src/mesa/drivers/dri/i965/gen6_constant_state.c

   1 /*
   2  * Copyright © 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 DEALINGS
  21  * IN THE SOFTWARE.
  22  */
  23
  24 #include "brw_context.h"
  25 #include "brw_state.h"
  26 #include "brw_defines.h"
  27 #include "intel_batchbuffer.h"
  28 #include "intel_buffer_objects.h"
  29 #include "program/prog_parameter.h"
  30
  31 /**
  32  * Creates a streamed BO containing the push constants for the VS or GS on
  33  * gen6+.
  34  *
  35  * Push constants are constant values (such as GLSL uniforms) that are
  36  * pre-loaded into a shader stage's register space at thread spawn time.
  37  *
  38  * Not all GLSL uniforms will be uploaded as push constants: The hardware has
  39  * a limitation of 32 or 64 EU registers (256 or 512 floats) per stage to be
  40  * uploaded as push constants, while GL 4.4 requires at least 1024 components
  41  * to be usable for the VS.  Plus, currently we always use pull constants
  42  * instead of push constants when doing variable-index array access.
  43  *
  44  * See brw_curbe.c for the equivalent gen4/5 code.
  45  */
  46 void
  47 gen6_upload_push_constants(struct brw_context *brw,
  48                            const struct gl_program *prog,
  49                            const struct brw_stage_prog_data *prog_data,
  50                            struct brw_stage_state *stage_state)
  51 {
  52    const struct gen_device_info *devinfo = &brw->screen->devinfo;
  53    struct gl_context *ctx = &brw->ctx;
  54
  55    if (prog_data->nr_params == 0) {
  56       stage_state->push_const_size = 0;
  57    } else {
  58       /* Updates the ParamaterValues[i] pointers for all parameters of the
  59        * basic type of PROGRAM_STATE_VAR.
  60        */
  61       /* XXX: Should this happen somewhere before to get our state flag set? */
  62       if (prog)
  63          _mesa_load_state_parameters(ctx, prog->Parameters);
  64
  65       int i;
  66       const int size = prog_data->nr_params * sizeof(gl_constant_value);
  67       gl_constant_value *param;
  68       if (devinfo->gen >= 8 || devinfo->is_haswell) {
  69          param = intel_upload_space(brw, size, 32,
  70                                     &stage_state->push_const_bo,
  71                                     &stage_state->push_const_offset);
  72       } else {
  73          param = brw_state_batch(brw, size, 32,
  74                                  &stage_state->push_const_offset);
  75       }
  76
  77       STATIC_ASSERT(sizeof(gl_constant_value) == sizeof(float));
  78
  79       /* _NEW_PROGRAM_CONSTANTS
  80        *
  81        * Also _NEW_TRANSFORM -- we may reference clip planes other than as a
  82        * side effect of dereferencing uniforms, so _NEW_PROGRAM_CONSTANTS
  83        * wouldn't be set for them.
  84        */
  85       for (i = 0; i < prog_data->nr_params; i++) {
  86          param[i] = *prog_data->param[i];
  87       }
  88
  89       if (0) {
  90          fprintf(stderr, "%s constants:\n",
  91                  _mesa_shader_stage_to_string(stage_state->stage));
  92          for (i = 0; i < prog_data->nr_params; i++) {
  93             if ((i & 7) == 0)
  94                fprintf(stderr, "g%d: ",
  95                        prog_data->dispatch_grf_start_reg + i / 8);
  96             fprintf(stderr, "%8f ", param[i].f);
  97             if ((i & 7) == 7)
  98                fprintf(stderr, "\n");
  99          }
 100          if ((i & 7) != 0)
 101             fprintf(stderr, "\n");
 102          fprintf(stderr, "\n");
 103       }
 104
 105       stage_state->push_const_size = ALIGN(prog_data->nr_params, 8) / 8;
 106       /* We can only push 32 registers of constants at a time. */
 107
 108       /* From the SNB PRM (vol2, part 1, section 3.2.1.4: 3DSTATE_CONSTANT_VS:
 109        *
 110        *     "The sum of all four read length fields (each incremented to
 111        *      represent the actual read length) must be less than or equal to
 112        *      32"
 113        *
 114        * From the IVB PRM (vol2, part 1, section 3.2.1.3: 3DSTATE_CONSTANT_VS:
 115        *
 116        *     "The sum of all four read length fields must be less than or
 117        *      equal to the size of 64"
 118        *
 119        * The other shader stages all match the VS's limits.
 120        */
 121       assert(stage_state->push_const_size <= 32);
 122    }
 123
 124    stage_state->push_constants_dirty = true;
 125 }
 126
 127
 128 /**
 129  * Creates a temporary BO containing the pull constant data for the shader
 130  * stage, and the SURFACE_STATE struct that points at it.
 131  *
 132  * Pull constants are GLSL uniforms (and other constant data) beyond what we
 133  * could fit as push constants, or that have variable-index array access
 134  * (which is easiest to support using pull constants, and avoids filling
 135  * register space with mostly-unused data).
 136  *
 137  * Compare this path to brw_curbe.c for gen4/5 push constants, and
 138  * gen6_vs_state.c for gen6+ push constants.
 139  */
 140 void
 141 brw_upload_pull_constants(struct brw_context *brw,
 142                           GLbitfield64 brw_new_constbuf,
 143                           const struct gl_program *prog,
 144                           struct brw_stage_state *stage_state,
 145                           const struct brw_stage_prog_data *prog_data)
 146 {
 147    unsigned i;
 148    uint32_t surf_index = prog_data->binding_table.pull_constants_start;
 149
 150    if (!prog_data->nr_pull_params) {
 151       if (stage_state->surf_offset[surf_index]) {
 152          stage_state->surf_offset[surf_index] = 0;
 153          brw->ctx.NewDriverState |= brw_new_constbuf;
 154       }
 155       return;
 156    }
 157
 158    /* Updates the ParamaterValues[i] pointers for all parameters of the
 159     * basic type of PROGRAM_STATE_VAR.
 160     */
 161    _mesa_load_state_parameters(&brw->ctx, prog->Parameters);
 162
 163    /* BRW_NEW_*_PROG_DATA | _NEW_PROGRAM_CONSTANTS */
 164    uint32_t size = prog_data->nr_pull_params * 4;
 165    struct brw_bo *const_bo = NULL;
 166    uint32_t const_offset;
 167    gl_constant_value *constants = intel_upload_space(brw, size, 64,
 168                                                      &const_bo, &const_offset);
 169
 170    STATIC_ASSERT(sizeof(gl_constant_value) == sizeof(float));
 171
 172    for (i = 0; i < prog_data->nr_pull_params; i++) {
 173       constants[i] = *prog_data->pull_param[i];
 174    }
 175
 176    if (0) {
 177       for (i = 0; i < ALIGN(prog_data->nr_pull_params, 4) / 4; i++) {
 178          const gl_constant_value *row = &constants[i * 4];
 179          fprintf(stderr, "const surface %3d: %4.3f %4.3f %4.3f %4.3f\n",
 180                  i, row[0].f, row[1].f, row[2].f, row[3].f);
 181       }
 182    }
 183
 184    brw_create_constant_surface(brw, const_bo, const_offset, size,
 185                                &stage_state->surf_offset[surf_index]);
 186    brw_bo_unreference(const_bo);
 187
 188    brw->ctx.NewDriverState |= brw_new_constbuf;
 189 }