src/mesa/drivers/dri/i965/gen6_gs_state.c

   1 /*
   2  * Copyright © 2009 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  * Authors:
  24  *    Eric Anholt <eric@anholt.net>
  25  *
  26  */
  27
  28 #include "brw_context.h"
  29 #include "brw_state.h"
  30 #include "brw_defines.h"
  31 #include "intel_batchbuffer.h"
  32 #include "main/shaderapi.h"
  33
  34 static void
  35 gen6_upload_gs_push_constants(struct brw_context *brw)
  36 {
  37    struct brw_stage_state *stage_state = &brw->gs.base;
  38
  39    /* BRW_NEW_GEOMETRY_PROGRAM */
  40    const struct brw_program *gp = brw_program_const(brw->geometry_program);
  41
  42    if (gp) {
  43       /* BRW_NEW_GS_PROG_DATA */
  44       struct brw_stage_prog_data *prog_data = brw->gs.base.prog_data;
  45
  46       _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_GEOMETRY);
  47       gen6_upload_push_constants(brw, &gp->program, prog_data, stage_state,
  48                                  AUB_TRACE_VS_CONSTANTS);
  49    }
  50
  51    if (brw->gen >= 7)
  52       gen7_upload_constant_state(brw, stage_state, gp, _3DSTATE_CONSTANT_GS);
  53 }
  54
  55 const struct brw_tracked_state gen6_gs_push_constants = {
  56    .dirty = {
  57       .mesa  = _NEW_PROGRAM_CONSTANTS |
  58                _NEW_TRANSFORM,
  59       .brw   = BRW_NEW_BATCH |
  60                BRW_NEW_BLORP |
  61                BRW_NEW_GEOMETRY_PROGRAM |
  62                BRW_NEW_GS_PROG_DATA |
  63                BRW_NEW_PUSH_CONSTANT_ALLOCATION,
  64    },
  65    .emit = gen6_upload_gs_push_constants,
  66 };
  67
  68 static void
  69 upload_gs_state_for_tf(struct brw_context *brw)
  70 {
  71    const struct gen_device_info *devinfo = &brw->screen->devinfo;
  72
  73    BEGIN_BATCH(7);
  74    OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
  75    OUT_BATCH(brw->ff_gs.prog_offset);
  76    OUT_BATCH(GEN6_GS_SPF_MODE | GEN6_GS_VECTOR_MASK_ENABLE);
  77    OUT_BATCH(0); /* no scratch space */
  78    OUT_BATCH((2 << GEN6_GS_DISPATCH_START_GRF_SHIFT) |
  79              (brw->ff_gs.prog_data->urb_read_length << GEN6_GS_URB_READ_LENGTH_SHIFT));
  80    OUT_BATCH(((devinfo->max_gs_threads - 1) << GEN6_GS_MAX_THREADS_SHIFT) |
  81              GEN6_GS_STATISTICS_ENABLE |
  82              GEN6_GS_SO_STATISTICS_ENABLE |
  83              GEN6_GS_RENDERING_ENABLE);
  84    OUT_BATCH(GEN6_GS_SVBI_PAYLOAD_ENABLE |
  85              GEN6_GS_SVBI_POSTINCREMENT_ENABLE |
  86              (brw->ff_gs.prog_data->svbi_postincrement_value <<
  87               GEN6_GS_SVBI_POSTINCREMENT_VALUE_SHIFT) |
  88              GEN6_GS_ENABLE);
  89    ADVANCE_BATCH();
  90 }
  91
  92 static void
  93 upload_gs_state(struct brw_context *brw)
  94 {
  95    const struct gen_device_info *devinfo = &brw->screen->devinfo;
  96    /* BRW_NEW_GEOMETRY_PROGRAM */
  97    bool active = brw->geometry_program;
  98    /* BRW_NEW_GS_PROG_DATA */
  99    const struct brw_stage_state *stage_state = &brw->gs.base;
 100    const struct brw_stage_prog_data *prog_data = stage_state->prog_data;
 101    const struct brw_vue_prog_data *vue_prog_data =
 102       brw_vue_prog_data(stage_state->prog_data);
 103
 104    if (!active || stage_state->push_const_size == 0) {
 105       /* Disable the push constant buffers. */
 106       BEGIN_BATCH(5);
 107       OUT_BATCH(_3DSTATE_CONSTANT_GS << 16 | (5 - 2));
 108       OUT_BATCH(0);
 109       OUT_BATCH(0);
 110       OUT_BATCH(0);
 111       OUT_BATCH(0);
 112       ADVANCE_BATCH();
 113    } else {
 114       BEGIN_BATCH(5);
 115       OUT_BATCH(_3DSTATE_CONSTANT_GS << 16 |
 116                 GEN6_CONSTANT_BUFFER_0_ENABLE |
 117                 (5 - 2));
 118       /* Pointer to the GS constant buffer.  Covered by the set of
 119        * state flags from gen6_upload_vs_constants
 120        */
 121       OUT_BATCH(stage_state->push_const_offset +
 122                 stage_state->push_const_size - 1);
 123       OUT_BATCH(0);
 124       OUT_BATCH(0);
 125       OUT_BATCH(0);
 126       ADVANCE_BATCH();
 127    }
 128
 129    if (active) {
 130       BEGIN_BATCH(7);
 131       OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
 132       OUT_BATCH(stage_state->prog_offset);
 133
 134       /* GEN6_GS_SPF_MODE and GEN6_GS_VECTOR_MASK_ENABLE are enabled as it
 135        * was previously done for gen6.
 136        *
 137        * TODO: test with both disabled to see if the HW is behaving
 138        * as expected, like in gen7.
 139        */
 140       OUT_BATCH(GEN6_GS_SPF_MODE | GEN6_GS_VECTOR_MASK_ENABLE |
 141                 ((ALIGN(stage_state->sampler_count, 4)/4) <<
 142                  GEN6_GS_SAMPLER_COUNT_SHIFT) |
 143                 ((prog_data->binding_table.size_bytes / 4) <<
 144                  GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
 145
 146       if (prog_data->total_scratch) {
 147          OUT_RELOC(stage_state->scratch_bo,
 148                    I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
 149                    ffs(stage_state->per_thread_scratch) - 11);
 150       } else {
 151          OUT_BATCH(0); /* no scratch space */
 152       }
 153
 154       OUT_BATCH((vue_prog_data->urb_read_length <<
 155                  GEN6_GS_URB_READ_LENGTH_SHIFT) |
 156                 (0 << GEN6_GS_URB_ENTRY_READ_OFFSET_SHIFT) |
 157                 (prog_data->dispatch_grf_start_reg <<
 158                  GEN6_GS_DISPATCH_START_GRF_SHIFT));
 159
 160       OUT_BATCH(((devinfo->max_gs_threads - 1) << GEN6_GS_MAX_THREADS_SHIFT) |
 161                 GEN6_GS_STATISTICS_ENABLE |
 162                 GEN6_GS_SO_STATISTICS_ENABLE |
 163                 GEN6_GS_RENDERING_ENABLE);
 164
 165       if (brw->geometry_program->info.has_transform_feedback_varyings) {
 166          /* GEN6_GS_REORDER is equivalent to GEN7_GS_REORDER_TRAILING
 167           * in gen7. SNB and IVB specs are the same regarding the reordering of
 168           * TRISTRIP/TRISTRIP_REV vertices and triangle orientation, so we do
 169           * the same thing in both generations. For more details, see the
 170           * comment in gen7_gs_state.c
 171           */
 172          OUT_BATCH(GEN6_GS_REORDER |
 173                    GEN6_GS_SVBI_PAYLOAD_ENABLE |
 174                    GEN6_GS_ENABLE);
 175       } else {
 176          OUT_BATCH(GEN6_GS_REORDER | GEN6_GS_ENABLE);
 177       }
 178       ADVANCE_BATCH();
 179    } else if (brw->ff_gs.prog_active) {
 180       /* In gen6, transform feedback for the VS stage is done with an ad-hoc GS
 181        * program. This function provides the needed 3DSTATE_GS for this.
 182        */
 183       upload_gs_state_for_tf(brw);
 184    } else {
 185       /* No GS function required */
 186       BEGIN_BATCH(7);
 187       OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
 188       OUT_BATCH(0); /* prog_bo */
 189       OUT_BATCH((0 << GEN6_GS_SAMPLER_COUNT_SHIFT) |
 190                 (0 << GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
 191       OUT_BATCH(0); /* scratch space base offset */
 192       OUT_BATCH((1 << GEN6_GS_DISPATCH_START_GRF_SHIFT) |
 193                 (0 << GEN6_GS_URB_READ_LENGTH_SHIFT) |
 194                 (0 << GEN6_GS_URB_ENTRY_READ_OFFSET_SHIFT));
 195       OUT_BATCH((0 << GEN6_GS_MAX_THREADS_SHIFT) |
 196                 GEN6_GS_STATISTICS_ENABLE |
 197                 GEN6_GS_RENDERING_ENABLE);
 198                 OUT_BATCH(0);
 199       ADVANCE_BATCH();
 200    }
 201    brw->gs.enabled = active;
 202 }
 203
 204 const struct brw_tracked_state gen6_gs_state = {
 205    .dirty = {
 206       .mesa  = _NEW_PROGRAM_CONSTANTS |
 207                _NEW_TRANSFORM,
 208       .brw   = BRW_NEW_BATCH |
 209                BRW_NEW_BLORP |
 210                BRW_NEW_CONTEXT |
 211                BRW_NEW_FF_GS_PROG_DATA |
 212                BRW_NEW_GEOMETRY_PROGRAM |
 213                BRW_NEW_GS_PROG_DATA |
 214                BRW_NEW_PUSH_CONSTANT_ALLOCATION,
 215    },
 216    .emit = upload_gs_state,
 217 };