1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -mattr=+sve -force-streaming-compatible < %s | FileCheck %s
3 ; RUN: llc -mattr=+sme -force-streaming < %s | FileCheck %s
4 ; RUN: llc -force-streaming-compatible < %s | FileCheck %s --check-prefix=NONEON-NOSVE
7 target triple = "aarch64-unknown-linux-gnu"
9 define void @hang_when_merging_stores_after_legalisation(ptr %a, <2 x i32> %b) {
10 ; CHECK-LABEL: hang_when_merging_stores_after_legalisation:
12 ; CHECK-NEXT: // kill: def $d0 killed $d0 def $z0
13 ; CHECK-NEXT: ptrue p0.s, vl4
14 ; CHECK-NEXT: mov z0.s, s0
15 ; CHECK-NEXT: mov z1.d, z0.d
16 ; CHECK-NEXT: st2w { z0.s, z1.s }, p0, [x0]
19 ; NONEON-NOSVE-LABEL: hang_when_merging_stores_after_legalisation:
20 ; NONEON-NOSVE: // %bb.0:
21 ; NONEON-NOSVE-NEXT: sub sp, sp, #32
22 ; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 32
23 ; NONEON-NOSVE-NEXT: str d0, [sp, #8]
24 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #8]
25 ; NONEON-NOSVE-NEXT: stp w8, w8, [sp, #24]
26 ; NONEON-NOSVE-NEXT: stp w8, w8, [sp, #16]
27 ; NONEON-NOSVE-NEXT: ldr q0, [sp, #16]
28 ; NONEON-NOSVE-NEXT: stp q0, q0, [x0]
29 ; NONEON-NOSVE-NEXT: add sp, sp, #32
30 ; NONEON-NOSVE-NEXT: ret
31 %splat = shufflevector <2 x i32> %b, <2 x i32> undef, <8 x i32> zeroinitializer
32 %interleaved.vec = shufflevector <8 x i32> %splat, <8 x i32> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
33 store <8 x i32> %interleaved.vec, ptr %a, align 4
37 define void @interleave_store_without_splat(ptr %a, <4 x i32> %v1, <4 x i32> %v2) {
38 ; CHECK-LABEL: interleave_store_without_splat:
40 ; CHECK-NEXT: ptrue p0.s, vl4
41 ; CHECK-NEXT: // kill: def $q1 killed $q1 killed $z0_z1 def $z0_z1
42 ; CHECK-NEXT: // kill: def $q0 killed $q0 killed $z0_z1 def $z0_z1
43 ; CHECK-NEXT: st2w { z0.s, z1.s }, p0, [x0]
46 ; NONEON-NOSVE-LABEL: interleave_store_without_splat:
47 ; NONEON-NOSVE: // %bb.0:
48 ; NONEON-NOSVE-NEXT: sub sp, sp, #64
49 ; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 64
50 ; NONEON-NOSVE-NEXT: stp q0, q1, [sp, #16]
51 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #36]
52 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
53 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #8]
54 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #32]
55 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
56 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp]
57 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #44]
58 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #28]
59 ; NONEON-NOSVE-NEXT: ldr q1, [sp]
60 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #56]
61 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #40]
62 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #24]
63 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #48]
64 ; NONEON-NOSVE-NEXT: ldr q0, [sp, #48]
65 ; NONEON-NOSVE-NEXT: stp q1, q0, [x0]
66 ; NONEON-NOSVE-NEXT: add sp, sp, #64
67 ; NONEON-NOSVE-NEXT: ret
68 %shuffle = shufflevector <4 x i32> %v1, <4 x i32> %v2, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
69 %interleaved = shufflevector <8 x i32> %shuffle, <8 x i32> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
70 store <8 x i32> %interleaved, ptr %a, align 1
74 define void @interleave_store_legalization(ptr %a, <8 x i32> %v1, <8 x i32> %v2) {
75 ; CHECK-LABEL: interleave_store_legalization:
77 ; CHECK-NEXT: mov z5.d, z2.d
78 ; CHECK-NEXT: // kill: def $q3 killed $q3 def $z2_z3
79 ; CHECK-NEXT: ptrue p0.s, vl4
80 ; CHECK-NEXT: mov x8, #8 // =0x8
81 ; CHECK-NEXT: mov z4.d, z0.d
82 ; CHECK-NEXT: mov z2.d, z1.d
83 ; CHECK-NEXT: st2w { z4.s, z5.s }, p0, [x0]
84 ; CHECK-NEXT: st2w { z2.s, z3.s }, p0, [x0, x8, lsl #2]
87 ; NONEON-NOSVE-LABEL: interleave_store_legalization:
88 ; NONEON-NOSVE: // %bb.0:
89 ; NONEON-NOSVE-NEXT: sub sp, sp, #128
90 ; NONEON-NOSVE-NEXT: .cfi_def_cfa_offset 128
91 ; NONEON-NOSVE-NEXT: stp q1, q3, [sp, #16]
92 ; NONEON-NOSVE-NEXT: stp q0, q2, [sp, #80]
93 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #100]
94 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #84]
95 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #72]
96 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #96]
97 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #80]
98 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #64]
99 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #108]
100 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #92]
101 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #120]
102 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #104]
103 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #88]
104 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #112]
105 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #36]
106 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #20]
107 ; NONEON-NOSVE-NEXT: ldr q3, [sp, #112]
108 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #8]
109 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #32]
110 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #16]
111 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp]
112 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #44]
113 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #28]
114 ; NONEON-NOSVE-NEXT: ldr q1, [sp]
115 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #56]
116 ; NONEON-NOSVE-NEXT: ldr w9, [sp, #40]
117 ; NONEON-NOSVE-NEXT: ldr w8, [sp, #24]
118 ; NONEON-NOSVE-NEXT: stp w8, w9, [sp, #48]
119 ; NONEON-NOSVE-NEXT: ldp q0, q2, [sp, #48]
120 ; NONEON-NOSVE-NEXT: stp q2, q3, [x0]
121 ; NONEON-NOSVE-NEXT: stp q1, q0, [x0, #32]
122 ; NONEON-NOSVE-NEXT: add sp, sp, #128
123 ; NONEON-NOSVE-NEXT: ret
124 %interleaved.vec = shufflevector <8 x i32> %v1, <8 x i32> %v2, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11,
125 i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
126 store <16 x i32> %interleaved.vec, ptr %a, align 4
130 ; Ensure we don't crash when trying to lower a shuffle via an extract
131 define void @crash_when_lowering_extract_shuffle(ptr %dst, i1 %cond) {
132 ; CHECK-LABEL: crash_when_lowering_extract_shuffle:
136 ; NONEON-NOSVE-LABEL: crash_when_lowering_extract_shuffle:
137 ; NONEON-NOSVE: // %bb.0:
138 ; NONEON-NOSVE-NEXT: ret
139 %broadcast.splat = shufflevector <32 x i1> zeroinitializer, <32 x i1> zeroinitializer, <32 x i32> zeroinitializer
140 br i1 %cond, label %exit, label %vector.body
143 %1 = load <32 x i32>, ptr %dst, align 16
144 %predphi = select <32 x i1> %broadcast.splat, <32 x i32> zeroinitializer, <32 x i32> %1
145 store <32 x i32> %predphi, ptr %dst, align 16