1edcbd46d721dc1f26ccab379682b77efe867a3d
[mesa.git] / src / gallium / drivers / radeon / radeon_llvm.h
1 /*
2 * Copyright 2011 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 * 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 * Authors: Tom Stellard <thomas.stellard@amd.com>
24 *
25 */
26
27 #ifndef RADEON_LLVM_H
28 #define RADEON_LLVM_H
29
30 #include <llvm-c/Core.h>
31 #include "gallivm/lp_bld_init.h"
32 #include "gallivm/lp_bld_tgsi.h"
33
34 #define RADEON_LLVM_MAX_INPUTS 16 * 4
35 #define RADEON_LLVM_MAX_OUTPUTS 16 * 4
36 #define RADEON_LLVM_MAX_BRANCH_DEPTH 16
37 #define RADEON_LLVM_MAX_LOOP_DEPTH 16
38
39 #define RADEON_LLVM_MAX_SYSTEM_VALUES 4
40
41 struct radeon_llvm_branch {
42 LLVMBasicBlockRef endif_block;
43 LLVMBasicBlockRef if_block;
44 LLVMBasicBlockRef else_block;
45 unsigned has_else;
46 };
47
48 struct radeon_llvm_loop {
49 LLVMBasicBlockRef loop_block;
50 LLVMBasicBlockRef endloop_block;
51 };
52
53 struct radeon_llvm_context {
54
55 struct lp_build_tgsi_soa_context soa;
56
57 unsigned chip_class;
58 unsigned type;
59 unsigned face_input;
60 unsigned two_side;
61 unsigned clip_vertex;
62 struct r600_shader_io * r600_inputs;
63 struct r600_shader_io * r600_outputs;
64 struct pipe_stream_output_info *stream_outputs;
65 unsigned color_buffer_count;
66 unsigned fs_color_all;
67
68 /*=== Front end configuration ===*/
69
70 /* Special Intrinsics */
71
72 /** Write to an output register: float store_output(float, i32) */
73 const char * store_output_intr;
74
75 /** Swizzle a vector value: <4 x float> swizzle(<4 x float>, i32)
76 * The swizzle is an unsigned integer that encodes a TGSI_SWIZZLE_* value
77 * in 2-bits.
78 * Swizzle{0-1} = X Channel
79 * Swizzle{2-3} = Y Channel
80 * Swizzle{4-5} = Z Channel
81 * Swizzle{6-7} = W Channel
82 */
83 const char * swizzle_intr;
84
85 /* Instructions that are not described by any of the TGSI opcodes. */
86
87 /** This function is responsible for initilizing the inputs array and will be
88 * called once for each input declared in the TGSI shader.
89 */
90 void (*load_input)(struct radeon_llvm_context *,
91 unsigned input_index,
92 const struct tgsi_full_declaration *decl);
93
94 void (*load_system_value)(struct radeon_llvm_context *,
95 unsigned index,
96 const struct tgsi_full_declaration *decl);
97
98 /** User data to use with the callbacks */
99 void * userdata;
100
101 /** This array contains the input values for the shader. Typically these
102 * values will be in the form of a target intrinsic that will inform the
103 * backend how to load the actual inputs to the shader.
104 */
105 LLVMValueRef inputs[RADEON_LLVM_MAX_INPUTS];
106 LLVMValueRef outputs[RADEON_LLVM_MAX_OUTPUTS][TGSI_NUM_CHANNELS];
107 unsigned output_reg_count;
108
109 LLVMValueRef system_values[RADEON_LLVM_MAX_SYSTEM_VALUES];
110
111 unsigned reserved_reg_count;
112 /*=== Private Members ===*/
113
114 struct radeon_llvm_branch branch[RADEON_LLVM_MAX_BRANCH_DEPTH];
115 struct radeon_llvm_loop loop[RADEON_LLVM_MAX_LOOP_DEPTH];
116
117 unsigned branch_depth;
118 unsigned loop_depth;
119
120
121 LLVMValueRef main_fn;
122
123 struct gallivm_state gallivm;
124 };
125
126 static inline LLVMValueRef bitcast(
127 struct lp_build_tgsi_context * bld_base,
128 enum tgsi_opcode_type type,
129 LLVMValueRef value
130 )
131 {
132 LLVMBuilderRef builder = bld_base->base.gallivm->builder;
133 LLVMContextRef ctx = bld_base->base.gallivm->context;
134 LLVMTypeRef dst_type;
135
136 switch (type) {
137 case TGSI_TYPE_UNSIGNED:
138 case TGSI_TYPE_SIGNED:
139 dst_type = LLVMInt32TypeInContext(ctx);
140 break;
141 case TGSI_TYPE_UNTYPED:
142 case TGSI_TYPE_FLOAT:
143 dst_type = LLVMFloatTypeInContext(ctx);
144 break;
145 default:
146 dst_type = 0;
147 break;
148 }
149
150 if (dst_type)
151 return LLVMBuildBitCast(builder, value, dst_type, "");
152 else
153 return value;
154 }
155
156
157 void radeon_llvm_emit_prepare_cube_coords(struct lp_build_tgsi_context * bld_base,
158 struct lp_build_emit_data * emit_data,
159 unsigned coord_arg);
160
161 void radeon_llvm_context_init(struct radeon_llvm_context * ctx);
162
163 void radeon_llvm_dispose(struct radeon_llvm_context * ctx);
164
165 inline static struct radeon_llvm_context * radeon_llvm_context(
166 struct lp_build_tgsi_context * bld_base)
167 {
168 return (struct radeon_llvm_context*)bld_base;
169 }
170
171 unsigned radeon_llvm_reg_index_soa(unsigned index, unsigned chan);
172
173 void radeon_llvm_finalize_module(struct radeon_llvm_context * ctx);
174
175 LLVMValueRef
176 build_intrinsic(LLVMBuilderRef builder,
177 const char *name,
178 LLVMTypeRef ret_type,
179 LLVMValueRef *args,
180 unsigned num_args,
181 LLVMAttribute attr);
182
183 void
184 build_tgsi_intrinsic_nomem(
185 const struct lp_build_tgsi_action * action,
186 struct lp_build_tgsi_context * bld_base,
187 struct lp_build_emit_data * emit_data);
188
189
190
191 #endif /* RADEON_LLVM_H */