src/mesa/drivers/dri/i965/gen7_gs_state.c

   1 /*
   2  * Copyright © 2013 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
  21  * DEALINGS IN THE SOFTWARE.
  22  */
  23
  24 #include "brw_context.h"
  25 #include "brw_state.h"
  26 #include "brw_defines.h"
  27 #include "intel_batchbuffer.h"
  28
  29 static void
  30 upload_gs_state(struct brw_context *brw)
  31 {
  32    const struct brw_stage_state *stage_state = &brw->gs.base;
  33    const int max_threads_shift = brw->is_haswell ?
  34       HSW_GS_MAX_THREADS_SHIFT : GEN6_GS_MAX_THREADS_SHIFT;
  35    /* BRW_NEW_GEOMETRY_PROGRAM */
  36    bool active = brw->geometry_program;
  37    /* BRW_NEW_GS_PROG_DATA */
  38    const struct brw_vec4_prog_data *prog_data = &brw->gs.prog_data->base;
  39
  40    /**
  41     * From Graphics BSpec: 3D-Media-GPGPU Engine > 3D Pipeline Stages >
  42     * Geometry > Geometry Shader > State:
  43     *
  44     *     "Note: Because of corruption in IVB:GT2, software needs to flush the
  45     *     whole fixed function pipeline when the GS enable changes value in
  46     *     the 3DSTATE_GS."
  47     *
  48     * The hardware architects have clarified that in this context "flush the
  49     * whole fixed function pipeline" means to emit a PIPE_CONTROL with the "CS
  50     * Stall" bit set.
  51     */
  52    if (!brw->is_haswell && brw->gt == 2 && brw->gs.enabled != active)
  53       gen7_emit_cs_stall_flush(brw);
  54
  55    if (active) {
  56       BEGIN_BATCH(7);
  57       OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
  58       OUT_BATCH(stage_state->prog_offset);
  59       OUT_BATCH(((ALIGN(stage_state->sampler_count, 4)/4) <<
  60                  GEN6_GS_SAMPLER_COUNT_SHIFT) |
  61                 ((brw->gs.prog_data->base.base.binding_table.size_bytes / 4) <<
  62                  GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
  63
  64       if (brw->gs.prog_data->base.base.total_scratch) {
  65          OUT_RELOC(stage_state->scratch_bo,
  66                    I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
  67                    ffs(brw->gs.prog_data->base.base.total_scratch) - 11);
  68       } else {
  69          OUT_BATCH(0);
  70       }
  71
  72       uint32_t dw4 =
  73          ((brw->gs.prog_data->output_vertex_size_hwords * 2 - 1) <<
  74           GEN7_GS_OUTPUT_VERTEX_SIZE_SHIFT) |
  75          (brw->gs.prog_data->output_topology <<
  76           GEN7_GS_OUTPUT_TOPOLOGY_SHIFT) |
  77          (prog_data->urb_read_length <<
  78           GEN6_GS_URB_READ_LENGTH_SHIFT) |
  79          (0 << GEN6_GS_URB_ENTRY_READ_OFFSET_SHIFT) |
  80          (prog_data->base.dispatch_grf_start_reg <<
  81           GEN6_GS_DISPATCH_START_GRF_SHIFT);
  82
  83       /* Note: the meaning of the GEN7_GS_REORDER_TRAILING bit changes between
  84        * Ivy Bridge and Haswell.
  85        *
  86        * On Ivy Bridge, setting this bit causes the vertices of a triangle
  87        * strip to be delivered to the geometry shader in an order that does
  88        * not strictly follow the OpenGL spec, but preserves triangle
  89        * orientation.  For example, if the vertices are (1, 2, 3, 4, 5), then
  90        * the geometry shader sees triangles:
  91        *
  92        * (1, 2, 3), (2, 4, 3), (3, 4, 5)
  93        *
  94        * (Clearing the bit is even worse, because it fails to preserve
  95        * orientation).
  96        *
  97        * Triangle strips with adjacency always ordered in a way that preserves
  98        * triangle orientation but does not strictly follow the OpenGL spec,
  99        * regardless of the setting of this bit.
 100        *
 101        * On Haswell, both triangle strips and triangle strips with adjacency
 102        * are always ordered in a way that preserves triangle orientation.
 103        * Setting this bit causes the ordering to strictly follow the OpenGL
 104        * spec.
 105        *
 106        * So in either case we want to set the bit.  Unfortunately on Ivy
 107        * Bridge this will get the order close to correct but not perfect.
 108        */
 109       uint32_t dw5 =
 110          ((brw->max_gs_threads - 1) << max_threads_shift) |
 111          (brw->gs.prog_data->control_data_header_size_hwords <<
 112           GEN7_GS_CONTROL_DATA_HEADER_SIZE_SHIFT) |
 113          ((brw->gs.prog_data->invocations - 1) <<
 114           GEN7_GS_INSTANCE_CONTROL_SHIFT) |
 115          brw->gs.prog_data->dispatch_mode |
 116          GEN6_GS_STATISTICS_ENABLE |
 117          (brw->gs.prog_data->include_primitive_id ?
 118           GEN7_GS_INCLUDE_PRIMITIVE_ID : 0) |
 119          GEN7_GS_REORDER_TRAILING |
 120          GEN7_GS_ENABLE;
 121       uint32_t dw6 = 0;
 122
 123       if (brw->is_haswell) {
 124          dw6 |= brw->gs.prog_data->control_data_format <<
 125             HSW_GS_CONTROL_DATA_FORMAT_SHIFT;
 126       } else {
 127          dw5 |= brw->gs.prog_data->control_data_format <<
 128             IVB_GS_CONTROL_DATA_FORMAT_SHIFT;
 129       }
 130
 131       OUT_BATCH(dw4);
 132       OUT_BATCH(dw5);
 133       OUT_BATCH(dw6);
 134       ADVANCE_BATCH();
 135    } else {
 136       BEGIN_BATCH(7);
 137       OUT_BATCH(_3DSTATE_GS << 16 | (7 - 2));
 138       OUT_BATCH(0); /* prog_bo */
 139       OUT_BATCH((0 << GEN6_GS_SAMPLER_COUNT_SHIFT) |
 140                 (0 << GEN6_GS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
 141       OUT_BATCH(0); /* scratch space base offset */
 142       OUT_BATCH((1 << GEN6_GS_DISPATCH_START_GRF_SHIFT) |
 143                 (0 << GEN6_GS_URB_READ_LENGTH_SHIFT) |
 144                 GEN7_GS_INCLUDE_VERTEX_HANDLES |
 145                 (0 << GEN6_GS_URB_ENTRY_READ_OFFSET_SHIFT));
 146       OUT_BATCH((0 << GEN6_GS_MAX_THREADS_SHIFT) |
 147                 GEN6_GS_STATISTICS_ENABLE);
 148       OUT_BATCH(0);
 149       ADVANCE_BATCH();
 150    }
 151    brw->gs.enabled = active;
 152 }
 153
 154 const struct brw_tracked_state gen7_gs_state = {
 155    .dirty = {
 156       .mesa  = _NEW_TRANSFORM,
 157       .brw   = BRW_NEW_BATCH |
 158                BRW_NEW_CONTEXT |
 159                BRW_NEW_GEOMETRY_PROGRAM,
 160       .cache = BRW_NEW_GS_PROG_DATA
 161    },
 162    .emit = upload_gs_state,
 163 };