src/mesa/drivers/dri/i965/gen8_draw_upload.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 #include "main/glheader.h"
  25 #include "main/bufferobj.h"
  26 #include "main/context.h"
  27 #include "main/enums.h"
  28 #include "main/macros.h"
  29
  30 #include "brw_draw.h"
  31 #include "brw_defines.h"
  32 #include "brw_context.h"
  33 #include "brw_state.h"
  34
  35 #include "intel_batchbuffer.h"
  36 #include "intel_buffer_objects.h"
  37
  38 static void
  39 gen8_emit_vertices(struct brw_context *brw)
  40 {
  41    struct gl_context *ctx = &brw->ctx;
  42
  43    brw_prepare_vertices(brw);
  44
  45    if (brw->vs.prog_data->uses_vertexid) {
  46       unsigned vue = brw->vb.nr_enabled;
  47
  48       WARN_ONCE(brw->vs.prog_data->inputs_read & VERT_BIT_EDGEFLAG,
  49                 "Using VID/IID with edgeflags, need to reorder the "
  50                 "vertex attributes");
  51       WARN_ONCE(vue >= 33,
  52                 "Trying to insert VID/IID past 33rd vertex element, "
  53                 "need to reorder the vertex attrbutes.");
  54
  55       BEGIN_BATCH(2);
  56       OUT_BATCH(_3DSTATE_VF_SGVS << 16 | (2 - 2));
  57       OUT_BATCH(GEN8_SGVS_ENABLE_VERTEX_ID |
  58                 (0 << GEN8_SGVS_VERTEX_ID_COMPONENT_SHIFT) |   /* .x channel */
  59                 (vue << GEN8_SGVS_VERTEX_ID_ELEMENT_OFFSET_SHIFT) |
  60                 GEN8_SGVS_ENABLE_INSTANCE_ID |
  61                 (1 << GEN8_SGVS_INSTANCE_ID_COMPONENT_SHIFT) | /* .y channel */
  62                 (vue << GEN8_SGVS_INSTANCE_ID_ELEMENT_OFFSET_SHIFT));
  63       ADVANCE_BATCH();
  64    } else {
  65       BEGIN_BATCH(2);
  66       OUT_BATCH(_3DSTATE_VF_SGVS << 16 | (2 - 2));
  67       OUT_BATCH(0);
  68       ADVANCE_BATCH();
  69    }
  70
  71    /* If the VS doesn't read any inputs (calculating vertex position from
  72     * a state variable for some reason, for example), emit a single pad
  73     * VERTEX_ELEMENT struct and bail.
  74     *
  75     * The stale VB state stays in place, but they don't do anything unless
  76     * a VE loads from them.
  77     */
  78    if (brw->vb.nr_enabled == 0) {
  79       BEGIN_BATCH(3);
  80       OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (3 - 2));
  81       OUT_BATCH((0 << GEN6_VE0_INDEX_SHIFT) |
  82                 GEN6_VE0_VALID |
  83                 (BRW_SURFACEFORMAT_R32G32B32A32_FLOAT << BRW_VE0_FORMAT_SHIFT) |
  84                 (0 << BRW_VE0_SRC_OFFSET_SHIFT));
  85       OUT_BATCH((BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_0_SHIFT) |
  86                 (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
  87                 (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
  88                 (BRW_VE1_COMPONENT_STORE_1_FLT << BRW_VE1_COMPONENT_3_SHIFT));
  89       ADVANCE_BATCH();
  90       return;
  91    }
  92
  93    /* Now emit 3DSTATE_VERTEX_BUFFERS and 3DSTATE_VERTEX_ELEMENTS packets. */
  94    if (brw->vb.nr_buffers) {
  95       assert(brw->vb.nr_buffers <= 33);
  96
  97       perf_debug("Missing MOCS setup for 3DSTATE_VERTEX_BUFFERS.");
  98
  99       BEGIN_BATCH(1 + 4*brw->vb.nr_buffers);
 100       OUT_BATCH((_3DSTATE_VERTEX_BUFFERS << 16) | (4*brw->vb.nr_buffers - 1));
 101       for (unsigned i = 0; i < brw->vb.nr_buffers; i++) {
 102          struct brw_vertex_buffer *buffer = &brw->vb.buffers[i];
 103          uint32_t dw0 = 0;
 104
 105          dw0 |= i << GEN6_VB0_INDEX_SHIFT;
 106          dw0 |= GEN7_VB0_ADDRESS_MODIFYENABLE;
 107          dw0 |= buffer->stride << BRW_VB0_PITCH_SHIFT;
 108
 109          OUT_BATCH(dw0);
 110          OUT_RELOC64(buffer->bo, I915_GEM_DOMAIN_VERTEX, 0, buffer->offset);
 111          OUT_BATCH(buffer->bo->size);
 112       }
 113       ADVANCE_BATCH();
 114    }
 115
 116    unsigned nr_elements = brw->vb.nr_enabled;
 117
 118    /* The hardware allows one more VERTEX_ELEMENTS than VERTEX_BUFFERS,
 119     * presumably for VertexID/InstanceID.
 120     */
 121    assert(nr_elements <= 34);
 122
 123    struct brw_vertex_element *gen6_edgeflag_input = NULL;
 124
 125    BEGIN_BATCH(1 + nr_elements * 2);
 126    OUT_BATCH((_3DSTATE_VERTEX_ELEMENTS << 16) | (2 * nr_elements - 1));
 127    for (unsigned i = 0; i < brw->vb.nr_enabled; i++) {
 128       struct brw_vertex_element *input = brw->vb.enabled[i];
 129       uint32_t format = brw_get_vertex_surface_type(brw, input->glarray);
 130       uint32_t comp0 = BRW_VE1_COMPONENT_STORE_SRC;
 131       uint32_t comp1 = BRW_VE1_COMPONENT_STORE_SRC;
 132       uint32_t comp2 = BRW_VE1_COMPONENT_STORE_SRC;
 133       uint32_t comp3 = BRW_VE1_COMPONENT_STORE_SRC;
 134
 135       /* The gen4 driver expects edgeflag to come in as a float, and passes
 136        * that float on to the tests in the clipper.  Mesa's current vertex
 137        * attribute value for EdgeFlag is stored as a float, which works out.
 138        * glEdgeFlagPointer, on the other hand, gives us an unnormalized
 139        * integer ubyte.  Just rewrite that to convert to a float.
 140        */
 141       if (input->attrib == VERT_ATTRIB_EDGEFLAG) {
 142          /* Gen6+ passes edgeflag as sideband along with the vertex, instead
 143           * of in the VUE.  We have to upload it sideband as the last vertex
 144           * element according to the B-Spec.
 145           */
 146          gen6_edgeflag_input = input;
 147          continue;
 148       }
 149
 150       switch (input->glarray->Size) {
 151       case 0: comp0 = BRW_VE1_COMPONENT_STORE_0;
 152       case 1: comp1 = BRW_VE1_COMPONENT_STORE_0;
 153       case 2: comp2 = BRW_VE1_COMPONENT_STORE_0;
 154       case 3: comp3 = input->glarray->Integer ? BRW_VE1_COMPONENT_STORE_1_INT
 155                                               : BRW_VE1_COMPONENT_STORE_1_FLT;
 156          break;
 157       }
 158
 159       OUT_BATCH((input->buffer << GEN6_VE0_INDEX_SHIFT) |
 160                 GEN6_VE0_VALID |
 161                 (format << BRW_VE0_FORMAT_SHIFT) |
 162                 (input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
 163
 164       OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) |
 165                 (comp1 << BRW_VE1_COMPONENT_1_SHIFT) |
 166                 (comp2 << BRW_VE1_COMPONENT_2_SHIFT) |
 167                 (comp3 << BRW_VE1_COMPONENT_3_SHIFT));
 168    }
 169
 170    if (gen6_edgeflag_input) {
 171       uint32_t format =
 172          brw_get_vertex_surface_type(brw, gen6_edgeflag_input->glarray);
 173
 174       OUT_BATCH((gen6_edgeflag_input->buffer << GEN6_VE0_INDEX_SHIFT) |
 175                 GEN6_VE0_VALID |
 176                 GEN6_VE0_EDGE_FLAG_ENABLE |
 177                 (format << BRW_VE0_FORMAT_SHIFT) |
 178                 (gen6_edgeflag_input->offset << BRW_VE0_SRC_OFFSET_SHIFT));
 179       OUT_BATCH((BRW_VE1_COMPONENT_STORE_SRC << BRW_VE1_COMPONENT_0_SHIFT) |
 180                 (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_1_SHIFT) |
 181                 (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_2_SHIFT) |
 182                 (BRW_VE1_COMPONENT_STORE_0 << BRW_VE1_COMPONENT_3_SHIFT));
 183    }
 184    ADVANCE_BATCH();
 185
 186    for (unsigned i = 0; i < brw->vb.nr_enabled; i++) {
 187       const struct brw_vertex_element *input = brw->vb.enabled[i];
 188       const struct brw_vertex_buffer *buffer = &brw->vb.buffers[input->buffer];
 189
 190       BEGIN_BATCH(3);
 191       OUT_BATCH(_3DSTATE_VF_INSTANCING << 16 | (3 - 2));
 192       OUT_BATCH(i | (buffer->step_rate ? GEN8_VF_INSTANCING_ENABLE : 0));
 193       OUT_BATCH(buffer->step_rate);
 194       ADVANCE_BATCH();
 195    }
 196 }
 197
 198 const struct brw_tracked_state gen8_vertices = {
 199    .dirty = {
 200       .mesa = _NEW_POLYGON,
 201       .brw = BRW_NEW_BATCH | BRW_NEW_VERTICES,
 202       .cache = CACHE_NEW_VS_PROG,
 203    },
 204    .emit = gen8_emit_vertices,
 205 };
 206
 207 static void
 208 gen8_emit_index_buffer(struct brw_context *brw)
 209 {
 210    const struct _mesa_index_buffer *index_buffer = brw->ib.ib;
 211
 212    if (index_buffer == NULL)
 213       return;
 214
 215    perf_debug("Missing MOCS setup for 3DSTATE_INDEX_BUFFER.");
 216
 217    BEGIN_BATCH(5);
 218    OUT_BATCH(CMD_INDEX_BUFFER << 16 | (5 - 2));
 219    OUT_BATCH(brw_get_index_type(index_buffer->type) << 8);
 220    OUT_RELOC64(brw->ib.bo, I915_GEM_DOMAIN_VERTEX, 0, 0);
 221    OUT_BATCH(brw->ib.bo->size);
 222    ADVANCE_BATCH();
 223 }
 224
 225 const struct brw_tracked_state gen8_index_buffer = {
 226    .dirty = {
 227       .mesa = 0,
 228       .brw = BRW_NEW_BATCH | BRW_NEW_INDEX_BUFFER,
 229       .cache = 0,
 230    },
 231    .emit = gen8_emit_index_buffer,
 232 };
 233
 234 static void
 235 gen8_emit_vf_topology(struct brw_context *brw)
 236 {
 237    BEGIN_BATCH(2);
 238    OUT_BATCH(_3DSTATE_VF_TOPOLOGY << 16 | (2 - 2));
 239    OUT_BATCH(brw->primitive);
 240    ADVANCE_BATCH();
 241 }
 242
 243 const struct brw_tracked_state gen8_vf_topology = {
 244    .dirty = {
 245       .mesa = 0,
 246       .brw = BRW_NEW_PRIMITIVE,
 247       .cache = 0,
 248    },
 249    .emit = gen8_emit_vf_topology,
 250 };