src/mesa/drivers/dri/i965/brw_cs.c

   1 /*
   2  * Copyright (c) 2014 - 2015 Intel Corporation
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
  21  * DEALINGS IN THE SOFTWARE.
  22  */
  23
  24 #include "util/ralloc.h"
  25 #include "brw_context.h"
  26 #include "brw_cs.h"
  27 #include "brw_eu.h"
  28 #include "brw_wm.h"
  29 #include "brw_shader.h"
  30 #include "intel_mipmap_tree.h"
  31 #include "brw_state.h"
  32 #include "intel_batchbuffer.h"
  33 #include "glsl/nir/nir.h"
  34
  35 static bool
  36 brw_codegen_cs_prog(struct brw_context *brw,
  37                     struct gl_shader_program *prog,
  38                     struct brw_compute_program *cp,
  39                     struct brw_cs_prog_key *key)
  40 {
  41    struct gl_context *ctx = &brw->ctx;
  42    const GLuint *program;
  43    void *mem_ctx = ralloc_context(NULL);
  44    GLuint program_size;
  45    struct brw_cs_prog_data prog_data;
  46    bool start_busy = false;
  47    double start_time = 0;
  48
  49    struct brw_shader *cs =
  50       (struct brw_shader *) prog->_LinkedShaders[MESA_SHADER_COMPUTE];
  51    assert (cs);
  52
  53    memset(&prog_data, 0, sizeof(prog_data));
  54
  55    /* Allocate the references to the uniforms that will end up in the
  56     * prog_data associated with the compiled program, and which will be freed
  57     * by the state cache.
  58     */
  59    int param_count = cp->program.Base.nir->num_uniforms;
  60
  61    /* The backend also sometimes adds params for texture size. */
  62    param_count += 2 * ctx->Const.Program[MESA_SHADER_COMPUTE].MaxTextureImageUnits;
  63    prog_data.base.param =
  64       rzalloc_array(NULL, const gl_constant_value *, param_count);
  65    prog_data.base.pull_param =
  66       rzalloc_array(NULL, const gl_constant_value *, param_count);
  67    prog_data.base.image_param =
  68       rzalloc_array(NULL, struct brw_image_param, cs->base.NumImages);
  69    prog_data.base.nr_params = param_count;
  70    prog_data.base.nr_image_params = cs->base.NumImages;
  71
  72    if (unlikely(brw->perf_debug)) {
  73       start_busy = (brw->batch.last_bo &&
  74                     drm_intel_bo_busy(brw->batch.last_bo));
  75       start_time = get_time();
  76    }
  77
  78    program = brw_cs_emit(brw, mem_ctx, key, &prog_data,
  79                          &cp->program, prog, &program_size);
  80    if (program == NULL) {
  81       ralloc_free(mem_ctx);
  82       return false;
  83    }
  84
  85    if (unlikely(brw->perf_debug) && cs) {
  86       if (cs->compiled_once) {
  87          _mesa_problem(&brw->ctx, "CS programs shouldn't need recompiles");
  88       }
  89       cs->compiled_once = true;
  90
  91       if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) {
  92          perf_debug("CS compile took %.03f ms and stalled the GPU\n",
  93                     (get_time() - start_time) * 1000);
  94       }
  95    }
  96
  97    if (prog_data.base.total_scratch) {
  98       brw_get_scratch_bo(brw, &brw->cs.base.scratch_bo,
  99                          prog_data.base.total_scratch * brw->max_cs_threads);
 100    }
 101
 102    if (unlikely(INTEL_DEBUG & DEBUG_CS))
 103       fprintf(stderr, "\n");
 104
 105    brw_upload_cache(&brw->cache, BRW_CACHE_CS_PROG,
 106                     key, sizeof(*key),
 107                     program, program_size,
 108                     &prog_data, sizeof(prog_data),
 109                     &brw->cs.base.prog_offset, &brw->cs.prog_data);
 110    ralloc_free(mem_ctx);
 111
 112    return true;
 113 }
 114
 115
 116 static void
 117 brw_cs_populate_key(struct brw_context *brw, struct brw_cs_prog_key *key)
 118 {
 119    struct gl_context *ctx = &brw->ctx;
 120    /* BRW_NEW_COMPUTE_PROGRAM */
 121    const struct brw_compute_program *cp =
 122       (struct brw_compute_program *) brw->compute_program;
 123    const struct gl_program *prog = (struct gl_program *) cp;
 124
 125    memset(key, 0, sizeof(*key));
 126
 127    /* _NEW_TEXTURE */
 128    brw_populate_sampler_prog_key_data(ctx, prog, brw->cs.base.sampler_count,
 129                                       &key->tex);
 130
 131    /* The unique compute program ID */
 132    key->program_string_id = cp->id;
 133 }
 134
 135
 136 void
 137 brw_upload_cs_prog(struct brw_context *brw)
 138 {
 139    struct gl_context *ctx = &brw->ctx;
 140    struct brw_cs_prog_key key;
 141    struct brw_compute_program *cp = (struct brw_compute_program *)
 142       brw->compute_program;
 143
 144    if (!cp)
 145       return;
 146
 147    if (!brw_state_dirty(brw, _NEW_TEXTURE, BRW_NEW_COMPUTE_PROGRAM))
 148       return;
 149
 150    brw->cs.base.sampler_count =
 151       _mesa_fls(ctx->ComputeProgram._Current->Base.SamplersUsed);
 152
 153    brw_cs_populate_key(brw, &key);
 154
 155    if (!brw_search_cache(&brw->cache, BRW_CACHE_CS_PROG,
 156                          &key, sizeof(key),
 157                          &brw->cs.base.prog_offset, &brw->cs.prog_data)) {
 158       bool success =
 159          brw_codegen_cs_prog(brw,
 160                              ctx->Shader.CurrentProgram[MESA_SHADER_COMPUTE],
 161                              cp, &key);
 162       (void) success;
 163       assert(success);
 164    }
 165    brw->cs.base.prog_data = &brw->cs.prog_data->base;
 166 }
 167
 168
 169 bool
 170 brw_cs_precompile(struct gl_context *ctx,
 171                   struct gl_shader_program *shader_prog,
 172                   struct gl_program *prog)
 173 {
 174    struct brw_context *brw = brw_context(ctx);
 175    struct brw_cs_prog_key key;
 176
 177    struct gl_compute_program *cp = (struct gl_compute_program *) prog;
 178    struct brw_compute_program *bcp = brw_compute_program(cp);
 179
 180    memset(&key, 0, sizeof(key));
 181    key.program_string_id = bcp->id;
 182
 183    brw_setup_tex_for_precompile(brw, &key.tex, prog);
 184
 185    uint32_t old_prog_offset = brw->cs.base.prog_offset;
 186    struct brw_cs_prog_data *old_prog_data = brw->cs.prog_data;
 187
 188    bool success = brw_codegen_cs_prog(brw, shader_prog, bcp, &key);
 189
 190    brw->cs.base.prog_offset = old_prog_offset;
 191    brw->cs.prog_data = old_prog_data;
 192
 193    return success;
 194 }