src/gallium/drivers/panfrost/pan_invocation.c

   1 /*
   2  * Copyright (C) 2019 Collabora, Ltd.
   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 FROM,
  20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  21  * SOFTWARE.
  22  *
  23  * Authors (Collabora):
  24  *   Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
  25  *
  26  */
  27
  28 #include "pan_context.h"
  29
  30 /* Compute shaders are invoked with a gl_NumWorkGroups X/Y/Z triplet. Vertex
  31  * shaders, it turns out, are invoked with the same mechanism, with the triplet
  32  * (1, vertex_count, instance_count).
  33  *
  34  * Alongside this triplet is the gl_WorkGroupSize X/Y/Z triplet.
  35  *
  36  * Unfortunately, the packing for these triplet into the
  37  * mali_vertex_tiler_prefix is a little funky, using a dynamic bitfield. The
  38  * routines here exist to pack this */
  39
  40 void
  41 panfrost_pack_work_groups_compute(
  42                 struct mali_vertex_tiler_prefix *out,
  43                 unsigned num_x,
  44                 unsigned num_y,
  45                 unsigned num_z,
  46                 unsigned size_x,
  47                 unsigned size_y,
  48                 unsigned size_z)
  49 {
  50         /* First of all, all 6 values are off-by-one (strictly positive).
  51          * Account for that, first by ensuring all values are strictly positive
  52          * and then by offsetting */
  53
  54         assert(num_x > 0);
  55         assert(num_y > 0);
  56         assert(num_z > 0);
  57
  58         assert(size_x > 0);
  59         assert(size_y > 0);
  60         assert(size_z > 0);
  61
  62         num_x = MALI_POSITIVE(num_x);
  63         num_y = MALI_POSITIVE(num_y);
  64         num_z = MALI_POSITIVE(num_z);
  65
  66         size_x = MALI_POSITIVE(size_x);
  67         size_y = MALI_POSITIVE(size_y);
  68         size_z = MALI_POSITIVE(size_z);
  69
  70         /* Next up is to pack in order */
  71
  72         uint32_t packed = 0;
  73
  74         /* The values needing packing, in order, and the corresponding shifts.
  75          * Indicies into shift are off-by-one to make the logic easier */
  76
  77         unsigned shifts[7] = { 0 };
  78         unsigned values[6] = { size_x, size_y, size_z, num_x, num_y, num_z };
  79
  80         for (unsigned i = 0; i < 6; ++i) {
  81                 /* OR it in, shifting as required */
  82                 packed |= (values[i] << shifts[i]);
  83
  84                 /* How many bits did we use? */
  85                 unsigned bit_count = util_logbase2_ceil(values[i] + 1);
  86
  87                 /* Set the next shift accordingly */
  88                 shifts[i + 1] = shifts[i] + bit_count;
  89         }
  90
  91         /* We're packed, so upload everything */
  92         out->invocation_count = packed;
  93         out->size_y_shift = shifts[1];
  94         out->size_z_shift = shifts[2];
  95         out->workgroups_x_shift = shifts[3];
  96         out->workgroups_y_shift = shifts[4];
  97         out->workgroups_z_shift = shifts[5];
  98
  99         /* Special fields */
 100         out->workgroups_x_shift_2 = MAX2(out->workgroups_x_shift, 2);
 101         out->workgroups_x_shift_3 = out->workgroups_x_shift_2;
 102 }
 103
 104 /* Packs vertex/tiler descriptors simultaneously */
 105 void
 106 panfrost_pack_work_groups_fused(
 107                 struct mali_vertex_tiler_prefix *vertex,
 108                 struct mali_vertex_tiler_prefix *tiler,
 109                 unsigned num_x,
 110                 unsigned num_y,
 111                 unsigned num_z,
 112                 unsigned size_x,
 113                 unsigned size_y,
 114                 unsigned size_z)
 115 {
 116         panfrost_pack_work_groups_compute(vertex, num_x, num_y, num_z, size_x, size_y, size_z);
 117
 118         /* Copy results over */
 119         tiler->invocation_count = vertex->invocation_count;
 120         tiler->size_y_shift = vertex->size_y_shift;
 121         tiler->size_z_shift = vertex->size_z_shift;
 122         tiler->workgroups_x_shift = vertex->workgroups_x_shift;
 123         tiler->workgroups_x_shift_2 = vertex->workgroups_x_shift_2;
 124         tiler->workgroups_y_shift = vertex->workgroups_y_shift;
 125         tiler->workgroups_z_shift = vertex->workgroups_z_shift;
 126
 127         /* Set special fields for each */
 128         vertex->workgroups_x_shift_3 = 5;
 129         tiler->workgroups_x_shift_3 = 6;
 130 }
 131