LINEAR
;
470 swdd
->ReadDepthPixels
= radeonReadDepthPixels_16_LINEAR
;
471 swdd
->WriteDepthPixels
= radeonWriteDepthPixels_16_LINEAR
;
477 swdd
->ReadDepthSpan
= radeonReadDepthSpan_24_8_LINEAR
;
478 swdd
->WriteDepthSpan
= radeonWriteDepthSpan_24_8_LINEAR
;
479 swdd
->WriteMonoDepthSpan
= radeonWriteMonoDepthSpan_24_8_LINEAR
;
480 swdd
->ReadDepthPixels
= radeonReadDepthPixels_24_8_LINEAR
;
481 swdd
->WriteDepthPixels
= radeonWriteDepthPixels_24_8_LINEAR
;
483 swdd
->ReadStencilSpan
= radeonReadStencilSpan_24_8_LINEAR
;
484 swdd
->WriteStencilSpan
= radeonWriteStencilSpan_24_8_LINEAR
;
485 swdd
->ReadStencilPixels
= radeonReadStencilPixels_24_8_LINEAR
;
486 swdd
->WriteStencilPixels
= radeonWriteStencilPixels_24_8_LINEAR
;
496 switch (radeon
->glCtx
->Visual
.depthBits
) {
499 swdd
->ReadDepthSpan
= radeonReadDepthSpan_16_TILE
;
500 swdd
->WriteDepthSpan
= radeonWriteDepthSpan_16_TILE
;
501 swdd
->WriteMonoDepthSpan
= radeonWriteMonoDepthSpan_16_TILE
;
502 swdd
->ReadDepthPixels
= radeonReadDepthPixels_16_TILE
;
503 swdd
->WriteDepthPixels
= radeonWriteDepthPixels_16_TILE
;
509 swdd
->ReadDepthSpan
= radeonReadDepthSpan_24_8_TILE
;
510 swdd
->WriteDepthSpan
= radeonWriteDepthSpan_24_8_TILE
;
511 swdd
->WriteMonoDepthSpan
= radeonWriteMonoDepthSpan_24_8_TILE
;
512 swdd
->ReadDepthPixels
= radeonReadDepthPixels_24_8_TILE
;
513 swdd
->WriteDepthPixels
= radeonWriteDepthPixels_24_8_TILE
;
515 swdd
->ReadStencilSpan
= radeonReadStencilSpan_24_8_TILE
;
516 swdd
->WriteStencilSpan
= radeonWriteStencilSpan_24_8_TILE
;
517 swdd
->ReadStencilPixels
= radeonReadStencilPixels_24_8_TILE
;
518 swdd
->WriteStencilPixels
= radeonWriteStencilPixels_24_8_TILE
;
527 swdd
->SpanRenderStart
= radeonSpanRenderStart
;
528 swdd
->SpanRenderFinish
= radeonSpanRenderFinish
;
532 * Plug in the Get/Put routines for the given driRenderbuffer.
534 void radeonSetSpanFunctions(driRenderbuffer
*drb
, const GLvisual
*vis
)
536 if (drb
->Base
.InternalFormat
== GL_RGBA
) {
537 if (vis
->redBits
== 5 && vis
->greenBits
== 6 && vis
->blueBits
== 5) {
538 drb
->Base
.GetRow
= radeonReadRGBASpan_RGB565
;
539 drb
->Base
.GetValues
= radeonReadRGBAPixels_RGB565
;
540 drb
->Base
.PutRow
= radeonWriteRGBASpan_RGB565
;
541 drb
->Base
.PutRowRGB
= radeonWriteRGBSpan_RGB565
;
542 drb
->Base
.PutMonoRow
= radeonWriteMonoRGBASpan_RGB565
;
543 drb
->Base
.PutValues
= radeonWriteRGBAPixels_RGB565
;
544 drb
->Base
.PutMonoValues
= radeonWriteMonoRGBAPixels_RGB565
;
547 drb
->Base
.GetRow
= radeonReadRGBASpan_ARGB8888
;
548 drb
->Base
.GetValues
= radeonReadRGBAPixels_ARGB8888
;
549 drb
->Base
.PutRow
= radeonWriteRGBASpan_ARGB8888
;
550 drb
->Base
.PutRowRGB
= radeonWriteRGBSpan_ARGB8888
;
551 drb
->Base
.PutMonoRow
= radeonWriteMonoRGBASpan_ARGB8888
;
552 drb
->Base
.PutValues
= radeonWriteRGBAPixels_ARGB8888
;
553 drb
->Base
.PutMonoValues
= radeonWriteMonoRGBAPixels_ARGB8888
;
556 else if (drb
->Base
.InternalFormat
== GL_DEPTH_COMPONENT16
) {
557 drb
->Base
.GetRow
= radeonReadDepthSpan_16_LINEAR
;
558 drb
->Base
.GetValues
= radeonReadDepthPixels_16_LINEAR
;
559 drb
->Base
.PutRow
= radeonWriteDepthSpan_16_LINEAR
;
560 drb
->Base
.PutMonoRow
= radeonWriteMonoDepthSpan_16_LINEAR
;
561 drb
->Base
.PutValues
= radeonWriteDepthPixels_16_LINEAR
;
562 drb
->Base
.PutMonoValues
= NULL
;
564 else if (drb
->Base
.InternalFormat
== GL_DEPTH_COMPONENT24
) {
565 drb
->Base
.GetRow
= radeonReadDepthSpan_24_8_LINEAR
;
566 drb
->Base
.GetValues
= radeonReadDepthPixels_24_8_LINEAR
;
567 drb
->Base
.PutRow
= radeonWriteDepthSpan_24_8_LINEAR
;
568 drb
->Base
.PutMonoRow
= radeonWriteMonoDepthSpan_24_8_LINEAR
;
569 drb
->Base
.PutValues
= radeonWriteDepthPixels_24_8_LINEAR
;
570 drb
->Base
.PutMonoValues
= NULL
;
572 else if (drb
->Base
.InternalFormat
== GL_STENCIL_INDEX8_EXT
) {
573 drb
->Base
.GetRow
= radeonReadStencilSpan_24_8_LINEAR
;
574 drb
->Base
.GetValues
= radeonReadStencilPixels_24_8_LINEAR
;
575 drb
->Base
.PutRow
= radeonWriteStencilSpan_24_8_LINEAR
;
576 drb
->Base
.PutMonoRow
= radeonWriteMonoStencilSpan_24_8_LINEAR
;
577 drb
->Base
.PutValues
= radeonWriteStencilPixels_24_8_LINEAR
;
578 drb
->Base
.PutMonoValues
= NULL
;