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
45struct 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
60fail:
61   return NULL;
62}
63
64void 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
86void 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
94void svga_hwtnl_set_unfilled( struct svga_hwtnl *hwtnl,
95                              unsigned mode )
96{
97   hwtnl->api_fillmode = mode;
98}
99
100void 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
116void 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 * Determine whether the specified buffer is referred in the primitive queue,
133 * for which no commands have been written yet.
134 */
135boolean
136svga_hwtnl_is_buffer_referred( struct svga_hwtnl *hwtnl,
137                               struct pipe_resource *buffer)
138{
139   unsigned i;
140
141   if (svga_buffer_is_user_buffer(buffer)) {
142      return FALSE;
143   }
144
145   if (!hwtnl->cmd.prim_count) {
146      return FALSE;
147   }
148
149   for (i = 0; i < hwtnl->cmd.vdecl_count; ++i) {
150      if (hwtnl->cmd.vdecl_vb[i] == buffer) {
151         return TRUE;
152      }
153   }
154
155   for (i = 0; i < hwtnl->cmd.prim_count; ++i) {
156      if (hwtnl->cmd.prim_ib[i] == buffer) {
157         return TRUE;
158      }
159   }
160
161   return FALSE;
162}
163
164
165enum pipe_error
166svga_hwtnl_flush( struct svga_hwtnl *hwtnl )
167{
168   struct svga_winsys_context *swc = hwtnl->cmd.swc;
169   struct svga_context *svga = hwtnl->svga;
170   enum pipe_error ret;
171
172   if (hwtnl->cmd.prim_count) {
173      struct svga_winsys_surface *vb_handle[SVGA3D_INPUTREG_MAX];
174      struct svga_winsys_surface *ib_handle[QSZ];
175      struct svga_winsys_surface *handle;
176      SVGA3dVertexDecl *vdecl;
177      SVGA3dPrimitiveRange *prim;
178      unsigned i;
179
180      /* Unmap upload manager vertex buffers */
181      u_upload_unmap(svga->upload_vb);
182
183      for (i = 0; i < hwtnl->cmd.vdecl_count; i++) {
184         handle = svga_buffer_handle(svga, hwtnl->cmd.vdecl_vb[i]);
185         if (handle == NULL)
186            return PIPE_ERROR_OUT_OF_MEMORY;
187
188         vb_handle[i] = handle;
189      }
190
191      /* Unmap upload manager index buffers */
192      u_upload_unmap(svga->upload_ib);
193
194      for (i = 0; i < hwtnl->cmd.prim_count; i++) {
195         if (hwtnl->cmd.prim_ib[i]) {
196            handle = svga_buffer_handle(svga, hwtnl->cmd.prim_ib[i]);
197            if (handle == NULL)
198               return PIPE_ERROR_OUT_OF_MEMORY;
199         }
200         else
201            handle = NULL;
202
203         ib_handle[i] = handle;
204      }
205
206      if (svga->rebind.rendertargets) {
207         ret = svga_reemit_framebuffer_bindings(svga);
208         if (ret != PIPE_OK) {
209            return ret;
210         }
211      }
212
213      if (svga->rebind.texture_samplers) {
214         ret = svga_reemit_tss_bindings(svga);
215         if (ret != PIPE_OK) {
216            return ret;
217         }
218      }
219
220      SVGA_DBG(DEBUG_DMA, "draw to sid %p, %d prims\n",
221               svga->curr.framebuffer.cbufs[0] ?
222               svga_surface(svga->curr.framebuffer.cbufs[0])->handle : NULL,
223               hwtnl->cmd.prim_count);
224
225      ret = SVGA3D_BeginDrawPrimitives(swc,
226                                       &vdecl,
227                                       hwtnl->cmd.vdecl_count,
228                                       &prim,
229                                       hwtnl->cmd.prim_count);
230      if (ret != PIPE_OK)
231         return ret;
232
233
234      memcpy( vdecl,
235              hwtnl->cmd.vdecl,
236              hwtnl->cmd.vdecl_count * sizeof hwtnl->cmd.vdecl[0]);
237
238      for (i = 0; i < hwtnl->cmd.vdecl_count; i++) {
239         /* Given rangeHint is considered to be relative to indexBias, and
240          * indexBias varies per primitive, we cannot accurately supply an
241          * rangeHint when emitting more than one primitive per draw command.
242          */
243         if (hwtnl->cmd.prim_count == 1) {
244            vdecl[i].rangeHint.first = hwtnl->cmd.min_index[0];
245            vdecl[i].rangeHint.last = hwtnl->cmd.max_index[0] + 1;
246         }
247         else {
248            vdecl[i].rangeHint.first = 0;
249            vdecl[i].rangeHint.last = 0;
250         }
251
252         swc->surface_relocation(swc,
253                                 &vdecl[i].array.surfaceId,
254                                 vb_handle[i],
255                                 SVGA_RELOC_READ);
256      }
257
258      memcpy( prim,
259              hwtnl->cmd.prim,
260              hwtnl->cmd.prim_count * sizeof hwtnl->cmd.prim[0]);
261
262      for (i = 0; i < hwtnl->cmd.prim_count; i++) {
263         swc->surface_relocation(swc,
264                                 &prim[i].indexArray.surfaceId,
265                                 ib_handle[i],
266                                 SVGA_RELOC_READ);
267         pipe_resource_reference(&hwtnl->cmd.prim_ib[i], NULL);
268      }
269
270      SVGA_FIFOCommitAll( swc );
271      hwtnl->cmd.prim_count = 0;
272   }
273
274   return PIPE_OK;
275}
276
277
278void svga_hwtnl_set_index_bias( struct svga_hwtnl *hwtnl,
279				int index_bias)
280{
281   hwtnl->index_bias = index_bias;
282}
283
284
285
286/***********************************************************************
287 * Internal functions:
288 */
289
290enum pipe_error svga_hwtnl_prim( struct svga_hwtnl *hwtnl,
291                                 const SVGA3dPrimitiveRange *range,
292                                 unsigned min_index,
293                                 unsigned max_index,
294                                 struct pipe_resource *ib )
295{
296   enum pipe_error ret = PIPE_OK;
297
298#ifdef DEBUG
299   {
300      unsigned i;
301      for (i = 0; i < hwtnl->cmd.vdecl_count; i++) {
302         struct pipe_resource *vb = hwtnl->cmd.vdecl_vb[i];
303         unsigned size = vb ? vb->width0 : 0;
304         unsigned offset = hwtnl->cmd.vdecl[i].array.offset;
305         unsigned stride = hwtnl->cmd.vdecl[i].array.stride;
306         int index_bias = (int) range->indexBias + hwtnl->index_bias;
307         unsigned width;
308
309         assert(vb);
310         assert(size);
311         assert(offset < size);
312         assert(min_index <= max_index);
313
314         switch (hwtnl->cmd.vdecl[i].identity.type) {
315         case SVGA3D_DECLTYPE_FLOAT1:
316            width = 4;
317            break;
318         case SVGA3D_DECLTYPE_FLOAT2:
319            width = 4*2;
320            break;
321         case SVGA3D_DECLTYPE_FLOAT3:
322            width = 4*3;
323            break;
324         case SVGA3D_DECLTYPE_FLOAT4:
325            width = 4*4;
326            break;
327         case SVGA3D_DECLTYPE_D3DCOLOR:
328            width = 4;
329            break;
330         case SVGA3D_DECLTYPE_UBYTE4:
331            width = 1*4;
332            break;
333         case SVGA3D_DECLTYPE_SHORT2:
334            width = 2*2;
335            break;
336         case SVGA3D_DECLTYPE_SHORT4:
337            width = 2*4;
338            break;
339         case SVGA3D_DECLTYPE_UBYTE4N:
340            width = 1*4;
341            break;
342         case SVGA3D_DECLTYPE_SHORT2N:
343            width = 2*2;
344            break;
345         case SVGA3D_DECLTYPE_SHORT4N:
346            width = 2*4;
347            break;
348         case SVGA3D_DECLTYPE_USHORT2N:
349            width = 2*2;
350            break;
351         case SVGA3D_DECLTYPE_USHORT4N:
352            width = 2*4;
353            break;
354         case SVGA3D_DECLTYPE_UDEC3:
355            width = 4;
356            break;
357         case SVGA3D_DECLTYPE_DEC3N:
358            width = 4;
359            break;
360         case SVGA3D_DECLTYPE_FLOAT16_2:
361            width = 2*2;
362            break;
363         case SVGA3D_DECLTYPE_FLOAT16_4:
364            width = 2*4;
365            break;
366         default:
367            assert(0);
368            width = 0;
369            break;
370         }
371
372         if (index_bias >= 0) {
373            assert(offset + index_bias*stride + width <= size);
374         }
375
376         /*
377          * min_index/max_index are merely conservative guesses, so we can't
378          * make buffer overflow detection based on their values.
379          */
380      }
381
382      assert(range->indexWidth == range->indexArray.stride);
383
384      if(ib) {
385         unsigned size = ib->width0;
386         unsigned offset = range->indexArray.offset;
387         unsigned stride = range->indexArray.stride;
388         unsigned count;
389
390         assert(size);
391         assert(offset < size);
392         assert(stride);
393
394         switch (range->primType) {
395         case SVGA3D_PRIMITIVE_POINTLIST:
396            count = range->primitiveCount;
397            break;
398         case SVGA3D_PRIMITIVE_LINELIST:
399            count = range->primitiveCount * 2;
400            break;
401         case SVGA3D_PRIMITIVE_LINESTRIP:
402            count = range->primitiveCount + 1;
403            break;
404         case SVGA3D_PRIMITIVE_TRIANGLELIST:
405            count = range->primitiveCount * 3;
406            break;
407         case SVGA3D_PRIMITIVE_TRIANGLESTRIP:
408            count = range->primitiveCount + 2;
409            break;
410         case SVGA3D_PRIMITIVE_TRIANGLEFAN:
411            count = range->primitiveCount + 2;
412            break;
413         default:
414            assert(0);
415            count = 0;
416            break;
417         }
418
419         assert(offset + count*stride <= size);
420      }
421   }
422#endif
423
424   if (hwtnl->cmd.prim_count+1 >= QSZ) {
425      ret = svga_hwtnl_flush( hwtnl );
426      if (ret != PIPE_OK)
427         return ret;
428   }
429
430   /* min/max indices are relative to bias */
431   hwtnl->cmd.min_index[hwtnl->cmd.prim_count] = min_index;
432   hwtnl->cmd.max_index[hwtnl->cmd.prim_count] = max_index;
433
434   hwtnl->cmd.prim[hwtnl->cmd.prim_count] = *range;
435   hwtnl->cmd.prim[hwtnl->cmd.prim_count].indexBias += hwtnl->index_bias;
436
437   pipe_resource_reference(&hwtnl->cmd.prim_ib[hwtnl->cmd.prim_count], ib);
438   hwtnl->cmd.prim_count++;
439
440   return ret;
441}
442