xref: /llvm-project/llvm/test/CodeGen/AArch64/scalar-mla-mls.ll (revision a110a1c0ed9850be168cd0e29f05179e80941b04)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -verify-machineinstrs -mtriple=aarch64-none-linux-gnu -mattr=+neon | FileCheck %s
3
4define ptr @test_scalar_msub(ptr %a, ptr %b) {
5; CHECK-LABEL: test_scalar_msub:
6; CHECK:       // %bb.0: // %entry
7; CHECK-NEXT:    ldp w8, w11, [x1]
8; CHECK-NEXT:    ldp w9, w10, [x0]
9; CHECK-NEXT:    mul w12, w8, w9
10; CHECK-NEXT:    mul w8, w10, w8
11; CHECK-NEXT:    madd w8, w11, w9, w8
12; CHECK-NEXT:    msub w9, w11, w10, w12
13; CHECK-NEXT:    stp w9, w8, [x0]
14; CHECK-NEXT:    ret
15entry:
16  %0 = load i32, ptr %a, align 4
17  %1 = load i32, ptr %b, align 4
18  %mul = mul nsw i32 %1, %0
19  %_M_imag = getelementptr inbounds i8, ptr %a, i64 4
20  %2 = load i32, ptr %_M_imag, align 4
21  %_M_imag.i = getelementptr inbounds i8, ptr %b, i64 4
22  %3 = load i32, ptr %_M_imag.i, align 4
23  %mul3 = mul nsw i32 %3, %2
24  %sub = sub nsw i32 %mul, %mul3
25  %mul6 = mul nsw i32 %3, %0
26  %mul9 = mul nsw i32 %2, %1
27  %add = add nsw i32 %mul6, %mul9
28  store i32 %add, ptr %_M_imag, align 4
29  store i32 %sub, ptr %a, align 4
30  ret ptr %a
31}
32
33define ptr @test_scalar_msub_i64(ptr %a, ptr %b) {
34; CHECK-LABEL: test_scalar_msub_i64:
35; CHECK:       // %bb.0: // %entry
36; CHECK-NEXT:    ldr x8, [x1]
37; CHECK-NEXT:    ldur x9, [x0, #4]
38; CHECK-NEXT:    ldr x10, [x0]
39; CHECK-NEXT:    ldur x12, [x1, #4]
40; CHECK-NEXT:    mul x11, x9, x8
41; CHECK-NEXT:    mul x8, x8, x10
42; CHECK-NEXT:    madd x10, x12, x10, x11
43; CHECK-NEXT:    msub x8, x12, x9, x8
44; CHECK-NEXT:    stur x10, [x0, #4]
45; CHECK-NEXT:    str x8, [x0]
46; CHECK-NEXT:    ret
47entry:
48  %0 = load i64, ptr %a, align 8
49  %1 = load i64, ptr %b, align 8
50  %mul = mul nsw i64 %1, %0
51  %_M_imag = getelementptr inbounds i8, ptr %a, i64 4
52  %2 = load i64, ptr %_M_imag, align 8
53  %_M_imag.i = getelementptr inbounds i8, ptr %b, i64 4
54  %3 = load i64, ptr %_M_imag.i, align 8
55  %mul3 = mul nsw i64 %3, %2
56  %sub = sub nsw i64 %mul, %mul3
57  %mul6 = mul nsw i64 %3, %0
58  %mul9 = mul nsw i64 %2, %1
59  %add = add nsw i64 %mul6, %mul9
60  store i64 %add, ptr %_M_imag, align 8
61  store i64 %sub, ptr %a, align 8
62  ret ptr %a
63}
64