src/gallium/auxiliary/draw/draw_vs_exec.c

   1 /**************************************************************************
   2  *
   3  * Copyright 2007 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   * Authors:
  30   *   Keith Whitwell <keithw@vmware.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
  47 struct exec_vertex_shader {
  48    struct draw_vertex_shader base;
  49    struct tgsi_exec_machine *machine;
  50 };
  51
  52
  53 static struct exec_vertex_shader *
  54 exec_vertex_shader(struct draw_vertex_shader *vs)
  55 {
  56    return (struct exec_vertex_shader *)vs;
  57 }
  58
  59
  60 /* Not required for run_linear.
  61  */
  62 static void
  63 vs_exec_prepare(struct draw_vertex_shader *shader,
  64                 struct draw_context *draw)
  65 {
  66    struct exec_vertex_shader *evs = exec_vertex_shader(shader);
  67
  68    debug_assert(!draw->llvm);
  69    /* Specify the vertex program to interpret/execute.
  70     * Avoid rebinding when possible.
  71     */
  72    if (evs->machine->Tokens != shader->state.tokens) {
  73       tgsi_exec_machine_bind_shader(evs->machine,
  74                                     shader->state.tokens,
  75                                     draw->vs.tgsi.sampler,
  76                                     draw->vs.tgsi.image,
  77                                     draw->vs.tgsi.buffer);
  78    }
  79 }
  80
  81
  82
  83 /**
  84  * Simplified vertex shader interface for the pt paths.  Given the
  85  * complexity of code-generating all the above operations together,
  86  * it's time to try doing all the other stuff separately.
  87  */
  88 static void
  89 vs_exec_run_linear(struct draw_vertex_shader *shader,
  90                    const float (*input)[4],
  91                    float (*output)[4],
  92                    const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
  93                     const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
  94                    unsigned count,
  95                    unsigned input_stride,
  96                    unsigned output_stride,
  97                    const unsigned *fetch_elts)
  98 {
  99    struct exec_vertex_shader *evs = exec_vertex_shader(shader);
 100    struct tgsi_exec_machine *machine = evs->machine;
 101    unsigned int i, j;
 102    unsigned slot;
 103    boolean clamp_vertex_color = shader->draw->rasterizer->clamp_vertex_color;
 104
 105    debug_assert(!shader->draw->llvm);
 106    tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
 107                                   constants, const_size);
 108
 109    if (shader->info.uses_instanceid) {
 110       unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INSTANCEID];
 111       assert(i < ARRAY_SIZE(machine->SystemValue));
 112       for (j = 0; j < TGSI_QUAD_SIZE; j++)
 113          machine->SystemValue[i].xyzw[0].i[j] = shader->draw->instance_id;
 114    }
 115
 116    for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
 117       unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
 118
 119       /* Swizzle inputs.
 120        */
 121       for (j = 0; j < max_vertices; j++) {
 122 #if 0
 123          debug_printf("%d) Input vert:\n", i + j);
 124          for (slot = 0; slot < shader->info.num_inputs; slot++) {
 125             debug_printf("\t%d: %f %f %f %f\n", slot,
 126                          input[slot][0],
 127                          input[slot][1],
 128                          input[slot][2],
 129                          input[slot][3]);
 130          }
 131 #endif
 132          int basevertex = shader->draw->pt.user.eltSize ? shader->draw->pt.user.eltBias : shader->draw->start_index;
 133
 134          if (shader->info.uses_vertexid) {
 135             unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_VERTEXID];
 136             assert(vid < ARRAY_SIZE(machine->SystemValue));
 137             machine->SystemValue[vid].xyzw[0].i[j] = fetch_elts ? fetch_elts[i + j] : (i + j + basevertex);
 138          }
 139          if (shader->info.uses_basevertex) {
 140             unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_BASEVERTEX];
 141             assert(vid < ARRAY_SIZE(machine->SystemValue));
 142             machine->SystemValue[vid].xyzw[0].i[j] = basevertex;
 143          }
 144          if (shader->info.uses_vertexid_nobase) {
 145             unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_VERTEXID_NOBASE];
 146             assert(vid < ARRAY_SIZE(machine->SystemValue));
 147             machine->SystemValue[vid].xyzw[0].i[j] = fetch_elts ? (fetch_elts[i + j] - basevertex) : (i + j);
 148          }
 149
 150          for (slot = 0; slot < shader->info.num_inputs; slot++) {
 151 #if 0
 152             assert(!util_is_inf_or_nan(input[slot][0]));
 153             assert(!util_is_inf_or_nan(input[slot][1]));
 154             assert(!util_is_inf_or_nan(input[slot][2]));
 155             assert(!util_is_inf_or_nan(input[slot][3]));
 156 #endif
 157             machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
 158             machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
 159             machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
 160             machine->Inputs[slot].xyzw[3].f[j] = input[slot][3];
 161          }
 162
 163          input = (const float (*)[4])((const char *)input + input_stride);
 164       }
 165
 166       machine->NonHelperMask = (1 << max_vertices) - 1;
 167       /* run interpreter */
 168       tgsi_exec_machine_run(machine, 0);
 169
 170       /* Unswizzle all output results.
 171        */
 172       for (j = 0; j < max_vertices; j++) {
 173          for (slot = 0; slot < shader->info.num_outputs; slot++) {
 174             enum tgsi_semantic name = shader->info.output_semantic_name[slot];
 175             if (clamp_vertex_color &&
 176                 (name == TGSI_SEMANTIC_COLOR || name == TGSI_SEMANTIC_BCOLOR)) {
 177                output[slot][0] = SATURATE(machine->Outputs[slot].xyzw[0].f[j]);
 178                output[slot][1] = SATURATE(machine->Outputs[slot].xyzw[1].f[j]);
 179                output[slot][2] = SATURATE(machine->Outputs[slot].xyzw[2].f[j]);
 180                output[slot][3] = SATURATE(machine->Outputs[slot].xyzw[3].f[j]);
 181             } else {
 182                output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
 183                output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
 184                output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
 185                output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
 186             }
 187          }
 188
 189 #if 0
 190          debug_printf("%d) Post xform vert:\n", i + j);
 191          for (slot = 0; slot < shader->info.num_outputs; slot++) {
 192             debug_printf("\t%d: %f %f %f %f\n", slot,
 193                          output[slot][0],
 194                          output[slot][1],
 195                          output[slot][2],
 196                          output[slot][3]);
 197             assert(!util_is_inf_or_nan(output[slot][0]));
 198          }
 199 #endif
 200
 201          output = (float (*)[4])((char *)output + output_stride);
 202       }
 203    }
 204 }
 205
 206
 207 static void
 208 vs_exec_delete(struct draw_vertex_shader *dvs)
 209 {
 210    FREE((void*) dvs->state.tokens);
 211    FREE(dvs);
 212 }
 213
 214
 215 struct draw_vertex_shader *
 216 draw_create_vs_exec(struct draw_context *draw,
 217                     const struct pipe_shader_state *state)
 218 {
 219    struct exec_vertex_shader *vs = CALLOC_STRUCT(exec_vertex_shader);
 220
 221    if (!vs)
 222       return NULL;
 223
 224    /* we make a private copy of the tokens */
 225    vs->base.state.tokens = tgsi_dup_tokens(state->tokens);
 226    if (!vs->base.state.tokens) {
 227       FREE(vs);
 228       return NULL;
 229    }
 230
 231    tgsi_scan_shader(state->tokens, &vs->base.info);
 232
 233    vs->base.state.type = state->type;
 234    vs->base.state.stream_output = state->stream_output;
 235    vs->base.draw = draw;
 236    vs->base.prepare = vs_exec_prepare;
 237    vs->base.run_linear = vs_exec_run_linear;
 238    vs->base.delete = vs_exec_delete;
 239    vs->base.create_variant = draw_vs_create_variant_generic;
 240    vs->machine = draw->vs.tgsi.machine;
 241
 242    return &vs->base;
 243 }