1 ; RUN: llc < %s -mtriple=arm64-eabi -asm-verbose=false -verify-machineinstrs -mcpu=cyclone | FileCheck -enable-var-scope %s
3 define i32 @val_compare_and_swap(i32* %p, i32 %cmp, i32 %new) #0 {
4 ; CHECK-LABEL: val_compare_and_swap:
5 ; CHECK-NEXT: mov x[[ADDR:[0-9]+]], x0
6 ; CHECK-NEXT: [[TRYBB:.?LBB[0-9_]+]]:
7 ; CHECK-NEXT: ldaxr [[RESULT:w[0-9]+]], [x[[ADDR]]]
8 ; CHECK-NEXT: cmp [[RESULT]], w1
9 ; CHECK-NEXT: b.ne [[FAILBB:.?LBB[0-9_]+]]
10 ; CHECK-NEXT: stxr [[SCRATCH_REG:w[0-9]+]], w2, [x[[ADDR]]]
11 ; CHECK-NEXT: cbnz [[SCRATCH_REG]], [[TRYBB]]
13 ; CHECK-NEXT: [[FAILBB]]:
16 %pair = cmpxchg i32* %p, i32 %cmp, i32 %new acquire acquire
17 %val = extractvalue { i32, i1 } %pair, 0
21 define i32 @val_compare_and_swap_from_load(i32* %p, i32 %cmp, i32* %pnew) #0 {
22 ; CHECK-LABEL: val_compare_and_swap_from_load:
23 ; CHECK-NEXT: ldr [[NEW:w[0-9]+]], [x2]
24 ; CHECK-NEXT: [[TRYBB:.?LBB[0-9_]+]]:
25 ; CHECK-NEXT: ldaxr w[[RESULT:[0-9]+]], [x0]
26 ; CHECK-NEXT: cmp w[[RESULT]], w1
27 ; CHECK-NEXT: b.ne [[FAILBB:.?LBB[0-9_]+]]
28 ; CHECK-NEXT: stxr [[SCRATCH_REG:w[0-9]+]], [[NEW]], [x0]
29 ; CHECK-NEXT: cbnz [[SCRATCH_REG]], [[TRYBB]]
30 ; CHECK-NEXT: mov x0, x[[RESULT]]
32 ; CHECK-NEXT: [[FAILBB]]:
34 ; CHECK-NEXT: mov x0, x[[RESULT]]
36 %new = load i32, i32* %pnew
37 %pair = cmpxchg i32* %p, i32 %cmp, i32 %new acquire acquire
38 %val = extractvalue { i32, i1 } %pair, 0
42 define i32 @val_compare_and_swap_rel(i32* %p, i32 %cmp, i32 %new) #0 {
43 ; CHECK-LABEL: val_compare_and_swap_rel:
44 ; CHECK-NEXT: mov x[[ADDR:[0-9]+]], x0
45 ; CHECK-NEXT: [[TRYBB:.?LBB[0-9_]+]]:
46 ; CHECK-NEXT: ldaxr [[RESULT:w[0-9]+]], [x[[ADDR]]]
47 ; CHECK-NEXT: cmp [[RESULT]], w1
48 ; CHECK-NEXT: b.ne [[FAILBB:.?LBB[0-9_]+]]
49 ; CHECK-NEXT: stlxr [[SCRATCH_REG:w[0-9]+]], w2, [x[[ADDR]]]
50 ; CHECK-NEXT: cbnz [[SCRATCH_REG]], [[TRYBB]]
52 ; CHECK-NEXT: [[FAILBB]]:
55 %pair = cmpxchg i32* %p, i32 %cmp, i32 %new acq_rel monotonic
56 %val = extractvalue { i32, i1 } %pair, 0
60 define i64 @val_compare_and_swap_64(i64* %p, i64 %cmp, i64 %new) #0 {
61 ; CHECK-LABEL: val_compare_and_swap_64:
62 ; CHECK-NEXT: mov x[[ADDR:[0-9]+]], x0
63 ; CHECK-NEXT: [[TRYBB:.?LBB[0-9_]+]]:
64 ; CHECK-NEXT: ldxr [[RESULT:x[0-9]+]], [x[[ADDR]]]
65 ; CHECK-NEXT: cmp [[RESULT]], x1
66 ; CHECK-NEXT: b.ne [[FAILBB:.?LBB[0-9_]+]]
67 ; CHECK-NEXT: stxr [[SCRATCH_REG:w[0-9]+]], x2, [x[[ADDR]]]
68 ; CHECK-NEXT: cbnz [[SCRATCH_REG]], [[TRYBB]]
70 ; CHECK-NEXT: [[FAILBB]]:
73 %pair = cmpxchg i64* %p, i64 %cmp, i64 %new monotonic monotonic
74 %val = extractvalue { i64, i1 } %pair, 0
78 define i32 @fetch_and_nand(i32* %p) #0 {
79 ; CHECK-LABEL: fetch_and_nand:
80 ; CHECK: [[TRYBB:.?LBB[0-9_]+]]:
81 ; CHECK: ldxr w[[DEST_REG:[0-9]+]], [x0]
82 ; CHECK: mvn [[TMP_REG:w[0-9]+]], w[[DEST_REG]]
83 ; CHECK: orr [[SCRATCH2_REG:w[0-9]+]], [[TMP_REG]], #0xfffffff8
84 ; CHECK-NOT: stlxr [[SCRATCH2_REG]], [[SCRATCH2_REG]]
85 ; CHECK: stlxr [[SCRATCH_REG:w[0-9]+]], [[SCRATCH2_REG]], [x0]
86 ; CHECK: cbnz [[SCRATCH_REG]], [[TRYBB]]
87 ; CHECK: mov x0, x[[DEST_REG]]
88 %val = atomicrmw nand i32* %p, i32 7 release
92 define i64 @fetch_and_nand_64(i64* %p) #0 {
93 ; CHECK-LABEL: fetch_and_nand_64:
94 ; CHECK: mov x[[ADDR:[0-9]+]], x0
95 ; CHECK: [[TRYBB:.?LBB[0-9_]+]]:
96 ; CHECK: ldaxr x[[DEST_REG:[0-9]+]], [x[[ADDR]]]
97 ; CHECK: mvn w[[TMP_REG:[0-9]+]], w[[DEST_REG]]
98 ; CHECK: orr [[SCRATCH2_REG:x[0-9]+]], x[[TMP_REG]], #0xfffffffffffffff8
99 ; CHECK: stlxr [[SCRATCH_REG:w[0-9]+]], [[SCRATCH2_REG]], [x[[ADDR]]]
100 ; CHECK: cbnz [[SCRATCH_REG]], [[TRYBB]]
102 %val = atomicrmw nand i64* %p, i64 7 acq_rel
106 define i32 @fetch_and_or(i32* %p) #0 {
107 ; CHECK-LABEL: fetch_and_or:
108 ; CHECK: mov [[OLDVAL_REG:w[0-9]+]], #5
109 ; CHECK: [[TRYBB:.?LBB[0-9_]+]]:
110 ; CHECK: ldaxr w[[DEST_REG:[0-9]+]], [x0]
111 ; CHECK: orr [[SCRATCH2_REG:w[0-9]+]], w[[DEST_REG]], [[OLDVAL_REG]]
112 ; CHECK-NOT: stlxr [[SCRATCH2_REG]], [[SCRATCH2_REG]]
113 ; CHECK: stlxr [[SCRATCH_REG:w[0-9]+]], [[SCRATCH2_REG]], [x0]
114 ; CHECK: cbnz [[SCRATCH_REG]], [[TRYBB]]
115 ; CHECK: mov x0, x[[DEST_REG]]
116 %val = atomicrmw or i32* %p, i32 5 seq_cst
120 define i64 @fetch_and_or_64(i64* %p) #0 {
121 ; CHECK: fetch_and_or_64:
122 ; CHECK: mov x[[ADDR:[0-9]+]], x0
123 ; CHECK: [[TRYBB:.?LBB[0-9_]+]]:
124 ; CHECK: ldxr [[DEST_REG:x[0-9]+]], [x[[ADDR]]]
125 ; CHECK: orr [[SCRATCH2_REG:x[0-9]+]], [[DEST_REG]], #0x7
126 ; CHECK: stxr [[SCRATCH_REG:w[0-9]+]], [[SCRATCH2_REG]], [x[[ADDR]]]
127 ; CHECK: cbnz [[SCRATCH_REG]], [[TRYBB]]
128 %val = atomicrmw or i64* %p, i64 7 monotonic
132 define void @acquire_fence() #0 {
135 ; CHECK-LABEL: acquire_fence:
139 define void @release_fence() #0 {
142 ; CHECK-LABEL: release_fence:
143 ; CHECK: dmb ish{{$}}
146 define void @seq_cst_fence() #0 {
149 ; CHECK-LABEL: seq_cst_fence:
150 ; CHECK: dmb ish{{$}}
153 define i32 @atomic_load(i32* %p) #0 {
154 %r = load atomic i32, i32* %p seq_cst, align 4
156 ; CHECK-LABEL: atomic_load:
160 define i8 @atomic_load_relaxed_8(i8* %p, i32 %off32) #0 {
161 ; CHECK-LABEL: atomic_load_relaxed_8:
162 %ptr_unsigned = getelementptr i8, i8* %p, i32 4095
163 %val_unsigned = load atomic i8, i8* %ptr_unsigned monotonic, align 1
164 ; CHECK: ldrb {{w[0-9]+}}, [x0, #4095]
166 %ptr_regoff = getelementptr i8, i8* %p, i32 %off32
167 %val_regoff = load atomic i8, i8* %ptr_regoff unordered, align 1
168 %tot1 = add i8 %val_unsigned, %val_regoff
169 ; CHECK: ldrb {{w[0-9]+}}, [x0, w1, sxtw]
171 %ptr_unscaled = getelementptr i8, i8* %p, i32 -256
172 %val_unscaled = load atomic i8, i8* %ptr_unscaled monotonic, align 1
173 %tot2 = add i8 %tot1, %val_unscaled
174 ; CHECK: ldurb {{w[0-9]+}}, [x0, #-256]
176 %ptr_random = getelementptr i8, i8* %p, i32 1191936 ; 0x123000 (i.e. ADD imm)
177 %val_random = load atomic i8, i8* %ptr_random unordered, align 1
178 %tot3 = add i8 %tot2, %val_random
179 ; CHECK: add x[[ADDR:[0-9]+]], x0, #291, lsl #12
180 ; CHECK: ldrb {{w[0-9]+}}, [x[[ADDR]]]
185 define i16 @atomic_load_relaxed_16(i16* %p, i32 %off32) #0 {
186 ; CHECK-LABEL: atomic_load_relaxed_16:
187 %ptr_unsigned = getelementptr i16, i16* %p, i32 4095
188 %val_unsigned = load atomic i16, i16* %ptr_unsigned monotonic, align 2
189 ; CHECK: ldrh {{w[0-9]+}}, [x0, #8190]
191 %ptr_regoff = getelementptr i16, i16* %p, i32 %off32
192 %val_regoff = load atomic i16, i16* %ptr_regoff unordered, align 2
193 %tot1 = add i16 %val_unsigned, %val_regoff
194 ; CHECK: ldrh {{w[0-9]+}}, [x0, w1, sxtw #1]
196 %ptr_unscaled = getelementptr i16, i16* %p, i32 -128
197 %val_unscaled = load atomic i16, i16* %ptr_unscaled monotonic, align 2
198 %tot2 = add i16 %tot1, %val_unscaled
199 ; CHECK: ldurh {{w[0-9]+}}, [x0, #-256]
201 %ptr_random = getelementptr i16, i16* %p, i32 595968 ; 0x123000/2 (i.e. ADD imm)
202 %val_random = load atomic i16, i16* %ptr_random unordered, align 2
203 %tot3 = add i16 %tot2, %val_random
204 ; CHECK: add x[[ADDR:[0-9]+]], x0, #291, lsl #12
205 ; CHECK: ldrh {{w[0-9]+}}, [x[[ADDR]]]
210 define i32 @atomic_load_relaxed_32(i32* %p, i32 %off32) #0 {
211 ; CHECK-LABEL: atomic_load_relaxed_32:
212 %ptr_unsigned = getelementptr i32, i32* %p, i32 4095
213 %val_unsigned = load atomic i32, i32* %ptr_unsigned monotonic, align 4
214 ; CHECK: ldr {{w[0-9]+}}, [x0, #16380]
216 %ptr_regoff = getelementptr i32, i32* %p, i32 %off32
217 %val_regoff = load atomic i32, i32* %ptr_regoff unordered, align 4
218 %tot1 = add i32 %val_unsigned, %val_regoff
219 ; CHECK: ldr {{w[0-9]+}}, [x0, w1, sxtw #2]
221 %ptr_unscaled = getelementptr i32, i32* %p, i32 -64
222 %val_unscaled = load atomic i32, i32* %ptr_unscaled monotonic, align 4
223 %tot2 = add i32 %tot1, %val_unscaled
224 ; CHECK: ldur {{w[0-9]+}}, [x0, #-256]
226 %ptr_random = getelementptr i32, i32* %p, i32 297984 ; 0x123000/4 (i.e. ADD imm)
227 %val_random = load atomic i32, i32* %ptr_random unordered, align 4
228 %tot3 = add i32 %tot2, %val_random
229 ; CHECK: add x[[ADDR:[0-9]+]], x0, #291, lsl #12
230 ; CHECK: ldr {{w[0-9]+}}, [x[[ADDR]]]
235 define i64 @atomic_load_relaxed_64(i64* %p, i32 %off32) #0 {
236 ; CHECK-LABEL: atomic_load_relaxed_64:
237 %ptr_unsigned = getelementptr i64, i64* %p, i32 4095
238 %val_unsigned = load atomic i64, i64* %ptr_unsigned monotonic, align 8
239 ; CHECK: ldr {{x[0-9]+}}, [x0, #32760]
241 %ptr_regoff = getelementptr i64, i64* %p, i32 %off32
242 %val_regoff = load atomic i64, i64* %ptr_regoff unordered, align 8
243 %tot1 = add i64 %val_unsigned, %val_regoff
244 ; CHECK: ldr {{x[0-9]+}}, [x0, w1, sxtw #3]
246 %ptr_unscaled = getelementptr i64, i64* %p, i32 -32
247 %val_unscaled = load atomic i64, i64* %ptr_unscaled monotonic, align 8
248 %tot2 = add i64 %tot1, %val_unscaled
249 ; CHECK: ldur {{x[0-9]+}}, [x0, #-256]
251 %ptr_random = getelementptr i64, i64* %p, i32 148992 ; 0x123000/8 (i.e. ADD imm)
252 %val_random = load atomic i64, i64* %ptr_random unordered, align 8
253 %tot3 = add i64 %tot2, %val_random
254 ; CHECK: add x[[ADDR:[0-9]+]], x0, #291, lsl #12
255 ; CHECK: ldr {{x[0-9]+}}, [x[[ADDR]]]
261 define void @atomc_store(i32* %p) #0 {
262 store atomic i32 4, i32* %p seq_cst, align 4
264 ; CHECK-LABEL: atomc_store:
268 define void @atomic_store_relaxed_8(i8* %p, i32 %off32, i8 %val) #0 {
269 ; CHECK-LABEL: atomic_store_relaxed_8:
270 %ptr_unsigned = getelementptr i8, i8* %p, i32 4095
271 store atomic i8 %val, i8* %ptr_unsigned monotonic, align 1
272 ; CHECK: strb {{w[0-9]+}}, [x0, #4095]
274 %ptr_regoff = getelementptr i8, i8* %p, i32 %off32
275 store atomic i8 %val, i8* %ptr_regoff unordered, align 1
276 ; CHECK: strb {{w[0-9]+}}, [x0, w1, sxtw]
278 %ptr_unscaled = getelementptr i8, i8* %p, i32 -256
279 store atomic i8 %val, i8* %ptr_unscaled monotonic, align 1
280 ; CHECK: sturb {{w[0-9]+}}, [x0, #-256]
282 %ptr_random = getelementptr i8, i8* %p, i32 1191936 ; 0x123000 (i.e. ADD imm)
283 store atomic i8 %val, i8* %ptr_random unordered, align 1
284 ; CHECK: add x[[ADDR:[0-9]+]], x0, #291, lsl #12
285 ; CHECK: strb {{w[0-9]+}}, [x[[ADDR]]]
290 define void @atomic_store_relaxed_16(i16* %p, i32 %off32, i16 %val) #0 {
291 ; CHECK-LABEL: atomic_store_relaxed_16:
292 %ptr_unsigned = getelementptr i16, i16* %p, i32 4095
293 store atomic i16 %val, i16* %ptr_unsigned monotonic, align 2
294 ; CHECK: strh {{w[0-9]+}}, [x0, #8190]
296 %ptr_regoff = getelementptr i16, i16* %p, i32 %off32
297 store atomic i16 %val, i16* %ptr_regoff unordered, align 2
298 ; CHECK: strh {{w[0-9]+}}, [x0, w1, sxtw #1]
300 %ptr_unscaled = getelementptr i16, i16* %p, i32 -128
301 store atomic i16 %val, i16* %ptr_unscaled monotonic, align 2
302 ; CHECK: sturh {{w[0-9]+}}, [x0, #-256]
304 %ptr_random = getelementptr i16, i16* %p, i32 595968 ; 0x123000/2 (i.e. ADD imm)
305 store atomic i16 %val, i16* %ptr_random unordered, align 2
306 ; CHECK: add x[[ADDR:[0-9]+]], x0, #291, lsl #12
307 ; CHECK: strh {{w[0-9]+}}, [x[[ADDR]]]
312 define void @atomic_store_relaxed_32(i32* %p, i32 %off32, i32 %val) #0 {
313 ; CHECK-LABEL: atomic_store_relaxed_32:
314 %ptr_unsigned = getelementptr i32, i32* %p, i32 4095
315 store atomic i32 %val, i32* %ptr_unsigned monotonic, align 4
316 ; CHECK: str {{w[0-9]+}}, [x0, #16380]
318 %ptr_regoff = getelementptr i32, i32* %p, i32 %off32
319 store atomic i32 %val, i32* %ptr_regoff unordered, align 4
320 ; CHECK: str {{w[0-9]+}}, [x0, w1, sxtw #2]
322 %ptr_unscaled = getelementptr i32, i32* %p, i32 -64
323 store atomic i32 %val, i32* %ptr_unscaled monotonic, align 4
324 ; CHECK: stur {{w[0-9]+}}, [x0, #-256]
326 %ptr_random = getelementptr i32, i32* %p, i32 297984 ; 0x123000/4 (i.e. ADD imm)
327 store atomic i32 %val, i32* %ptr_random unordered, align 4
328 ; CHECK: add x[[ADDR:[0-9]+]], x0, #291, lsl #12
329 ; CHECK: str {{w[0-9]+}}, [x[[ADDR]]]
334 define void @atomic_store_relaxed_64(i64* %p, i32 %off32, i64 %val) #0 {
335 ; CHECK-LABEL: atomic_store_relaxed_64:
336 %ptr_unsigned = getelementptr i64, i64* %p, i32 4095
337 store atomic i64 %val, i64* %ptr_unsigned monotonic, align 8
338 ; CHECK: str {{x[0-9]+}}, [x0, #32760]
340 %ptr_regoff = getelementptr i64, i64* %p, i32 %off32
341 store atomic i64 %val, i64* %ptr_regoff unordered, align 8
342 ; CHECK: str {{x[0-9]+}}, [x0, w1, sxtw #3]
344 %ptr_unscaled = getelementptr i64, i64* %p, i32 -32
345 store atomic i64 %val, i64* %ptr_unscaled monotonic, align 8
346 ; CHECK: stur {{x[0-9]+}}, [x0, #-256]
348 %ptr_random = getelementptr i64, i64* %p, i32 148992 ; 0x123000/8 (i.e. ADD imm)
349 store atomic i64 %val, i64* %ptr_random unordered, align 8
350 ; CHECK: add x[[ADDR:[0-9]+]], x0, #291, lsl #12
351 ; CHECK: str {{x[0-9]+}}, [x[[ADDR]]]
359 %"class.X::Atomic" = type { %struct.x_atomic_t }
360 %struct.x_atomic_t = type { i32 }
362 @counter = external hidden global %"class.X::Atomic", align 4
364 define i32 @next_id() nounwind optsize ssp align 2 {
366 %0 = atomicrmw add i32* getelementptr inbounds (%"class.X::Atomic", %"class.X::Atomic"* @counter, i64 0, i32 0, i32 0), i32 1 seq_cst
367 %add.i = add i32 %0, 1
368 %tobool = icmp eq i32 %add.i, 0
369 br i1 %tobool, label %if.else, label %return
371 if.else: ; preds = %entry
372 %1 = atomicrmw add i32* getelementptr inbounds (%"class.X::Atomic", %"class.X::Atomic"* @counter, i64 0, i32 0, i32 0), i32 1 seq_cst
373 %add.i2 = add i32 %1, 1
376 return: ; preds = %if.else, %entry
377 %retval.0 = phi i32 [ %add.i2, %if.else ], [ %add.i, %entry ]
381 attributes #0 = { nounwind }