1; RUN: llc < %s -debug-only=legalize-types 2>&1 | FileCheck %s --check-prefix=CHECK-LEGALIZATION
2; RUN: llc < %s | FileCheck %s
3; REQUIRES: asserts
4
5target triple = "aarch64-unknown-linux-gnu"
6attributes #0 = {"target-features"="+sve"}
7
8declare <vscale x 2 x i64> @llvm.experimental.vector.insert.nxv2i64.v8i64(<vscale x 2 x i64>, <8 x i64>, i64)
9declare <vscale x 2 x double> @llvm.experimental.vector.insert.nxv2f64.v8f64(<vscale x 2 x double>, <8 x double>, i64)
10
11define <vscale x 2 x i64> @test_nxv2i64_v8i64(<vscale x 2 x i64> %a, <8 x i64> %b) #0 {
12; CHECK-LEGALIZATION: Legally typed node: [[T1:t[0-9]+]]: nxv2i64 = insert_subvector {{t[0-9]+}}, {{t[0-9]+}}, Constant:i64<0>
13; CHECK-LEGALIZATION: Legally typed node: [[T2:t[0-9]+]]: nxv2i64 = insert_subvector [[T1]], {{t[0-9]+}}, Constant:i64<2>
14; CHECK-LEGALIZATION: Legally typed node: [[T3:t[0-9]+]]: nxv2i64 = insert_subvector [[T2]], {{t[0-9]+}}, Constant:i64<4>
15; CHECK-LEGALIZATION: Legally typed node: [[T4:t[0-9]+]]: nxv2i64 = insert_subvector [[T3]], {{t[0-9]+}}, Constant:i64<6>
16
17; CHECK-LABEL: test_nxv2i64_v8i64:
18; CHECK:       // %bb.0:
19; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
20; CHECK-NEXT:    addvl sp, sp, #-3
21; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 24 * VG
22; CHECK-NEXT:    .cfi_offset w29, -16
23; CHECK-NEXT:    cntd x8
24; CHECK-NEXT:    mov w9, #2
25; CHECK-NEXT:    sub x8, x8, #2
26; CHECK-NEXT:    ptrue p0.d, vl2
27; CHECK-NEXT:    cmp x8, #2
28; CHECK-NEXT:    mov x10, sp
29; CHECK-NEXT:    csel x9, x8, x9, lo
30; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
31; CHECK-NEXT:    mov z0.d, p0/m, z1.d
32; CHECK-NEXT:    lsl x9, x9, #3
33; CHECK-NEXT:    ptrue p0.d
34; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
35; CHECK-NEXT:    cmp x8, #4
36; CHECK-NEXT:    str q2, [x10, x9]
37; CHECK-NEXT:    mov w9, #4
38; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp]
39; CHECK-NEXT:    csel x9, x8, x9, lo
40; CHECK-NEXT:    lsl x9, x9, #3
41; CHECK-NEXT:    addvl x10, sp, #1
42; CHECK-NEXT:    cmp x8, #6
43; CHECK-NEXT:    st1d { z0.d }, p0, [sp, #1, mul vl]
44; CHECK-NEXT:    str q3, [x10, x9]
45; CHECK-NEXT:    mov w9, #6
46; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #1, mul vl]
47; CHECK-NEXT:    csel x8, x8, x9, lo
48; CHECK-NEXT:    addvl x9, sp, #2
49; CHECK-NEXT:    lsl x8, x8, #3
50; CHECK-NEXT:    st1d { z0.d }, p0, [sp, #2, mul vl]
51; CHECK-NEXT:    str q4, [x9, x8]
52; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #2, mul vl]
53; CHECK-NEXT:    addvl sp, sp, #3
54; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
55; CHECK-NEXT:    ret
56
57
58
59
60  %r = call <vscale x 2 x i64> @llvm.experimental.vector.insert.nxv2i64.v8i64(<vscale x 2 x i64> %a, <8 x i64> %b, i64 0)
61  ret <vscale x 2 x i64> %r
62}
63
64define <vscale x 2 x double> @test_nxv2f64_v8f64(<vscale x 2 x double> %a, <8 x double> %b) #0 {
65; CHECK-LEGALIZATION: Legally typed node: [[T1:t[0-9]+]]: nxv2f64 = insert_subvector {{t[0-9]+}}, {{t[0-9]+}}, Constant:i64<0>
66; CHECK-LEGALIZATION: Legally typed node: [[T2:t[0-9]+]]: nxv2f64 = insert_subvector [[T1]], {{t[0-9]+}}, Constant:i64<2>
67; CHECK-LEGALIZATION: Legally typed node: [[T3:t[0-9]+]]: nxv2f64 = insert_subvector [[T2]], {{t[0-9]+}}, Constant:i64<4>
68; CHECK-LEGALIZATION: Legally typed node: [[T4:t[0-9]+]]: nxv2f64 = insert_subvector [[T3]], {{t[0-9]+}}, Constant:i64<6>
69
70; CHECK-LABEL: test_nxv2f64_v8f64:
71; CHECK:       // %bb.0:
72; CHECK-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
73; CHECK-NEXT:    addvl sp, sp, #-3
74; CHECK-NEXT:    .cfi_escape 0x0f, 0x0c, 0x8f, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0x2e, 0x00, 0x1e, 0x22 // sp + 16 + 24 * VG
75; CHECK-NEXT:    .cfi_offset w29, -16
76; CHECK-NEXT:    cntd x8
77; CHECK-NEXT:    mov w9, #2
78; CHECK-NEXT:    sub x8, x8, #2
79; CHECK-NEXT:    ptrue p0.d, vl2
80; CHECK-NEXT:    cmp x8, #2
81; CHECK-NEXT:    mov x10, sp
82; CHECK-NEXT:    csel x9, x8, x9, lo
83; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
84; CHECK-NEXT:    mov z0.d, p0/m, z1.d
85; CHECK-NEXT:    lsl x9, x9, #3
86; CHECK-NEXT:    ptrue p0.d
87; CHECK-NEXT:    st1d { z0.d }, p0, [sp]
88; CHECK-NEXT:    cmp x8, #4
89; CHECK-NEXT:    str q2, [x10, x9]
90; CHECK-NEXT:    mov w9, #4
91; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp]
92; CHECK-NEXT:    csel x9, x8, x9, lo
93; CHECK-NEXT:    lsl x9, x9, #3
94; CHECK-NEXT:    addvl x10, sp, #1
95; CHECK-NEXT:    cmp x8, #6
96; CHECK-NEXT:    st1d { z0.d }, p0, [sp, #1, mul vl]
97; CHECK-NEXT:    str q3, [x10, x9]
98; CHECK-NEXT:    mov w9, #6
99; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #1, mul vl]
100; CHECK-NEXT:    csel x8, x8, x9, lo
101; CHECK-NEXT:    addvl x9, sp, #2
102; CHECK-NEXT:    lsl x8, x8, #3
103; CHECK-NEXT:    st1d { z0.d }, p0, [sp, #2, mul vl]
104; CHECK-NEXT:    str q4, [x9, x8]
105; CHECK-NEXT:    ld1d { z0.d }, p0/z, [sp, #2, mul vl]
106; CHECK-NEXT:    addvl sp, sp, #3
107; CHECK-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
108; CHECK-NEXT:    ret
109
110
111
112
113  %r = call <vscale x 2 x double> @llvm.experimental.vector.insert.nxv2f64.v8f64(<vscale x 2 x double> %a, <8 x double> %b, i64 0)
114  ret <vscale x 2 x double> %r
115}
116