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
|
; RUN: llc < %s -mcpu=cortex-a57 -aarch64-a57-fp-load-balancing-override=1 -aarch64-a57-fp-load-balancing-force-all | FileCheck %s --check-prefix CHECK --check-prefix CHECK-A57 --check-prefix CHECK-EVEN
; RUN: llc < %s -mcpu=cortex-a57 -aarch64-a57-fp-load-balancing-override=2 -aarch64-a57-fp-load-balancing-force-all | FileCheck %s --check-prefix CHECK --check-prefix CHECK-A57 --check-prefix CHECK-ODD
; RUN: llc < %s -mcpu=cortex-a53 -aarch64-a57-fp-load-balancing-override=1 -aarch64-a57-fp-load-balancing-force-all | FileCheck %s --check-prefix CHECK --check-prefix CHECK-A53 --check-prefix CHECK-EVEN
; RUN: llc < %s -mcpu=cortex-a53 -aarch64-a57-fp-load-balancing-override=2 -aarch64-a57-fp-load-balancing-force-all | FileCheck %s --check-prefix CHECK --check-prefix CHECK-A53 --check-prefix CHECK-ODD
; Test the AArch64A57FPLoadBalancing pass. This pass relies heavily on register allocation, so
; our test strategy is to:
; * Force the pass to always perform register swapping even if the dest register is of the
; correct color already (-force-all)
; * Force the pass to ignore all hints it obtained from regalloc (-deterministic-balance),
; and run it twice, once where it always hints odd, and once where it always hints even.
;
; We then use regex magic to check that in the two cases the register allocation is
; different; this is what gives us the testing coverage and distinguishes cases where
; the pass has done some work versus accidental regalloc.
target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
target triple = "aarch64"
; Non-overlapping groups - shouldn't need any changing at all.
; CHECK-LABEL: f1:
; CHECK-EVEN: fmadd [[x:d[0-9]*[02468]]]
; CHECK-ODD: fmadd [[x:d[0-9]*[13579]]]
; CHECK: fmadd [[x]]
; CHECK: fmsub [[x]]
; CHECK: fmadd [[x]]
; CHECK: str [[x]]
define void @f1(double* nocapture readonly %p, double* nocapture %q) #0 {
entry:
%0 = load double, double* %p, align 8
%arrayidx1 = getelementptr inbounds double, double* %p, i64 1
%1 = load double, double* %arrayidx1, align 8
%arrayidx2 = getelementptr inbounds double, double* %p, i64 2
%2 = load double, double* %arrayidx2, align 8
%arrayidx3 = getelementptr inbounds double, double* %p, i64 3
%3 = load double, double* %arrayidx3, align 8
%arrayidx4 = getelementptr inbounds double, double* %p, i64 4
%4 = load double, double* %arrayidx4, align 8
%mul = fmul fast double %0, %1
%add = fadd fast double %mul, %4
%mul5 = fmul fast double %1, %2
%add6 = fadd fast double %mul5, %add
%mul7 = fmul fast double %1, %3
%sub = fsub fast double %add6, %mul7
%mul8 = fmul fast double %2, %3
%add9 = fadd fast double %mul8, %sub
store double %add9, double* %q, align 8
%arrayidx11 = getelementptr inbounds double, double* %p, i64 5
%5 = load double, double* %arrayidx11, align 8
%arrayidx12 = getelementptr inbounds double, double* %p, i64 6
%6 = load double, double* %arrayidx12, align 8
%arrayidx13 = getelementptr inbounds double, double* %p, i64 7
%7 = load double, double* %arrayidx13, align 8
%mul15 = fmul fast double %6, %7
%mul16 = fmul fast double %0, %5
%add17 = fadd fast double %mul16, %mul15
%mul18 = fmul fast double %5, %6
%add19 = fadd fast double %mul18, %add17
%arrayidx20 = getelementptr inbounds double, double* %q, i64 1
store double %add19, double* %arrayidx20, align 8
ret void
}
; Overlapping groups - coloring needed.
; CHECK-LABEL: f2:
; CHECK-EVEN: fmadd [[x:d[0-9]*[02468]]]
; CHECK-EVEN: fmul [[y:d[0-9]*[13579]]]
; CHECK-ODD: fmadd [[x:d[0-9]*[13579]]]
; CHECK-ODD: fmul [[y:d[0-9]*[02468]]]
; CHECK: fmadd [[x]]
; CHECK: fmadd [[y]]
; CHECK: fmsub [[x]]
; CHECK: fmadd [[y]]
; CHECK: fmadd [[x]]
; CHECK-A57: stp [[x]], [[y]]
; CHECK-A53-DAG: str [[x]]
; CHECK-A53-DAG: str [[y]]
define void @f2(double* nocapture readonly %p, double* nocapture %q) #0 {
entry:
%0 = load double, double* %p, align 8
%arrayidx1 = getelementptr inbounds double, double* %p, i64 1
%1 = load double, double* %arrayidx1, align 8
%arrayidx2 = getelementptr inbounds double, double* %p, i64 2
%2 = load double, double* %arrayidx2, align 8
%arrayidx3 = getelementptr inbounds double, double* %p, i64 3
%3 = load double, double* %arrayidx3, align 8
%arrayidx4 = getelementptr inbounds double, double* %p, i64 4
%4 = load double, double* %arrayidx4, align 8
%arrayidx5 = getelementptr inbounds double, double* %p, i64 5
%5 = load double, double* %arrayidx5, align 8
%arrayidx6 = getelementptr inbounds double, double* %p, i64 6
%6 = load double, double* %arrayidx6, align 8
%arrayidx7 = getelementptr inbounds double, double* %p, i64 7
%7 = load double, double* %arrayidx7, align 8
%mul = fmul fast double %0, %1
%add = fadd fast double %mul, %7
%mul8 = fmul fast double %5, %6
%mul9 = fmul fast double %1, %2
%add10 = fadd fast double %mul9, %add
%mul11 = fmul fast double %3, %4
%add12 = fadd fast double %mul11, %mul8
%mul13 = fmul fast double %1, %3
%sub = fsub fast double %add10, %mul13
%mul14 = fmul fast double %4, %5
%add15 = fadd fast double %mul14, %add12
%mul16 = fmul fast double %2, %3
%add17 = fadd fast double %mul16, %sub
store double %add17, double* %q, align 8
%arrayidx19 = getelementptr inbounds double, double* %q, i64 1
store double %add15, double* %arrayidx19, align 8
ret void
}
; Dest register is live on block exit - fixup needed.
; CHECK-LABEL: f3:
; CHECK-EVEN: fmadd [[x:d[0-9]*[02468]]]
; CHECK-ODD: fmadd [[x:d[0-9]*[13579]]]
; CHECK: fmadd [[x]]
; CHECK: fmsub [[x]]
; CHECK: fmadd [[y:d[0-9]+]], {{.*}}, [[x]]
; CHECK: str [[y]]
define void @f3(double* nocapture readonly %p, double* nocapture %q) #0 {
entry:
%0 = load double, double* %p, align 8
%arrayidx1 = getelementptr inbounds double, double* %p, i64 1
%1 = load double, double* %arrayidx1, align 8
%arrayidx2 = getelementptr inbounds double, double* %p, i64 2
%2 = load double, double* %arrayidx2, align 8
%arrayidx3 = getelementptr inbounds double, double* %p, i64 3
%3 = load double, double* %arrayidx3, align 8
%arrayidx4 = getelementptr inbounds double, double* %p, i64 4
%4 = load double, double* %arrayidx4, align 8
%mul = fmul fast double %0, %1
%add = fadd fast double %mul, %4
%mul5 = fmul fast double %1, %2
%add6 = fadd fast double %mul5, %add
%mul7 = fmul fast double %1, %3
%sub = fsub fast double %add6, %mul7
%mul8 = fmul fast double %2, %3
%add9 = fadd fast double %mul8, %sub
%cmp = fcmp oeq double %3, 0.000000e+00
br i1 %cmp, label %if.then, label %if.end
if.then: ; preds = %entry
tail call void bitcast (void (...)* @g to void ()*)() #2
br label %if.end
if.end: ; preds = %if.then, %entry
store double %add9, double* %q, align 8
ret void
}
declare void @g(...) #1
; Single precision version of f2.
; CHECK-LABEL: f4:
; CHECK-EVEN: fmadd [[x:s[0-9]*[02468]]]
; CHECK-EVEN: fmul [[y:s[0-9]*[13579]]]
; CHECK-ODD: fmadd [[x:s[0-9]*[13579]]]
; CHECK-ODD: fmul [[y:s[0-9]*[02468]]]
; CHECK: fmadd [[x]]
; CHECK: fmadd [[y]]
; CHECK: fmsub [[x]]
; CHECK: fmadd [[y]]
; CHECK: fmadd [[x]]
; CHECK-A57: stp [[x]], [[y]]
; CHECK-A53-DAG: str [[x]]
; CHECK-A53-DAG: str [[y]]
define void @f4(float* nocapture readonly %p, float* nocapture %q) #0 {
entry:
%0 = load float, float* %p, align 4
%arrayidx1 = getelementptr inbounds float, float* %p, i64 1
%1 = load float, float* %arrayidx1, align 4
%arrayidx2 = getelementptr inbounds float, float* %p, i64 2
%2 = load float, float* %arrayidx2, align 4
%arrayidx3 = getelementptr inbounds float, float* %p, i64 3
%3 = load float, float* %arrayidx3, align 4
%arrayidx4 = getelementptr inbounds float, float* %p, i64 4
%4 = load float, float* %arrayidx4, align 4
%arrayidx5 = getelementptr inbounds float, float* %p, i64 5
%5 = load float, float* %arrayidx5, align 4
%arrayidx6 = getelementptr inbounds float, float* %p, i64 6
%6 = load float, float* %arrayidx6, align 4
%arrayidx7 = getelementptr inbounds float, float* %p, i64 7
%7 = load float, float* %arrayidx7, align 4
%mul = fmul fast float %0, %1
%add = fadd fast float %mul, %7
%mul8 = fmul fast float %5, %6
%mul9 = fmul fast float %1, %2
%add10 = fadd fast float %mul9, %add
%mul11 = fmul fast float %3, %4
%add12 = fadd fast float %mul11, %mul8
%mul13 = fmul fast float %1, %3
%sub = fsub fast float %add10, %mul13
%mul14 = fmul fast float %4, %5
%add15 = fadd fast float %mul14, %add12
%mul16 = fmul fast float %2, %3
%add17 = fadd fast float %mul16, %sub
store float %add17, float* %q, align 4
%arrayidx19 = getelementptr inbounds float, float* %q, i64 1
store float %add15, float* %arrayidx19, align 4
ret void
}
; Single precision version of f3
; CHECK-LABEL: f5:
; CHECK-EVEN: fmadd [[x:s[0-9]*[02468]]]
; CHECK-ODD: fmadd [[x:s[0-9]*[13579]]]
; CHECK: fmadd [[x]]
; CHECK: fmsub [[x]]
; CHECK: fmadd [[y:s[0-9]+]], {{.*}}, [[x]]
; CHECK: str [[y]]
define void @f5(float* nocapture readonly %p, float* nocapture %q) #0 {
entry:
%0 = load float, float* %p, align 4
%arrayidx1 = getelementptr inbounds float, float* %p, i64 1
%1 = load float, float* %arrayidx1, align 4
%arrayidx2 = getelementptr inbounds float, float* %p, i64 2
%2 = load float, float* %arrayidx2, align 4
%arrayidx3 = getelementptr inbounds float, float* %p, i64 3
%3 = load float, float* %arrayidx3, align 4
%arrayidx4 = getelementptr inbounds float, float* %p, i64 4
%4 = load float, float* %arrayidx4, align 4
%mul = fmul fast float %0, %1
%add = fadd fast float %mul, %4
%mul5 = fmul fast float %1, %2
%add6 = fadd fast float %mul5, %add
%mul7 = fmul fast float %1, %3
%sub = fsub fast float %add6, %mul7
%mul8 = fmul fast float %2, %3
%add9 = fadd fast float %mul8, %sub
%cmp = fcmp oeq float %3, 0.000000e+00
br i1 %cmp, label %if.then, label %if.end
if.then: ; preds = %entry
tail call void bitcast (void (...)* @g to void ()*)() #2
br label %if.end
if.end: ; preds = %if.then, %entry
store float %add9, float* %q, align 4
ret void
}
; Test that regmask clobbering stops a chain sequence.
; CHECK-LABEL: f6:
; CHECK-EVEN: fmadd [[x:d[0-9]*[02468]]]
; CHECK-ODD: fmadd [[x:d[0-9]*[13579]]]
; CHECK: fmadd [[x]]
; CHECK: fmsub [[x]]
; CHECK: fmadd d0, {{.*}}, [[x]]
; CHECK: bl hh
; CHECK: str d0
define void @f6(double* nocapture readonly %p, double* nocapture %q) #0 {
entry:
%0 = load double, double* %p, align 8
%arrayidx1 = getelementptr inbounds double, double* %p, i64 1
%1 = load double, double* %arrayidx1, align 8
%arrayidx2 = getelementptr inbounds double, double* %p, i64 2
%2 = load double, double* %arrayidx2, align 8
%arrayidx3 = getelementptr inbounds double, double* %p, i64 3
%3 = load double, double* %arrayidx3, align 8
%arrayidx4 = getelementptr inbounds double, double* %p, i64 4
%4 = load double, double* %arrayidx4, align 8
%mul = fmul fast double %0, %1
%add = fadd fast double %mul, %4
%mul5 = fmul fast double %1, %2
%add6 = fadd fast double %mul5, %add
%mul7 = fmul fast double %1, %3
%sub = fsub fast double %add6, %mul7
%mul8 = fmul fast double %2, %3
%add9 = fadd fast double %mul8, %sub
%call = tail call double @hh(double %add9) #2
store double %call, double* %q, align 8
ret void
}
declare double @hh(double) #1
; Check that we correctly deal with repeated operands.
; The following testcase creates:
; %D1<def> = FADDDrr %D0<kill>, %D0
; We'll get a crash if we naively look at the first operand, remove it
; from the substitution list then look at the second operand.
; CHECK: fmadd [[x:d[0-9]+]]
; CHECK: fadd d1, [[x]], [[x]]
define void @f7(double* nocapture readonly %p, double* nocapture %q) #0 {
entry:
%0 = load double, double* %p, align 8
%arrayidx1 = getelementptr inbounds double, double* %p, i64 1
%1 = load double, double* %arrayidx1, align 8
%arrayidx2 = getelementptr inbounds double, double* %p, i64 2
%2 = load double, double* %arrayidx2, align 8
%arrayidx3 = getelementptr inbounds double, double* %p, i64 3
%3 = load double, double* %arrayidx3, align 8
%arrayidx4 = getelementptr inbounds double, double* %p, i64 4
%4 = load double, double* %arrayidx4, align 8
%mul = fmul fast double %0, %1
%add = fadd fast double %mul, %4
%mul5 = fmul fast double %1, %2
%add6 = fadd fast double %mul5, %add
%mul7 = fmul fast double %1, %3
%sub = fsub fast double %add6, %mul7
%mul8 = fmul fast double %2, %3
%add9 = fadd fast double %mul8, %sub
%add10 = fadd fast double %add9, %add9
call void @hhh(double 0.0, double %add10)
ret void
}
declare void @hhh(double, double)
attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "unsafe-fp-math"="true" "use-soft-float"="false" }
attributes #1 = { "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "unsafe-fp-math"="true" "use-soft-float"="false" }
attributes #2 = { nounwind }
|
