1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=riscv32 -mattr=+m,+v < %s | FileCheck %s --check-prefixes=CHECK,RV32
3; RUN: llc -mtriple=riscv64 -mattr=+m,+v < %s | FileCheck %s --check-prefixes=CHECK,RV64
4
5; Check that we correctly scale the split part indirect offsets by VSCALE.
6define <vscale x 32 x i32> @callee_scalable_vector_split_indirect(<vscale x 32 x i32> %x, <vscale x 32 x i32> %y) {
7; CHECK-LABEL: callee_scalable_vector_split_indirect:
8; CHECK:       # %bb.0:
9; CHECK-NEXT:    csrr a1, vlenb
10; CHECK-NEXT:    slli a1, a1, 3
11; CHECK-NEXT:    add a1, a0, a1
12; CHECK-NEXT:    vl8re32.v v24, (a0)
13; CHECK-NEXT:    vl8re32.v v0, (a1)
14; CHECK-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
15; CHECK-NEXT:    vadd.vv v8, v8, v24
16; CHECK-NEXT:    vadd.vv v16, v16, v0
17; CHECK-NEXT:    ret
18  %a = add <vscale x 32 x i32> %x, %y
19  ret <vscale x 32 x i32> %a
20}
21
22; Call the function above. Check that we set the arguments correctly.
23define <vscale x 32 x i32> @caller_scalable_vector_split_indirect(<vscale x 32 x i32> %x) {
24; RV32-LABEL: caller_scalable_vector_split_indirect:
25; RV32:       # %bb.0:
26; RV32-NEXT:    addi sp, sp, -144
27; RV32-NEXT:    .cfi_def_cfa_offset 144
28; RV32-NEXT:    sw ra, 140(sp) # 4-byte Folded Spill
29; RV32-NEXT:    .cfi_offset ra, -4
30; RV32-NEXT:    addi s0, sp, 144
31; RV32-NEXT:    .cfi_def_cfa s0, 0
32; RV32-NEXT:    csrr a0, vlenb
33; RV32-NEXT:    slli a0, a0, 4
34; RV32-NEXT:    sub sp, sp, a0
35; RV32-NEXT:    andi sp, sp, -128
36; RV32-NEXT:    addi a0, sp, 128
37; RV32-NEXT:    vs8r.v v8, (a0)
38; RV32-NEXT:    csrr a1, vlenb
39; RV32-NEXT:    slli a1, a1, 3
40; RV32-NEXT:    add a0, a0, a1
41; RV32-NEXT:    vs8r.v v16, (a0)
42; RV32-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
43; RV32-NEXT:    vmv.v.i v8, 0
44; RV32-NEXT:    addi a0, sp, 128
45; RV32-NEXT:    vmv.v.i v16, 0
46; RV32-NEXT:    call callee_scalable_vector_split_indirect@plt
47; RV32-NEXT:    addi sp, s0, -144
48; RV32-NEXT:    lw ra, 140(sp) # 4-byte Folded Reload
49; RV32-NEXT:    addi sp, sp, 144
50; RV32-NEXT:    ret
51;
52; RV64-LABEL: caller_scalable_vector_split_indirect:
53; RV64:       # %bb.0:
54; RV64-NEXT:    addi sp, sp, -144
55; RV64-NEXT:    .cfi_def_cfa_offset 144
56; RV64-NEXT:    sd ra, 136(sp) # 8-byte Folded Spill
57; RV64-NEXT:    .cfi_offset ra, -8
58; RV64-NEXT:    addi s0, sp, 144
59; RV64-NEXT:    .cfi_def_cfa s0, 0
60; RV64-NEXT:    csrr a0, vlenb
61; RV64-NEXT:    slli a0, a0, 4
62; RV64-NEXT:    sub sp, sp, a0
63; RV64-NEXT:    andi sp, sp, -128
64; RV64-NEXT:    addi a0, sp, 128
65; RV64-NEXT:    vs8r.v v8, (a0)
66; RV64-NEXT:    csrr a1, vlenb
67; RV64-NEXT:    slli a1, a1, 3
68; RV64-NEXT:    add a0, a0, a1
69; RV64-NEXT:    vs8r.v v16, (a0)
70; RV64-NEXT:    vsetvli a0, zero, e32, m8, ta, mu
71; RV64-NEXT:    vmv.v.i v8, 0
72; RV64-NEXT:    addi a0, sp, 128
73; RV64-NEXT:    vmv.v.i v16, 0
74; RV64-NEXT:    call callee_scalable_vector_split_indirect@plt
75; RV64-NEXT:    addi sp, s0, -144
76; RV64-NEXT:    ld ra, 136(sp) # 8-byte Folded Reload
77; RV64-NEXT:    addi sp, sp, 144
78; RV64-NEXT:    ret
79  %c = alloca i64
80  %a = call <vscale x 32 x i32> @callee_scalable_vector_split_indirect(<vscale x 32 x i32> zeroinitializer, <vscale x 32 x i32> %x)
81  ret <vscale x 32 x i32> %a
82}
83