radeonsi: rename prefixes from radeon to si
[mesa.git] / src / gallium / drivers / radeonsi / si_shader_internal.h
1 /*
2 * Copyright 2016 Advanced Micro Devices, Inc.
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 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the 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 NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
22 */
23
24 #ifndef SI_SHADER_PRIVATE_H
25 #define SI_SHADER_PRIVATE_H
26
27 #include "si_shader.h"
28 #include "gallivm/lp_bld_init.h"
29 #include "gallivm/lp_bld_tgsi.h"
30 #include "tgsi/tgsi_parse.h"
31
32 #include <llvm-c/Core.h>
33 #include <llvm-c/TargetMachine.h>
34
35 struct pipe_debug_callback;
36 struct radeon_shader_binary;
37
38 #define RADEON_LLVM_MAX_INPUT_SLOTS 32
39 #define RADEON_LLVM_MAX_INPUTS 32 * 4
40 #define RADEON_LLVM_MAX_OUTPUTS 32 * 4
41
42 #define RADEON_LLVM_INITIAL_CF_DEPTH 4
43
44 #define RADEON_LLVM_MAX_SYSTEM_VALUES 4
45
46 struct si_llvm_flow;
47
48 struct si_shader_context {
49 struct lp_build_tgsi_soa_context soa;
50 struct gallivm_state gallivm;
51 struct si_shader *shader;
52 struct si_screen *screen;
53
54 unsigned type; /* PIPE_SHADER_* specifies the type of shader. */
55 bool is_gs_copy_shader;
56 /* Whether to generate the optimized shader variant compiled as a whole
57 * (without a prolog and epilog)
58 */
59 bool is_monolithic;
60
61 /** This function is responsible for initilizing the inputs array and will be
62 * called once for each input declared in the TGSI shader.
63 */
64 void (*load_input)(struct si_shader_context *,
65 unsigned input_index,
66 const struct tgsi_full_declaration *decl,
67 LLVMValueRef out[4]);
68
69 void (*load_system_value)(struct si_shader_context *,
70 unsigned index,
71 const struct tgsi_full_declaration *decl);
72
73 void (*declare_memory_region)(struct si_shader_context *,
74 const struct tgsi_full_declaration *decl);
75
76 /** This array contains the input values for the shader. Typically these
77 * values will be in the form of a target intrinsic that will inform the
78 * backend how to load the actual inputs to the shader.
79 */
80 struct tgsi_full_declaration input_decls[RADEON_LLVM_MAX_INPUT_SLOTS];
81 LLVMValueRef inputs[RADEON_LLVM_MAX_INPUTS];
82 LLVMValueRef outputs[RADEON_LLVM_MAX_OUTPUTS][TGSI_NUM_CHANNELS];
83
84 /** This pointer is used to contain the temporary values.
85 * The amount of temporary used in tgsi can't be bound to a max value and
86 * thus we must allocate this array at runtime.
87 */
88 LLVMValueRef *temps;
89 unsigned temps_count;
90 LLVMValueRef system_values[RADEON_LLVM_MAX_SYSTEM_VALUES];
91
92 struct si_llvm_flow *flow;
93 unsigned flow_depth;
94 unsigned flow_depth_max;
95
96 struct tgsi_array_info *temp_arrays;
97 LLVMValueRef *temp_array_allocas;
98
99 LLVMValueRef undef_alloca;
100
101 LLVMValueRef main_fn;
102 LLVMTypeRef return_type;
103
104 int param_streamout_config;
105 int param_streamout_write_index;
106 int param_streamout_offset[4];
107 int param_vertex_id;
108 int param_rel_auto_id;
109 int param_vs_prim_id;
110 int param_instance_id;
111 int param_vertex_index0;
112 int param_tes_u;
113 int param_tes_v;
114 int param_tes_rel_patch_id;
115 int param_tes_patch_id;
116 int param_es2gs_offset;
117 int param_oc_lds;
118
119 /* Sets a bit if the dynamic HS control word was 0x80000000. The bit is
120 * 0x800000 for VS, 0x1 for ES.
121 */
122 int param_tess_offchip;
123
124 LLVMTargetMachineRef tm;
125
126 unsigned invariant_load_md_kind;
127 unsigned range_md_kind;
128 unsigned uniform_md_kind;
129 unsigned fpmath_md_kind;
130 LLVMValueRef fpmath_md_2p5_ulp;
131 LLVMValueRef empty_md;
132
133 /* Preloaded descriptors. */
134 LLVMValueRef esgs_ring;
135 LLVMValueRef gsvs_ring[4];
136
137 LLVMValueRef lds;
138 LLVMValueRef gs_next_vertex[4];
139 LLVMValueRef return_value;
140
141 LLVMTypeRef voidt;
142 LLVMTypeRef i1;
143 LLVMTypeRef i8;
144 LLVMTypeRef i32;
145 LLVMTypeRef i64;
146 LLVMTypeRef i128;
147 LLVMTypeRef f32;
148 LLVMTypeRef v16i8;
149 LLVMTypeRef v2i32;
150 LLVMTypeRef v4i32;
151 LLVMTypeRef v4f32;
152 LLVMTypeRef v8i32;
153
154 LLVMValueRef shared_memory;
155 };
156
157 static inline struct si_shader_context *
158 si_shader_context(struct lp_build_tgsi_context *bld_base)
159 {
160 return (struct si_shader_context*)bld_base;
161 }
162
163 void si_llvm_add_attribute(LLVMValueRef F, const char *name, int value);
164 void si_llvm_shader_type(LLVMValueRef F, unsigned type);
165
166 LLVMTargetRef si_llvm_get_amdgpu_target(const char *triple);
167
168 unsigned si_llvm_compile(LLVMModuleRef M, struct radeon_shader_binary *binary,
169 LLVMTargetMachineRef tm,
170 struct pipe_debug_callback *debug);
171
172 LLVMTypeRef tgsi2llvmtype(struct lp_build_tgsi_context *bld_base,
173 enum tgsi_opcode_type type);
174
175 LLVMValueRef bitcast(struct lp_build_tgsi_context *bld_base,
176 enum tgsi_opcode_type type, LLVMValueRef value);
177
178 LLVMValueRef si_llvm_bound_index(struct si_shader_context *ctx,
179 LLVMValueRef index,
180 unsigned num);
181
182 void si_llvm_context_init(struct si_shader_context *ctx,
183 const char *triple,
184 const struct tgsi_shader_info *info,
185 const struct tgsi_token *tokens);
186
187 void si_llvm_create_func(struct si_shader_context *ctx,
188 LLVMTypeRef *return_types, unsigned num_return_elems,
189 LLVMTypeRef *ParamTypes, unsigned ParamCount);
190
191 void si_llvm_dispose(struct si_shader_context *ctx);
192
193 void si_llvm_finalize_module(struct si_shader_context *ctx,
194 bool run_verifier);
195
196 LLVMValueRef si_llvm_emit_fetch_64bit(struct lp_build_tgsi_context *bld_base,
197 enum tgsi_opcode_type type,
198 LLVMValueRef ptr,
199 LLVMValueRef ptr2);
200
201 LLVMValueRef si_llvm_saturate(struct lp_build_tgsi_context *bld_base,
202 LLVMValueRef value);
203
204 LLVMValueRef si_llvm_emit_fetch(struct lp_build_tgsi_context *bld_base,
205 const struct tgsi_full_src_register *reg,
206 enum tgsi_opcode_type type,
207 unsigned swizzle);
208
209 void si_llvm_emit_store(struct lp_build_tgsi_context *bld_base,
210 const struct tgsi_full_instruction *inst,
211 const struct tgsi_opcode_info *info,
212 LLVMValueRef dst[4]);
213
214 void si_shader_context_init_alu(struct lp_build_tgsi_context *bld_base);
215 void si_prepare_cube_coords(struct lp_build_tgsi_context *bld_base,
216 struct lp_build_emit_data *emit_data,
217 LLVMValueRef *coords_arg,
218 LLVMValueRef *derivs_arg);
219
220 #endif