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_geometry_program *gp =
  41       (struct brw_geometry_program *) brw->geometry_program;
  42
  43    if (gp) {
  44       /* BRW_NEW_GS_PROG_DATA */
  45       struct brw_stage_prog_data *prog_data = brw->gs.base.prog_data;
  46
  47       _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_GEOMETRY);
  48       gen6_upload_push_constants(brw, &gp->program, prog_data, stage_state,
  49                                  AUB_TRACE_VS_CONSTANTS);
  50    }
  51
  52    if (brw->gen >= 7)
  53       gen7_upload_constant_state(brw, stage_state, gp, _3DSTATE_CONSTANT_GS);
  54 }
  55
  56 const struct brw_tracked_state gen6_gs_push_constants = {
  57    .dirty = {
  58       .mesa  = _NEW_PROGRAM_CONSTANTS |
  59                _NEW_TRANSFORM,
  60       .brw   = BRW_NEW_BATCH |
  61                BRW_NEW_BLORP |
  62                BRW_NEW_GEOMETRY_PROGRAM |
  63                BRW_NEW_GS_PROG_DATA |
  64                BRW_NEW_PUSH_CONSTANT_ALLOCATION,
  65    },
  66    .emit = gen6_upload_gs_push_constants,
  67 };
  68
  69 static void
  70 upload_gs_state_for_tf(struct brw_context *brw)
  71 {
  72    const struct gen_device_info *devinfo = &brw->screen->devinfo;
  73
  74    BEGIN_BATCH(7);
  75    OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
  76    OUT_BATCH(brw->ff_gs.prog_offset);
  77    OUT_BATCH(GEN6_GS_SPF_MODE | GEN6_GS_VECTOR_MASK_ENABLE);
  78    OUT_BATCH(0); /* no scratch space */
  79    OUT_BATCH((2 << GEN6_GS_DISPATCH_START_GRF_SHIFT) |
  80              (brw->ff_gs.prog_data->urb_read_length << GEN6_GS_URB_READ_LENGTH_SHIFT));
  81    OUT_BATCH(((devinfo->max_gs_threads - 1) << GEN6_GS_MAX_THREADS_SHIFT) |
  82              GEN6_GS_STATISTICS_ENABLE |
  83              GEN6_GS_SO_STATISTICS_ENABLE |
  84              GEN6_GS_RENDERING_ENABLE);
  85    OUT_BATCH(GEN6_GS_SVBI_PAYLOAD_ENABLE |
  86              GEN6_GS_SVBI_POSTINCREMENT_ENABLE |
  87              (brw->ff_gs.prog_data->svbi_postincrement_value <<
  88               GEN6_GS_SVBI_POSTINCREMENT_VALUE_SHIFT) |
  89              GEN6_GS_ENABLE);
  90    ADVANCE_BATCH();
  91 }
  92
  93 static void
  94 upload_gs_state(struct brw_context *brw)
  95 {
  96    const struct gen_device_info *devinfo = &brw->screen->devinfo;
  97    /* BRW_NEW_GEOMETRY_PROGRAM */
  98    bool active = brw->geometry_program;
  99    /* BRW_NEW_GS_PROG_DATA */
 100    const struct brw_stage_state *stage_state = &brw->gs.base;
 101    const struct brw_stage_prog_data *prog_data = stage_state->prog_data;
 102    const struct brw_vue_prog_data *vue_prog_data =
 103       brw_vue_prog_data(stage_state->prog_data);
 104    const struct brw_gs_prog_data *gs_prog_data =
 105       brw_gs_prog_data(stage_state->prog_data);
 106
 107    if (!active || stage_state->push_const_size == 0) {
 108       /* Disable the push constant buffers. */
 109       BEGIN_BATCH(5);
 110       OUT_BATCH(_3DSTATE_CONSTANT_GS << 16 | (5 - 2));
 111       OUT_BATCH(0);
 112       OUT_BATCH(0);
 113       OUT_BATCH(0);
 114       OUT_BATCH(0);
 115       ADVANCE_BATCH();
 116    } else {
 117       BEGIN_BATCH(5);
 118       OUT_BATCH(_3DSTATE_CONSTANT_GS << 16 |
 119                 GEN6_CONSTANT_BUFFER_0_ENABLE |
 120                 (5 - 2));
 121       /* Pointer to the GS constant buffer.  Covered by the set of
 122        * state flags from gen6_upload_vs_constants
 123        */
 124       OUT_BATCH(stage_state->push_const_offset +
 125                 stage_state->push_const_size - 1);
 126       OUT_BATCH(0);
 127       OUT_BATCH(0);
 128       OUT_BATCH(0);
 129       ADVANCE_BATCH();
 130    }
 131
 132    if (active) {
 133       BEGIN_BATCH(7);
 134       OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
 135       OUT_BATCH(stage_state->prog_offset);
 136
 137       /* GEN6_GS_SPF_MODE and GEN6_GS_VECTOR_MASK_ENABLE are enabled as it
 138        * was previously done for gen6.
 139        *
 140        * TODO: test with both disabled to see if the HW is behaving
 141        * as expected, like in gen7.
 142        */
 143       OUT_BATCH(GEN6_GS_SPF_MODE | GEN6_GS_VECTOR_MASK_ENABLE |
 144                 ((ALIGN(stage_state->sampler_count, 4)/4) <<
 145                  GEN6_GS_SAMPLER_COUNT_SHIFT) |
 146                 ((prog_data->binding_table.size_bytes / 4) <<
 147                  GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
 148
 149       if (prog_data->total_scratch) {
 150          OUT_RELOC(stage_state->scratch_bo,
 151                    I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
 152                    ffs(stage_state->per_thread_scratch) - 11);
 153       } else {
 154          OUT_BATCH(0); /* no scratch space */
 155       }
 156
 157       OUT_BATCH((vue_prog_data->urb_read_length <<
 158                  GEN6_GS_URB_READ_LENGTH_SHIFT) |
 159                 (0 << GEN6_GS_URB_ENTRY_READ_OFFSET_SHIFT) |
 160                 (prog_data->dispatch_grf_start_reg <<
 161                  GEN6_GS_DISPATCH_START_GRF_SHIFT));
 162
 163       OUT_BATCH(((devinfo->max_gs_threads - 1) << GEN6_GS_MAX_THREADS_SHIFT) |
 164                 GEN6_GS_STATISTICS_ENABLE |
 165                 GEN6_GS_SO_STATISTICS_ENABLE |
 166                 GEN6_GS_RENDERING_ENABLE);
 167
 168       if (gs_prog_data->gen6_xfb_enabled) {
 169          /* GEN6_GS_REORDER is equivalent to GEN7_GS_REORDER_TRAILING
 170           * in gen7. SNB and IVB specs are the same regarding the reordering of
 171           * TRISTRIP/TRISTRIP_REV vertices and triangle orientation, so we do
 172           * the same thing in both generations. For more details, see the
 173           * comment in gen7_gs_state.c
 174           */
 175          OUT_BATCH(GEN6_GS_REORDER |
 176                    GEN6_GS_SVBI_PAYLOAD_ENABLE |
 177                    GEN6_GS_ENABLE);
 178       } else {
 179          OUT_BATCH(GEN6_GS_REORDER | GEN6_GS_ENABLE);
 180       }
 181       ADVANCE_BATCH();
 182    } else if (brw->ff_gs.prog_active) {
 183       /* In gen6, transform feedback for the VS stage is done with an ad-hoc GS
 184        * program. This function provides the needed 3DSTATE_GS for this.
 185        */
 186       upload_gs_state_for_tf(brw);
 187    } else {
 188       /* No GS function required */
 189       BEGIN_BATCH(7);
 190       OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
 191       OUT_BATCH(0); /* prog_bo */
 192       OUT_BATCH((0 << GEN6_GS_SAMPLER_COUNT_SHIFT) |
 193                 (0 << GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
 194       OUT_BATCH(0); /* scratch space base offset */
 195       OUT_BATCH((1 << GEN6_GS_DISPATCH_START_GRF_SHIFT) |
 196                 (0 << GEN6_GS_URB_READ_LENGTH_SHIFT) |
 197                 (0 << GEN6_GS_URB_ENTRY_READ_OFFSET_SHIFT));
 198       OUT_BATCH((0 << GEN6_GS_MAX_THREADS_SHIFT) |
 199                 GEN6_GS_STATISTICS_ENABLE |
 200                 GEN6_GS_RENDERING_ENABLE);
 201                 OUT_BATCH(0);
 202       ADVANCE_BATCH();
 203    }
 204    brw->gs.enabled = active;
 205 }
 206
 207 const struct brw_tracked_state gen6_gs_state = {
 208    .dirty = {
 209       .mesa  = _NEW_PROGRAM_CONSTANTS |
 210                _NEW_TRANSFORM,
 211       .brw   = BRW_NEW_BATCH |
 212                BRW_NEW_BLORP |
 213                BRW_NEW_CONTEXT |
 214                BRW_NEW_FF_GS_PROG_DATA |
 215                BRW_NEW_GEOMETRY_PROGRAM |
 216                BRW_NEW_GS_PROG_DATA |
 217                BRW_NEW_PUSH_CONSTANT_ALLOCATION,
 218    },
 219    .emit = upload_gs_state,
 220 };