src/mesa/drivers/dri/i965/gen6_sol.c

   1 /*
   2  * Copyright © 2011 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 /** \file gen6_sol.c
  25  *
  26  * Code to initialize the binding table entries used by transform feedback.
  27  */
  28
  29 #include "main/bufferobj.h"
  30 #include "main/macros.h"
  31 #include "brw_context.h"
  32 #include "intel_batchbuffer.h"
  33 #include "brw_defines.h"
  34 #include "brw_state.h"
  35 #include "main/transformfeedback.h"
  36
  37 static void
  38 gen6_update_sol_surfaces(struct brw_context *brw)
  39 {
  40    struct gl_context *ctx = &brw->ctx;
  41    /* BRW_NEW_TRANSFORM_FEEDBACK */
  42    struct gl_transform_feedback_object *xfb_obj =
  43       ctx->TransformFeedback.CurrentObject;
  44    /* BRW_NEW_VERTEX_PROGRAM */
  45    const struct gl_shader_program *shaderprog =
  46       ctx->Shader.CurrentVertexProgram;
  47    const struct gl_transform_feedback_info *linked_xfb_info =
  48       &shaderprog->LinkedTransformFeedback;
  49    int i;
  50
  51    for (i = 0; i < BRW_MAX_SOL_BINDINGS; ++i) {
  52       const int surf_index = SURF_INDEX_GEN6_SOL_BINDING(i);
  53       if (_mesa_is_xfb_active_and_unpaused(ctx) &&
  54           i < linked_xfb_info->NumOutputs) {
  55          unsigned buffer = linked_xfb_info->Outputs[i].OutputBuffer;
  56          unsigned buffer_offset =
  57             xfb_obj->Offset[buffer] / 4 +
  58             linked_xfb_info->Outputs[i].DstOffset;
  59          brw_update_sol_surface(
  60             brw, xfb_obj->Buffers[buffer], &brw->ff_gs.surf_offset[surf_index],
  61             linked_xfb_info->Outputs[i].NumComponents,
  62             linked_xfb_info->BufferStride[buffer], buffer_offset);
  63       } else {
  64          brw->ff_gs.surf_offset[surf_index] = 0;
  65       }
  66    }
  67
  68    brw->state.dirty.brw |= BRW_NEW_SURFACES;
  69 }
  70
  71 const struct brw_tracked_state gen6_sol_surface = {
  72    .dirty = {
  73       .mesa = 0,
  74       .brw = (BRW_NEW_BATCH |
  75               BRW_NEW_VERTEX_PROGRAM |
  76               BRW_NEW_TRANSFORM_FEEDBACK),
  77       .cache = 0
  78    },
  79    .emit = gen6_update_sol_surfaces,
  80 };
  81
  82 /**
  83  * Constructs the binding table for the WM surface state, which maps unit
  84  * numbers to surface state objects.
  85  */
  86 static void
  87 brw_gs_upload_binding_table(struct brw_context *brw)
  88 {
  89    struct gl_context *ctx = &brw->ctx;
  90    /* BRW_NEW_VERTEX_PROGRAM */
  91    const struct gl_shader_program *shaderprog =
  92       ctx->Shader.CurrentVertexProgram;
  93    bool has_surfaces = false;
  94    uint32_t *bind;
  95
  96    if (shaderprog) {
  97       const struct gl_transform_feedback_info *linked_xfb_info =
  98          &shaderprog->LinkedTransformFeedback;
  99       /* Currently we only ever upload surfaces for SOL. */
 100       has_surfaces = linked_xfb_info->NumOutputs != 0;
 101    }
 102
 103    /* Skip making a binding table if we don't have anything to put in it. */
 104    if (!has_surfaces) {
 105       if (brw->ff_gs.bind_bo_offset != 0) {
 106          brw->state.dirty.brw |= BRW_NEW_GS_BINDING_TABLE;
 107          brw->ff_gs.bind_bo_offset = 0;
 108       }
 109       return;
 110    }
 111
 112    /* Might want to calculate nr_surfaces first, to avoid taking up so much
 113     * space for the binding table.
 114     */
 115    bind = brw_state_batch(brw, AUB_TRACE_BINDING_TABLE,
 116                           sizeof(uint32_t) * BRW_MAX_GEN6_GS_SURFACES,
 117                           32, &brw->ff_gs.bind_bo_offset);
 118
 119    /* BRW_NEW_SURFACES */
 120    memcpy(bind, brw->ff_gs.surf_offset, BRW_MAX_GEN6_GS_SURFACES * sizeof(uint32_t));
 121
 122    brw->state.dirty.brw |= BRW_NEW_GS_BINDING_TABLE;
 123 }
 124
 125 const struct brw_tracked_state gen6_gs_binding_table = {
 126    .dirty = {
 127       .mesa = 0,
 128       .brw = (BRW_NEW_BATCH |
 129               BRW_NEW_VERTEX_PROGRAM |
 130               BRW_NEW_SURFACES),
 131       .cache = 0
 132    },
 133    .emit = brw_gs_upload_binding_table,
 134 };
 135
 136 struct gl_transform_feedback_object *
 137 brw_new_transform_feedback(struct gl_context *ctx, GLuint name)
 138 {
 139    struct brw_context *brw = brw_context(ctx);
 140    struct brw_transform_feedback_object *brw_obj =
 141       CALLOC_STRUCT(brw_transform_feedback_object);
 142    if (!brw_obj)
 143       return NULL;
 144
 145    _mesa_init_transform_feedback_object(&brw_obj->base, name);
 146
 147    brw_obj->offset_bo =
 148       drm_intel_bo_alloc(brw->bufmgr, "transform feedback offsets", 16, 64);
 149
 150    return &brw_obj->base;
 151 }
 152
 153 void
 154 brw_delete_transform_feedback(struct gl_context *ctx,
 155                               struct gl_transform_feedback_object *obj)
 156 {
 157    struct brw_transform_feedback_object *brw_obj =
 158       (struct brw_transform_feedback_object *) obj;
 159
 160    for (unsigned i = 0; i < Elements(obj->Buffers); i++) {
 161       _mesa_reference_buffer_object(ctx, &obj->Buffers[i], NULL);
 162    }
 163
 164    drm_intel_bo_unreference(brw_obj->offset_bo);
 165
 166    free(brw_obj);
 167 }
 168
 169 void
 170 brw_begin_transform_feedback(struct gl_context *ctx, GLenum mode,
 171                              struct gl_transform_feedback_object *obj)
 172 {
 173    struct brw_context *brw = brw_context(ctx);
 174    const struct gl_shader_program *vs_prog =
 175       ctx->Shader.CurrentVertexProgram;
 176    const struct gl_transform_feedback_info *linked_xfb_info =
 177       &vs_prog->LinkedTransformFeedback;
 178    struct gl_transform_feedback_object *xfb_obj =
 179       ctx->TransformFeedback.CurrentObject;
 180
 181    assert(brw->gen == 6);
 182
 183    /* Compute the maximum number of vertices that we can write without
 184     * overflowing any of the buffers currently being used for feedback.
 185     */
 186    unsigned max_index
 187       = _mesa_compute_max_transform_feedback_vertices(xfb_obj,
 188                                                       linked_xfb_info);
 189
 190    /* 3DSTATE_GS_SVB_INDEX is non-pipelined. */
 191    intel_emit_post_sync_nonzero_flush(brw);
 192
 193    /* Initialize the SVBI 0 register to zero and set the maximum index. */
 194    BEGIN_BATCH(4);
 195    OUT_BATCH(_3DSTATE_GS_SVB_INDEX << 16 | (4 - 2));
 196    OUT_BATCH(0); /* SVBI 0 */
 197    OUT_BATCH(0); /* starting index */
 198    OUT_BATCH(max_index);
 199    ADVANCE_BATCH();
 200
 201    /* Initialize the rest of the unused streams to sane values.  Otherwise,
 202     * they may indicate that there is no room to write data and prevent
 203     * anything from happening at all.
 204     */
 205    for (int i = 1; i < 4; i++) {
 206       BEGIN_BATCH(4);
 207       OUT_BATCH(_3DSTATE_GS_SVB_INDEX << 16 | (4 - 2));
 208       OUT_BATCH(i << SVB_INDEX_SHIFT);
 209       OUT_BATCH(0); /* starting index */
 210       OUT_BATCH(0xffffffff);
 211       ADVANCE_BATCH();
 212    }
 213 }
 214
 215 void
 216 brw_end_transform_feedback(struct gl_context *ctx,
 217                            struct gl_transform_feedback_object *obj)
 218 {
 219    /* After EndTransformFeedback, it's likely that the client program will try
 220     * to draw using the contents of the transform feedback buffer as vertex
 221     * input.  In order for this to work, we need to flush the data through at
 222     * least the GS stage of the pipeline, and flush out the render cache.  For
 223     * simplicity, just do a full flush.
 224     */
 225    struct brw_context *brw = brw_context(ctx);
 226    intel_batchbuffer_emit_mi_flush(brw);
 227 }