draw_vs_exec.c revision 5f996e2b1d09dad64c088ccabb1a4a53ebfb8102 
1/**************************************************************************
2 *
3 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
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 TUNGSTEN GRAPHICS 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  * Authors:
30  *   Keith Whitwell <keith@tungstengraphics.com>
31  *   Brian Paul
32  */
33
34#include "util/u_math.h"
35#include "util/u_memory.h"
36#include "pipe/p_shader_tokens.h"
37
38#include "draw_private.h"
39#include "draw_context.h"
40#include "draw_vs.h"
41
42#include "tgsi/tgsi_parse.h"
43#include "tgsi/tgsi_scan.h"
44#include "tgsi/tgsi_exec.h"
45
46
47struct exec_vertex_shader {
48   struct draw_vertex_shader base;
49   struct tgsi_exec_machine *machine;
50};
51
52static struct exec_vertex_shader *exec_vertex_shader( struct draw_vertex_shader *vs )
53{
54   return (struct exec_vertex_shader *)vs;
55}
56
57
58/* Not required for run_linear.
59 */
60static void
61vs_exec_prepare( struct draw_vertex_shader *shader,
62		 struct draw_context *draw )
63{
64   struct exec_vertex_shader *evs = exec_vertex_shader(shader);
65
66   /* Specify the vertex program to interpret/execute.
67    * Avoid rebinding when possible.
68    */
69   if (evs->machine->Tokens != shader->state.tokens) {
70      tgsi_exec_machine_bind_shader(evs->machine,
71                                    shader->state.tokens,
72                                    draw->vs.num_samplers,
73                                    draw->vs.samplers);
74   }
75}
76
77
78
79
80/* Simplified vertex shader interface for the pt paths.  Given the
81 * complexity of code-generating all the above operations together,
82 * it's time to try doing all the other stuff separately.
83 */
84static void
85vs_exec_run_linear( struct draw_vertex_shader *shader,
86		    const float (*input)[4],
87		    float (*output)[4],
88                    const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
89                    const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
90		    unsigned count,
91		    unsigned input_stride,
92		    unsigned output_stride )
93{
94   struct exec_vertex_shader *evs = exec_vertex_shader(shader);
95   struct tgsi_exec_machine *machine = evs->machine;
96   unsigned int i, j;
97   unsigned slot;
98   boolean clamp_vertex_color = shader->draw->rasterizer->clamp_vertex_color;
99
100   tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
101                                  constants, const_size);
102
103   if (shader->info.uses_instanceid) {
104      unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INSTANCEID];
105      assert(i < Elements(machine->SystemValue));
106      machine->SystemValue[i][0] = shader->draw->instance_id;
107   }
108
109   for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
110      unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
111
112      /* Swizzle inputs.
113       */
114      for (j = 0; j < max_vertices; j++) {
115#if 0
116         debug_printf("%d) Input vert:\n", i + j);
117         for (slot = 0; slot < shader->info.num_inputs; slot++) {
118            debug_printf("\t%d: %f %f %f %f\n", slot,
119			 input[slot][0],
120			 input[slot][1],
121			 input[slot][2],
122			 input[slot][3]);
123         }
124#endif
125
126         for (slot = 0; slot < shader->info.num_inputs; slot++) {
127#if 0
128            assert(!util_is_inf_or_nan(input[slot][0]));
129            assert(!util_is_inf_or_nan(input[slot][1]));
130            assert(!util_is_inf_or_nan(input[slot][2]));
131            assert(!util_is_inf_or_nan(input[slot][3]));
132#endif
133            machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
134            machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
135            machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
136            machine->Inputs[slot].xyzw[3].f[j] = input[slot][3];
137         }
138
139	 input = (const float (*)[4])((const char *)input + input_stride);
140      }
141
142      tgsi_set_exec_mask(machine,
143                         1,
144                         max_vertices > 1,
145                         max_vertices > 2,
146                         max_vertices > 3);
147
148      /* run interpreter */
149      tgsi_exec_machine_run( machine );
150
151      /* Unswizzle all output results.
152       */
153      for (j = 0; j < max_vertices; j++) {
154         for (slot = 0; slot < shader->info.num_outputs; slot++) {
155            unsigned name = shader->info.output_semantic_name[slot];
156            if(clamp_vertex_color &&
157                  (name == TGSI_SEMANTIC_COLOR || name == TGSI_SEMANTIC_BCOLOR))
158            {
159               output[slot][0] = CLAMP(machine->Outputs[slot].xyzw[0].f[j], 0.0f, 1.0f);
160               output[slot][1] = CLAMP(machine->Outputs[slot].xyzw[1].f[j], 0.0f, 1.0f);
161               output[slot][2] = CLAMP(machine->Outputs[slot].xyzw[2].f[j], 0.0f, 1.0f);
162               output[slot][3] = CLAMP(machine->Outputs[slot].xyzw[3].f[j], 0.0f, 1.0f);
163            }
164            else
165            {
166               output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
167               output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
168               output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
169               output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
170            }
171         }
172
173#if 0
174	 debug_printf("%d) Post xform vert:\n", i + j);
175	 for (slot = 0; slot < shader->info.num_outputs; slot++) {
176	    debug_printf("\t%d: %f %f %f %f\n", slot,
177			 output[slot][0],
178			 output[slot][1],
179			 output[slot][2],
180			 output[slot][3]);
181            assert(!util_is_inf_or_nan(output[slot][0]));
182         }
183#endif
184
185	 output = (float (*)[4])((char *)output + output_stride);
186      }
187
188   }
189}
190
191
192
193
194static void
195vs_exec_delete( struct draw_vertex_shader *dvs )
196{
197   FREE((void*) dvs->state.tokens);
198   FREE( dvs );
199}
200
201
202struct draw_vertex_shader *
203draw_create_vs_exec(struct draw_context *draw,
204		    const struct pipe_shader_state *state)
205{
206   struct exec_vertex_shader *vs = CALLOC_STRUCT( exec_vertex_shader );
207
208   if (vs == NULL)
209      return NULL;
210
211   /* we make a private copy of the tokens */
212   vs->base.state.tokens = tgsi_dup_tokens(state->tokens);
213   if (!vs->base.state.tokens) {
214      FREE(vs);
215      return NULL;
216   }
217
218   tgsi_scan_shader(state->tokens, &vs->base.info);
219
220   vs->base.draw = draw;
221   vs->base.prepare = vs_exec_prepare;
222   vs->base.run_linear = vs_exec_run_linear;
223   vs->base.delete = vs_exec_delete;
224   vs->base.create_variant = draw_vs_create_variant_generic;
225   vs->machine = draw->vs.machine;
226
227   return &vs->base;
228}
229