1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
|
/*
* Copyright 2014 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sub license, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
* DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
* OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
* USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* The above copyright notice and this permission notice (including the
* next paragraph) shall be included in all copies or substantial portions
* of the Software.
*
*/
#include <cstring>
#include "ac_binary.h"
#include "ac_llvm_util.h"
#include "ac_llvm_build.h"
#include "util/macros.h"
#include <llvm-c/Core.h>
#include <llvm/Target/TargetMachine.h>
#include <llvm/IR/IRBuilder.h>
#include <llvm/Analysis/TargetLibraryInfo.h>
#include <llvm/Transforms/IPO.h>
#include <llvm/IR/LegacyPassManager.h>
void ac_add_attr_dereferenceable(LLVMValueRef val, uint64_t bytes)
{
llvm::Argument *A = llvm::unwrap<llvm::Argument>(val);
A->addAttr(llvm::Attribute::getWithDereferenceableBytes(A->getContext(), bytes));
}
bool ac_is_sgpr_param(LLVMValueRef arg)
{
llvm::Argument *A = llvm::unwrap<llvm::Argument>(arg);
llvm::AttributeList AS = A->getParent()->getAttributes();
unsigned ArgNo = A->getArgNo();
return AS.hasAttribute(ArgNo + 1, llvm::Attribute::InReg);
}
LLVMValueRef ac_llvm_get_called_value(LLVMValueRef call)
{
return LLVMGetCalledValue(call);
}
bool ac_llvm_is_function(LLVMValueRef v)
{
return LLVMGetValueKind(v) == LLVMFunctionValueKind;
}
LLVMModuleRef ac_create_module(LLVMTargetMachineRef tm, LLVMContextRef ctx)
{
llvm::TargetMachine *TM = reinterpret_cast<llvm::TargetMachine*>(tm);
LLVMModuleRef module = LLVMModuleCreateWithNameInContext("mesa-shader", ctx);
llvm::unwrap(module)->setTargetTriple(TM->getTargetTriple().getTriple());
llvm::unwrap(module)->setDataLayout(TM->createDataLayout());
return module;
}
LLVMBuilderRef ac_create_builder(LLVMContextRef ctx,
enum ac_float_mode float_mode)
{
LLVMBuilderRef builder = LLVMCreateBuilderInContext(ctx);
llvm::FastMathFlags flags;
switch (float_mode) {
case AC_FLOAT_MODE_DEFAULT:
case AC_FLOAT_MODE_DENORM_FLUSH_TO_ZERO:
break;
case AC_FLOAT_MODE_NO_SIGNED_ZEROS_FP_MATH:
flags.setNoSignedZeros();
llvm::unwrap(builder)->setFastMathFlags(flags);
break;
case AC_FLOAT_MODE_UNSAFE_FP_MATH:
flags.setFast();
llvm::unwrap(builder)->setFastMathFlags(flags);
break;
}
return builder;
}
LLVMTargetLibraryInfoRef
ac_create_target_library_info(const char *triple)
{
return reinterpret_cast<LLVMTargetLibraryInfoRef>(new llvm::TargetLibraryInfoImpl(llvm::Triple(triple)));
}
void
ac_dispose_target_library_info(LLVMTargetLibraryInfoRef library_info)
{
delete reinterpret_cast<llvm::TargetLibraryInfoImpl *>(library_info);
}
/* Implementation of raw_pwrite_stream that works on malloc()ed memory for
* better compatibility with C code. */
struct raw_memory_ostream : public llvm::raw_pwrite_stream {
char *buffer;
size_t written;
size_t bufsize;
raw_memory_ostream()
{
buffer = NULL;
written = 0;
bufsize = 0;
SetUnbuffered();
}
~raw_memory_ostream()
{
free(buffer);
}
void clear()
{
written = 0;
}
void take(char *&out_buffer, size_t &out_size)
{
out_buffer = buffer;
out_size = written;
buffer = NULL;
written = 0;
bufsize = 0;
}
void flush() = delete;
void write_impl(const char *ptr, size_t size) override
{
if (unlikely(written + size < written))
abort();
if (written + size > bufsize) {
bufsize = MAX3(1024, written + size, bufsize / 3 * 4);
buffer = (char *)realloc(buffer, bufsize);
if (!buffer) {
fprintf(stderr, "amd: out of memory allocating ELF buffer\n");
abort();
}
}
memcpy(buffer + written, ptr, size);
written += size;
}
void pwrite_impl(const char *ptr, size_t size, uint64_t offset) override
{
assert(offset == (size_t)offset &&
offset + size >= offset && offset + size <= written);
memcpy(buffer + offset, ptr, size);
}
uint64_t current_pos() const override
{
return written;
}
};
/* The LLVM compiler is represented as a pass manager containing passes for
* optimizations, instruction selection, and code generation.
*/
struct ac_compiler_passes {
raw_memory_ostream ostream; /* ELF shader binary stream */
llvm::legacy::PassManager passmgr; /* list of passes */
};
struct ac_compiler_passes *ac_create_llvm_passes(LLVMTargetMachineRef tm)
{
struct ac_compiler_passes *p = new ac_compiler_passes();
if (!p)
return NULL;
llvm::TargetMachine *TM = reinterpret_cast<llvm::TargetMachine*>(tm);
if (TM->addPassesToEmitFile(p->passmgr, p->ostream,
nullptr,
llvm::TargetMachine::CGFT_ObjectFile)) {
fprintf(stderr, "amd: TargetMachine can't emit a file of this type!\n");
delete p;
return NULL;
}
return p;
}
void ac_destroy_llvm_passes(struct ac_compiler_passes *p)
{
delete p;
}
/* This returns false on failure. */
bool ac_compile_module_to_elf(struct ac_compiler_passes *p, LLVMModuleRef module,
char **pelf_buffer, size_t *pelf_size)
{
p->passmgr.run(*llvm::unwrap(module));
p->ostream.take(*pelf_buffer, *pelf_size);
return true;
}
void ac_llvm_add_barrier_noop_pass(LLVMPassManagerRef passmgr)
{
llvm::unwrap(passmgr)->add(llvm::createBarrierNoopPass());
}
void ac_enable_global_isel(LLVMTargetMachineRef tm)
{
reinterpret_cast<llvm::TargetMachine*>(tm)->setGlobalISel(true);
}
LLVMValueRef ac_build_atomic_rmw(struct ac_llvm_context *ctx, LLVMAtomicRMWBinOp op,
LLVMValueRef ptr, LLVMValueRef val,
const char *sync_scope) {
llvm::AtomicRMWInst::BinOp binop;
switch (op) {
case LLVMAtomicRMWBinOpXchg:
binop = llvm::AtomicRMWInst::Xchg;
break;
case LLVMAtomicRMWBinOpAdd:
binop = llvm::AtomicRMWInst::Add;
break;
case LLVMAtomicRMWBinOpSub:
binop = llvm::AtomicRMWInst::Sub;
break;
case LLVMAtomicRMWBinOpAnd:
binop = llvm::AtomicRMWInst::And;
break;
case LLVMAtomicRMWBinOpNand:
binop = llvm::AtomicRMWInst::Nand;
break;
case LLVMAtomicRMWBinOpOr:
binop = llvm::AtomicRMWInst::Or;
break;
case LLVMAtomicRMWBinOpXor:
binop = llvm::AtomicRMWInst::Xor;
break;
case LLVMAtomicRMWBinOpMax:
binop = llvm::AtomicRMWInst::Max;
break;
case LLVMAtomicRMWBinOpMin:
binop = llvm::AtomicRMWInst::Min;
break;
case LLVMAtomicRMWBinOpUMax:
binop = llvm::AtomicRMWInst::UMax;
break;
case LLVMAtomicRMWBinOpUMin:
binop = llvm::AtomicRMWInst::UMin;
break;
default:
unreachable(!"invalid LLVMAtomicRMWBinOp");
break;
}
unsigned SSID = llvm::unwrap(ctx->context)->getOrInsertSyncScopeID(sync_scope);
return llvm::wrap(llvm::unwrap(ctx->builder)->CreateAtomicRMW(
binop, llvm::unwrap(ptr), llvm::unwrap(val),
llvm::AtomicOrdering::SequentiallyConsistent, SSID));
}
LLVMValueRef ac_build_atomic_cmp_xchg(struct ac_llvm_context *ctx, LLVMValueRef ptr,
LLVMValueRef cmp, LLVMValueRef val,
const char *sync_scope) {
unsigned SSID = llvm::unwrap(ctx->context)->getOrInsertSyncScopeID(sync_scope);
return llvm::wrap(llvm::unwrap(ctx->builder)->CreateAtomicCmpXchg(
llvm::unwrap(ptr), llvm::unwrap(cmp), llvm::unwrap(val),
llvm::AtomicOrdering::SequentiallyConsistent,
llvm::AtomicOrdering::SequentiallyConsistent, SSID));
}
|