panfrost: Set the MALI_WRITES_{Z,S} flags when needed
[mesa.git] / src / gallium / drivers / panfrost / pan_assemble.c
1 /*
2 * © Copyright 2018 Alyssa Rosenzweig
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 */
24
25 #include <stdio.h>
26 #include <stdlib.h>
27 #include <string.h>
28 #include "pan_bo.h"
29 #include "pan_context.h"
30 #include "pan_util.h"
31
32 #include "compiler/nir/nir.h"
33 #include "nir/tgsi_to_nir.h"
34 #include "midgard/midgard_compile.h"
35 #include "util/u_dynarray.h"
36
37 #include "tgsi/tgsi_dump.h"
38
39 void
40 panfrost_shader_compile(
41 struct panfrost_context *ctx,
42 struct mali_shader_meta *meta,
43 enum pipe_shader_ir ir_type,
44 const void *ir,
45 gl_shader_stage stage,
46 struct panfrost_shader_state *state,
47 uint64_t *outputs_written)
48 {
49 struct panfrost_screen *screen = pan_screen(ctx->base.screen);
50 uint8_t *dst;
51
52 nir_shader *s;
53
54 if (ir_type == PIPE_SHADER_IR_NIR) {
55 s = nir_shader_clone(NULL, ir);
56 } else {
57 assert (ir_type == PIPE_SHADER_IR_TGSI);
58 s = tgsi_to_nir(ir, ctx->base.screen);
59 }
60
61 s->info.stage = stage;
62
63 /* Call out to Midgard compiler given the above NIR */
64
65 midgard_program program = {
66 .alpha_ref = state->alpha_state.ref_value
67 };
68
69 midgard_compile_shader_nir(s, &program, false, 0, screen->gpu_id,
70 pan_debug & PAN_DBG_PRECOMPILE);
71
72 /* Prepare the compiled binary for upload */
73 int size = program.compiled.size;
74 dst = program.compiled.data;
75
76 /* Upload the shader. The lookahead tag is ORed on as a tagged pointer.
77 * I bet someone just thought that would be a cute pun. At least,
78 * that's how I'd do it. */
79
80 if (size) {
81 state->bo = panfrost_bo_create(screen, size, PAN_BO_EXECUTE);
82 memcpy(state->bo->cpu, dst, size);
83 meta->shader = state->bo->gpu | program.first_tag;
84 } else {
85 /* no shader */
86 meta->shader = 0x0;
87 }
88
89 util_dynarray_fini(&program.compiled);
90
91 /* Sysvals are prepended */
92 program.uniform_count += program.sysval_count;
93 state->sysval_count = program.sysval_count;
94 memcpy(state->sysval, program.sysvals, sizeof(state->sysval[0]) * state->sysval_count);
95
96 meta->midgard1.uniform_count = MIN2(program.uniform_count, program.uniform_cutoff);
97 meta->midgard1.work_count = program.work_register_count;
98
99 bool vertex_id = s->info.system_values_read & (1 << SYSTEM_VALUE_VERTEX_ID);
100 bool instance_id = s->info.system_values_read & (1 << SYSTEM_VALUE_INSTANCE_ID);
101
102 switch (stage) {
103 case MESA_SHADER_VERTEX:
104 meta->attribute_count = util_bitcount64(s->info.inputs_read);
105 meta->varying_count = util_bitcount64(s->info.outputs_written);
106
107 if (vertex_id)
108 meta->attribute_count = MAX2(meta->attribute_count, PAN_VERTEX_ID + 1);
109
110 if (instance_id)
111 meta->attribute_count = MAX2(meta->attribute_count, PAN_INSTANCE_ID + 1);
112
113 break;
114 case MESA_SHADER_FRAGMENT:
115 meta->attribute_count = 0;
116 meta->varying_count = util_bitcount64(s->info.inputs_read);
117 if (s->info.outputs_written & BITFIELD64_BIT(FRAG_RESULT_DEPTH))
118 state->writes_depth = true;
119 if (s->info.outputs_written & BITFIELD64_BIT(FRAG_RESULT_STENCIL))
120 state->writes_stencil = true;
121 break;
122 case MESA_SHADER_COMPUTE:
123 /* TODO: images */
124 meta->attribute_count = 0;
125 meta->varying_count = 0;
126 break;
127 default:
128 unreachable("Unknown shader state");
129 }
130
131 state->can_discard = s->info.fs.uses_discard;
132 state->writes_point_size = program.writes_point_size;
133 state->reads_point_coord = false;
134 state->helper_invocations = s->info.fs.needs_helper_invocations;
135 state->stack_size = program.tls_size;
136
137 if (outputs_written)
138 *outputs_written = s->info.outputs_written;
139
140 /* Separate as primary uniform count is truncated */
141 state->uniform_count = program.uniform_count;
142
143 meta->midgard1.flags_hi = 8; /* XXX */
144
145 unsigned default_vec1_swizzle = panfrost_get_default_swizzle(1);
146 unsigned default_vec2_swizzle = panfrost_get_default_swizzle(2);
147 unsigned default_vec4_swizzle = panfrost_get_default_swizzle(4);
148
149 /* Iterate the varyings and emit the corresponding descriptor */
150 for (unsigned i = 0; i < meta->varying_count; ++i) {
151 unsigned location = program.varyings[i];
152
153 /* Default to a vec4 varying */
154 struct mali_attr_meta v = {
155 .format = program.varying_type[i],
156 .swizzle = default_vec4_swizzle,
157 .unknown1 = 0x2,
158 };
159
160 /* Check for special cases, otherwise assume general varying */
161
162 if (location == VARYING_SLOT_POS) {
163 if (stage == MESA_SHADER_FRAGMENT)
164 state->reads_frag_coord = true;
165 else
166 v.format = MALI_VARYING_POS;
167 } else if (location == VARYING_SLOT_PSIZ) {
168 v.format = MALI_R16F;
169 v.swizzle = default_vec1_swizzle;
170
171 state->writes_point_size = true;
172 } else if (location == VARYING_SLOT_PNTC) {
173 v.format = MALI_RG16F;
174 v.swizzle = default_vec2_swizzle;
175
176 state->reads_point_coord = true;
177 } else if (location == VARYING_SLOT_FACE) {
178 v.format = MALI_R32I;
179 v.swizzle = default_vec1_swizzle;
180
181 state->reads_face = true;
182 }
183
184 state->varyings[i] = v;
185 state->varyings_loc[i] = location;
186 }
187 }