AMDGPUGenRegisterBankInfo.def
5.88 KB
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
//===- AMDGPUGenRegisterBankInfo.def -----------------------------*- C++ -*-==//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
/// \file
/// This file defines all the static objects used by AMDGPURegisterBankInfo.
/// \todo This should be generated by TableGen.
//===----------------------------------------------------------------------===//
namespace llvm {
namespace AMDGPU {
enum PartialMappingIdx {
None = - 1,
PM_SGPR1 = 1,
PM_SGPR16 = 5,
PM_SGPR32 = 6,
PM_SGPR64 = 7,
PM_SGPR128 = 8,
PM_SGPR256 = 9,
PM_SGPR512 = 10,
PM_SGPR1024 = 11,
PM_VGPR1 = 12,
PM_VGPR16 = 16,
PM_VGPR32 = 17,
PM_VGPR64 = 18,
PM_VGPR128 = 19,
PM_VGPR256 = 20,
PM_VGPR512 = 21,
PM_VGPR1024 = 22,
PM_SGPR96 = 23,
PM_VGPR96 = 24,
PM_AGPR96 = 25,
PM_AGPR32 = 31,
PM_AGPR64 = 32,
PM_AGPR128 = 33,
PM_AGPR256 = 34,
PM_AGPR512 = 35,
PM_AGPR1024 = 36
};
const RegisterBankInfo::PartialMapping PartMappings[] {
// StartIdx, Length, RegBank
{0, 1, VCCRegBank},
{0, 1, SGPRRegBank}, // SGPR begin
{0, 16, SGPRRegBank},
{0, 32, SGPRRegBank},
{0, 64, SGPRRegBank},
{0, 128, SGPRRegBank},
{0, 256, SGPRRegBank},
{0, 512, SGPRRegBank},
{0, 1024, SGPRRegBank},
{0, 1, VGPRRegBank}, // VGPR begin
{0, 16, VGPRRegBank},
{0, 32, VGPRRegBank},
{0, 64, VGPRRegBank},
{0, 128, VGPRRegBank},
{0, 256, VGPRRegBank},
{0, 512, VGPRRegBank},
{0, 1024, VGPRRegBank},
{0, 96, SGPRRegBank},
{0, 96, VGPRRegBank},
{0, 96, AGPRRegBank},
{0, 32, AGPRRegBank}, // AGPR begin
{0, 64, AGPRRegBank},
{0, 128, AGPRRegBank},
{0, 256, AGPRRegBank},
{0, 512, AGPRRegBank},
{0, 1024, AGPRRegBank}
};
const RegisterBankInfo::ValueMapping ValMappings[] {
// VCC
{&PartMappings[0], 1},
// SGPRs
{&PartMappings[1], 1}, // 1
{nullptr, 0}, // Illegal power of 2 sizes
{nullptr, 0},
{nullptr, 0},
{&PartMappings[2], 1}, // 16
{&PartMappings[3], 1}, // 32
{&PartMappings[4], 1}, // 64
{&PartMappings[5], 1}, // 128
{&PartMappings[6], 1}, // 256
{&PartMappings[7], 1}, // 512
{&PartMappings[8], 1}, // 1024
// VGPRs
{&PartMappings[9], 1}, // 1
{nullptr, 0},
{nullptr, 0},
{nullptr, 0},
{&PartMappings[10], 1}, // 16
{&PartMappings[11], 1}, // 32
{&PartMappings[12], 1}, // 64
{&PartMappings[13], 1}, // 128
{&PartMappings[14], 1}, // 256
{&PartMappings[15], 1}, // 512
{&PartMappings[16], 1}, // 1024
{&PartMappings[17], 1},
{&PartMappings[18], 1},
{&PartMappings[19], 1},
// AGPRs
{nullptr, 0},
{nullptr, 0},
{nullptr, 0},
{nullptr, 0},
{nullptr, 0},
{&PartMappings[20], 1}, // 32
{&PartMappings[21], 1}, // 64
{&PartMappings[22], 1}, // 128
{&PartMappings[23], 1}, // 256
{&PartMappings[24], 1}, // 512
{&PartMappings[25], 1} // 1024
};
const RegisterBankInfo::PartialMapping SGPROnly64BreakDown[] {
{0, 32, SGPRRegBank}, // 32-bit op
{0, 32, SGPRRegBank}, // 2x32-bit op
{32, 32, SGPRRegBank},
{0, 64, SGPRRegBank}, // <2x32-bit> op
{0, 32, VGPRRegBank}, // 32-bit op
{0, 32, VGPRRegBank}, // 2x32-bit op
{32, 32, VGPRRegBank},
};
// For some instructions which can operate 64-bit only for the scalar
// version. Otherwise, these need to be split into 2 32-bit operations.
const RegisterBankInfo::ValueMapping ValMappingsSGPR64OnlyVGPR32[] {
/*32-bit sgpr*/ {&SGPROnly64BreakDown[0], 1},
/*2 x 32-bit sgpr*/ {&SGPROnly64BreakDown[1], 2},
/*64-bit sgpr */ {&SGPROnly64BreakDown[3], 1},
/*32-bit vgpr*/ {&SGPROnly64BreakDown[4], 1},
/*2 x 32-bit vgpr*/ {&SGPROnly64BreakDown[5], 2}
};
enum ValueMappingIdx {
SGPRStartIdx = 1,
VGPRStartIdx = 12,
AGPRStartIdx = 26
};
const RegisterBankInfo::ValueMapping *getValueMapping(unsigned BankID,
unsigned Size) {
unsigned Idx;
switch (Size) {
case 1:
if (BankID == AMDGPU::VCCRegBankID)
return &ValMappings[0];
Idx = BankID == AMDGPU::SGPRRegBankID ? PM_SGPR1 : PM_VGPR1;
break;
case 96:
switch (BankID) {
case AMDGPU::VGPRRegBankID:
Idx = PM_VGPR96;
break;
case AMDGPU::SGPRRegBankID:
Idx = PM_SGPR96;
break;
case AMDGPU::AGPRRegBankID:
Idx = PM_AGPR96;
break;
default: llvm_unreachable("Invalid register bank");
}
break;
default:
switch (BankID) {
case AMDGPU::VGPRRegBankID:
Idx = VGPRStartIdx;
break;
case AMDGPU::SGPRRegBankID:
Idx = SGPRStartIdx;
break;
case AMDGPU::AGPRRegBankID:
Idx = AGPRStartIdx;
break;
default: llvm_unreachable("Invalid register bank");
}
Idx += Log2_32_Ceil(Size);
break;
}
assert(Log2_32_Ceil(Size) == Log2_32_Ceil(ValMappings[Idx].BreakDown->Length));
assert(BankID == ValMappings[Idx].BreakDown->RegBank->getID());
return &ValMappings[Idx];
}
const RegisterBankInfo::ValueMapping *getValueMappingSGPR64Only(unsigned BankID,
unsigned Size) {
if (Size != 64)
return getValueMapping(BankID, Size);
if (BankID == AMDGPU::VGPRRegBankID)
return &ValMappingsSGPR64OnlyVGPR32[4];
assert(BankID == AMDGPU::SGPRRegBankID);
return &ValMappingsSGPR64OnlyVGPR32[2];
}
/// Split any 64-bit value into 2 32-bit pieces. Unlike
/// getValueMappingSGPR64Only, this splits both VGPRs and SGPRs.
const RegisterBankInfo::ValueMapping *getValueMappingSplit64(unsigned BankID,
unsigned Size) {
assert(Size == 64);
if (BankID == AMDGPU::VGPRRegBankID)
return &ValMappingsSGPR64OnlyVGPR32[4];
assert(BankID == AMDGPU::SGPRRegBankID);
return &ValMappingsSGPR64OnlyVGPR32[1];
}
} // End AMDGPU namespace.
} // End llvm namespace.