Merge branch 'new-frag-attribs'
[mesa.git] / src / mesa / drivers / dri / r300 / r300_draw.c
1 /**************************************************************************
2 *
3 * Copyright 2009 Maciej Cencora
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the
7 * "Software"), to deal in the Software without restriction, including
8 * without limitation the rights to use, copy, modify, merge, publish,
9 * distribute, sub license, and/or sell copies of the Software, and to
10 * permit persons to whom the Software is furnished to do so, subject to
11 * the following conditions:
12 *
13 * The above copyright notice and this permission notice (including the
14 * next paragraph) shall be included in all copies or substantial portions
15 * of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
19 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
20 * IN NO EVENT SHALL THE AUTHOR(S) AND/OR ITS SUPPLIERS BE LIABLE FOR
21 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
22 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
23 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
24 *
25 **************************************************************************/
26
27 #include <stdlib.h>
28
29 #include "main/glheader.h"
30 #include "main/context.h"
31 #include "main/state.h"
32 #include "main/api_validate.h"
33 #include "main/enums.h"
34
35 #include "r300_reg.h"
36 #include "r300_context.h"
37 #include "r300_emit.h"
38 #include "r300_render.h"
39 #include "r300_state.h"
40 #include "r300_tex.h"
41
42 #include "tnl/tnl.h"
43 #include "tnl/t_vp_build.h"
44 #include "vbo/vbo_context.h"
45 #include "swrast/swrast.h"
46 #include "swrast_setup/swrast_setup.h"
47
48 static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf, struct gl_buffer_object **bo, GLuint *nr_bo)
49 {
50 r300ContextPtr r300 = R300_CONTEXT(ctx);
51 struct r300_index_buffer *ind_buf = &r300->ind_buf;
52 GLvoid *src_ptr;
53
54 if (!mesa_ind_buf) {
55 ind_buf->ptr = NULL;
56 return;
57 }
58
59 ind_buf->count = mesa_ind_buf->count;
60 if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) {
61 bo[*nr_bo] = mesa_ind_buf->obj;
62 (*nr_bo)++;
63 ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj);
64 assert(mesa_ind_buf->obj->Pointer != NULL);
65 }
66 src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr);
67
68 if (mesa_ind_buf->type == GL_UNSIGNED_BYTE) {
69 GLubyte *in = (GLubyte *)src_ptr;
70 GLuint *out = _mesa_malloc(sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1));
71 int i;
72
73 ind_buf->ptr = out;
74
75 for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
76 *out++ = in[i] | in[i + 1] << 16;
77 }
78
79 if (i < mesa_ind_buf->count) {
80 *out++ = in[i];
81 }
82
83 ind_buf->free_needed = GL_TRUE;
84 ind_buf->is_32bit = GL_FALSE;
85 } else if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) {
86 #if MESA_BIG_ENDIAN
87 GLushort *in = (GLushort *)src_ptr;
88 GLuint *out = _mesa_malloc(sizeof(GLushort) *
89 ((mesa_ind_buf->count + 1) & ~1));
90 int i;
91
92 ind_buf->ptr = out;
93
94 for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
95 *out++ = in[i] | in[i + 1] << 16;
96 }
97
98 if (i < mesa_ind_buf->count) {
99 *out++ = in[i];
100 }
101
102 ind_buf->free_needed = GL_TRUE;
103 #else
104 ind_buf->ptr = src_ptr;
105 ind_buf->free_needed = GL_FALSE;
106 #endif
107 ind_buf->is_32bit = GL_FALSE;
108 } else {
109 ind_buf->ptr = src_ptr;
110 ind_buf->free_needed = GL_FALSE;
111 ind_buf->is_32bit = GL_TRUE;
112 }
113 }
114
115 static int getTypeSize(GLenum type)
116 {
117 switch (type) {
118 case GL_DOUBLE:
119 return sizeof(GLdouble);
120 case GL_FLOAT:
121 return sizeof(GLfloat);
122 case GL_INT:
123 return sizeof(GLint);
124 case GL_UNSIGNED_INT:
125 return sizeof(GLuint);
126 case GL_SHORT:
127 return sizeof(GLshort);
128 case GL_UNSIGNED_SHORT:
129 return sizeof(GLushort);
130 case GL_BYTE:
131 return sizeof(GLbyte);
132 case GL_UNSIGNED_BYTE:
133 return sizeof(GLubyte);
134 default:
135 assert(0);
136 return 0;
137 }
138 }
139
140 #define CONVERT( TYPE, MACRO ) do { \
141 GLuint i, j, sz; \
142 sz = input->Size; \
143 if (input->Normalized) { \
144 for (i = 0; i < count; i++) { \
145 const TYPE *in = (TYPE *)src_ptr; \
146 for (j = 0; j < sz; j++) { \
147 *dst_ptr++ = MACRO(*in); \
148 in++; \
149 } \
150 src_ptr += stride; \
151 } \
152 } else { \
153 for (i = 0; i < count; i++) { \
154 const TYPE *in = (TYPE *)src_ptr; \
155 for (j = 0; j < sz; j++) { \
156 *dst_ptr++ = (GLfloat)(*in); \
157 in++; \
158 } \
159 src_ptr += stride; \
160 } \
161 } \
162 } while (0)
163
164 static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const struct gl_client_array *input, struct gl_buffer_object **bo, GLuint *nr_bo)
165 {
166 r300ContextPtr r300 = R300_CONTEXT(ctx);
167 struct r300_vertex_buffer *vbuf = &r300->vbuf;
168 struct vertex_attribute r300_attr;
169 const void *src_ptr;
170 GLenum type;
171 GLuint stride;
172
173 if (input->BufferObj->Name) {
174 if (!input->BufferObj->Pointer) {
175 bo[*nr_bo] = input->BufferObj;
176 (*nr_bo)++;
177 ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj);
178 assert(input->BufferObj->Pointer != NULL);
179 }
180
181 src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr);
182 } else
183 src_ptr = input->Ptr;
184
185 stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB;
186
187 if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT ||
188 #if MESA_BIG_ENDIAN
189 getTypeSize(input->Type) != 4 ||
190 #endif
191 stride < 4) {
192 if (RADEON_DEBUG & DEBUG_FALLBACKS) {
193 fprintf(stderr, "%s: Converting vertex attributes, attribute data format %x,", __FUNCTION__, input->Type);
194 fprintf(stderr, "stride %d, components %d\n", stride, input->Size);
195 }
196
197 GLfloat *dst_ptr, *tmp;
198
199 /* Convert value for first element only */
200 if (input->StrideB == 0)
201 count = 1;
202
203 tmp = dst_ptr = _mesa_malloc(sizeof(GLfloat) * input->Size * count);
204
205 switch (input->Type) {
206 case GL_DOUBLE:
207 CONVERT(GLdouble, (GLfloat));
208 break;
209 case GL_UNSIGNED_INT:
210 CONVERT(GLuint, UINT_TO_FLOAT);
211 break;
212 case GL_INT:
213 CONVERT(GLint, INT_TO_FLOAT);
214 break;
215 case GL_UNSIGNED_SHORT:
216 CONVERT(GLushort, USHORT_TO_FLOAT);
217 break;
218 case GL_SHORT:
219 CONVERT(GLshort, SHORT_TO_FLOAT);
220 break;
221 case GL_UNSIGNED_BYTE:
222 assert(input->Format != GL_BGRA);
223 CONVERT(GLubyte, UBYTE_TO_FLOAT);
224 break;
225 case GL_BYTE:
226 CONVERT(GLbyte, BYTE_TO_FLOAT);
227 break;
228 default:
229 assert(0);
230 break;
231 }
232
233 type = GL_FLOAT;
234 r300_attr.free_needed = GL_TRUE;
235 r300_attr.data = tmp;
236 if (input->StrideB == 0) {
237 r300_attr.stride = 0;
238 } else {
239 r300_attr.stride = sizeof(GLfloat) * input->Size;
240 }
241 r300_attr.dwords = input->Size;
242 } else {
243 type = input->Type;
244 r300_attr.free_needed = GL_FALSE;
245 r300_attr.data = (GLvoid *)src_ptr;
246 r300_attr.stride = input->StrideB;
247 r300_attr.dwords = (getTypeSize(type) * input->Size + 3)/ 4;
248 }
249
250 r300_attr.size = input->Size;
251 r300_attr.element = attr;
252 r300_attr.dst_loc = vbuf->num_attribs;
253
254 switch (type) {
255 case GL_FLOAT:
256 switch (input->Size) {
257 case 1: r300_attr.data_type = R300_DATA_TYPE_FLOAT_1; break;
258 case 2: r300_attr.data_type = R300_DATA_TYPE_FLOAT_2; break;
259 case 3: r300_attr.data_type = R300_DATA_TYPE_FLOAT_3; break;
260 case 4: r300_attr.data_type = R300_DATA_TYPE_FLOAT_4; break;
261 }
262 r300_attr._signed = 0;
263 r300_attr.normalize = 0;
264 break;
265 case GL_SHORT:
266 r300_attr._signed = 1;
267 r300_attr.normalize = input->Normalized;
268 switch (input->Size) {
269 case 1:
270 case 2:
271 r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
272 break;
273 case 3:
274 case 4:
275 r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
276 break;
277 }
278 break;
279 case GL_BYTE:
280 r300_attr._signed = 1;
281 r300_attr.normalize = input->Normalized;
282 r300_attr.data_type = R300_DATA_TYPE_BYTE;
283 break;
284 case GL_UNSIGNED_SHORT:
285 r300_attr._signed = 0;
286 r300_attr.normalize = input->Normalized;
287 switch (input->Size) {
288 case 1:
289 case 2:
290 r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
291 break;
292 case 3:
293 case 4:
294 r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
295 break;
296 }
297 break;
298 case GL_UNSIGNED_BYTE:
299 r300_attr._signed = 0;
300 r300_attr.normalize = input->Normalized;
301 if (input->Format == GL_BGRA)
302 r300_attr.data_type = R300_DATA_TYPE_D3DCOLOR;
303 else
304 r300_attr.data_type = R300_DATA_TYPE_BYTE;
305 break;
306
307 default:
308 case GL_DOUBLE:
309 case GL_INT:
310 case GL_UNSIGNED_INT:
311 assert(0);
312 break;
313 }
314
315 switch (input->Size) {
316 case 4:
317 r300_attr.swizzle = SWIZZLE_XYZW;
318 break;
319 case 3:
320 r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
321 break;
322 case 2:
323 r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
324 break;
325 case 1:
326 r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
327 break;
328 }
329
330 r300_attr.write_mask = MASK_XYZW;
331
332 vbuf->attribs[vbuf->num_attribs] = r300_attr;
333 ++vbuf->num_attribs;
334 }
335
336 static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count, struct gl_buffer_object **bo, GLuint *nr_bo)
337 {
338 r300ContextPtr r300 = R300_CONTEXT(ctx);
339 struct r300_vertex_buffer *vbuf = &r300->vbuf;
340
341 {
342 int i, tmp;
343
344 tmp = r300->selected_vp->code.InputsRead;
345 i = 0;
346 vbuf->num_attribs = 0;
347 while (tmp) {
348 /* find first enabled bit */
349 while (!(tmp & 1)) {
350 tmp >>= 1;
351 ++i;
352 }
353
354 r300TranslateAttrib(ctx, i, count, arrays[i], bo, nr_bo);
355
356 tmp >>= 1;
357 ++i;
358 }
359 }
360
361 r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
362 if (r300->fallback)
363 return;
364
365 {
366 int i;
367
368 for (i = 0; i < vbuf->num_attribs; i++) {
369 rcommon_emit_vector(ctx, &r300->radeon.tcl.aos[i],
370 vbuf->attribs[i].data, vbuf->attribs[i].dwords,
371 vbuf->attribs[i].stride, count);
372 }
373
374 r300->radeon.tcl.aos_count = vbuf->num_attribs;
375 }
376 }
377
378 static void r300FreeData(GLcontext *ctx, struct gl_buffer_object **bo, GLuint nr_bo)
379 {
380 {
381 struct r300_vertex_buffer *vbuf = &R300_CONTEXT(ctx)->vbuf;
382 int i;
383
384 for (i = 0; i < vbuf->num_attribs; i++) {
385 if (vbuf->attribs[i].free_needed)
386 _mesa_free(vbuf->attribs[i].data);
387 }
388 }
389
390 {
391 struct r300_index_buffer *ind_buf = &R300_CONTEXT(ctx)->ind_buf;
392 if (ind_buf->free_needed)
393 _mesa_free(ind_buf->ptr);
394 }
395
396 {
397 int i;
398
399 for (i = 0; i < nr_bo; ++i) {
400 ctx->Driver.UnmapBuffer(ctx, 0, bo[i]);
401 }
402 }
403 }
404
405 static GLboolean r300TryDrawPrims(GLcontext *ctx,
406 const struct gl_client_array *arrays[],
407 const struct _mesa_prim *prim,
408 GLuint nr_prims,
409 const struct _mesa_index_buffer *ib,
410 GLuint min_index,
411 GLuint max_index )
412 {
413 struct r300_context *r300 = R300_CONTEXT(ctx);
414 struct gl_buffer_object *bo[VERT_ATTRIB_MAX+1];
415 GLuint i, nr_bo = 0;
416
417 if (ctx->NewState)
418 _mesa_update_state( ctx );
419
420 if (r300->options.hw_tcl_enabled)
421 _tnl_UpdateFixedFunctionProgram(ctx);
422
423 r300UpdateShaders(r300);
424
425 r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, !r300ValidateBuffers(ctx));
426
427 r300FixupIndexBuffer(ctx, ib, bo, &nr_bo);
428
429 /* ensure we have the cmd buf space in advance to cover
430 * the state + DMA AOS pointers */
431 rcommonEnsureCmdBufSpace(&r300->radeon,
432 r300->radeon.hw.max_state_size + (50*sizeof(int)),
433 __FUNCTION__);
434
435 r300SetVertexFormat(ctx, arrays, max_index + 1, bo, &nr_bo);
436
437 if (r300->fallback)
438 return GL_FALSE;
439
440 r300SetupVAP(ctx, r300->selected_vp->code.InputsRead, r300->selected_vp->code.OutputsWritten);
441
442 r300UpdateShaderStates(r300);
443
444 r300EmitCacheFlush(r300);
445 radeonEmitState(&r300->radeon);
446
447 for (i = 0; i < nr_prims; ++i) {
448 r300RunRenderPrimitive(ctx, prim[i].start, prim[i].start + prim[i].count, prim[i].mode);
449 }
450
451 r300EmitCacheFlush(r300);
452
453 radeonReleaseArrays(ctx, ~0);
454
455 r300FreeData(ctx, bo, nr_bo);
456
457 return GL_TRUE;
458 }
459
460 static void r300DrawPrims(GLcontext *ctx,
461 const struct gl_client_array *arrays[],
462 const struct _mesa_prim *prim,
463 GLuint nr_prims,
464 const struct _mesa_index_buffer *ib,
465 GLboolean index_bounds_valid,
466 GLuint min_index,
467 GLuint max_index)
468 {
469 struct split_limits limits;
470 GLboolean retval;
471
472 if (ib)
473 limits.max_verts = 0xffffffff;
474 else
475 limits.max_verts = 65535;
476
477 limits.max_indices = 65535;
478 limits.max_vb_size = 1024*1024;
479
480 /* This check should get folded into just the places that
481 * min/max index are really needed.
482 */
483 if (!index_bounds_valid)
484 vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
485
486 if (min_index) {
487 vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims );
488 return;
489 }
490 if ((ib && ib->count > 65535)) {
491 vbo_split_prims (ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims, &limits);
492 return;
493 }
494
495 /* Make an attempt at drawing */
496 retval = r300TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
497
498 /* If failed run tnl pipeline - it should take care of fallbacks */
499 if (!retval)
500 _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
501 }
502
503 void r300InitDraw(GLcontext *ctx)
504 {
505 struct vbo_context *vbo = vbo_context(ctx);
506
507 vbo->draw_prims = r300DrawPrims;
508 }