float-bit-preserving-dagcombines.ll
11.6 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
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
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \
; RUN: | FileCheck -check-prefix=RV32F %s
; RUN: llc -mtriple=riscv32 -mattr=+f,+d -verify-machineinstrs < %s \
; RUN: | FileCheck -check-prefix=RV32FD %s
; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
; RUN: | FileCheck -check-prefix=RV64F %s
; RUN: llc -mtriple=riscv64 -mattr=+f,+d -verify-machineinstrs < %s \
; RUN: | FileCheck -check-prefix=RV64FD %s
; These functions perform extra work to ensure that `%a3` starts in a
; floating-point register, if the machine has them, and the result of
; the bitwise operation is then needed in a floating-point register.
; This should mean the optimisations will fire even if you're using the
; soft-float ABI on a machine with hardware floating-point support.
define float @bitcast_and(float %a1, float %a2) nounwind {
; RV32F-LABEL: bitcast_and:
; RV32F: # %bb.0:
; RV32F-NEXT: fmv.w.x ft0, a1
; RV32F-NEXT: fmv.w.x ft1, a0
; RV32F-NEXT: fadd.s ft0, ft1, ft0
; RV32F-NEXT: fabs.s ft0, ft0
; RV32F-NEXT: fadd.s ft0, ft1, ft0
; RV32F-NEXT: fmv.x.w a0, ft0
; RV32F-NEXT: ret
;
; RV32FD-LABEL: bitcast_and:
; RV32FD: # %bb.0:
; RV32FD-NEXT: fmv.w.x ft0, a1
; RV32FD-NEXT: fmv.w.x ft1, a0
; RV32FD-NEXT: fadd.s ft0, ft1, ft0
; RV32FD-NEXT: fabs.s ft0, ft0
; RV32FD-NEXT: fadd.s ft0, ft1, ft0
; RV32FD-NEXT: fmv.x.w a0, ft0
; RV32FD-NEXT: ret
;
; RV64F-LABEL: bitcast_and:
; RV64F: # %bb.0:
; RV64F-NEXT: fmv.w.x ft0, a1
; RV64F-NEXT: fmv.w.x ft1, a0
; RV64F-NEXT: fadd.s ft0, ft1, ft0
; RV64F-NEXT: fabs.s ft0, ft0
; RV64F-NEXT: fadd.s ft0, ft1, ft0
; RV64F-NEXT: fmv.x.w a0, ft0
; RV64F-NEXT: ret
;
; RV64FD-LABEL: bitcast_and:
; RV64FD: # %bb.0:
; RV64FD-NEXT: fmv.w.x ft0, a1
; RV64FD-NEXT: fmv.w.x ft1, a0
; RV64FD-NEXT: fadd.s ft0, ft1, ft0
; RV64FD-NEXT: fabs.s ft0, ft0
; RV64FD-NEXT: fadd.s ft0, ft1, ft0
; RV64FD-NEXT: fmv.x.w a0, ft0
; RV64FD-NEXT: ret
%a3 = fadd float %a1, %a2
%bc1 = bitcast float %a3 to i32
%and = and i32 %bc1, 2147483647
%bc2 = bitcast i32 %and to float
%a4 = fadd float %a1, %bc2
ret float %a4
}
define double @bitcast_double_and(double %a1, double %a2) nounwind {
; RV32F-LABEL: bitcast_double_and:
; RV32F: # %bb.0:
; RV32F-NEXT: addi sp, sp, -16
; RV32F-NEXT: sw ra, 12(sp)
; RV32F-NEXT: sw s0, 8(sp)
; RV32F-NEXT: sw s1, 4(sp)
; RV32F-NEXT: mv s0, a1
; RV32F-NEXT: mv s1, a0
; RV32F-NEXT: call __adddf3
; RV32F-NEXT: mv a2, a0
; RV32F-NEXT: lui a0, 524288
; RV32F-NEXT: addi a0, a0, -1
; RV32F-NEXT: and a3, a1, a0
; RV32F-NEXT: mv a0, s1
; RV32F-NEXT: mv a1, s0
; RV32F-NEXT: call __adddf3
; RV32F-NEXT: lw s1, 4(sp)
; RV32F-NEXT: lw s0, 8(sp)
; RV32F-NEXT: lw ra, 12(sp)
; RV32F-NEXT: addi sp, sp, 16
; RV32F-NEXT: ret
;
; RV32FD-LABEL: bitcast_double_and:
; RV32FD: # %bb.0:
; RV32FD-NEXT: addi sp, sp, -16
; RV32FD-NEXT: sw a2, 8(sp)
; RV32FD-NEXT: sw a3, 12(sp)
; RV32FD-NEXT: fld ft0, 8(sp)
; RV32FD-NEXT: sw a0, 8(sp)
; RV32FD-NEXT: sw a1, 12(sp)
; RV32FD-NEXT: fld ft1, 8(sp)
; RV32FD-NEXT: fadd.d ft0, ft1, ft0
; RV32FD-NEXT: fabs.d ft0, ft0
; RV32FD-NEXT: fadd.d ft0, ft1, ft0
; RV32FD-NEXT: fsd ft0, 8(sp)
; RV32FD-NEXT: lw a0, 8(sp)
; RV32FD-NEXT: lw a1, 12(sp)
; RV32FD-NEXT: addi sp, sp, 16
; RV32FD-NEXT: ret
;
; RV64F-LABEL: bitcast_double_and:
; RV64F: # %bb.0:
; RV64F-NEXT: addi sp, sp, -16
; RV64F-NEXT: sd ra, 8(sp)
; RV64F-NEXT: sd s0, 0(sp)
; RV64F-NEXT: mv s0, a0
; RV64F-NEXT: call __adddf3
; RV64F-NEXT: addi a1, zero, -1
; RV64F-NEXT: slli a1, a1, 63
; RV64F-NEXT: addi a1, a1, -1
; RV64F-NEXT: and a1, a0, a1
; RV64F-NEXT: mv a0, s0
; RV64F-NEXT: call __adddf3
; RV64F-NEXT: ld s0, 0(sp)
; RV64F-NEXT: ld ra, 8(sp)
; RV64F-NEXT: addi sp, sp, 16
; RV64F-NEXT: ret
;
; RV64FD-LABEL: bitcast_double_and:
; RV64FD: # %bb.0:
; RV64FD-NEXT: fmv.d.x ft0, a1
; RV64FD-NEXT: fmv.d.x ft1, a0
; RV64FD-NEXT: fadd.d ft0, ft1, ft0
; RV64FD-NEXT: fabs.d ft0, ft0
; RV64FD-NEXT: fadd.d ft0, ft1, ft0
; RV64FD-NEXT: fmv.x.d a0, ft0
; RV64FD-NEXT: ret
%a3 = fadd double %a1, %a2
%bc1 = bitcast double %a3 to i64
%and = and i64 %bc1, 9223372036854775807
%bc2 = bitcast i64 %and to double
%a4 = fadd double %a1, %bc2
ret double %a4
}
define float @bitcast_xor(float %a1, float %a2) nounwind {
; RV32F-LABEL: bitcast_xor:
; RV32F: # %bb.0:
; RV32F-NEXT: fmv.w.x ft0, a1
; RV32F-NEXT: fmv.w.x ft1, a0
; RV32F-NEXT: fmul.s ft0, ft1, ft0
; RV32F-NEXT: fneg.s ft0, ft0
; RV32F-NEXT: fmul.s ft0, ft1, ft0
; RV32F-NEXT: fmv.x.w a0, ft0
; RV32F-NEXT: ret
;
; RV32FD-LABEL: bitcast_xor:
; RV32FD: # %bb.0:
; RV32FD-NEXT: fmv.w.x ft0, a1
; RV32FD-NEXT: fmv.w.x ft1, a0
; RV32FD-NEXT: fmul.s ft0, ft1, ft0
; RV32FD-NEXT: fneg.s ft0, ft0
; RV32FD-NEXT: fmul.s ft0, ft1, ft0
; RV32FD-NEXT: fmv.x.w a0, ft0
; RV32FD-NEXT: ret
;
; RV64F-LABEL: bitcast_xor:
; RV64F: # %bb.0:
; RV64F-NEXT: fmv.w.x ft0, a1
; RV64F-NEXT: fmv.w.x ft1, a0
; RV64F-NEXT: fmul.s ft0, ft1, ft0
; RV64F-NEXT: fneg.s ft0, ft0
; RV64F-NEXT: fmul.s ft0, ft1, ft0
; RV64F-NEXT: fmv.x.w a0, ft0
; RV64F-NEXT: ret
;
; RV64FD-LABEL: bitcast_xor:
; RV64FD: # %bb.0:
; RV64FD-NEXT: fmv.w.x ft0, a1
; RV64FD-NEXT: fmv.w.x ft1, a0
; RV64FD-NEXT: fmul.s ft0, ft1, ft0
; RV64FD-NEXT: fneg.s ft0, ft0
; RV64FD-NEXT: fmul.s ft0, ft1, ft0
; RV64FD-NEXT: fmv.x.w a0, ft0
; RV64FD-NEXT: ret
%a3 = fmul float %a1, %a2
%bc1 = bitcast float %a3 to i32
%and = xor i32 %bc1, 2147483648
%bc2 = bitcast i32 %and to float
%a4 = fmul float %a1, %bc2
ret float %a4
}
define double @bitcast_double_xor(double %a1, double %a2) nounwind {
; RV32F-LABEL: bitcast_double_xor:
; RV32F: # %bb.0:
; RV32F-NEXT: addi sp, sp, -16
; RV32F-NEXT: sw ra, 12(sp)
; RV32F-NEXT: sw s0, 8(sp)
; RV32F-NEXT: sw s1, 4(sp)
; RV32F-NEXT: mv s0, a1
; RV32F-NEXT: mv s1, a0
; RV32F-NEXT: call __muldf3
; RV32F-NEXT: mv a2, a0
; RV32F-NEXT: lui a0, 524288
; RV32F-NEXT: xor a3, a1, a0
; RV32F-NEXT: mv a0, s1
; RV32F-NEXT: mv a1, s0
; RV32F-NEXT: call __muldf3
; RV32F-NEXT: lw s1, 4(sp)
; RV32F-NEXT: lw s0, 8(sp)
; RV32F-NEXT: lw ra, 12(sp)
; RV32F-NEXT: addi sp, sp, 16
; RV32F-NEXT: ret
;
; RV32FD-LABEL: bitcast_double_xor:
; RV32FD: # %bb.0:
; RV32FD-NEXT: addi sp, sp, -16
; RV32FD-NEXT: sw a2, 8(sp)
; RV32FD-NEXT: sw a3, 12(sp)
; RV32FD-NEXT: fld ft0, 8(sp)
; RV32FD-NEXT: sw a0, 8(sp)
; RV32FD-NEXT: sw a1, 12(sp)
; RV32FD-NEXT: fld ft1, 8(sp)
; RV32FD-NEXT: fmul.d ft0, ft1, ft0
; RV32FD-NEXT: fneg.d ft0, ft0
; RV32FD-NEXT: fmul.d ft0, ft1, ft0
; RV32FD-NEXT: fsd ft0, 8(sp)
; RV32FD-NEXT: lw a0, 8(sp)
; RV32FD-NEXT: lw a1, 12(sp)
; RV32FD-NEXT: addi sp, sp, 16
; RV32FD-NEXT: ret
;
; RV64F-LABEL: bitcast_double_xor:
; RV64F: # %bb.0:
; RV64F-NEXT: addi sp, sp, -16
; RV64F-NEXT: sd ra, 8(sp)
; RV64F-NEXT: sd s0, 0(sp)
; RV64F-NEXT: mv s0, a0
; RV64F-NEXT: call __muldf3
; RV64F-NEXT: addi a1, zero, -1
; RV64F-NEXT: slli a1, a1, 63
; RV64F-NEXT: xor a1, a0, a1
; RV64F-NEXT: mv a0, s0
; RV64F-NEXT: call __muldf3
; RV64F-NEXT: ld s0, 0(sp)
; RV64F-NEXT: ld ra, 8(sp)
; RV64F-NEXT: addi sp, sp, 16
; RV64F-NEXT: ret
;
; RV64FD-LABEL: bitcast_double_xor:
; RV64FD: # %bb.0:
; RV64FD-NEXT: fmv.d.x ft0, a1
; RV64FD-NEXT: fmv.d.x ft1, a0
; RV64FD-NEXT: fmul.d ft0, ft1, ft0
; RV64FD-NEXT: fneg.d ft0, ft0
; RV64FD-NEXT: fmul.d ft0, ft1, ft0
; RV64FD-NEXT: fmv.x.d a0, ft0
; RV64FD-NEXT: ret
%a3 = fmul double %a1, %a2
%bc1 = bitcast double %a3 to i64
%and = xor i64 %bc1, 9223372036854775808
%bc2 = bitcast i64 %and to double
%a4 = fmul double %a1, %bc2
ret double %a4
}
define float @bitcast_or(float %a1, float %a2) nounwind {
; RV32F-LABEL: bitcast_or:
; RV32F: # %bb.0:
; RV32F-NEXT: fmv.w.x ft0, a1
; RV32F-NEXT: fmv.w.x ft1, a0
; RV32F-NEXT: fmul.s ft0, ft1, ft0
; RV32F-NEXT: fabs.s ft0, ft0
; RV32F-NEXT: fneg.s ft0, ft0
; RV32F-NEXT: fmul.s ft0, ft1, ft0
; RV32F-NEXT: fmv.x.w a0, ft0
; RV32F-NEXT: ret
;
; RV32FD-LABEL: bitcast_or:
; RV32FD: # %bb.0:
; RV32FD-NEXT: fmv.w.x ft0, a1
; RV32FD-NEXT: fmv.w.x ft1, a0
; RV32FD-NEXT: fmul.s ft0, ft1, ft0
; RV32FD-NEXT: fabs.s ft0, ft0
; RV32FD-NEXT: fneg.s ft0, ft0
; RV32FD-NEXT: fmul.s ft0, ft1, ft0
; RV32FD-NEXT: fmv.x.w a0, ft0
; RV32FD-NEXT: ret
;
; RV64F-LABEL: bitcast_or:
; RV64F: # %bb.0:
; RV64F-NEXT: fmv.w.x ft0, a1
; RV64F-NEXT: fmv.w.x ft1, a0
; RV64F-NEXT: fmul.s ft0, ft1, ft0
; RV64F-NEXT: fabs.s ft0, ft0
; RV64F-NEXT: fneg.s ft0, ft0
; RV64F-NEXT: fmul.s ft0, ft1, ft0
; RV64F-NEXT: fmv.x.w a0, ft0
; RV64F-NEXT: ret
;
; RV64FD-LABEL: bitcast_or:
; RV64FD: # %bb.0:
; RV64FD-NEXT: fmv.w.x ft0, a1
; RV64FD-NEXT: fmv.w.x ft1, a0
; RV64FD-NEXT: fmul.s ft0, ft1, ft0
; RV64FD-NEXT: fabs.s ft0, ft0
; RV64FD-NEXT: fneg.s ft0, ft0
; RV64FD-NEXT: fmul.s ft0, ft1, ft0
; RV64FD-NEXT: fmv.x.w a0, ft0
; RV64FD-NEXT: ret
%a3 = fmul float %a1, %a2
%bc1 = bitcast float %a3 to i32
%and = or i32 %bc1, 2147483648
%bc2 = bitcast i32 %and to float
%a4 = fmul float %a1, %bc2
ret float %a4
}
define double @bitcast_double_or(double %a1, double %a2) nounwind {
; RV32F-LABEL: bitcast_double_or:
; RV32F: # %bb.0:
; RV32F-NEXT: addi sp, sp, -16
; RV32F-NEXT: sw ra, 12(sp)
; RV32F-NEXT: sw s0, 8(sp)
; RV32F-NEXT: sw s1, 4(sp)
; RV32F-NEXT: mv s0, a1
; RV32F-NEXT: mv s1, a0
; RV32F-NEXT: call __muldf3
; RV32F-NEXT: mv a2, a0
; RV32F-NEXT: lui a0, 524288
; RV32F-NEXT: or a3, a1, a0
; RV32F-NEXT: mv a0, s1
; RV32F-NEXT: mv a1, s0
; RV32F-NEXT: call __muldf3
; RV32F-NEXT: lw s1, 4(sp)
; RV32F-NEXT: lw s0, 8(sp)
; RV32F-NEXT: lw ra, 12(sp)
; RV32F-NEXT: addi sp, sp, 16
; RV32F-NEXT: ret
;
; RV32FD-LABEL: bitcast_double_or:
; RV32FD: # %bb.0:
; RV32FD-NEXT: addi sp, sp, -16
; RV32FD-NEXT: sw a2, 8(sp)
; RV32FD-NEXT: sw a3, 12(sp)
; RV32FD-NEXT: fld ft0, 8(sp)
; RV32FD-NEXT: sw a0, 8(sp)
; RV32FD-NEXT: sw a1, 12(sp)
; RV32FD-NEXT: fld ft1, 8(sp)
; RV32FD-NEXT: fmul.d ft0, ft1, ft0
; RV32FD-NEXT: fabs.d ft0, ft0
; RV32FD-NEXT: fneg.d ft0, ft0
; RV32FD-NEXT: fmul.d ft0, ft1, ft0
; RV32FD-NEXT: fsd ft0, 8(sp)
; RV32FD-NEXT: lw a0, 8(sp)
; RV32FD-NEXT: lw a1, 12(sp)
; RV32FD-NEXT: addi sp, sp, 16
; RV32FD-NEXT: ret
;
; RV64F-LABEL: bitcast_double_or:
; RV64F: # %bb.0:
; RV64F-NEXT: addi sp, sp, -16
; RV64F-NEXT: sd ra, 8(sp)
; RV64F-NEXT: sd s0, 0(sp)
; RV64F-NEXT: mv s0, a0
; RV64F-NEXT: call __muldf3
; RV64F-NEXT: addi a1, zero, -1
; RV64F-NEXT: slli a1, a1, 63
; RV64F-NEXT: or a1, a0, a1
; RV64F-NEXT: mv a0, s0
; RV64F-NEXT: call __muldf3
; RV64F-NEXT: ld s0, 0(sp)
; RV64F-NEXT: ld ra, 8(sp)
; RV64F-NEXT: addi sp, sp, 16
; RV64F-NEXT: ret
;
; RV64FD-LABEL: bitcast_double_or:
; RV64FD: # %bb.0:
; RV64FD-NEXT: fmv.d.x ft0, a1
; RV64FD-NEXT: fmv.d.x ft1, a0
; RV64FD-NEXT: fmul.d ft0, ft1, ft0
; RV64FD-NEXT: fabs.d ft0, ft0
; RV64FD-NEXT: fneg.d ft0, ft0
; RV64FD-NEXT: fmul.d ft0, ft1, ft0
; RV64FD-NEXT: fmv.x.d a0, ft0
; RV64FD-NEXT: ret
%a3 = fmul double %a1, %a2
%bc1 = bitcast double %a3 to i64
%and = or i64 %bc1, 9223372036854775808
%bc2 = bitcast i64 %and to double
%a4 = fmul double %a1, %bc2
ret double %a4
}