1 /* -*- mesa-c++ -*-
2 *
3 * Copyright (c) 2022 Collabora LTD
4 *
5 * Author: Gert Wollny <gert.wollny@collabora.com>
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * on the rights to use, copy, modify, merge, publish, distribute, sub
11 * license, and/or sell copies of the Software, and to permit persons to whom
12 * the Software is furnished to do so, subject to the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the next
15 * paragraph) shall be included in all copies or substantial portions of the
16 * Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
19 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
20 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
21 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
22 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
23 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
24 * USE OR OTHER DEALINGS IN THE SOFTWARE.
25 */
26
27 #include "sfn_shader_cs.h"
28 #include "sfn_instr_fetch.h"
29
30
31 namespace r600 {
32
ComputeShader(UNUSED const r600_shader_key & key)33 ComputeShader::ComputeShader(UNUSED const r600_shader_key& key):
34 Shader("CS")
35 {
36
37 }
38
do_scan_instruction(UNUSED nir_instr * instr)39 bool ComputeShader::do_scan_instruction(UNUSED nir_instr *instr)
40 {
41 return false;
42 }
43
do_allocate_reserved_registers()44 int ComputeShader::do_allocate_reserved_registers()
45 {
46 auto& vf = value_factory();
47
48 const int thread_id_sel = 0;
49 const int wg_id_sel = 1;
50
51 for (int i = 0; i < 3; ++i) {
52 m_local_invocation_id[i] = vf.allocate_pinned_register(thread_id_sel, i);
53 m_local_invocation_id[i]->pin_live_range(true);
54
55 m_workgroup_id[i] = vf.allocate_pinned_register(wg_id_sel, i);
56 m_workgroup_id[i]->pin_live_range(true);
57 }
58 return 2;
59 }
60
process_stage_intrinsic(nir_intrinsic_instr * instr)61 bool ComputeShader::process_stage_intrinsic(nir_intrinsic_instr *instr)
62 {
63 switch (instr->intrinsic) {
64 case nir_intrinsic_load_local_invocation_id:
65 return emit_load_3vec(instr, m_local_invocation_id);
66 case nir_intrinsic_load_workgroup_id:
67 return emit_load_3vec(instr, m_workgroup_id);
68 case nir_intrinsic_load_num_workgroups:
69 return emit_load_num_workgroups(instr);
70 default:
71 return false;
72 }
73 }
74
do_get_shader_info(r600_shader * sh_info)75 void ComputeShader::do_get_shader_info(r600_shader *sh_info)
76 {
77 sh_info->processor_type = PIPE_SHADER_COMPUTE;
78 }
79
read_prop(UNUSED std::istream & is)80 bool ComputeShader::read_prop(UNUSED std::istream& is)
81 {
82 return true;
83 }
84
do_print_properties(UNUSED std::ostream & os) const85 void ComputeShader::do_print_properties(UNUSED std::ostream& os) const
86 {
87
88 }
89
emit_load_num_workgroups(nir_intrinsic_instr * instr)90 bool ComputeShader::emit_load_num_workgroups(nir_intrinsic_instr* instr)
91 {
92 auto zero = value_factory().temp_register();
93
94 emit_instruction(new AluInstr(op1_mov, zero, value_factory().inline_const(ALU_SRC_0, 0),
95 AluInstr::last_write));
96 auto dest = value_factory().dest_vec4(instr->dest, pin_group);
97
98 auto ir = new LoadFromBuffer(dest, {0,1,2,7}, zero, 16,
99 R600_BUFFER_INFO_CONST_BUFFER,
100 nullptr, fmt_32_32_32_32);
101
102 ir->set_fetch_flag(LoadFromBuffer::srf_mode);
103 ir->reset_fetch_flag(LoadFromBuffer::format_comp_signed);
104 ir->set_num_format(vtx_nf_int);
105 emit_instruction(ir);
106 return true;
107
108 }
109
emit_load_3vec(nir_intrinsic_instr * instr,const std::array<PRegister,3> & src)110 bool ComputeShader::emit_load_3vec(nir_intrinsic_instr* instr, const std::array<PRegister,3>& src)
111 {
112 auto& vf = value_factory();
113
114 for (int i = 0; i < 3; ++i) {
115 auto dest = vf.dest(instr->dest, i, pin_none);
116 emit_instruction(new AluInstr(op1_mov, dest, src[i], i == 2 ? AluInstr::last_write : AluInstr::write));
117 }
118 return true;
119 }
120
121 }
122