1 /**************************************************************************
2 *
3 * Copyright 2003 VMware, Inc.
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28 /*
29 * Render unclipped vertex buffers by emitting vertices directly to
30 * dma buffers. Use strip/fan hardware acceleration where possible.
31 *
32 */
33 #include "main/glheader.h"
34 #include "main/context.h"
35 #include "main/macros.h"
36 #include "main/imports.h"
37 #include "main/mtypes.h"
38 #include "main/enums.h"
39
40 #include "math/m_xform.h"
41
42 #include "tnl/t_context.h"
43 #include "tnl/t_vertex.h"
44 #include "tnl/t_pipeline.h"
45
46 #include "intel_screen.h"
47 #include "intel_context.h"
48 #include "intel_tris.h"
49 #include "intel_batchbuffer.h"
50 #include "intel_reg.h"
51
52 /*
53 * Render unclipped vertex buffers by emitting vertices directly to
54 * dma buffers. Use strip/fan hardware primitives where possible.
55 * Try to simulate missing primitives with indexed vertices.
56 */
57 #define HAVE_POINTS 1
58 #define HAVE_LINES 1
59 #define HAVE_LINE_STRIPS 1
60 #define HAVE_TRIANGLES 1
61 #define HAVE_TRI_STRIPS 1
62 #define HAVE_TRI_FANS 1
63 #define HAVE_POLYGONS 1
64
65 #define HAVE_ELTS 0
66
67 static const uint32_t hw_prim[GL_POLYGON + 1] = {
68 [GL_POINTS] = PRIM3D_POINTLIST,
69 [GL_LINES ] = PRIM3D_LINELIST,
70 [GL_LINE_LOOP] = PRIM3D_LINESTRIP,
71 [GL_LINE_STRIP] = PRIM3D_LINESTRIP,
72 [GL_TRIANGLES] = PRIM3D_TRILIST,
73 [GL_TRIANGLE_STRIP] = PRIM3D_TRISTRIP,
74 [GL_TRIANGLE_FAN] = PRIM3D_TRIFAN,
75 [GL_QUADS] = 0,
76 [GL_QUAD_STRIP] = 0,
77 [GL_POLYGON] = PRIM3D_POLY,
78 };
79
80 static const GLenum reduced_prim[GL_POLYGON + 1] = {
81 [GL_POINTS] = GL_POINTS,
82 [GL_LINES] = GL_LINES,
83 [GL_LINE_LOOP] = GL_LINES,
84 [GL_LINE_STRIP] = GL_LINES,
85 [GL_TRIANGLES] = GL_TRIANGLES,
86 [GL_TRIANGLE_STRIP] = GL_TRIANGLES,
87 [GL_TRIANGLE_FAN] = GL_TRIANGLES,
88 [GL_QUADS] = GL_TRIANGLES,
89 [GL_QUAD_STRIP] = GL_TRIANGLES,
90 [GL_POLYGON] = GL_TRIANGLES,
91 };
92
93 static const int scale_prim[GL_POLYGON + 1] = {
94 [GL_POINTS] = 1,
95 [GL_LINES] = 1,
96 [GL_LINE_LOOP] = 2,
97 [GL_LINE_STRIP] = 2,
98 [GL_TRIANGLES] = 1,
99 [GL_TRIANGLE_STRIP] = 3,
100 [GL_TRIANGLE_FAN] = 3,
101 [GL_QUADS] = 0, /* fallback case */
102 [GL_QUAD_STRIP] = 0, /* fallback case */
103 [GL_POLYGON] = 3,
104 };
105
106
107 static void
intelDmaPrimitive(struct intel_context * intel,GLenum prim)108 intelDmaPrimitive(struct intel_context *intel, GLenum prim)
109 {
110 if (0)
111 fprintf(stderr, "%s %s\n", __func__, _mesa_enum_to_string(prim));
112 INTEL_FIREVERTICES(intel);
113 intel->vtbl.reduced_primitive_state(intel, reduced_prim[prim]);
114 intel_set_prim(intel, hw_prim[prim]);
115 }
116
117 #define INTEL_NO_VBO_STATE_RESERVED 1500
118
intel_get_vb_max(struct intel_context * intel)119 static inline GLuint intel_get_vb_max(struct intel_context *intel)
120 {
121 GLuint ret;
122
123 if (intel->intelScreen->no_vbo) {
124 ret = intel->batch.bo->size - INTEL_NO_VBO_STATE_RESERVED;
125 } else
126 ret = INTEL_VB_SIZE;
127 ret /= (intel->vertex_size * 4);
128 return ret;
129 }
130
intel_get_current_max(struct intel_context * intel)131 static inline GLuint intel_get_current_max(struct intel_context *intel)
132 {
133 GLuint ret;
134
135 if (intel->intelScreen->no_vbo) {
136 ret = intel_batchbuffer_space(intel);
137 ret = ret <= INTEL_NO_VBO_STATE_RESERVED ? 0 : ret - INTEL_NO_VBO_STATE_RESERVED;
138 } else
139 ret = (INTEL_VB_SIZE - intel->prim.current_offset);
140
141 return ret / (intel->vertex_size * 4);
142 }
143
144 #define LOCAL_VARS struct intel_context *intel = intel_context(ctx)
145 #define INIT( prim ) \
146 do { \
147 intelDmaPrimitive( intel, prim ); \
148 } while (0)
149
150 #define FLUSH() INTEL_FIREVERTICES(intel)
151
152 #define GET_SUBSEQUENT_VB_MAX_VERTS() intel_get_vb_max(intel)
153 #define GET_CURRENT_VB_MAX_VERTS() intel_get_current_max(intel)
154
155 #define ALLOC_VERTS(nr) intel_get_prim_space(intel, nr)
156
157 #define EMIT_VERTS( ctx, j, nr, buf ) \
158 _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf )
159
160 #define TAG(x) intel_##x
161 #include "tnl_dd/t_dd_dmatmp.h"
162
163
164 /**********************************************************************/
165 /* Render pipeline stage */
166 /**********************************************************************/
167
168 /* Heuristic to choose between the two render paths:
169 */
170 static bool
choose_render(struct intel_context * intel,struct vertex_buffer * VB)171 choose_render(struct intel_context *intel, struct vertex_buffer *VB)
172 {
173 int vertsz = intel->vertex_size;
174 int cost_render = 0;
175 int cost_fallback = 0;
176 int nr_prims = 0;
177 int nr_rprims = 0;
178 int nr_rverts = 0;
179 int rprim = intel->reduced_primitive;
180 int i = 0;
181
182 for (i = 0; i < VB->PrimitiveCount; i++) {
183 GLuint prim = VB->Primitive[i].mode;
184 GLuint length = VB->Primitive[i].count;
185
186 if (!length)
187 continue;
188
189 nr_prims++;
190 nr_rverts += length * scale_prim[prim & PRIM_MODE_MASK];
191
192 if (reduced_prim[prim & PRIM_MODE_MASK] != rprim) {
193 nr_rprims++;
194 rprim = reduced_prim[prim & PRIM_MODE_MASK];
195 }
196 }
197
198 /* One point for each generated primitive:
199 */
200 cost_render = nr_prims;
201 cost_fallback = nr_rprims;
202
203 /* One point for every 1024 dwords (4k) of dma:
204 */
205 cost_render += (vertsz * i) / 1024;
206 cost_fallback += (vertsz * nr_rverts) / 1024;
207
208 if (0)
209 fprintf(stderr, "cost render: %d fallback: %d\n",
210 cost_render, cost_fallback);
211
212 if (cost_render > cost_fallback)
213 return false;
214
215 return true;
216 }
217
218
219 static GLboolean
intel_run_render(struct gl_context * ctx,struct tnl_pipeline_stage * stage)220 intel_run_render(struct gl_context * ctx, struct tnl_pipeline_stage *stage)
221 {
222 struct intel_context *intel = intel_context(ctx);
223 TNLcontext *tnl = TNL_CONTEXT(ctx);
224 struct vertex_buffer *VB = &tnl->vb;
225 GLuint i;
226
227 intel->vtbl.render_prevalidate( intel );
228
229 /* Don't handle clipping or indexed vertices.
230 */
231 if (intel->RenderIndex != 0 ||
232 !intel_validate_render(ctx, VB) || !choose_render(intel, VB)) {
233 return true;
234 }
235
236 tnl->clipspace.new_inputs |= VERT_BIT_POS;
237
238 tnl->Driver.Render.Start(ctx);
239
240 for (i = 0; i < VB->PrimitiveCount; i++) {
241 GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
242 GLuint start = VB->Primitive[i].start;
243 GLuint length = VB->Primitive[i].count;
244
245 if (!length)
246 continue;
247
248 intel_render_tab_verts[prim & PRIM_MODE_MASK] (ctx, start,
249 length, prim);
250 }
251
252 tnl->Driver.Render.Finish(ctx);
253
254 INTEL_FIREVERTICES(intel);
255
256 return false; /* finished the pipe */
257 }
258
259 static const struct tnl_pipeline_stage _intel_render_stage = {
260 "intel render",
261 NULL,
262 NULL,
263 NULL,
264 NULL,
265 intel_run_render /* run */
266 };
267
268 const struct tnl_pipeline_stage *intel_pipeline[] = {
269 &_tnl_vertex_transform_stage,
270 &_tnl_normal_transform_stage,
271 &_tnl_lighting_stage,
272 &_tnl_fog_coordinate_stage,
273 &_tnl_texgen_stage,
274 &_tnl_texture_transform_stage,
275 &_tnl_point_attenuation_stage,
276 &_tnl_vertex_program_stage,
277 #if 1
278 &_intel_render_stage, /* ADD: unclipped rastersetup-to-dma */
279 #endif
280 &_tnl_render_stage,
281 0,
282 };
283