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
|
; Test 32-bit subtraction in which the second operand is variable.
;
; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
declare i32 @foo()
; Check SLR.
define zeroext i1 @f1(i32 %dummy, i32 %a, i32 %b, i32 *%res) {
; CHECK-LABEL: f1:
; CHECK: slr %r3, %r4
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check using the overflow result for a branch.
define void @f2(i32 %dummy, i32 %a, i32 %b, i32 *%res) {
; CHECK-LABEL: f2:
; CHECK: slr %r3, %r4
; CHECK: st %r3, 0(%r5)
; CHECK: jgle foo@PLT
; CHECK: br %r14
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
br i1 %obit, label %call, label %exit
call:
tail call i32 @foo()
br label %exit
exit:
ret void
}
; ... and the same with the inverted direction.
define void @f3(i32 %dummy, i32 %a, i32 %b, i32 *%res) {
; CHECK-LABEL: f3:
; CHECK: slr %r3, %r4
; CHECK: st %r3, 0(%r5)
; CHECK: jgnle foo@PLT
; CHECK: br %r14
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
br i1 %obit, label %exit, label %call
call:
tail call i32 @foo()
br label %exit
exit:
ret void
}
; Check the low end of the SL range.
define zeroext i1 @f4(i32 %dummy, i32 %a, i32 *%src, i32 *%res) {
; CHECK-LABEL: f4:
; CHECK: sl %r3, 0(%r4)
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%b = load i32, i32 *%src
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check the high end of the aligned SL range.
define zeroext i1 @f5(i32 %dummy, i32 %a, i32 *%src, i32 *%res) {
; CHECK-LABEL: f5:
; CHECK: sl %r3, 4092(%r4)
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%ptr = getelementptr i32, i32 *%src, i64 1023
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check the next word up, which should use SLY instead of SL.
define zeroext i1 @f6(i32 %dummy, i32 %a, i32 *%src, i32 *%res) {
; CHECK-LABEL: f6:
; CHECK: sly %r3, 4096(%r4)
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%ptr = getelementptr i32, i32 *%src, i64 1024
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check the high end of the aligned SLY range.
define zeroext i1 @f7(i32 %dummy, i32 %a, i32 *%src, i32 *%res) {
; CHECK-LABEL: f7:
; CHECK: sly %r3, 524284(%r4)
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%ptr = getelementptr i32, i32 *%src, i64 131071
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check the next word up, which needs separate address logic.
; Other sequences besides this one would be OK.
define zeroext i1 @f8(i32 %dummy, i32 %a, i32 *%src, i32 *%res) {
; CHECK-LABEL: f8:
; CHECK: agfi %r4, 524288
; CHECK: sl %r3, 0(%r4)
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%ptr = getelementptr i32, i32 *%src, i64 131072
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check the high end of the negative aligned SLY range.
define zeroext i1 @f9(i32 %dummy, i32 %a, i32 *%src, i32 *%res) {
; CHECK-LABEL: f9:
; CHECK: sly %r3, -4(%r4)
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%ptr = getelementptr i32, i32 *%src, i64 -1
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check the low end of the SLY range.
define zeroext i1 @f10(i32 %dummy, i32 %a, i32 *%src, i32 *%res) {
; CHECK-LABEL: f10:
; CHECK: sly %r3, -524288(%r4)
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%ptr = getelementptr i32, i32 *%src, i64 -131072
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check the next word down, which needs separate address logic.
; Other sequences besides this one would be OK.
define zeroext i1 @f11(i32 %dummy, i32 %a, i32 *%src, i32 *%res) {
; CHECK-LABEL: f11:
; CHECK: agfi %r4, -524292
; CHECK: sl %r3, 0(%r4)
; CHECK-DAG: st %r3, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%ptr = getelementptr i32, i32 *%src, i64 -131073
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check that SL allows an index.
define zeroext i1 @f12(i64 %src, i64 %index, i32 %a, i32 *%res) {
; CHECK-LABEL: f12:
; CHECK: sl %r4, 4092({{%r3,%r2|%r2,%r3}})
; CHECK-DAG: st %r4, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%add1 = add i64 %src, %index
%add2 = add i64 %add1, 4092
%ptr = inttoptr i64 %add2 to i32 *
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check that SLY allows an index.
define zeroext i1 @f13(i64 %src, i64 %index, i32 %a, i32 *%res) {
; CHECK-LABEL: f13:
; CHECK: sly %r4, 4096({{%r3,%r2|%r2,%r3}})
; CHECK-DAG: st %r4, 0(%r5)
; CHECK-DAG: ipm [[REG:%r[0-5]]]
; CHECK-DAG: afi [[REG]], -536870912
; CHECK-DAG: risbg %r2, [[REG]], 63, 191, 33
; CHECK: br %r14
%add1 = add i64 %src, %index
%add2 = add i64 %add1, 4096
%ptr = inttoptr i64 %add2 to i32 *
%b = load i32, i32 *%ptr
%t = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %a, i32 %b)
%val = extractvalue {i32, i1} %t, 0
%obit = extractvalue {i32, i1} %t, 1
store i32 %val, i32 *%res
ret i1 %obit
}
; Check that subtractions of spilled values can use SL rather than SLR.
define zeroext i1 @f14(i32 *%ptr0) {
; CHECK-LABEL: f14:
; CHECK: brasl %r14, foo@PLT
; CHECK: sl %r2, 16{{[04]}}(%r15)
; CHECK: br %r14
%ptr1 = getelementptr i32, i32 *%ptr0, i64 2
%ptr2 = getelementptr i32, i32 *%ptr0, i64 4
%ptr3 = getelementptr i32, i32 *%ptr0, i64 6
%ptr4 = getelementptr i32, i32 *%ptr0, i64 8
%ptr5 = getelementptr i32, i32 *%ptr0, i64 10
%ptr6 = getelementptr i32, i32 *%ptr0, i64 12
%ptr7 = getelementptr i32, i32 *%ptr0, i64 14
%ptr8 = getelementptr i32, i32 *%ptr0, i64 16
%ptr9 = getelementptr i32, i32 *%ptr0, i64 18
%val0 = load i32, i32 *%ptr0
%val1 = load i32, i32 *%ptr1
%val2 = load i32, i32 *%ptr2
%val3 = load i32, i32 *%ptr3
%val4 = load i32, i32 *%ptr4
%val5 = load i32, i32 *%ptr5
%val6 = load i32, i32 *%ptr6
%val7 = load i32, i32 *%ptr7
%val8 = load i32, i32 *%ptr8
%val9 = load i32, i32 *%ptr9
%ret = call i32 @foo()
%t0 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %ret, i32 %val0)
%add0 = extractvalue {i32, i1} %t0, 0
%obit0 = extractvalue {i32, i1} %t0, 1
%t1 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add0, i32 %val1)
%add1 = extractvalue {i32, i1} %t1, 0
%obit1 = extractvalue {i32, i1} %t1, 1
%res1 = or i1 %obit0, %obit1
%t2 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add1, i32 %val2)
%add2 = extractvalue {i32, i1} %t2, 0
%obit2 = extractvalue {i32, i1} %t2, 1
%res2 = or i1 %res1, %obit2
%t3 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add2, i32 %val3)
%add3 = extractvalue {i32, i1} %t3, 0
%obit3 = extractvalue {i32, i1} %t3, 1
%res3 = or i1 %res2, %obit3
%t4 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add3, i32 %val4)
%add4 = extractvalue {i32, i1} %t4, 0
%obit4 = extractvalue {i32, i1} %t4, 1
%res4 = or i1 %res3, %obit4
%t5 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add4, i32 %val5)
%add5 = extractvalue {i32, i1} %t5, 0
%obit5 = extractvalue {i32, i1} %t5, 1
%res5 = or i1 %res4, %obit5
%t6 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add5, i32 %val6)
%add6 = extractvalue {i32, i1} %t6, 0
%obit6 = extractvalue {i32, i1} %t6, 1
%res6 = or i1 %res5, %obit6
%t7 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add6, i32 %val7)
%add7 = extractvalue {i32, i1} %t7, 0
%obit7 = extractvalue {i32, i1} %t7, 1
%res7 = or i1 %res6, %obit7
%t8 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add7, i32 %val8)
%add8 = extractvalue {i32, i1} %t8, 0
%obit8 = extractvalue {i32, i1} %t8, 1
%res8 = or i1 %res7, %obit8
%t9 = call {i32, i1} @llvm.usub.with.overflow.i32(i32 %add8, i32 %val9)
%add9 = extractvalue {i32, i1} %t9, 0
%obit9 = extractvalue {i32, i1} %t9, 1
%res9 = or i1 %res8, %obit9
ret i1 %res9
}
declare {i32, i1} @llvm.usub.with.overflow.i32(i32, i32) nounwind readnone
|
