src/mesa/drivers/dri/i965/brw_vue_map.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 /**
  25  * @file brw_vue_map.c
  26  *
  27  * This file computes the "VUE map" for a (non-fragment) shader stage, which
  28  * describes the layout of its output varyings.  The VUE map is used to match
  29  * outputs from one stage with the inputs of the next.
  30  *
  31  * Largely, varyings can be placed however we like - producers/consumers simply
  32  * have to agree on the layout.  However, there is also a "VUE Header" that
  33  * prescribes a fixed-layout for items that interact with fixed function
  34  * hardware, such as the clipper and rasterizer.
  35  *
  36  * Authors:
  37  *   Paul Berry <stereotype441@gmail.com>
  38  *   Chris Forbes <chrisf@ijw.co.nz>
  39  *   Eric Anholt <eric@anholt.net>
  40  */
  41
  42
  43 #include "main/compiler.h"
  44 #include "brw_context.h"
  45
  46 static inline void
  47 assign_vue_slot(struct brw_vue_map *vue_map, int varying, int slot)
  48 {
  49    /* Make sure this varying hasn't been assigned a slot already */
  50    assert (vue_map->varying_to_slot[varying] == -1);
  51
  52    vue_map->varying_to_slot[varying] = slot;
  53    vue_map->slot_to_varying[slot] = varying;
  54 }
  55
  56 /**
  57  * Compute the VUE map for a shader stage.
  58  */
  59 void
  60 brw_compute_vue_map(const struct brw_device_info *devinfo,
  61                     struct brw_vue_map *vue_map,
  62                     GLbitfield64 slots_valid,
  63                     bool separate)
  64 {
  65    /* Keep using the packed/contiguous layout on old hardware - we only need
  66     * the SSO layout when using geometry/tessellation shaders or 32 FS input
  67     * varyings, which only exist on Gen >= 6.  It's also a bit more efficient.
  68     */
  69    if (devinfo->gen < 6)
  70       separate = false;
  71
  72    vue_map->slots_valid = slots_valid;
  73    vue_map->separate = separate;
  74
  75    /* gl_Layer and gl_ViewportIndex don't get their own varying slots -- they
  76     * are stored in the first VUE slot (VARYING_SLOT_PSIZ).
  77     */
  78    slots_valid &= ~(VARYING_BIT_LAYER | VARYING_BIT_VIEWPORT);
  79
  80    /* Make sure that the values we store in vue_map->varying_to_slot and
  81     * vue_map->slot_to_varying won't overflow the signed chars that are used
  82     * to store them.  Note that since vue_map->slot_to_varying sometimes holds
  83     * values equal to BRW_VARYING_SLOT_COUNT, we need to ensure that
  84     * BRW_VARYING_SLOT_COUNT is <= 127, not 128.
  85     */
  86    STATIC_ASSERT(BRW_VARYING_SLOT_COUNT <= 127);
  87
  88    for (int i = 0; i < BRW_VARYING_SLOT_COUNT; ++i) {
  89       vue_map->varying_to_slot[i] = -1;
  90       vue_map->slot_to_varying[i] = BRW_VARYING_SLOT_PAD;
  91    }
  92
  93    int slot = 0;
  94
  95    /* VUE header: format depends on chip generation and whether clipping is
  96     * enabled.
  97     *
  98     * See the Sandybridge PRM, Volume 2 Part 1, section 1.5.1 (page 30),
  99     * "Vertex URB Entry (VUE) Formats" which describes the VUE header layout.
 100     */
 101    if (devinfo->gen < 6) {
 102       /* There are 8 dwords in VUE header pre-Ironlake:
 103        * dword 0-3 is indices, point width, clip flags.
 104        * dword 4-7 is ndc position
 105        * dword 8-11 is the first vertex data.
 106        *
 107        * On Ironlake the VUE header is nominally 20 dwords, but the hardware
 108        * will accept the same header layout as Gen4 [and should be a bit faster]
 109        */
 110       assign_vue_slot(vue_map, VARYING_SLOT_PSIZ, slot++);
 111       assign_vue_slot(vue_map, BRW_VARYING_SLOT_NDC, slot++);
 112       assign_vue_slot(vue_map, VARYING_SLOT_POS, slot++);
 113    } else {
 114       /* There are 8 or 16 DWs (D0-D15) in VUE header on Sandybridge:
 115        * dword 0-3 of the header is indices, point width, clip flags.
 116        * dword 4-7 is the 4D space position
 117        * dword 8-15 of the vertex header is the user clip distance if
 118        * enabled.
 119        * dword 8-11 or 16-19 is the first vertex element data we fill.
 120        */
 121       assign_vue_slot(vue_map, VARYING_SLOT_PSIZ, slot++);
 122       assign_vue_slot(vue_map, VARYING_SLOT_POS, slot++);
 123       if (slots_valid & BITFIELD64_BIT(VARYING_SLOT_CLIP_DIST0))
 124          assign_vue_slot(vue_map, VARYING_SLOT_CLIP_DIST0, slot++);
 125       if (slots_valid & BITFIELD64_BIT(VARYING_SLOT_CLIP_DIST1))
 126          assign_vue_slot(vue_map, VARYING_SLOT_CLIP_DIST1, slot++);
 127
 128       /* front and back colors need to be consecutive so that we can use
 129        * ATTRIBUTE_SWIZZLE_INPUTATTR_FACING to swizzle them when doing
 130        * two-sided color.
 131        */
 132       if (slots_valid & BITFIELD64_BIT(VARYING_SLOT_COL0))
 133          assign_vue_slot(vue_map, VARYING_SLOT_COL0, slot++);
 134       if (slots_valid & BITFIELD64_BIT(VARYING_SLOT_BFC0))
 135          assign_vue_slot(vue_map, VARYING_SLOT_BFC0, slot++);
 136       if (slots_valid & BITFIELD64_BIT(VARYING_SLOT_COL1))
 137          assign_vue_slot(vue_map, VARYING_SLOT_COL1, slot++);
 138       if (slots_valid & BITFIELD64_BIT(VARYING_SLOT_BFC1))
 139          assign_vue_slot(vue_map, VARYING_SLOT_BFC1, slot++);
 140    }
 141
 142    /* The hardware doesn't care about the rest of the vertex outputs, so we
 143     * can assign them however we like.  For normal programs, we simply assign
 144     * them contiguously.
 145     *
 146     * For separate shader pipelines, we first assign built-in varyings
 147     * contiguous slots.  This works because ARB_separate_shader_objects
 148     * requires that all shaders have matching built-in varying interface
 149     * blocks.  Next, we assign generic varyings based on their location
 150     * (either explicit or linker assigned).  This guarantees a fixed layout.
 151     *
 152     * We generally don't need to assign a slot for VARYING_SLOT_CLIP_VERTEX,
 153     * since it's encoded as the clip distances by emit_clip_distances().
 154     * However, it may be output by transform feedback, and we'd rather not
 155     * recompute state when TF changes, so we just always include it.
 156     */
 157    GLbitfield64 builtins = slots_valid & BITFIELD64_MASK(VARYING_SLOT_VAR0);
 158    while (builtins != 0) {
 159       const int varying = ffsll(builtins) - 1;
 160       if (vue_map->varying_to_slot[varying] == -1) {
 161          assign_vue_slot(vue_map, varying, slot++);
 162       }
 163       builtins &= ~BITFIELD64_BIT(varying);
 164    }
 165
 166    const int first_generic_slot = slot;
 167    GLbitfield64 generics = slots_valid & ~BITFIELD64_MASK(VARYING_SLOT_VAR0);
 168    while (generics != 0) {
 169       const int varying = ffsll(generics) - 1;
 170       if (separate) {
 171          slot = first_generic_slot + varying - VARYING_SLOT_VAR0;
 172          assign_vue_slot(vue_map, varying, slot);
 173       } else {
 174          assign_vue_slot(vue_map, varying, slot++);
 175       }
 176       generics &= ~BITFIELD64_BIT(varying);
 177    }
 178
 179    vue_map->num_slots = separate ? slot + 1 : slot;
 180 }