1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=aarch64-none-linux-gnu -mattr=+neon | FileCheck %s --check-prefix=CHECK
4 ; Same as vecreduce-fmul-legalization.ll, but without fmf.
6 declare half @llvm.vector.reduce.fmul.f16.v1f16(half, <1 x half>)
7 declare float @llvm.vector.reduce.fmul.f32.v1f32(float, <1 x float>)
8 declare double @llvm.vector.reduce.fmul.f64.v1f64(double, <1 x double>)
9 declare fp128 @llvm.vector.reduce.fmul.f128.v1f128(fp128, <1 x fp128>)
11 declare float @llvm.vector.reduce.fmul.f32.v3f32(float, <3 x float>)
12 declare fp128 @llvm.vector.reduce.fmul.f128.v2f128(fp128, <2 x fp128>)
13 declare float @llvm.vector.reduce.fmul.f32.v16f32(float, <16 x float>)
15 define half @test_v1f16(<1 x half> %a) nounwind {
16 ; CHECK-LABEL: test_v1f16:
19 %b = call half @llvm.vector.reduce.fmul.f16.v1f16(half 1.0, <1 x half> %a)
23 define float @test_v1f32(<1 x float> %a) nounwind {
24 ; CHECK-LABEL: test_v1f32:
26 ; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0
27 ; CHECK-NEXT: // kill: def $s0 killed $s0 killed $q0
29 %b = call float @llvm.vector.reduce.fmul.f32.v1f32(float 1.0, <1 x float> %a)
33 define double @test_v1f64(<1 x double> %a) nounwind {
34 ; CHECK-LABEL: test_v1f64:
37 %b = call double @llvm.vector.reduce.fmul.f64.v1f64(double 1.0, <1 x double> %a)
41 define fp128 @test_v1f128(<1 x fp128> %a) nounwind {
42 ; CHECK-LABEL: test_v1f128:
45 %b = call fp128 @llvm.vector.reduce.fmul.f128.v1f128(fp128 0xL00000000000000003fff00000000000000, <1 x fp128> %a)
49 define float @test_v3f32(<3 x float> %a) nounwind {
50 ; CHECK-LABEL: test_v3f32:
52 ; CHECK-NEXT: fmul s1, s0, v0.s[1]
53 ; CHECK-NEXT: fmul s0, s1, v0.s[2]
55 %b = call float @llvm.vector.reduce.fmul.f32.v3f32(float 1.0, <3 x float> %a)
59 define fp128 @test_v2f128(<2 x fp128> %a) nounwind {
60 ; CHECK-LABEL: test_v2f128:
62 ; CHECK-NEXT: b __multf3
63 %b = call fp128 @llvm.vector.reduce.fmul.f128.v2f128(fp128 0xL00000000000000003fff00000000000000, <2 x fp128> %a)
67 define float @test_v16f32(<16 x float> %a) nounwind {
68 ; CHECK-LABEL: test_v16f32:
70 ; CHECK-NEXT: fmul s4, s0, v0.s[1]
71 ; CHECK-NEXT: fmul s4, s4, v0.s[2]
72 ; CHECK-NEXT: fmul s0, s4, v0.s[3]
73 ; CHECK-NEXT: fmul s0, s0, s1
74 ; CHECK-NEXT: fmul s0, s0, v1.s[1]
75 ; CHECK-NEXT: fmul s0, s0, v1.s[2]
76 ; CHECK-NEXT: fmul s0, s0, v1.s[3]
77 ; CHECK-NEXT: fmul s0, s0, s2
78 ; CHECK-NEXT: fmul s0, s0, v2.s[1]
79 ; CHECK-NEXT: fmul s0, s0, v2.s[2]
80 ; CHECK-NEXT: fmul s0, s0, v2.s[3]
81 ; CHECK-NEXT: fmul s0, s0, s3
82 ; CHECK-NEXT: fmul s0, s0, v3.s[1]
83 ; CHECK-NEXT: fmul s0, s0, v3.s[2]
84 ; CHECK-NEXT: fmul s0, s0, v3.s[3]
86 %b = call float @llvm.vector.reduce.fmul.f32.v16f32(float 1.0, <16 x float> %a)