src/compiler/nir/nir_gather_xfb_info.c

   1 /*
   2  * Copyright © 2018 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 "nir_xfb_info.h"
  25
  26 #include <util/u_math.h>
  27
  28 static void
  29 add_var_xfb_outputs(nir_xfb_info *xfb,
  30                     nir_variable *var,
  31                     unsigned *location,
  32                     unsigned *offset,
  33                     const struct glsl_type *type)
  34 {
  35    if (glsl_type_is_array(type) || glsl_type_is_matrix(type)) {
  36       unsigned length = glsl_get_length(type);
  37       const struct glsl_type *child_type = glsl_get_array_element(type);
  38       for (unsigned i = 0; i < length; i++)
  39          add_var_xfb_outputs(xfb, var, location, offset, child_type);
  40    } else if (glsl_type_is_struct(type)) {
  41       unsigned length = glsl_get_length(type);
  42       for (unsigned i = 0; i < length; i++) {
  43          const struct glsl_type *child_type = glsl_get_struct_field(type, i);
  44          add_var_xfb_outputs(xfb, var, location, offset, child_type);
  45       }
  46    } else {
  47       assert(var->data.xfb_buffer < NIR_MAX_XFB_BUFFERS);
  48       if (xfb->buffers_written & (1 << var->data.xfb_buffer)) {
  49          assert(xfb->strides[var->data.xfb_buffer] == var->data.xfb_stride);
  50          assert(xfb->buffer_to_stream[var->data.xfb_buffer] == var->data.stream);
  51       } else {
  52          xfb->buffers_written |= (1 << var->data.xfb_buffer);
  53          xfb->strides[var->data.xfb_buffer] = var->data.xfb_stride;
  54          xfb->buffer_to_stream[var->data.xfb_buffer] = var->data.stream;
  55       }
  56
  57       assert(var->data.stream < NIR_MAX_XFB_STREAMS);
  58       xfb->streams_written |= (1 << var->data.stream);
  59
  60       unsigned comp_slots = glsl_get_component_slots(type);
  61       unsigned attrib_slots = DIV_ROUND_UP(comp_slots, 4);
  62       assert(attrib_slots == glsl_count_attribute_slots(type, false));
  63
  64       /* Ensure that we don't have, for instance, a dvec2 with a location_frac
  65        * of 2 which would make it crass a location boundary even though it
  66        * fits in a single slot.  However, you can have a dvec3 which crosses
  67        * the slot boundary with a location_frac of 2.
  68        */
  69       assert(DIV_ROUND_UP(var->data.location_frac + comp_slots, 4) == attrib_slots);
  70
  71       assert(var->data.location_frac + comp_slots <= 8);
  72       uint8_t comp_mask = ((1 << comp_slots) - 1) << var->data.location_frac;
  73
  74       assert(attrib_slots <= 2);
  75       for (unsigned s = 0; s < attrib_slots; s++) {
  76          nir_xfb_output_info *output = &xfb->outputs[xfb->output_count++];
  77
  78          output->buffer = var->data.xfb_buffer;
  79          output->offset = *offset + s * 16;
  80          output->location = *location;
  81          output->component_mask = (comp_mask >> (s * 4)) & 0xf;
  82
  83          (*location)++;
  84       }
  85       *offset += comp_slots * 4;
  86    }
  87 }
  88
  89 static int
  90 compare_xfb_output_offsets(const void *_a, const void *_b)
  91 {
  92    const nir_xfb_output_info *a = _a, *b = _b;
  93    return a->offset - b->offset;
  94 }
  95
  96 nir_xfb_info *
  97 nir_gather_xfb_info(const nir_shader *shader, void *mem_ctx)
  98 {
  99    assert(shader->info.stage == MESA_SHADER_VERTEX ||
 100           shader->info.stage == MESA_SHADER_TESS_EVAL ||
 101           shader->info.stage == MESA_SHADER_GEOMETRY);
 102
 103    /* Compute the number of outputs we have.  This is simply the number of
 104     * cumulative locations consumed by all the variables.  If a location is
 105     * represented by multiple variables, then they each count separately in
 106     * number of outputs.
 107     */
 108    unsigned num_outputs = 0;
 109    nir_foreach_variable(var, &shader->outputs) {
 110       if (var->data.explicit_xfb_buffer &&
 111           var->data.explicit_offset) {
 112
 113          num_outputs += glsl_count_attribute_slots(var->type, false);
 114       }
 115    }
 116    if (num_outputs == 0)
 117       return NULL;
 118
 119    nir_xfb_info *xfb = rzalloc_size(mem_ctx, nir_xfb_info_size(num_outputs));
 120
 121    /* Walk the list of outputs and add them to the array */
 122    nir_foreach_variable(var, &shader->outputs) {
 123       if (var->data.explicit_xfb_buffer &&
 124           var->data.explicit_offset) {
 125
 126          unsigned location = var->data.location;
 127          unsigned offset = var->data.offset;
 128          add_var_xfb_outputs(xfb, var, &location, &offset, var->type);
 129       }
 130    }
 131    assert(xfb->output_count == num_outputs);
 132
 133    /* Everything is easier in the state setup code if the list is sorted in
 134     * order of output offset.
 135     */
 136    qsort(xfb->outputs, xfb->output_count, sizeof(xfb->outputs[0]),
 137          compare_xfb_output_offsets);
 138
 139    /* Finally, do a sanity check */
 140    unsigned max_offset[NIR_MAX_XFB_BUFFERS] = {0};
 141    for (unsigned i = 0; i < xfb->output_count; i++) {
 142       assert(xfb->outputs[i].offset >= max_offset[xfb->outputs[i].buffer]);
 143       assert(xfb->outputs[i].component_mask != 0);
 144       unsigned slots = util_bitcount(xfb->outputs[i].component_mask);
 145       max_offset[xfb->outputs[i].buffer] = xfb->outputs[i].offset + slots * 4;
 146    }
 147
 148    return xfb;
 149 }