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 "draw/draw_context.h"
27 #include "draw/draw_vbuf.h"
28 #include "util/u_bitmask.h"
29 #include "util/u_inlines.h"
30 #include "pipe/p_state.h"
31
32 #include "svga_cmd.h"
33 #include "svga_context.h"
34 #include "svga_shader.h"
35 #include "svga_swtnl.h"
36 #include "svga_state.h"
37 #include "svga_tgsi.h"
38 #include "svga_swtnl_private.h"
39
40
41 #define SVGA_POINT_ADJ_X -0.375f
42 #define SVGA_POINT_ADJ_Y -0.5f
43
44 #define SVGA_LINE_ADJ_X -0.5f
45 #define SVGA_LINE_ADJ_Y -0.5f
46
47 #define SVGA_TRIANGLE_ADJ_X -0.375f
48 #define SVGA_TRIANGLE_ADJ_Y -0.5f
49
50
set_draw_viewport(struct svga_context * svga)51 static void set_draw_viewport( struct svga_context *svga )
52 {
53 struct pipe_viewport_state vp = svga->curr.viewport;
54 float adjx = 0.0f;
55 float adjy = 0.0f;
56
57 if (svga_have_vgpu10(svga)) {
58 if (svga->curr.reduced_prim == PIPE_PRIM_TRIANGLES) {
59 adjy = 0.25;
60 }
61 }
62 else {
63 switch (svga->curr.reduced_prim) {
64 case PIPE_PRIM_POINTS:
65 adjx = SVGA_POINT_ADJ_X;
66 adjy = SVGA_POINT_ADJ_Y;
67 break;
68 case PIPE_PRIM_LINES:
69 /* XXX: This is to compensate for the fact that wide lines are
70 * going to be drawn with triangles, but we're not catching all
71 * cases where that will happen.
72 */
73 if (svga->curr.rast->need_pipeline & SVGA_PIPELINE_FLAG_LINES)
74 {
75 adjx = SVGA_LINE_ADJ_X + 0.175f;
76 adjy = SVGA_LINE_ADJ_Y - 0.175f;
77 }
78 else {
79 adjx = SVGA_LINE_ADJ_X;
80 adjy = SVGA_LINE_ADJ_Y;
81 }
82 break;
83 case PIPE_PRIM_TRIANGLES:
84 adjx += SVGA_TRIANGLE_ADJ_X;
85 adjy += SVGA_TRIANGLE_ADJ_Y;
86 break;
87 }
88 }
89
90 vp.translate[0] += adjx;
91 vp.translate[1] += adjy;
92
93 draw_set_viewport_states(svga->swtnl.draw, 0, 1, &vp);
94 }
95
96 static enum pipe_error
update_swtnl_draw(struct svga_context * svga,unsigned dirty)97 update_swtnl_draw( struct svga_context *svga,
98 unsigned dirty )
99 {
100 SVGA_STATS_TIME_PUSH(svga_sws(svga), SVGA_STATS_TIME_SWTNLUPDATEDRAW);
101
102 draw_flush( svga->swtnl.draw );
103
104 if (dirty & SVGA_NEW_VS)
105 draw_bind_vertex_shader(svga->swtnl.draw,
106 svga->curr.vs->draw_shader);
107
108 if (dirty & SVGA_NEW_FS)
109 draw_bind_fragment_shader(svga->swtnl.draw,
110 svga->curr.fs->draw_shader);
111
112 if (dirty & SVGA_NEW_VBUFFER)
113 draw_set_vertex_buffers(svga->swtnl.draw, 0,
114 svga->curr.num_vertex_buffers,
115 svga->curr.vb);
116
117 if (dirty & SVGA_NEW_VELEMENT)
118 draw_set_vertex_elements(svga->swtnl.draw,
119 svga->curr.velems->count,
120 svga->curr.velems->velem );
121
122 if (dirty & SVGA_NEW_CLIP)
123 draw_set_clip_state(svga->swtnl.draw,
124 &svga->curr.clip);
125
126 if (dirty & (SVGA_NEW_VIEWPORT |
127 SVGA_NEW_REDUCED_PRIMITIVE |
128 SVGA_NEW_RAST))
129 set_draw_viewport( svga );
130
131 if (dirty & SVGA_NEW_RAST)
132 draw_set_rasterizer_state(svga->swtnl.draw,
133 &svga->curr.rast->templ,
134 (void *) svga->curr.rast);
135
136 /* Tell the draw module how deep the Z/depth buffer is.
137 *
138 * If no depth buffer is bound, send the utility function the
139 * format for no bound depth (PIPE_FORMAT_NONE).
140 */
141 if (dirty & SVGA_NEW_FRAME_BUFFER)
142 draw_set_zs_format(svga->swtnl.draw,
143 (svga->curr.framebuffer.zsbuf) ?
144 svga->curr.framebuffer.zsbuf->format : PIPE_FORMAT_NONE);
145
146 SVGA_STATS_TIME_POP(svga_sws(svga));
147 return PIPE_OK;
148 }
149
150
151 struct svga_tracked_state svga_update_swtnl_draw =
152 {
153 "update draw module state",
154 (SVGA_NEW_VS |
155 SVGA_NEW_VBUFFER |
156 SVGA_NEW_VELEMENT |
157 SVGA_NEW_CLIP |
158 SVGA_NEW_VIEWPORT |
159 SVGA_NEW_RAST |
160 SVGA_NEW_FRAME_BUFFER |
161 SVGA_NEW_REDUCED_PRIMITIVE),
162 update_swtnl_draw
163 };
164
165
166 static SVGA3dSurfaceFormat
translate_vertex_format(SVGA3dDeclType format)167 translate_vertex_format(SVGA3dDeclType format)
168 {
169 switch (format) {
170 case SVGA3D_DECLTYPE_FLOAT1:
171 return SVGA3D_R32_FLOAT;
172 case SVGA3D_DECLTYPE_FLOAT2:
173 return SVGA3D_R32G32_FLOAT;
174 case SVGA3D_DECLTYPE_FLOAT3:
175 return SVGA3D_R32G32B32_FLOAT;
176 case SVGA3D_DECLTYPE_FLOAT4:
177 return SVGA3D_R32G32B32A32_FLOAT;
178 default:
179 assert(!"Unexpected format in translate_vertex_format()");
180 return SVGA3D_R32G32B32A32_FLOAT;
181 }
182 }
183
184
185 static SVGA3dElementLayoutId
svga_vdecl_to_input_element(struct svga_context * svga,const SVGA3dVertexDecl * vdecl,unsigned num_decls)186 svga_vdecl_to_input_element(struct svga_context *svga,
187 const SVGA3dVertexDecl *vdecl, unsigned num_decls)
188 {
189 SVGA3dElementLayoutId id;
190 SVGA3dInputElementDesc elements[PIPE_MAX_ATTRIBS];
191 enum pipe_error ret;
192 unsigned i;
193
194 assert(num_decls <= PIPE_MAX_ATTRIBS);
195 assert(svga_have_vgpu10(svga));
196
197 for (i = 0; i < num_decls; i++) {
198 elements[i].inputSlot = 0; /* vertex buffer index */
199 elements[i].alignedByteOffset = vdecl[i].array.offset;
200 elements[i].format = translate_vertex_format(vdecl[i].identity.type);
201 elements[i].inputSlotClass = SVGA3D_INPUT_PER_VERTEX_DATA;
202 elements[i].instanceDataStepRate = 0;
203 elements[i].inputRegister = i;
204 }
205
206 id = util_bitmask_add(svga->input_element_object_id_bm);
207
208 ret = SVGA3D_vgpu10_DefineElementLayout(svga->swc, num_decls, id, elements);
209 if (ret != PIPE_OK) {
210 svga_context_flush(svga, NULL);
211 ret = SVGA3D_vgpu10_DefineElementLayout(svga->swc, num_decls, id, elements);
212 assert(ret == PIPE_OK);
213 }
214
215 return id;
216 }
217
218
219 enum pipe_error
svga_swtnl_update_vdecl(struct svga_context * svga)220 svga_swtnl_update_vdecl( struct svga_context *svga )
221 {
222 struct svga_vbuf_render *svga_render = svga_vbuf_render(svga->swtnl.backend);
223 struct draw_context *draw = svga->swtnl.draw;
224 struct vertex_info *vinfo = &svga_render->vertex_info;
225 SVGA3dVertexDecl vdecl[PIPE_MAX_ATTRIBS];
226 struct svga_fragment_shader *fs = svga->curr.fs;
227 int offset = 0;
228 int nr_decls = 0;
229 int src;
230 unsigned i;
231 int any_change;
232
233 SVGA_STATS_TIME_PUSH(svga_sws(svga), SVGA_STATS_TIME_SWTNLUPDATEVDECL);
234
235 memset(vinfo, 0, sizeof(*vinfo));
236 memset(vdecl, 0, sizeof(vdecl));
237
238 draw_prepare_shader_outputs(draw);
239
240 /* always add position */
241 src = draw_find_shader_output(draw, TGSI_SEMANTIC_POSITION, 0);
242 draw_emit_vertex_attr(vinfo, EMIT_4F, src);
243 vinfo->attrib[0].emit = EMIT_4F;
244 vdecl[0].array.offset = offset;
245 vdecl[0].identity.method = SVGA3D_DECLMETHOD_DEFAULT;
246 vdecl[0].identity.type = SVGA3D_DECLTYPE_FLOAT4;
247 vdecl[0].identity.usage = SVGA3D_DECLUSAGE_POSITIONT;
248 vdecl[0].identity.usageIndex = 0;
249 offset += 16;
250 nr_decls++;
251
252 for (i = 0; i < fs->base.info.num_inputs; i++) {
253 const unsigned sem_name = fs->base.info.input_semantic_name[i];
254 const unsigned sem_index = fs->base.info.input_semantic_index[i];
255
256 src = draw_find_shader_output(draw, sem_name, sem_index);
257
258 vdecl[nr_decls].array.offset = offset;
259 vdecl[nr_decls].identity.usageIndex = sem_index;
260
261 switch (sem_name) {
262 case TGSI_SEMANTIC_COLOR:
263 draw_emit_vertex_attr(vinfo, EMIT_4F, src);
264 vdecl[nr_decls].identity.usage = SVGA3D_DECLUSAGE_COLOR;
265 vdecl[nr_decls].identity.type = SVGA3D_DECLTYPE_FLOAT4;
266 offset += 16;
267 nr_decls++;
268 break;
269 case TGSI_SEMANTIC_GENERIC:
270 draw_emit_vertex_attr(vinfo, EMIT_4F, src);
271 vdecl[nr_decls].identity.usage = SVGA3D_DECLUSAGE_TEXCOORD;
272 vdecl[nr_decls].identity.type = SVGA3D_DECLTYPE_FLOAT4;
273 vdecl[nr_decls].identity.usageIndex =
274 svga_remap_generic_index(fs->generic_remap_table, sem_index);
275 offset += 16;
276 nr_decls++;
277 break;
278 case TGSI_SEMANTIC_FOG:
279 draw_emit_vertex_attr(vinfo, EMIT_1F, src);
280 vdecl[nr_decls].identity.usage = SVGA3D_DECLUSAGE_TEXCOORD;
281 vdecl[nr_decls].identity.type = SVGA3D_DECLTYPE_FLOAT1;
282 assert(vdecl[nr_decls].identity.usageIndex == 0);
283 offset += 4;
284 nr_decls++;
285 break;
286 case TGSI_SEMANTIC_POSITION:
287 /* generated internally, not a vertex shader output */
288 break;
289 default:
290 assert(0);
291 }
292 }
293
294 draw_compute_vertex_size(vinfo);
295
296 svga_render->vdecl_count = nr_decls;
297 for (i = 0; i < svga_render->vdecl_count; i++) {
298 vdecl[i].array.stride = offset;
299 }
300
301 any_change = memcmp(svga_render->vdecl, vdecl, sizeof(vdecl));
302
303 if (svga_have_vgpu10(svga)) {
304 enum pipe_error ret;
305
306 if (!any_change && svga_render->layout_id != SVGA3D_INVALID_ID) {
307 goto done;
308 }
309
310 if (svga_render->layout_id != SVGA3D_INVALID_ID) {
311 /* destroy old */
312 ret = SVGA3D_vgpu10_DestroyElementLayout(svga->swc,
313 svga_render->layout_id);
314 if (ret != PIPE_OK) {
315 svga_context_flush(svga, NULL);
316 ret = SVGA3D_vgpu10_DestroyElementLayout(svga->swc,
317 svga_render->layout_id);
318 assert(ret == PIPE_OK);
319 }
320
321 /**
322 * reset current layout id state after the element layout is
323 * destroyed, so that if a new layout has the same layout id, we
324 * will know to re-issue the SetInputLayout command.
325 */
326 if (svga->state.hw_draw.layout_id == svga_render->layout_id)
327 svga->state.hw_draw.layout_id = SVGA3D_INVALID_ID;
328
329 util_bitmask_clear(svga->input_element_object_id_bm,
330 svga_render->layout_id);
331 }
332
333 svga_render->layout_id =
334 svga_vdecl_to_input_element(svga, vdecl, nr_decls);
335
336 /* bind new */
337 if (svga->state.hw_draw.layout_id != svga_render->layout_id) {
338 ret = SVGA3D_vgpu10_SetInputLayout(svga->swc, svga_render->layout_id);
339 if (ret != PIPE_OK) {
340 svga_context_flush(svga, NULL);
341 ret = SVGA3D_vgpu10_SetInputLayout(svga->swc,
342 svga_render->layout_id);
343 assert(ret == PIPE_OK);
344 }
345
346 svga->state.hw_draw.layout_id = svga_render->layout_id;
347 }
348 }
349 else {
350 if (!any_change)
351 goto done;
352 }
353
354 memcpy(svga_render->vdecl, vdecl, sizeof(vdecl));
355 svga->swtnl.new_vdecl = TRUE;
356
357 done:
358 SVGA_STATS_TIME_POP(svga_sws(svga));
359 return PIPE_OK;
360 }
361
362
363 static enum pipe_error
update_swtnl_vdecl(struct svga_context * svga,unsigned dirty)364 update_swtnl_vdecl( struct svga_context *svga,
365 unsigned dirty )
366 {
367 return svga_swtnl_update_vdecl( svga );
368 }
369
370
371 struct svga_tracked_state svga_update_swtnl_vdecl =
372 {
373 "update draw module vdecl",
374 (SVGA_NEW_VS |
375 SVGA_NEW_FS),
376 update_swtnl_vdecl
377 };
378