1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
|
/*
* Copyright (C) 2019 Collabora, Ltd.
* Copyright (C) 2019 Red Hat Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
* Authors (Collabora):
* Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
*
*/
#include "pan_context.h"
#include "util/u_memory.h"
#include "nir_serialize.h"
/* Compute CSOs are tracked like graphics shader CSOs, but are
* considerably simpler. We do not implement multiple
* variants/keying. So the CSO create function just goes ahead and
* compiles the thing. */
static void *
panfrost_create_compute_state(
struct pipe_context *pctx,
const struct pipe_compute_state *cso)
{
struct panfrost_context *ctx = pan_context(pctx);
struct panfrost_shader_variants *so = CALLOC_STRUCT(panfrost_shader_variants);
so->cbase = *cso;
so->is_compute = true;
struct panfrost_shader_state *v = &so->variants[0];
so->variant_count = 1;
so->active_variant = 0;
v->tripipe = malloc(sizeof(struct mali_shader_meta));
if (cso->ir_type == PIPE_SHADER_IR_NIR_SERIALIZED) {
struct blob_reader reader;
const struct pipe_binary_program_header *hdr = cso->prog;
blob_reader_init(&reader, hdr->blob, hdr->num_bytes);
so->cbase.prog = nir_deserialize(NULL, &midgard_nir_options, &reader);
so->cbase.ir_type = PIPE_SHADER_IR_NIR;
}
panfrost_shader_compile(ctx, v->tripipe,
so->cbase.ir_type, so->cbase.prog,
MESA_SHADER_COMPUTE, v, NULL);
return so;
}
static void
panfrost_bind_compute_state(struct pipe_context *pipe, void *cso)
{
struct panfrost_context *ctx = pan_context(pipe);
struct panfrost_shader_variants *variants =
(struct panfrost_shader_variants *) cso;
ctx->shader[PIPE_SHADER_COMPUTE] = variants;
}
static void
panfrost_delete_compute_state(struct pipe_context *pipe, void *cso)
{
free(cso);
}
/* Launch grid is the compute equivalent of draw_vbo, so in this routine, we
* construct the COMPUTE job and some of its payload.
*/
static void
panfrost_launch_grid(struct pipe_context *pipe,
const struct pipe_grid_info *info)
{
struct panfrost_context *ctx = pan_context(pipe);
/* TODO: Do we want a special compute-only batch? */
struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);
ctx->compute_grid = info;
struct mali_job_descriptor_header job = {
.job_type = JOB_TYPE_COMPUTE,
.job_descriptor_size = 1,
.job_barrier = 1
};
/* TODO: Stub */
struct midgard_payload_vertex_tiler *payload = &ctx->payloads[PIPE_SHADER_COMPUTE];
panfrost_emit_for_draw(ctx, false);
/* Compute jobs have a "compute FBD". It's not a real framebuffer
* descriptor - there is no framebuffer - but it takes the place of
* one. As far as I can tell, it's actually the beginning of a
* single-render-target framebuffer descriptor with almost everything
* zeroed out.
*/
struct mali_compute_fbd compute_fbd = {
.unknown1 = {
0, 0x1F, 0, 0, 0, 0, 0, 0
}
};
payload->postfix.framebuffer =
panfrost_upload_transient(batch, &compute_fbd, sizeof(compute_fbd));
/* Invoke according to the grid info */
panfrost_pack_work_groups_compute(&payload->prefix,
info->grid[0], info->grid[1], info->grid[2],
info->block[0], info->block[1], info->block[2], false);
/* Upload the payload */
struct panfrost_transfer transfer = panfrost_allocate_transient(batch, sizeof(job) + sizeof(*payload));
memcpy(transfer.cpu, &job, sizeof(job));
memcpy(transfer.cpu + sizeof(job), payload, sizeof(*payload));
/* Queue the job */
panfrost_scoreboard_queue_compute_job(batch, transfer);
panfrost_flush_all_batches(ctx, true);
}
void
panfrost_compute_context_init(struct pipe_context *pctx)
{
pctx->create_compute_state = panfrost_create_compute_state;
pctx->bind_compute_state = panfrost_bind_compute_state;
pctx->delete_compute_state = panfrost_delete_compute_state;
pctx->launch_grid = panfrost_launch_grid;
}
|