2 * Copyright 2016 Advanced Micro Devices, Inc.
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:
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
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.
24 #ifndef SI_SHADER_PRIVATE_H
25 #define SI_SHADER_PRIVATE_H
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 #include "ac_llvm_util.h"
33 #include <llvm-c/Core.h>
34 #include <llvm-c/TargetMachine.h>
36 struct pipe_debug_callback
;
37 struct radeon_shader_binary
;
39 #define RADEON_LLVM_MAX_INPUT_SLOTS 32
40 #define RADEON_LLVM_MAX_INPUTS 32 * 4
41 #define RADEON_LLVM_MAX_OUTPUTS 32 * 4
43 #define RADEON_LLVM_INITIAL_CF_DEPTH 4
45 #define RADEON_LLVM_MAX_SYSTEM_VALUES 4
46 #define RADEON_LLVM_MAX_ADDRS 16
50 struct si_shader_context
{
51 struct lp_build_tgsi_context bld_base
;
52 struct gallivm_state gallivm
;
53 struct ac_llvm_context ac
;
54 struct si_shader
*shader
;
55 struct si_screen
*screen
;
57 unsigned type
; /* PIPE_SHADER_* specifies the type of shader. */
59 /* Whether the prolog will be compiled separately. */
62 /** This function is responsible for initilizing the inputs array and will be
63 * called once for each input declared in the TGSI shader.
65 void (*load_input
)(struct si_shader_context
*,
67 const struct tgsi_full_declaration
*decl
,
70 void (*load_system_value
)(struct si_shader_context
*,
72 const struct tgsi_full_declaration
*decl
);
74 void (*declare_memory_region
)(struct si_shader_context
*,
75 const struct tgsi_full_declaration
*decl
);
77 /** This array contains the input values for the shader. Typically these
78 * values will be in the form of a target intrinsic that will inform the
79 * backend how to load the actual inputs to the shader.
81 struct tgsi_full_declaration input_decls
[RADEON_LLVM_MAX_INPUT_SLOTS
];
82 LLVMValueRef inputs
[RADEON_LLVM_MAX_INPUTS
];
83 LLVMValueRef outputs
[RADEON_LLVM_MAX_OUTPUTS
][TGSI_NUM_CHANNELS
];
84 LLVMValueRef addrs
[RADEON_LLVM_MAX_ADDRS
][TGSI_NUM_CHANNELS
];
86 /** This pointer is used to contain the temporary values.
87 * The amount of temporary used in tgsi can't be bound to a max value and
88 * thus we must allocate this array at runtime.
92 LLVMValueRef system_values
[RADEON_LLVM_MAX_SYSTEM_VALUES
];
97 struct si_llvm_flow
*flow
;
99 unsigned flow_depth_max
;
101 struct tgsi_array_info
*temp_arrays
;
102 LLVMValueRef
*temp_array_allocas
;
104 LLVMValueRef undef_alloca
;
106 LLVMValueRef main_fn
;
107 LLVMTypeRef return_type
;
109 int param_streamout_config
;
110 int param_streamout_write_index
;
111 int param_streamout_offset
[4];
113 int param_rel_auto_id
;
114 int param_vs_prim_id
;
115 int param_instance_id
;
116 int param_vertex_index0
;
119 int param_tes_rel_patch_id
;
120 int param_tes_patch_id
;
121 int param_es2gs_offset
;
124 LLVMTargetMachineRef tm
;
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
;
133 /* Preloaded descriptors. */
134 LLVMValueRef esgs_ring
;
135 LLVMValueRef gsvs_ring
[4];
138 LLVMValueRef gs_next_vertex
[4];
139 LLVMValueRef return_value
;
154 LLVMValueRef shared_memory
;
157 static inline struct si_shader_context
*
158 si_shader_context(struct lp_build_tgsi_context
*bld_base
)
160 return (struct si_shader_context
*)bld_base
;
163 void si_llvm_add_attribute(LLVMValueRef F
, const char *name
, int value
);
164 void si_llvm_shader_type(LLVMValueRef F
, unsigned type
);
166 LLVMTargetRef
si_llvm_get_amdgpu_target(const char *triple
);
168 unsigned si_llvm_compile(LLVMModuleRef M
, struct radeon_shader_binary
*binary
,
169 LLVMTargetMachineRef tm
,
170 struct pipe_debug_callback
*debug
);
172 LLVMTypeRef
tgsi2llvmtype(struct lp_build_tgsi_context
*bld_base
,
173 enum tgsi_opcode_type type
);
175 LLVMValueRef
bitcast(struct lp_build_tgsi_context
*bld_base
,
176 enum tgsi_opcode_type type
, LLVMValueRef value
);
178 LLVMValueRef
si_llvm_bound_index(struct si_shader_context
*ctx
,
182 void si_llvm_context_init(struct si_shader_context
*ctx
,
183 struct si_screen
*sscreen
,
184 struct si_shader
*shader
,
185 LLVMTargetMachineRef tm
,
186 const struct tgsi_shader_info
*info
,
187 const struct tgsi_token
*tokens
);
189 void si_llvm_create_func(struct si_shader_context
*ctx
,
191 LLVMTypeRef
*return_types
, unsigned num_return_elems
,
192 LLVMTypeRef
*ParamTypes
, unsigned ParamCount
);
194 void si_llvm_dispose(struct si_shader_context
*ctx
);
196 void si_llvm_finalize_module(struct si_shader_context
*ctx
,
199 LLVMValueRef
si_llvm_emit_fetch_64bit(struct lp_build_tgsi_context
*bld_base
,
200 enum tgsi_opcode_type type
,
204 LLVMValueRef
si_llvm_saturate(struct lp_build_tgsi_context
*bld_base
,
207 LLVMValueRef
si_llvm_emit_fetch(struct lp_build_tgsi_context
*bld_base
,
208 const struct tgsi_full_src_register
*reg
,
209 enum tgsi_opcode_type type
,
212 void si_llvm_emit_store(struct lp_build_tgsi_context
*bld_base
,
213 const struct tgsi_full_instruction
*inst
,
214 const struct tgsi_opcode_info
*info
,
215 LLVMValueRef dst
[4]);
217 void si_shader_context_init_alu(struct lp_build_tgsi_context
*bld_base
);