1 ; RUN: llc < %s -march=x86 -mcpu=corei7 | FileCheck %s
2 ; RUN: llc < %s -march=x86 -mcpu=core-avx-i | FileCheck %s --check-prefix=AVX
4 define <1 x float> @test1(<1 x double>* %p) nounwind {
7 ; CHECK-NEXT: pushl %eax
8 ; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
9 ; CHECK-NEXT: movsd (%eax), %xmm0
10 ; CHECK-NEXT: cvtsd2ss %xmm0, %xmm0
11 ; CHECK-NEXT: movss %xmm0, (%esp)
12 ; CHECK-NEXT: flds (%esp)
13 ; CHECK-NEXT: popl %eax
18 ; AVX-NEXT: pushl %eax
19 ; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
20 ; AVX-NEXT: vmovsd (%eax), %xmm0
21 ; AVX-NEXT: vcvtsd2ss %xmm0, %xmm0, %xmm0
22 ; AVX-NEXT: vmovss %xmm0, (%esp)
23 ; AVX-NEXT: flds (%esp)
26 %x = load <1 x double>* %p
27 %y = fptrunc <1 x double> %x to <1 x float>
31 define <2 x float> @test2(<2 x double>* %p) nounwind {
34 ; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
35 ; CHECK-NEXT: cvtpd2ps (%eax), %xmm0
40 ; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
41 ; AVX-NEXT: vcvtpd2psx (%eax), %xmm0
43 %x = load <2 x double>* %p
44 %y = fptrunc <2 x double> %x to <2 x float>
48 define <4 x float> @test3(<4 x double>* %p) nounwind {
51 ; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
52 ; CHECK-NEXT: cvtpd2ps 16(%eax), %xmm1
53 ; CHECK-NEXT: cvtpd2ps (%eax), %xmm0
54 ; CHECK-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
59 ; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
60 ; AVX-NEXT: vcvtpd2psy (%eax), %xmm0
62 %x = load <4 x double>* %p
63 %y = fptrunc <4 x double> %x to <4 x float>
67 define <8 x float> @test4(<8 x double>* %p) nounwind {
70 ; CHECK-NEXT: movl {{[0-9]+}}(%esp), %eax
71 ; CHECK-NEXT: cvtpd2ps 16(%eax), %xmm1
72 ; CHECK-NEXT: cvtpd2ps (%eax), %xmm0
73 ; CHECK-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
74 ; CHECK-NEXT: cvtpd2ps 48(%eax), %xmm2
75 ; CHECK-NEXT: cvtpd2ps 32(%eax), %xmm1
76 ; CHECK-NEXT: movlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0]
81 ; AVX-NEXT: movl {{[0-9]+}}(%esp), %eax
82 ; AVX-NEXT: vcvtpd2psy (%eax), %xmm0
83 ; AVX-NEXT: vcvtpd2psy 32(%eax), %xmm1
84 ; AVX-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
86 %x = load <8 x double>* %p
87 %y = fptrunc <8 x double> %x to <8 x float>