blob: f737519bd1536fbc4ccd6ddbfd6f5767736f1e3b [file] [log] [blame]
Jim Grosbachb6535c32014-07-18 00:40:56 +00001; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s
2
3define <4 x float> @foo(<4 x float> %val, <4 x float> %test) nounwind {
Jim Grosbach19dd3082014-07-23 20:41:31 +00004; CHECK-LABEL: LCPI0_0:
Jim Grosbachb6535c32014-07-18 00:40:56 +00005; CHECK-NEXT: .long 1065353216 ## float 1.000000e+00
6; CHECK-NEXT: .long 1065353216 ## float 1.000000e+00
7; CHECK-NEXT: .long 1065353216 ## float 1.000000e+00
8; CHECK-NEXT: .long 1065353216 ## float 1.000000e+00
9; CHECK-LABEL: foo:
10; CHECK: cmpeqps %xmm1, %xmm0
11; CHECK-NEXT: andps LCPI0_0(%rip), %xmm0
12; CHECK-NEXT: retq
13
14 %cmp = fcmp oeq <4 x float> %val, %test
15 %ext = zext <4 x i1> %cmp to <4 x i32>
16 %result = sitofp <4 x i32> %ext to <4 x float>
17 ret <4 x float> %result
18}
Jim Grosbach19dd3082014-07-23 20:41:31 +000019
Jim Grosbach8f6f0852014-07-23 20:41:38 +000020; Make sure the operation doesn't try to get folded when the sizes don't match,
21; as that ends up crashing later when trying to form a bitcast operation for
22; the folded nodes.
23define void @foo1(<4 x float> %val, <4 x float> %test, <4 x double>* %p) nounwind {
Jim Grosbach19dd3082014-07-23 20:41:31 +000024; CHECK-LABEL: LCPI1_0:
Jim Grosbach8f6f0852014-07-23 20:41:38 +000025; CHECK-NEXT: .long 1 ## 0x1
26; CHECK-NEXT: .long 1 ## 0x1
27; CHECK-NEXT: .long 1 ## 0x1
28; CHECK-NEXT: .long 1 ## 0x1
29; CHECK-LABEL: foo1:
30; FIXME: The operation gets scalarized. If/when the compiler learns to better
31; use [V]CVTDQ2PD, this will need updated.
32; CHECK: cvtsi2sdq
33; CHECK: cvtsi2sdq
34; CHECK: cvtsi2sdq
35; CHECK: cvtsi2sdq
36 %cmp = fcmp oeq <4 x float> %val, %test
37 %ext = zext <4 x i1> %cmp to <4 x i32>
38 %result = sitofp <4 x i32> %ext to <4 x double>
39 store <4 x double> %result, <4 x double>* %p
40 ret void
41}
42
43; Also test the general purpose constant folding of int->fp.
44define void @foo2(<4 x float>* noalias %result) nounwind {
45; CHECK-LABEL: LCPI2_0:
Jim Grosbach19dd3082014-07-23 20:41:31 +000046; CHECK-NEXT: .long 1082130432 ## float 4.000000e+00
47; CHECK-NEXT: .long 1084227584 ## float 5.000000e+00
48; CHECK-NEXT: .long 1086324736 ## float 6.000000e+00
49; CHECK-NEXT: .long 1088421888 ## float 7.000000e+00
Jim Grosbach8f6f0852014-07-23 20:41:38 +000050; CHECK-LABEL: foo2:
51; CHECK: movaps LCPI2_0(%rip), %xmm0
Jim Grosbach19dd3082014-07-23 20:41:31 +000052
53 %val = uitofp <4 x i32> <i32 4, i32 5, i32 6, i32 7> to <4 x float>
54 store <4 x float> %val, <4 x float>* %result
55 ret void
56}