r200: simplify / unify input map handling for vp and fftnl
[mesa.git] / src / mesa / drivers / dri / r200 / r200_maos_arrays.c
1 /* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_maos_arrays.c,v 1.3 2003/02/23 23:59:01 dawes Exp $ */
2 /*
3 Copyright (C) The Weather Channel, Inc. 2002. All Rights Reserved.
4
5 The Weather Channel (TM) funded Tungsten Graphics to develop the
6 initial release of the Radeon 8500 driver under the XFree86 license.
7 This notice must be preserved.
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 * Keith Whitwell <keith@tungstengraphics.com>
34 */
35
36 #include "glheader.h"
37 #include "mtypes.h"
38 #include "colormac.h"
39 #include "imports.h"
40 #include "macros.h"
41
42 #include "swrast_setup/swrast_setup.h"
43 #include "math/m_translate.h"
44 #include "tnl/tnl.h"
45 #include "tnl/t_context.h"
46
47 #include "r200_context.h"
48 #include "r200_ioctl.h"
49 #include "r200_state.h"
50 #include "r200_swtcl.h"
51 #include "r200_maos.h"
52 #include "r200_tcl.h"
53
54
55 #if 0
56 /* Usage:
57 * - from r200_tcl_render
58 * - call r200EmitArrays to ensure uptodate arrays in dma
59 * - emit primitives (new type?) which reference the data
60 * -- need to use elts for lineloop, quads, quadstrip/flat
61 * -- other primitives are all well-formed (need tristrip-1,fake-poly)
62 *
63 */
64 static void emit_ubyte_rgba3( GLcontext *ctx,
65 struct r200_dma_region *rvb,
66 char *data,
67 int stride,
68 int count )
69 {
70 int i;
71 r200_color_t *out = (r200_color_t *)(rvb->start + rvb->address);
72
73 if (R200_DEBUG & DEBUG_VERTS)
74 fprintf(stderr, "%s count %d stride %d out %p\n",
75 __FUNCTION__, count, stride, (void *)out);
76
77 for (i = 0; i < count; i++) {
78 out->red = *data;
79 out->green = *(data+1);
80 out->blue = *(data+2);
81 out->alpha = 0xFF;
82 out++;
83 data += stride;
84 }
85 }
86
87 static void emit_ubyte_rgba4( GLcontext *ctx,
88 struct r200_dma_region *rvb,
89 char *data,
90 int stride,
91 int count )
92 {
93 int i;
94 int *out = (int *)(rvb->address + rvb->start);
95
96 if (R200_DEBUG & DEBUG_VERTS)
97 fprintf(stderr, "%s count %d stride %d\n",
98 __FUNCTION__, count, stride);
99
100 if (stride == 4) {
101 for (i = 0; i < count; i++)
102 ((int *)out)[i] = LE32_TO_CPU(((int *)data)[i]);
103 } else {
104 for (i = 0; i < count; i++) {
105 *(int *)out++ = LE32_TO_CPU(*(int *)data);
106 data += stride;
107 }
108 }
109 }
110
111
112 static void emit_ubyte_rgba( GLcontext *ctx,
113 struct r200_dma_region *rvb,
114 char *data,
115 int size,
116 int stride,
117 int count )
118 {
119 r200ContextPtr rmesa = R200_CONTEXT(ctx);
120
121 if (R200_DEBUG & DEBUG_VERTS)
122 fprintf(stderr, "%s %d/%d\n", __FUNCTION__, count, size);
123
124 assert (!rvb->buf);
125
126 if (stride == 0) {
127 r200AllocDmaRegion( rmesa, rvb, 4, 4 );
128 count = 1;
129 rvb->aos_start = GET_START(rvb);
130 rvb->aos_stride = 0;
131 rvb->aos_size = 1;
132 }
133 else {
134 r200AllocDmaRegion( rmesa, rvb, 4 * count, 4 ); /* alignment? */
135 rvb->aos_start = GET_START(rvb);
136 rvb->aos_stride = 1;
137 rvb->aos_size = 1;
138 }
139
140 /* Emit the data
141 */
142 switch (size) {
143 case 3:
144 emit_ubyte_rgba3( ctx, rvb, data, stride, count );
145 break;
146 case 4:
147 emit_ubyte_rgba4( ctx, rvb, data, stride, count );
148 break;
149 default:
150 assert(0);
151 exit(1);
152 break;
153 }
154 }
155 #endif
156
157
158 #if defined(USE_X86_ASM)
159 #define COPY_DWORDS( dst, src, nr ) \
160 do { \
161 int __tmp; \
162 __asm__ __volatile__( "rep ; movsl" \
163 : "=%c" (__tmp), "=D" (dst), "=S" (__tmp) \
164 : "0" (nr), \
165 "D" ((long)dst), \
166 "S" ((long)src) ); \
167 } while (0)
168 #else
169 #define COPY_DWORDS( dst, src, nr ) \
170 do { \
171 int j; \
172 for ( j = 0 ; j < nr ; j++ ) \
173 dst[j] = ((int *)src)[j]; \
174 dst += nr; \
175 } while (0)
176 #endif
177
178
179 static void emit_vecfog( GLcontext *ctx,
180 struct r200_dma_region *rvb,
181 char *data,
182 int stride,
183 int count )
184 {
185 int i;
186 GLfloat *out;
187
188 r200ContextPtr rmesa = R200_CONTEXT(ctx);
189
190 if (R200_DEBUG & DEBUG_VERTS)
191 fprintf(stderr, "%s count %d stride %d\n",
192 __FUNCTION__, count, stride);
193
194 assert (!rvb->buf);
195
196 if (stride == 0) {
197 r200AllocDmaRegion( rmesa, rvb, 4, 4 );
198 count = 1;
199 rvb->aos_start = GET_START(rvb);
200 rvb->aos_stride = 0;
201 rvb->aos_size = 1;
202 }
203 else {
204 r200AllocDmaRegion( rmesa, rvb, count * 4, 4 ); /* alignment? */
205 rvb->aos_start = GET_START(rvb);
206 rvb->aos_stride = 1;
207 rvb->aos_size = 1;
208 }
209
210 /* Emit the data
211 */
212 out = (GLfloat *)(rvb->address + rvb->start);
213 for (i = 0; i < count; i++) {
214 out[0] = r200ComputeFogBlendFactor( ctx, *(GLfloat *)data );
215 out++;
216 data += stride;
217 }
218
219 }
220
221
222 static void emit_vec4( GLcontext *ctx,
223 struct r200_dma_region *rvb,
224 char *data,
225 int stride,
226 int count )
227 {
228 int i;
229 int *out = (int *)(rvb->address + rvb->start);
230
231 if (R200_DEBUG & DEBUG_VERTS)
232 fprintf(stderr, "%s count %d stride %d\n",
233 __FUNCTION__, count, stride);
234
235 if (stride == 4)
236 COPY_DWORDS( out, data, count );
237 else
238 for (i = 0; i < count; i++) {
239 out[0] = *(int *)data;
240 out++;
241 data += stride;
242 }
243 }
244
245
246 static void emit_vec8( GLcontext *ctx,
247 struct r200_dma_region *rvb,
248 char *data,
249 int stride,
250 int count )
251 {
252 int i;
253 int *out = (int *)(rvb->address + rvb->start);
254
255 if (R200_DEBUG & DEBUG_VERTS)
256 fprintf(stderr, "%s count %d stride %d\n",
257 __FUNCTION__, count, stride);
258
259 if (stride == 8)
260 COPY_DWORDS( out, data, count*2 );
261 else
262 for (i = 0; i < count; i++) {
263 out[0] = *(int *)data;
264 out[1] = *(int *)(data+4);
265 out += 2;
266 data += stride;
267 }
268 }
269
270 static void emit_vec12( GLcontext *ctx,
271 struct r200_dma_region *rvb,
272 char *data,
273 int stride,
274 int count )
275 {
276 int i;
277 int *out = (int *)(rvb->address + rvb->start);
278
279 if (R200_DEBUG & DEBUG_VERTS)
280 fprintf(stderr, "%s count %d stride %d out %p data %p\n",
281 __FUNCTION__, count, stride, (void *)out, (void *)data);
282
283 if (stride == 12)
284 COPY_DWORDS( out, data, count*3 );
285 else
286 for (i = 0; i < count; i++) {
287 out[0] = *(int *)data;
288 out[1] = *(int *)(data+4);
289 out[2] = *(int *)(data+8);
290 out += 3;
291 data += stride;
292 }
293 }
294
295 static void emit_vec16( GLcontext *ctx,
296 struct r200_dma_region *rvb,
297 char *data,
298 int stride,
299 int count )
300 {
301 int i;
302 int *out = (int *)(rvb->address + rvb->start);
303
304 if (R200_DEBUG & DEBUG_VERTS)
305 fprintf(stderr, "%s count %d stride %d\n",
306 __FUNCTION__, count, stride);
307
308 if (stride == 16)
309 COPY_DWORDS( out, data, count*4 );
310 else
311 for (i = 0; i < count; i++) {
312 out[0] = *(int *)data;
313 out[1] = *(int *)(data+4);
314 out[2] = *(int *)(data+8);
315 out[3] = *(int *)(data+12);
316 out += 4;
317 data += stride;
318 }
319 }
320
321
322 static void emit_vector( GLcontext *ctx,
323 struct r200_dma_region *rvb,
324 char *data,
325 int size,
326 int stride,
327 int count )
328 {
329 r200ContextPtr rmesa = R200_CONTEXT(ctx);
330
331 if (R200_DEBUG & DEBUG_VERTS)
332 fprintf(stderr, "%s count %d size %d stride %d\n",
333 __FUNCTION__, count, size, stride);
334
335 assert (!rvb->buf);
336
337 if (stride == 0) {
338 r200AllocDmaRegion( rmesa, rvb, size * 4, 4 );
339 count = 1;
340 rvb->aos_start = GET_START(rvb);
341 rvb->aos_stride = 0;
342 rvb->aos_size = size;
343 }
344 else {
345 r200AllocDmaRegion( rmesa, rvb, size * count * 4, 4 ); /* alignment? */
346 rvb->aos_start = GET_START(rvb);
347 rvb->aos_stride = size;
348 rvb->aos_size = size;
349 }
350
351 /* Emit the data
352 */
353 switch (size) {
354 case 1:
355 emit_vec4( ctx, rvb, data, stride, count );
356 break;
357 case 2:
358 emit_vec8( ctx, rvb, data, stride, count );
359 break;
360 case 3:
361 emit_vec12( ctx, rvb, data, stride, count );
362 break;
363 case 4:
364 emit_vec16( ctx, rvb, data, stride, count );
365 break;
366 default:
367 assert(0);
368 exit(1);
369 break;
370 }
371
372 }
373
374
375
376 /* Emit any changed arrays to new GART memory, re-emit a packet to
377 * update the arrays.
378 */
379 void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev )
380 {
381 r200ContextPtr rmesa = R200_CONTEXT( ctx );
382 struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
383 struct r200_dma_region **component = rmesa->tcl.aos_components;
384 GLuint nr = 0;
385 GLuint vfmt0 = 0, vfmt1 = 0;
386 GLuint count = VB->Count;
387 GLuint i, emitsize;
388
389 for ( i = 0; i < 15; i++ ) {
390 GLubyte attrib = vimap_rev[i];
391 if (attrib != 255) {
392 switch (i) {
393 case 0:
394 emitsize = (VB->AttribPtr[attrib]->size);
395 switch (emitsize) {
396 case 4:
397 vfmt0 |= R200_VTX_W0;
398 /* fallthrough */
399 case 3:
400 vfmt0 |= R200_VTX_Z0;
401 break;
402 case 2:
403 break;
404 default: assert(0);
405 }
406 break;
407 case 1:
408 assert(attrib == VERT_ATTRIB_WEIGHT);
409 emitsize = (VB->AttribPtr[attrib]->size);
410 vfmt0 |= emitsize << R200_VTX_WEIGHT_COUNT_SHIFT;
411 break;
412 case 2:
413 assert(attrib == VERT_ATTRIB_NORMAL);
414 emitsize = 3;
415 vfmt0 |= R200_VTX_N0;
416 break;
417 case 3:
418 /* special handling to fix up fog. Will get us into trouble with vbos...*/
419 assert(attrib == VERT_ATTRIB_FOG);
420 if (!rmesa->tcl.vertex_data[i].buf) {
421 if (ctx->VertexProgram._Enabled)
422 emit_vector( ctx,
423 &(rmesa->tcl.vertex_data[attrib]),
424 (char *)VB->AttribPtr[attrib]->data,
425 1,
426 VB->AttribPtr[attrib]->stride,
427 count);
428 else
429 emit_vecfog( ctx,
430 &(rmesa->tcl.vertex_data[attrib]),
431 (char *)VB->AttribPtr[attrib]->data,
432 VB->AttribPtr[attrib]->stride,
433 count);
434 }
435 vfmt0 |= R200_VTX_DISCRETE_FOG;
436 goto after_emit;
437 break;
438 case 4:
439 case 5:
440 case 6:
441 case 7:
442 if (VB->AttribPtr[attrib]->size == 4 &&
443 (VB->AttribPtr[attrib]->stride != 0 ||
444 VB->AttribPtr[attrib]->data[0][3] != 1.0)) emitsize = 4;
445 else emitsize = 3;
446 if (emitsize == 4)
447 vfmt0 |= R200_VTX_FP_RGBA << (R200_VTX_COLOR_0_SHIFT + (i - 4) * 2);
448 else {
449 vfmt0 |= R200_VTX_FP_RGB << (R200_VTX_COLOR_0_SHIFT + (i - 4) * 2);
450 }
451 break;
452 case 8:
453 case 9:
454 case 10:
455 case 11:
456 case 12:
457 case 13:
458 emitsize = VB->AttribPtr[attrib]->size;
459 vfmt1 |= emitsize << (R200_VTX_TEX0_COMP_CNT_SHIFT + (i - 8) * 3);
460 break;
461 case 14:
462 emitsize = VB->AttribPtr[attrib]->size >= 2 ? VB->AttribPtr[attrib]->size : 2;
463 switch (emitsize) {
464 case 2:
465 vfmt0 |= R200_VTX_XY1;
466 /* fallthrough */
467 case 3:
468 vfmt0 |= R200_VTX_Z1;
469 /* fallthrough */
470 case 4:
471 vfmt0 |= R200_VTX_W1;
472 break;
473 }
474 default:
475 assert(0);
476 }
477 if (!rmesa->tcl.vertex_data[i].buf) {
478 emit_vector( ctx,
479 &(rmesa->tcl.vertex_data[i]),
480 (char *)VB->AttribPtr[attrib]->data,
481 emitsize,
482 VB->AttribPtr[attrib]->stride,
483 count );
484 }
485 after_emit:
486 assert(nr < 12);
487 component[nr++] = &rmesa->tcl.vertex_data[i];
488 }
489 }
490
491 if (vfmt0 != rmesa->hw.vtx.cmd[VTX_VTXFMT_0] ||
492 vfmt1 != rmesa->hw.vtx.cmd[VTX_VTXFMT_1]) {
493 R200_STATECHANGE( rmesa, vtx );
494 rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = vfmt0;
495 rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = vfmt1;
496 }
497
498 rmesa->tcl.nr_aos_components = nr;
499 }
500
501
502 void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs )
503 {
504 r200ContextPtr rmesa = R200_CONTEXT( ctx );
505
506 /* only do it for changed inputs ? */
507 int i;
508 for (i = 0; i < 15; i++) {
509 if (newinputs & (1 << i))
510 r200ReleaseDmaRegion( rmesa,
511 &rmesa->tcl.vertex_data[i], __FUNCTION__ );
512 }
513 }