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
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
|
//==- SystemZInstrFP.td - Floating-point SystemZ instructions --*- tblgen-*-==//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
// Select instructions
//===----------------------------------------------------------------------===//
// C's ?: operator for floating-point operands.
def SelectF32 : SelectWrapper<FP32>;
def SelectF64 : SelectWrapper<FP64>;
def SelectF128 : SelectWrapper<FP128>;
defm CondStoreF32 : CondStores<FP32, nonvolatile_store,
nonvolatile_load, bdxaddr20only>;
defm CondStoreF64 : CondStores<FP64, nonvolatile_store,
nonvolatile_load, bdxaddr20only>;
//===----------------------------------------------------------------------===//
// Move instructions
//===----------------------------------------------------------------------===//
// Load zero.
let neverHasSideEffects = 1, isAsCheapAsAMove = 1, isMoveImm = 1 in {
def LZER : InherentRRE<"lzer", 0xB374, FP32, (fpimm0)>;
def LZDR : InherentRRE<"lzdr", 0xB375, FP64, (fpimm0)>;
def LZXR : InherentRRE<"lzxr", 0xB376, FP128, (fpimm0)>;
}
// Moves between two floating-point registers.
let neverHasSideEffects = 1 in {
def LER : UnaryRR <"le", 0x38, null_frag, FP32, FP32>;
def LDR : UnaryRR <"ld", 0x28, null_frag, FP64, FP64>;
def LXR : UnaryRRE<"lx", 0xB365, null_frag, FP128, FP128>;
}
// Moves between two floating-point registers that also set the condition
// codes.
let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0xF in {
defm LTEBR : LoadAndTestRRE<"lteb", 0xB302, FP32>;
defm LTDBR : LoadAndTestRRE<"ltdb", 0xB312, FP64>;
defm LTXBR : LoadAndTestRRE<"ltxb", 0xB342, FP128>;
}
def : CompareZeroFP<LTEBRCompare, FP32>;
def : CompareZeroFP<LTDBRCompare, FP64>;
def : CompareZeroFP<LTXBRCompare, FP128>;
// Moves between 64-bit integer and floating-point registers.
def LGDR : UnaryRRE<"lgd", 0xB3CD, bitconvert, GR64, FP64>;
def LDGR : UnaryRRE<"ldg", 0xB3C1, bitconvert, FP64, GR64>;
// fcopysign with an FP32 result.
let isCodeGenOnly = 1 in {
def CPSDRss : BinaryRRF<"cpsd", 0xB372, fcopysign, FP32, FP32>;
def CPSDRsd : BinaryRRF<"cpsd", 0xB372, fcopysign, FP32, FP64>;
}
// The sign of an FP128 is in the high register.
def : Pat<(fcopysign FP32:$src1, FP128:$src2),
(CPSDRsd FP32:$src1, (EXTRACT_SUBREG FP128:$src2, subreg_high))>;
// fcopysign with an FP64 result.
let isCodeGenOnly = 1 in
def CPSDRds : BinaryRRF<"cpsd", 0xB372, fcopysign, FP64, FP32>;
def CPSDRdd : BinaryRRF<"cpsd", 0xB372, fcopysign, FP64, FP64>;
// The sign of an FP128 is in the high register.
def : Pat<(fcopysign FP64:$src1, FP128:$src2),
(CPSDRdd FP64:$src1, (EXTRACT_SUBREG FP128:$src2, subreg_high))>;
// fcopysign with an FP128 result. Use "upper" as the high half and leave
// the low half as-is.
class CopySign128<RegisterOperand cls, dag upper>
: Pat<(fcopysign FP128:$src1, cls:$src2),
(INSERT_SUBREG FP128:$src1, upper, subreg_high)>;
def : CopySign128<FP32, (CPSDRds (EXTRACT_SUBREG FP128:$src1, subreg_high),
FP32:$src2)>;
def : CopySign128<FP64, (CPSDRdd (EXTRACT_SUBREG FP128:$src1, subreg_high),
FP64:$src2)>;
def : CopySign128<FP128, (CPSDRdd (EXTRACT_SUBREG FP128:$src1, subreg_high),
(EXTRACT_SUBREG FP128:$src2, subreg_high))>;
defm LoadStoreF32 : MVCLoadStore<load, store, f32, MVCWrapper, 4>;
defm LoadStoreF64 : MVCLoadStore<load, store, f64, MVCWrapper, 8>;
defm LoadStoreF128 : MVCLoadStore<load, store, f128, MVCWrapper, 16>;
//===----------------------------------------------------------------------===//
// Load instructions
//===----------------------------------------------------------------------===//
let canFoldAsLoad = 1, SimpleBDXLoad = 1 in {
defm LE : UnaryRXPair<"le", 0x78, 0xED64, load, FP32, 4>;
defm LD : UnaryRXPair<"ld", 0x68, 0xED65, load, FP64, 8>;
// These instructions are split after register allocation, so we don't
// want a custom inserter.
let Has20BitOffset = 1, HasIndex = 1, Is128Bit = 1 in {
def LX : Pseudo<(outs FP128:$dst), (ins bdxaddr20only128:$src),
[(set FP128:$dst, (load bdxaddr20only128:$src))]>;
}
}
//===----------------------------------------------------------------------===//
// Store instructions
//===----------------------------------------------------------------------===//
let SimpleBDXStore = 1 in {
defm STE : StoreRXPair<"ste", 0x70, 0xED66, store, FP32, 4>;
defm STD : StoreRXPair<"std", 0x60, 0xED67, store, FP64, 8>;
// These instructions are split after register allocation, so we don't
// want a custom inserter.
let Has20BitOffset = 1, HasIndex = 1, Is128Bit = 1 in {
def STX : Pseudo<(outs), (ins FP128:$src, bdxaddr20only128:$dst),
[(store FP128:$src, bdxaddr20only128:$dst)]>;
}
}
//===----------------------------------------------------------------------===//
// Conversion instructions
//===----------------------------------------------------------------------===//
// Convert floating-point values to narrower representations, rounding
// according to the current mode. The destination of LEXBR and LDXBR
// is a 128-bit value, but only the first register of the pair is used.
def LEDBR : UnaryRRE<"ledb", 0xB344, fround, FP32, FP64>;
def LEXBR : UnaryRRE<"lexb", 0xB346, null_frag, FP128, FP128>;
def LDXBR : UnaryRRE<"ldxb", 0xB345, null_frag, FP128, FP128>;
def : Pat<(f32 (fround FP128:$src)),
(EXTRACT_SUBREG (LEXBR FP128:$src), subreg_32bit)>;
def : Pat<(f64 (fround FP128:$src)),
(EXTRACT_SUBREG (LDXBR FP128:$src), subreg_high)>;
// Extend register floating-point values to wider representations.
def LDEBR : UnaryRRE<"ldeb", 0xB304, fextend, FP64, FP32>;
def LXEBR : UnaryRRE<"lxeb", 0xB306, fextend, FP128, FP32>;
def LXDBR : UnaryRRE<"lxdb", 0xB305, fextend, FP128, FP64>;
// Extend memory floating-point values to wider representations.
def LDEB : UnaryRXE<"ldeb", 0xED04, extloadf32, FP64, 4>;
def LXEB : UnaryRXE<"lxeb", 0xED06, extloadf32, FP128, 4>;
def LXDB : UnaryRXE<"lxdb", 0xED05, extloadf64, FP128, 8>;
// Convert a signed integer register value to a floating-point one.
def CEFBR : UnaryRRE<"cefb", 0xB394, sint_to_fp, FP32, GR32>;
def CDFBR : UnaryRRE<"cdfb", 0xB395, sint_to_fp, FP64, GR32>;
def CXFBR : UnaryRRE<"cxfb", 0xB396, sint_to_fp, FP128, GR32>;
def CEGBR : UnaryRRE<"cegb", 0xB3A4, sint_to_fp, FP32, GR64>;
def CDGBR : UnaryRRE<"cdgb", 0xB3A5, sint_to_fp, FP64, GR64>;
def CXGBR : UnaryRRE<"cxgb", 0xB3A6, sint_to_fp, FP128, GR64>;
// Convert a floating-point register value to a signed integer value,
// with the second operand (modifier M3) specifying the rounding mode.
let Defs = [CC] in {
def CFEBR : UnaryRRF<"cfeb", 0xB398, GR32, FP32>;
def CFDBR : UnaryRRF<"cfdb", 0xB399, GR32, FP64>;
def CFXBR : UnaryRRF<"cfxb", 0xB39A, GR32, FP128>;
def CGEBR : UnaryRRF<"cgeb", 0xB3A8, GR64, FP32>;
def CGDBR : UnaryRRF<"cgdb", 0xB3A9, GR64, FP64>;
def CGXBR : UnaryRRF<"cgxb", 0xB3AA, GR64, FP128>;
}
// fp_to_sint always rounds towards zero, which is modifier value 5.
def : Pat<(i32 (fp_to_sint FP32:$src)), (CFEBR 5, FP32:$src)>;
def : Pat<(i32 (fp_to_sint FP64:$src)), (CFDBR 5, FP64:$src)>;
def : Pat<(i32 (fp_to_sint FP128:$src)), (CFXBR 5, FP128:$src)>;
def : Pat<(i64 (fp_to_sint FP32:$src)), (CGEBR 5, FP32:$src)>;
def : Pat<(i64 (fp_to_sint FP64:$src)), (CGDBR 5, FP64:$src)>;
def : Pat<(i64 (fp_to_sint FP128:$src)), (CGXBR 5, FP128:$src)>;
//===----------------------------------------------------------------------===//
// Unary arithmetic
//===----------------------------------------------------------------------===//
// Negation (Load Complement).
let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0xF in {
def LCEBR : UnaryRRE<"lceb", 0xB303, fneg, FP32, FP32>;
def LCDBR : UnaryRRE<"lcdb", 0xB313, fneg, FP64, FP64>;
def LCXBR : UnaryRRE<"lcxb", 0xB343, fneg, FP128, FP128>;
}
// Absolute value (Load Positive).
let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0xF in {
def LPEBR : UnaryRRE<"lpeb", 0xB300, fabs, FP32, FP32>;
def LPDBR : UnaryRRE<"lpdb", 0xB310, fabs, FP64, FP64>;
def LPXBR : UnaryRRE<"lpxb", 0xB340, fabs, FP128, FP128>;
}
// Negative absolute value (Load Negative).
let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0xF in {
def LNEBR : UnaryRRE<"lneb", 0xB301, fnabs, FP32, FP32>;
def LNDBR : UnaryRRE<"lndb", 0xB311, fnabs, FP64, FP64>;
def LNXBR : UnaryRRE<"lnxb", 0xB341, fnabs, FP128, FP128>;
}
// Square root.
def SQEBR : UnaryRRE<"sqeb", 0xB314, fsqrt, FP32, FP32>;
def SQDBR : UnaryRRE<"sqdb", 0xB315, fsqrt, FP64, FP64>;
def SQXBR : UnaryRRE<"sqxb", 0xB316, fsqrt, FP128, FP128>;
def SQEB : UnaryRXE<"sqeb", 0xED14, loadu<fsqrt>, FP32, 4>;
def SQDB : UnaryRXE<"sqdb", 0xED15, loadu<fsqrt>, FP64, 8>;
// Round to an integer, with the second operand (modifier M3) specifying
// the rounding mode. These forms always check for inexact conditions.
def FIEBR : UnaryRRF<"fieb", 0xB357, FP32, FP32>;
def FIDBR : UnaryRRF<"fidb", 0xB35F, FP64, FP64>;
def FIXBR : UnaryRRF<"fixb", 0xB347, FP128, FP128>;
// Extended forms of the previous three instructions. M4 can be set to 4
// to suppress detection of inexact conditions.
def FIEBRA : UnaryRRF4<"fiebra", 0xB357, FP32, FP32>,
Requires<[FeatureFPExtension]>;
def FIDBRA : UnaryRRF4<"fidbra", 0xB35F, FP64, FP64>,
Requires<[FeatureFPExtension]>;
def FIXBRA : UnaryRRF4<"fixbra", 0xB347, FP128, FP128>,
Requires<[FeatureFPExtension]>;
// frint rounds according to the current mode (modifier 0) and detects
// inexact conditions.
def : Pat<(frint FP32:$src), (FIEBR 0, FP32:$src)>;
def : Pat<(frint FP64:$src), (FIDBR 0, FP64:$src)>;
def : Pat<(frint FP128:$src), (FIXBR 0, FP128:$src)>;
let Predicates = [FeatureFPExtension] in {
// fnearbyint is like frint but does not detect inexact conditions.
def : Pat<(fnearbyint FP32:$src), (FIEBRA 0, FP32:$src, 4)>;
def : Pat<(fnearbyint FP64:$src), (FIDBRA 0, FP64:$src, 4)>;
def : Pat<(fnearbyint FP128:$src), (FIXBRA 0, FP128:$src, 4)>;
// floor is no longer allowed to raise an inexact condition,
// so restrict it to the cases where the condition can be suppressed.
// Mode 7 is round towards -inf.
def : Pat<(ffloor FP32:$src), (FIEBRA 7, FP32:$src, 4)>;
def : Pat<(ffloor FP64:$src), (FIDBRA 7, FP64:$src, 4)>;
def : Pat<(ffloor FP128:$src), (FIXBRA 7, FP128:$src, 4)>;
// Same idea for ceil, where mode 6 is round towards +inf.
def : Pat<(fceil FP32:$src), (FIEBRA 6, FP32:$src, 4)>;
def : Pat<(fceil FP64:$src), (FIDBRA 6, FP64:$src, 4)>;
def : Pat<(fceil FP128:$src), (FIXBRA 6, FP128:$src, 4)>;
// Same idea for trunc, where mode 5 is round towards zero.
def : Pat<(ftrunc FP32:$src), (FIEBRA 5, FP32:$src, 4)>;
def : Pat<(ftrunc FP64:$src), (FIDBRA 5, FP64:$src, 4)>;
def : Pat<(ftrunc FP128:$src), (FIXBRA 5, FP128:$src, 4)>;
// Same idea for round, where mode 1 is round towards nearest with
// ties away from zero.
def : Pat<(frnd FP32:$src), (FIEBRA 1, FP32:$src, 4)>;
def : Pat<(frnd FP64:$src), (FIDBRA 1, FP64:$src, 4)>;
def : Pat<(frnd FP128:$src), (FIXBRA 1, FP128:$src, 4)>;
}
//===----------------------------------------------------------------------===//
// Binary arithmetic
//===----------------------------------------------------------------------===//
// Addition.
let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0xF in {
let isCommutable = 1 in {
def AEBR : BinaryRRE<"aeb", 0xB30A, fadd, FP32, FP32>;
def ADBR : BinaryRRE<"adb", 0xB31A, fadd, FP64, FP64>;
def AXBR : BinaryRRE<"axb", 0xB34A, fadd, FP128, FP128>;
}
def AEB : BinaryRXE<"aeb", 0xED0A, fadd, FP32, load, 4>;
def ADB : BinaryRXE<"adb", 0xED1A, fadd, FP64, load, 8>;
}
// Subtraction.
let Defs = [CC], CCValues = 0xF, CompareZeroCCMask = 0xF in {
def SEBR : BinaryRRE<"seb", 0xB30B, fsub, FP32, FP32>;
def SDBR : BinaryRRE<"sdb", 0xB31B, fsub, FP64, FP64>;
def SXBR : BinaryRRE<"sxb", 0xB34B, fsub, FP128, FP128>;
def SEB : BinaryRXE<"seb", 0xED0B, fsub, FP32, load, 4>;
def SDB : BinaryRXE<"sdb", 0xED1B, fsub, FP64, load, 8>;
}
// Multiplication.
let isCommutable = 1 in {
def MEEBR : BinaryRRE<"meeb", 0xB317, fmul, FP32, FP32>;
def MDBR : BinaryRRE<"mdb", 0xB31C, fmul, FP64, FP64>;
def MXBR : BinaryRRE<"mxb", 0xB34C, fmul, FP128, FP128>;
}
def MEEB : BinaryRXE<"meeb", 0xED17, fmul, FP32, load, 4>;
def MDB : BinaryRXE<"mdb", 0xED1C, fmul, FP64, load, 8>;
// f64 multiplication of two FP32 registers.
def MDEBR : BinaryRRE<"mdeb", 0xB30C, null_frag, FP64, FP32>;
def : Pat<(fmul (f64 (fextend FP32:$src1)), (f64 (fextend FP32:$src2))),
(MDEBR (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
FP32:$src1, subreg_32bit), FP32:$src2)>;
// f64 multiplication of an FP32 register and an f32 memory.
def MDEB : BinaryRXE<"mdeb", 0xED0C, null_frag, FP64, load, 4>;
def : Pat<(fmul (f64 (fextend FP32:$src1)),
(f64 (extloadf32 bdxaddr12only:$addr))),
(MDEB (INSERT_SUBREG (f64 (IMPLICIT_DEF)), FP32:$src1, subreg_32bit),
bdxaddr12only:$addr)>;
// f128 multiplication of two FP64 registers.
def MXDBR : BinaryRRE<"mxdb", 0xB307, null_frag, FP128, FP64>;
def : Pat<(fmul (f128 (fextend FP64:$src1)), (f128 (fextend FP64:$src2))),
(MXDBR (INSERT_SUBREG (f128 (IMPLICIT_DEF)),
FP64:$src1, subreg_high), FP64:$src2)>;
// f128 multiplication of an FP64 register and an f64 memory.
def MXDB : BinaryRXE<"mxdb", 0xED07, null_frag, FP128, load, 8>;
def : Pat<(fmul (f128 (fextend FP64:$src1)),
(f128 (extloadf64 bdxaddr12only:$addr))),
(MXDB (INSERT_SUBREG (f128 (IMPLICIT_DEF)), FP64:$src1, subreg_high),
bdxaddr12only:$addr)>;
// Fused multiply-add.
def MAEBR : TernaryRRD<"maeb", 0xB30E, z_fma, FP32>;
def MADBR : TernaryRRD<"madb", 0xB31E, z_fma, FP64>;
def MAEB : TernaryRXF<"maeb", 0xED0E, z_fma, FP32, load, 4>;
def MADB : TernaryRXF<"madb", 0xED1E, z_fma, FP64, load, 8>;
// Fused multiply-subtract.
def MSEBR : TernaryRRD<"mseb", 0xB30F, z_fms, FP32>;
def MSDBR : TernaryRRD<"msdb", 0xB31F, z_fms, FP64>;
def MSEB : TernaryRXF<"mseb", 0xED0F, z_fms, FP32, load, 4>;
def MSDB : TernaryRXF<"msdb", 0xED1F, z_fms, FP64, load, 8>;
// Division.
def DEBR : BinaryRRE<"deb", 0xB30D, fdiv, FP32, FP32>;
def DDBR : BinaryRRE<"ddb", 0xB31D, fdiv, FP64, FP64>;
def DXBR : BinaryRRE<"dxb", 0xB34D, fdiv, FP128, FP128>;
def DEB : BinaryRXE<"deb", 0xED0D, fdiv, FP32, load, 4>;
def DDB : BinaryRXE<"ddb", 0xED1D, fdiv, FP64, load, 8>;
//===----------------------------------------------------------------------===//
// Comparisons
//===----------------------------------------------------------------------===//
let Defs = [CC], CCValues = 0xF in {
def CEBR : CompareRRE<"ceb", 0xB309, z_cmp, FP32, FP32>;
def CDBR : CompareRRE<"cdb", 0xB319, z_cmp, FP64, FP64>;
def CXBR : CompareRRE<"cxb", 0xB349, z_cmp, FP128, FP128>;
def CEB : CompareRXE<"ceb", 0xED09, z_cmp, FP32, load, 4>;
def CDB : CompareRXE<"cdb", 0xED19, z_cmp, FP64, load, 8>;
}
//===----------------------------------------------------------------------===//
// Peepholes
//===----------------------------------------------------------------------===//
def : Pat<(f32 fpimmneg0), (LCEBR (LZER))>;
def : Pat<(f64 fpimmneg0), (LCDBR (LZDR))>;
def : Pat<(f128 fpimmneg0), (LCXBR (LZXR))>;
|
