xref: /llvm-project/llvm/test/CodeGen/ARM/vecreduce-fadd-legalization-soft-float.ll (revision 1cc4f8d17247cd9be88addd75d060f9321b6f8b0)
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=arm-none-eabi -mattr=-neon | FileCheck %s --check-prefix=CHECK
3
4declare float @llvm.experimental.vector.reduce.v2.fadd.f32.v4f32(float, <4 x float>)
5declare double @llvm.experimental.vector.reduce.v2.fadd.f64.v2f64(double, <2 x double>)
6declare fp128 @llvm.experimental.vector.reduce.v2.fadd.f128.v2f128(fp128, <2 x fp128>)
7
8define float @test_v4f32(<4 x float> %a) nounwind {
9; CHECK-LABEL: test_v4f32:
10; CHECK:       @ %bb.0:
11; CHECK-NEXT:    .save {r4, r5, r6, lr}
12; CHECK-NEXT:    push {r4, r5, r6, lr}
13; CHECK-NEXT:    mov r5, r1
14; CHECK-NEXT:    mov r1, r2
15; CHECK-NEXT:    mov r4, r3
16; CHECK-NEXT:    bl __aeabi_fadd
17; CHECK-NEXT:    mov r6, r0
18; CHECK-NEXT:    mov r0, r5
19; CHECK-NEXT:    mov r1, r4
20; CHECK-NEXT:    bl __aeabi_fadd
21; CHECK-NEXT:    mov r1, r0
22; CHECK-NEXT:    mov r0, r6
23; CHECK-NEXT:    bl __aeabi_fadd
24; CHECK-NEXT:    pop {r4, r5, r6, lr}
25; CHECK-NEXT:    mov pc, lr
26  %b = call fast float @llvm.experimental.vector.reduce.v2.fadd.f32.v4f32(float 0.0, <4 x float> %a)
27  ret float %b
28}
29
30define double @test_v2f64(<2 x double> %a) nounwind {
31; CHECK-LABEL: test_v2f64:
32; CHECK:       @ %bb.0:
33; CHECK-NEXT:    .save {r11, lr}
34; CHECK-NEXT:    push {r11, lr}
35; CHECK-NEXT:    bl __aeabi_dadd
36; CHECK-NEXT:    pop {r11, lr}
37; CHECK-NEXT:    mov pc, lr
38  %b = call fast double @llvm.experimental.vector.reduce.v2.fadd.f64.v2f64(double zeroinitializer, <2 x double> %a)
39  ret double %b
40}
41
42define fp128 @test_v2f128(<2 x fp128> %a) nounwind {
43; CHECK-LABEL: test_v2f128:
44; CHECK:       @ %bb.0:
45; CHECK-NEXT:    .save {r11, lr}
46; CHECK-NEXT:    push {r11, lr}
47; CHECK-NEXT:    .pad #16
48; CHECK-NEXT:    sub sp, sp, #16
49; CHECK-NEXT:    ldr r12, [sp, #36]
50; CHECK-NEXT:    str r12, [sp, #12]
51; CHECK-NEXT:    ldr r12, [sp, #32]
52; CHECK-NEXT:    str r12, [sp, #8]
53; CHECK-NEXT:    ldr r12, [sp, #28]
54; CHECK-NEXT:    str r12, [sp, #4]
55; CHECK-NEXT:    ldr r12, [sp, #24]
56; CHECK-NEXT:    str r12, [sp]
57; CHECK-NEXT:    bl __addtf3
58; CHECK-NEXT:    add sp, sp, #16
59; CHECK-NEXT:    pop {r11, lr}
60; CHECK-NEXT:    mov pc, lr
61  %b = call fast fp128 @llvm.experimental.vector.reduce.v2.fadd.f128.v2f128(fp128 zeroinitializer, <2 x fp128> %a)
62  ret fp128 %b
63}
64