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 <assert.h>
29 #include "util/u_math.h"
30 #include "pan_encoder.h"
31
32 /* Compute shaders are invoked with a gl_NumWorkGroups X/Y/Z triplet. Vertex
33 * shaders, it turns out, are invoked with the same mechanism, with the triplet
34 * (1, vertex_count, instance_count).
35 *
36 * Alongside this triplet is the gl_WorkGroupSize X/Y/Z triplet.
37 *
38 * Unfortunately, the packing for these triplet into the
39 * mali_vertex_tiler_prefix is a little funky, using a dynamic bitfield. The
40 * routines here exist to pack this */
41
42 void
panfrost_pack_work_groups_compute(struct mali_invocation_packed * out,unsigned num_x,unsigned num_y,unsigned num_z,unsigned size_x,unsigned size_y,unsigned size_z,bool quirk_graphics)43 panfrost_pack_work_groups_compute(
44 struct mali_invocation_packed *out,
45 unsigned num_x,
46 unsigned num_y,
47 unsigned num_z,
48 unsigned size_x,
49 unsigned size_y,
50 unsigned size_z,
51 bool quirk_graphics)
52 {
53 uint32_t packed = 0;
54
55 /* The values needing packing, in order, and the corresponding shifts.
56 * Indicies into shift are off-by-one to make the logic easier */
57
58 unsigned shifts[7] = { 0 };
59
60 unsigned values[6] = {
61 MALI_POSITIVE(size_x),
62 MALI_POSITIVE(size_y),
63 MALI_POSITIVE(size_z),
64 MALI_POSITIVE(num_x),
65 MALI_POSITIVE(num_y),
66 MALI_POSITIVE(num_z),
67 };
68
69 for (unsigned i = 0; i < 6; ++i) {
70 /* OR it in, shifting as required */
71 packed |= (values[i] << shifts[i]);
72
73 /* How many bits did we use? */
74 unsigned bit_count = util_logbase2_ceil(values[i] + 1);
75
76 /* Set the next shift accordingly */
77 shifts[i + 1] = shifts[i] + bit_count;
78 }
79
80 pan_pack(out, INVOCATION, cfg) {
81 cfg.invocations = packed;
82 cfg.size_y_shift = shifts[1];
83 cfg.size_z_shift = shifts[2];
84 cfg.workgroups_x_shift = shifts[3];
85 cfg.workgroups_y_shift = shifts[4];
86 cfg.workgroups_z_shift = shifts[5];
87
88 /* Quirk: for non-instanced graphics, the blob sets
89 * workgroups_z_shift = 32. This doesn't appear to matter to
90 * the hardware, but it's good to be bit-identical. */
91
92 if (quirk_graphics && (num_z <= 1))
93 cfg.workgroups_z_shift = 32;
94
95 /* Quirk: for graphics, >= 2. For compute, 2 without barriers
96 * but equal to workgroups_x_shift with barriers */
97
98 cfg.unknown_shift = quirk_graphics ? 2 : cfg.workgroups_x_shift;
99 }
100 }
101