src/mesa/drivers/dri/i965/gen6_vs_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 "brw_util.h"
  32 #include "program/prog_parameter.h"
  33 #include "program/prog_statevars.h"
  34 #include "main/shaderapi.h"
  35 #include "intel_batchbuffer.h"
  36
  37 static void
  38 gen6_upload_vs_push_constants(struct brw_context *brw)
  39 {
  40    struct brw_stage_state *stage_state = &brw->vs.base;
  41
  42    /* _BRW_NEW_VERTEX_PROGRAM */
  43    const struct brw_program *vp = brw_program_const(brw->vertex_program);
  44    /* BRW_NEW_VS_PROG_DATA */
  45    const struct brw_stage_prog_data *prog_data = brw->vs.base.prog_data;
  46
  47    _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_VERTEX);
  48    gen6_upload_push_constants(brw, &vp->program, prog_data, stage_state,
  49                               AUB_TRACE_VS_CONSTANTS);
  50
  51    if (brw->gen >= 7) {
  52       if (brw->gen == 7 && !brw->is_haswell && !brw->is_baytrail)
  53          gen7_emit_vs_workaround_flush(brw);
  54
  55       gen7_upload_constant_state(brw, stage_state, true /* active */,
  56                                  _3DSTATE_CONSTANT_VS);
  57    }
  58 }
  59
  60 const struct brw_tracked_state gen6_vs_push_constants = {
  61    .dirty = {
  62       .mesa  = _NEW_PROGRAM_CONSTANTS |
  63                _NEW_TRANSFORM,
  64       .brw   = BRW_NEW_BATCH |
  65                BRW_NEW_BLORP |
  66                BRW_NEW_PUSH_CONSTANT_ALLOCATION |
  67                BRW_NEW_VERTEX_PROGRAM |
  68                BRW_NEW_VS_PROG_DATA,
  69    },
  70    .emit = gen6_upload_vs_push_constants,
  71 };
  72
  73 static void
  74 upload_vs_state(struct brw_context *brw)
  75 {
  76    const struct gen_device_info *devinfo = &brw->screen->devinfo;
  77    const struct brw_stage_state *stage_state = &brw->vs.base;
  78    const struct brw_stage_prog_data *prog_data = stage_state->prog_data;
  79    const struct brw_vue_prog_data *vue_prog_data =
  80       brw_vue_prog_data(stage_state->prog_data);
  81    uint32_t floating_point_mode = 0;
  82
  83    /* From the BSpec, 3D Pipeline > Geometry > Vertex Shader > State,
  84     * 3DSTATE_VS, Dword 5.0 "VS Function Enable":
  85     *
  86     *   [DevSNB] A pipeline flush must be programmed prior to a 3DSTATE_VS
  87     *   command that causes the VS Function Enable to toggle. Pipeline
  88     *   flush can be executed by sending a PIPE_CONTROL command with CS
  89     *   stall bit set and a post sync operation.
  90     *
  91     * We've already done such a flush at the start of state upload, so we
  92     * don't need to do another one here.
  93     */
  94
  95    if (stage_state->push_const_size == 0) {
  96       /* Disable the push constant buffers. */
  97       BEGIN_BATCH(5);
  98       OUT_BATCH(_3DSTATE_CONSTANT_VS << 16 | (5 - 2));
  99       OUT_BATCH(0);
 100       OUT_BATCH(0);
 101       OUT_BATCH(0);
 102       OUT_BATCH(0);
 103       ADVANCE_BATCH();
 104    } else {
 105       BEGIN_BATCH(5);
 106       OUT_BATCH(_3DSTATE_CONSTANT_VS << 16 |
 107                 GEN6_CONSTANT_BUFFER_0_ENABLE |
 108                 (5 - 2));
 109       /* Pointer to the VS constant buffer.  Covered by the set of
 110        * state flags from gen6_upload_vs_constants
 111        */
 112       OUT_BATCH(stage_state->push_const_offset +
 113                 stage_state->push_const_size - 1);
 114       OUT_BATCH(0);
 115       OUT_BATCH(0);
 116       OUT_BATCH(0);
 117       ADVANCE_BATCH();
 118    }
 119
 120    if (prog_data->use_alt_mode)
 121       floating_point_mode = GEN6_VS_FLOATING_POINT_MODE_ALT;
 122
 123    BEGIN_BATCH(6);
 124    OUT_BATCH(_3DSTATE_VS << 16 | (6 - 2));
 125    OUT_BATCH(stage_state->prog_offset);
 126    OUT_BATCH(floating_point_mode |
 127              ((ALIGN(stage_state->sampler_count, 4)/4) << GEN6_VS_SAMPLER_COUNT_SHIFT) |
 128              ((prog_data->binding_table.size_bytes / 4) <<
 129               GEN6_VS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
 130
 131    if (prog_data->total_scratch) {
 132       OUT_RELOC(stage_state->scratch_bo,
 133                 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
 134                 ffs(stage_state->per_thread_scratch) - 11);
 135    } else {
 136       OUT_BATCH(0);
 137    }
 138
 139    OUT_BATCH((prog_data->dispatch_grf_start_reg <<
 140               GEN6_VS_DISPATCH_START_GRF_SHIFT) |
 141              (vue_prog_data->urb_read_length << GEN6_VS_URB_READ_LENGTH_SHIFT) |
 142              (0 << GEN6_VS_URB_ENTRY_READ_OFFSET_SHIFT));
 143
 144    OUT_BATCH(((devinfo->max_vs_threads - 1) << GEN6_VS_MAX_THREADS_SHIFT) |
 145              GEN6_VS_STATISTICS_ENABLE |
 146              GEN6_VS_ENABLE);
 147    ADVANCE_BATCH();
 148
 149    /* Based on my reading of the simulator, the VS constants don't get
 150     * pulled into the VS FF unit until an appropriate pipeline flush
 151     * happens, and instead the 3DSTATE_CONSTANT_VS packet just adds
 152     * references to them into a little FIFO.  The flushes are common,
 153     * but don't reliably happen between this and a 3DPRIMITIVE, causing
 154     * the primitive to use the wrong constants.  Then the FIFO
 155     * containing the constant setup gets added to again on the next
 156     * constants change, and eventually when a flush does happen the
 157     * unit is overwhelmed by constant changes and dies.
 158     *
 159     * To avoid this, send a PIPE_CONTROL down the line that will
 160     * update the unit immediately loading the constants.  The flush
 161     * type bits here were those set by the STATE_BASE_ADDRESS whose
 162     * move in a82a43e8d99e1715dd11c9c091b5ab734079b6a6 triggered the
 163     * bug reports that led to this workaround, and may be more than
 164     * what is strictly required to avoid the issue.
 165     */
 166    brw_emit_pipe_control_flush(brw,
 167                                PIPE_CONTROL_DEPTH_STALL |
 168                                PIPE_CONTROL_INSTRUCTION_INVALIDATE |
 169                                PIPE_CONTROL_STATE_CACHE_INVALIDATE);
 170 }
 171
 172 const struct brw_tracked_state gen6_vs_state = {
 173    .dirty = {
 174       .mesa  = _NEW_PROGRAM_CONSTANTS |
 175                _NEW_TRANSFORM,
 176       .brw   = BRW_NEW_BATCH |
 177                BRW_NEW_BLORP |
 178                BRW_NEW_CONTEXT |
 179                BRW_NEW_PUSH_CONSTANT_ALLOCATION |
 180                BRW_NEW_VERTEX_PROGRAM |
 181                BRW_NEW_VS_PROG_DATA,
 182    },
 183    .emit = upload_vs_state,
 184 };