Merge branch 'xa_branch'
[mesa.git] / src / gallium / drivers / svga / svga_draw.c
1 /**********************************************************
2 * Copyright 2008-2009 VMware, Inc. All rights reserved.
3 *
4 * Permission is hereby granted, free of charge, to any person
5 * obtaining a copy of this software and associated documentation
6 * files (the "Software"), to deal in the Software without
7 * restriction, including without limitation the rights to use, copy,
8 * modify, merge, publish, distribute, sublicense, and/or sell copies
9 * of the Software, and to permit persons to whom the Software is
10 * furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice shall be
13 * included in all copies or substantial portions of the Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
17 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
19 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
20 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
21 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 *
24 **********************************************************/
25
26 #include "pipe/p_compiler.h"
27 #include "util/u_inlines.h"
28 #include "pipe/p_defines.h"
29 #include "util/u_memory.h"
30 #include "util/u_math.h"
31 #include "util/u_upload_mgr.h"
32
33 #include "svga_context.h"
34 #include "svga_draw.h"
35 #include "svga_draw_private.h"
36 #include "svga_debug.h"
37 #include "svga_screen.h"
38 #include "svga_resource_buffer.h"
39 #include "svga_resource_texture.h"
40 #include "svga_surface.h"
41 #include "svga_winsys.h"
42 #include "svga_cmd.h"
43
44
45 struct svga_hwtnl *svga_hwtnl_create( struct svga_context *svga,
46 struct u_upload_mgr *upload_ib,
47 struct svga_winsys_context *swc )
48 {
49 struct svga_hwtnl *hwtnl = CALLOC_STRUCT(svga_hwtnl);
50 if (hwtnl == NULL)
51 goto fail;
52
53 hwtnl->svga = svga;
54 hwtnl->upload_ib = upload_ib;
55
56 hwtnl->cmd.swc = swc;
57
58 return hwtnl;
59
60 fail:
61 return NULL;
62 }
63
64 void svga_hwtnl_destroy( struct svga_hwtnl *hwtnl )
65 {
66 int i, j;
67
68 for (i = 0; i < PIPE_PRIM_MAX; i++) {
69 for (j = 0; j < IDX_CACHE_MAX; j++) {
70 pipe_resource_reference( &hwtnl->index_cache[i][j].buffer,
71 NULL );
72 }
73 }
74
75 for (i = 0; i < hwtnl->cmd.vdecl_count; i++)
76 pipe_resource_reference(&hwtnl->cmd.vdecl_vb[i], NULL);
77
78 for (i = 0; i < hwtnl->cmd.prim_count; i++)
79 pipe_resource_reference(&hwtnl->cmd.prim_ib[i], NULL);
80
81
82 FREE(hwtnl);
83 }
84
85
86 void svga_hwtnl_set_flatshade( struct svga_hwtnl *hwtnl,
87 boolean flatshade,
88 boolean flatshade_first )
89 {
90 hwtnl->hw_pv = PV_FIRST;
91 hwtnl->api_pv = (flatshade && !flatshade_first) ? PV_LAST : PV_FIRST;
92 }
93
94 void svga_hwtnl_set_unfilled( struct svga_hwtnl *hwtnl,
95 unsigned mode )
96 {
97 hwtnl->api_fillmode = mode;
98 }
99
100 void svga_hwtnl_reset_vdecl( struct svga_hwtnl *hwtnl,
101 unsigned count )
102 {
103 unsigned i;
104
105 assert(hwtnl->cmd.prim_count == 0);
106
107 for (i = count; i < hwtnl->cmd.vdecl_count; i++) {
108 pipe_resource_reference(&hwtnl->cmd.vdecl_vb[i],
109 NULL);
110 }
111
112 hwtnl->cmd.vdecl_count = count;
113 }
114
115
116 void svga_hwtnl_vdecl( struct svga_hwtnl *hwtnl,
117 unsigned i,
118 const SVGA3dVertexDecl *decl,
119 struct pipe_resource *vb)
120 {
121 assert(hwtnl->cmd.prim_count == 0);
122
123 assert( i < hwtnl->cmd.vdecl_count );
124
125 hwtnl->cmd.vdecl[i] = *decl;
126
127 pipe_resource_reference(&hwtnl->cmd.vdecl_vb[i], vb);
128 }
129
130
131
132 enum pipe_error
133 svga_hwtnl_flush( struct svga_hwtnl *hwtnl )
134 {
135 struct svga_winsys_context *swc = hwtnl->cmd.swc;
136 struct svga_context *svga = hwtnl->svga;
137 enum pipe_error ret;
138
139 if (hwtnl->cmd.prim_count) {
140 struct svga_winsys_surface *vb_handle[SVGA3D_INPUTREG_MAX];
141 struct svga_winsys_surface *ib_handle[QSZ];
142 struct svga_winsys_surface *handle;
143 SVGA3dVertexDecl *vdecl;
144 SVGA3dPrimitiveRange *prim;
145 unsigned i;
146
147 /* Unmap upload manager vertex buffers */
148 u_upload_unmap(svga->upload_vb);
149
150 for (i = 0; i < hwtnl->cmd.vdecl_count; i++) {
151 handle = svga_buffer_handle(svga, hwtnl->cmd.vdecl_vb[i]);
152 if (handle == NULL)
153 return PIPE_ERROR_OUT_OF_MEMORY;
154
155 vb_handle[i] = handle;
156 }
157
158 /* Unmap upload manager index buffers */
159 u_upload_unmap(svga->upload_ib);
160
161 for (i = 0; i < hwtnl->cmd.prim_count; i++) {
162 if (hwtnl->cmd.prim_ib[i]) {
163 handle = svga_buffer_handle(svga, hwtnl->cmd.prim_ib[i]);
164 if (handle == NULL)
165 return PIPE_ERROR_OUT_OF_MEMORY;
166 }
167 else
168 handle = NULL;
169
170 ib_handle[i] = handle;
171 }
172
173 if (svga->rebind.rendertargets) {
174 ret = svga_reemit_framebuffer_bindings(svga);
175 if (ret != PIPE_OK) {
176 return ret;
177 }
178 }
179
180 if (svga->rebind.texture_samplers) {
181 ret = svga_reemit_tss_bindings(svga);
182 if (ret != PIPE_OK) {
183 return ret;
184 }
185 }
186
187 SVGA_DBG(DEBUG_DMA, "draw to sid %p, %d prims\n",
188 svga->curr.framebuffer.cbufs[0] ?
189 svga_surface(svga->curr.framebuffer.cbufs[0])->handle : NULL,
190 hwtnl->cmd.prim_count);
191
192 ret = SVGA3D_BeginDrawPrimitives(swc,
193 &vdecl,
194 hwtnl->cmd.vdecl_count,
195 &prim,
196 hwtnl->cmd.prim_count);
197 if (ret != PIPE_OK)
198 return ret;
199
200
201 memcpy( vdecl,
202 hwtnl->cmd.vdecl,
203 hwtnl->cmd.vdecl_count * sizeof hwtnl->cmd.vdecl[0]);
204
205 for (i = 0; i < hwtnl->cmd.vdecl_count; i++) {
206 /* Given rangeHint is considered to be relative to indexBias, and
207 * indexBias varies per primitive, we cannot accurately supply an
208 * rangeHint when emitting more than one primitive per draw command.
209 */
210 if (hwtnl->cmd.prim_count == 1) {
211 vdecl[i].rangeHint.first = hwtnl->cmd.min_index[0];
212 vdecl[i].rangeHint.last = hwtnl->cmd.max_index[0] + 1;
213 }
214 else {
215 vdecl[i].rangeHint.first = 0;
216 vdecl[i].rangeHint.last = 0;
217 }
218
219 swc->surface_relocation(swc,
220 &vdecl[i].array.surfaceId,
221 vb_handle[i],
222 SVGA_RELOC_READ);
223 }
224
225 memcpy( prim,
226 hwtnl->cmd.prim,
227 hwtnl->cmd.prim_count * sizeof hwtnl->cmd.prim[0]);
228
229 for (i = 0; i < hwtnl->cmd.prim_count; i++) {
230 swc->surface_relocation(swc,
231 &prim[i].indexArray.surfaceId,
232 ib_handle[i],
233 SVGA_RELOC_READ);
234 pipe_resource_reference(&hwtnl->cmd.prim_ib[i], NULL);
235 }
236
237 SVGA_FIFOCommitAll( swc );
238 hwtnl->cmd.prim_count = 0;
239 }
240
241 return PIPE_OK;
242 }
243
244
245 void svga_hwtnl_set_index_bias( struct svga_hwtnl *hwtnl,
246 int index_bias)
247 {
248 hwtnl->index_bias = index_bias;
249 }
250
251
252
253 /***********************************************************************
254 * Internal functions:
255 */
256
257 enum pipe_error svga_hwtnl_prim( struct svga_hwtnl *hwtnl,
258 const SVGA3dPrimitiveRange *range,
259 unsigned min_index,
260 unsigned max_index,
261 struct pipe_resource *ib )
262 {
263 int ret = PIPE_OK;
264
265 #ifdef DEBUG
266 {
267 unsigned i;
268 for (i = 0; i < hwtnl->cmd.vdecl_count; i++) {
269 struct pipe_resource *vb = hwtnl->cmd.vdecl_vb[i];
270 unsigned size = vb ? vb->width0 : 0;
271 unsigned offset = hwtnl->cmd.vdecl[i].array.offset;
272 unsigned stride = hwtnl->cmd.vdecl[i].array.stride;
273 int index_bias = (int) range->indexBias + hwtnl->index_bias;
274 unsigned width;
275
276 assert(vb);
277 assert(size);
278 assert(offset < size);
279 assert(min_index <= max_index);
280 if (index_bias >= 0) {
281 assert(offset + index_bias*stride < size);
282 }
283 if (min_index != ~0) {
284 assert(offset + (index_bias + min_index) * stride < size);
285 }
286
287 switch (hwtnl->cmd.vdecl[i].identity.type) {
288 case SVGA3D_DECLTYPE_FLOAT1:
289 width = 4;
290 break;
291 case SVGA3D_DECLTYPE_FLOAT2:
292 width = 4*2;
293 break;
294 case SVGA3D_DECLTYPE_FLOAT3:
295 width = 4*3;
296 break;
297 case SVGA3D_DECLTYPE_FLOAT4:
298 width = 4*4;
299 break;
300 case SVGA3D_DECLTYPE_D3DCOLOR:
301 width = 4;
302 break;
303 case SVGA3D_DECLTYPE_UBYTE4:
304 width = 1*4;
305 break;
306 case SVGA3D_DECLTYPE_SHORT2:
307 width = 2*2;
308 break;
309 case SVGA3D_DECLTYPE_SHORT4:
310 width = 2*4;
311 break;
312 case SVGA3D_DECLTYPE_UBYTE4N:
313 width = 1*4;
314 break;
315 case SVGA3D_DECLTYPE_SHORT2N:
316 width = 2*2;
317 break;
318 case SVGA3D_DECLTYPE_SHORT4N:
319 width = 2*4;
320 break;
321 case SVGA3D_DECLTYPE_USHORT2N:
322 width = 2*2;
323 break;
324 case SVGA3D_DECLTYPE_USHORT4N:
325 width = 2*4;
326 break;
327 case SVGA3D_DECLTYPE_UDEC3:
328 width = 4;
329 break;
330 case SVGA3D_DECLTYPE_DEC3N:
331 width = 4;
332 break;
333 case SVGA3D_DECLTYPE_FLOAT16_2:
334 width = 2*2;
335 break;
336 case SVGA3D_DECLTYPE_FLOAT16_4:
337 width = 2*4;
338 break;
339 default:
340 assert(0);
341 width = 0;
342 break;
343 }
344
345 if (max_index != ~0) {
346 assert(offset + (index_bias + max_index) * stride + width <= size);
347 }
348 }
349
350 assert(range->indexWidth == range->indexArray.stride);
351
352 if(ib) {
353 unsigned size = ib->width0;
354 unsigned offset = range->indexArray.offset;
355 unsigned stride = range->indexArray.stride;
356 unsigned count;
357
358 assert(size);
359 assert(offset < size);
360 assert(stride);
361
362 switch (range->primType) {
363 case SVGA3D_PRIMITIVE_POINTLIST:
364 count = range->primitiveCount;
365 break;
366 case SVGA3D_PRIMITIVE_LINELIST:
367 count = range->primitiveCount * 2;
368 break;
369 case SVGA3D_PRIMITIVE_LINESTRIP:
370 count = range->primitiveCount + 1;
371 break;
372 case SVGA3D_PRIMITIVE_TRIANGLELIST:
373 count = range->primitiveCount * 3;
374 break;
375 case SVGA3D_PRIMITIVE_TRIANGLESTRIP:
376 count = range->primitiveCount + 2;
377 break;
378 case SVGA3D_PRIMITIVE_TRIANGLEFAN:
379 count = range->primitiveCount + 2;
380 break;
381 default:
382 assert(0);
383 count = 0;
384 break;
385 }
386
387 assert(offset + count*stride <= size);
388 }
389 }
390 #endif
391
392 if (hwtnl->cmd.prim_count+1 >= QSZ) {
393 ret = svga_hwtnl_flush( hwtnl );
394 if (ret != PIPE_OK)
395 return ret;
396 }
397
398 /* min/max indices are relative to bias */
399 hwtnl->cmd.min_index[hwtnl->cmd.prim_count] = min_index;
400 hwtnl->cmd.max_index[hwtnl->cmd.prim_count] = max_index;
401
402 hwtnl->cmd.prim[hwtnl->cmd.prim_count] = *range;
403 hwtnl->cmd.prim[hwtnl->cmd.prim_count].indexBias += hwtnl->index_bias;
404
405 pipe_resource_reference(&hwtnl->cmd.prim_ib[hwtnl->cmd.prim_count], ib);
406 hwtnl->cmd.prim_count++;
407
408 return ret;
409 }