src/mesa/drivers/dri/i965/gen8_sol_state.c

   1 /*
   2  * Copyright © 2012 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 /**
  25  * @file gen8_sol_state.c
  26  *
  27  * Controls the stream output logic (SOL) stage of the gen8 hardware, which is
  28  * used to implement GL_EXT_transform_feedback.
  29  */
  30
  31 #include "brw_context.h"
  32 #include "brw_state.h"
  33 #include "brw_defines.h"
  34 #include "intel_batchbuffer.h"
  35 #include "intel_buffer_objects.h"
  36 #include "main/transformfeedback.h"
  37
  38 static void
  39 gen8_upload_3dstate_so_buffers(struct brw_context *brw)
  40 {
  41    struct gl_context *ctx = &brw->ctx;
  42    /* BRW_NEW_TRANSFORM_FEEDBACK */
  43    struct gl_transform_feedback_object *xfb_obj =
  44       ctx->TransformFeedback.CurrentObject;
  45    struct brw_transform_feedback_object *brw_obj =
  46       (struct brw_transform_feedback_object *) xfb_obj;
  47
  48    /* Set up the up to 4 output buffers.  These are the ranges defined in the
  49     * gl_transform_feedback_object.
  50     */
  51    for (int i = 0; i < 4; i++) {
  52       struct intel_buffer_object *bufferobj =
  53          intel_buffer_object(xfb_obj->Buffers[i]);
  54
  55       if (!bufferobj) {
  56          BEGIN_BATCH(8);
  57          OUT_BATCH(_3DSTATE_SO_BUFFER << 16 | (8 - 2));
  58          OUT_BATCH((i << SO_BUFFER_INDEX_SHIFT));
  59          OUT_BATCH(0);
  60          OUT_BATCH(0);
  61          OUT_BATCH(0);
  62          OUT_BATCH(0);
  63          OUT_BATCH(0);
  64          OUT_BATCH(0);
  65          ADVANCE_BATCH();
  66          continue;
  67       }
  68
  69       uint32_t start = xfb_obj->Offset[i];
  70       assert(start % 4 == 0);
  71       uint32_t end = ALIGN(start + xfb_obj->Size[i], 4);
  72       drm_intel_bo *bo =
  73          intel_bufferobj_buffer(brw, bufferobj, start, end - start);
  74       assert(end <= bo->size);
  75
  76       perf_debug("Missing MOCS setup for 3DSTATE_SO_BUFFER.");
  77
  78       BEGIN_BATCH(8);
  79       OUT_BATCH(_3DSTATE_SO_BUFFER << 16 | (8 - 2));
  80       OUT_BATCH(GEN8_SO_BUFFER_ENABLE | (i << SO_BUFFER_INDEX_SHIFT) |
  81                 GEN8_SO_BUFFER_OFFSET_WRITE_ENABLE |
  82                 GEN8_SO_BUFFER_OFFSET_ADDRESS_ENABLE);
  83       OUT_RELOC64(bo, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, start);
  84       OUT_BATCH(xfb_obj->Size[i] / 4 - 1);
  85       OUT_RELOC64(brw_obj->offset_bo,
  86                   I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
  87                   i * sizeof(uint32_t));
  88       if (brw_obj->zero_offsets)
  89          OUT_BATCH(0); /* Zero out the offset and write that to offset_bo */
  90       else
  91          OUT_BATCH(0xFFFFFFFF); /* Use offset_bo as the "Stream Offset." */
  92       ADVANCE_BATCH();
  93    }
  94    brw_obj->zero_offsets = false;
  95 }
  96
  97 static void
  98 gen8_upload_3dstate_streamout(struct brw_context *brw, bool active,
  99                               struct brw_vue_map *vue_map)
 100 {
 101    struct gl_context *ctx = &brw->ctx;
 102
 103    /* BRW_NEW_VERTEX_PROGRAM */
 104    const struct gl_shader_program *vs_prog =
 105       ctx->Shader.CurrentProgram[MESA_SHADER_VERTEX];
 106    /* BRW_NEW_TRANSFORM_FEEDBACK */
 107    const struct gl_transform_feedback_info *linked_xfb_info =
 108       &vs_prog->LinkedTransformFeedback;
 109    struct gl_transform_feedback_object *xfb_obj =
 110       ctx->TransformFeedback.CurrentObject;
 111    uint32_t dw1 = 0, dw2 = 0, dw3 = 0, dw4 = 0;
 112
 113    if (active) {
 114       int urb_entry_read_offset = 0;
 115       int urb_entry_read_length = (vue_map->num_slots + 1) / 2 -
 116          urb_entry_read_offset;
 117
 118       dw1 |= SO_FUNCTION_ENABLE;
 119       dw1 |= SO_STATISTICS_ENABLE;
 120
 121       /* _NEW_LIGHT */
 122       if (ctx->Light.ProvokingVertex != GL_FIRST_VERTEX_CONVENTION)
 123          dw1 |= SO_REORDER_TRAILING;
 124
 125       /* We always read the whole vertex.  This could be reduced at some
 126        * point by reading less and offsetting the register index in the
 127        * SO_DECLs.
 128        */
 129       dw2 |= urb_entry_read_offset << SO_STREAM_0_VERTEX_READ_OFFSET_SHIFT;
 130       dw2 |= (urb_entry_read_length - 1) << SO_STREAM_0_VERTEX_READ_LENGTH_SHIFT;
 131
 132       /* Set buffer pitches; 0 means unbound. */
 133       if (xfb_obj->Buffers[0])
 134          dw3 |= linked_xfb_info->BufferStride[0] * 4;
 135       if (xfb_obj->Buffers[1])
 136          dw3 |= (linked_xfb_info->BufferStride[1] * 4) << 16;
 137       if (xfb_obj->Buffers[2])
 138          dw4 |= linked_xfb_info->BufferStride[2] * 4;
 139       if (xfb_obj->Buffers[3])
 140          dw4 |= (linked_xfb_info->BufferStride[3] * 4) << 16;
 141    }
 142
 143    BEGIN_BATCH(5);
 144    OUT_BATCH(_3DSTATE_STREAMOUT << 16 | (5 - 2));
 145    OUT_BATCH(dw1);
 146    OUT_BATCH(dw2);
 147    OUT_BATCH(dw3);
 148    OUT_BATCH(dw4);
 149    ADVANCE_BATCH();
 150 }
 151
 152 static void
 153 upload_sol_state(struct brw_context *brw)
 154 {
 155    struct gl_context *ctx = &brw->ctx;
 156    /* BRW_NEW_TRANSFORM_FEEDBACK */
 157    bool active = _mesa_is_xfb_active_and_unpaused(ctx);
 158
 159    if (active) {
 160       gen8_upload_3dstate_so_buffers(brw);
 161       /* BRW_NEW_VUE_MAP_GEOM_OUT */
 162       gen7_upload_3dstate_so_decl_list(brw, &brw->vue_map_geom_out);
 163    }
 164
 165    gen8_upload_3dstate_streamout(brw, active, &brw->vue_map_geom_out);
 166 }
 167
 168 const struct brw_tracked_state gen8_sol_state = {
 169    .dirty = {
 170       .mesa  = _NEW_LIGHT,
 171       .brw   = BRW_NEW_BATCH |
 172                BRW_NEW_RASTERIZER_DISCARD |
 173                BRW_NEW_TRANSFORM_FEEDBACK |
 174                BRW_NEW_VERTEX_PROGRAM |
 175                BRW_NEW_VUE_MAP_GEOM_OUT,
 176       .cache = 0,
 177    },
 178    .emit = upload_sol_state,
 179 };