1 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx,+f16c | FileCheck %s
2 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+f16c | FileCheck %s
4 define <4 x float> @test_x86_vcvtph2ps_128(<8 x i16> %a0) {
5 ; CHECK-LABEL: test_x86_vcvtph2ps_128:
8 %res = call <4 x float> @llvm.x86.vcvtph2ps.128(<8 x i16> %a0) ; <<4 x float>> [#uses=1]
11 declare <4 x float> @llvm.x86.vcvtph2ps.128(<8 x i16>) nounwind readonly
14 define <8 x float> @test_x86_vcvtph2ps_256(<8 x i16> %a0) {
15 ; CHECK-LABEL: test_x86_vcvtph2ps_256:
18 %res = call <8 x float> @llvm.x86.vcvtph2ps.256(<8 x i16> %a0) ; <<8 x float>> [#uses=1]
21 declare <8 x float> @llvm.x86.vcvtph2ps.256(<8 x i16>) nounwind readonly
23 define <8 x float> @test_x86_vcvtph2ps_256_m(<8 x i16>* nocapture %a) nounwind {
25 ; CHECK-LABEL: test_x86_vcvtph2ps_256_m:
28 %tmp1 = load <8 x i16>, <8 x i16>* %a, align 16
29 %0 = tail call <8 x float> @llvm.x86.vcvtph2ps.256(<8 x i16> %tmp1)
33 define <8 x i16> @test_x86_vcvtps2ph_128(<4 x float> %a0) {
34 ; CHECK-LABEL: test_x86_vcvtps2ph_128:
37 %res = call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %a0, i32 0) ; <<8 x i16>> [#uses=1]
40 declare <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float>, i32) nounwind readonly
42 define <8 x i16> @test_x86_vcvtps2ph_256(<8 x float> %a0) {
43 ; CHECK-LABEL: test_x86_vcvtps2ph_256:
46 %res = call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %a0, i32 0) ; <<8 x i16>> [#uses=1]
49 declare <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float>, i32) nounwind readonly
51 define <4 x float> @test_x86_vcvtps2ph_128_scalar(i64* %ptr) {
52 ; CHECK-LABEL: test_x86_vcvtps2ph_128_scalar:
56 %load = load i64, i64* %ptr
57 %ins1 = insertelement <2 x i64> undef, i64 %load, i32 0
58 %ins2 = insertelement <2 x i64> %ins1, i64 0, i32 1
59 %bc = bitcast <2 x i64> %ins2 to <8 x i16>
60 %res = tail call <4 x float> @llvm.x86.vcvtph2ps.128(<8 x i16> %bc) #2
64 define <4 x float> @test_x86_vcvtps2ph_128_scalar2(i64* %ptr) {
65 ; CHECK-LABEL: test_x86_vcvtps2ph_128_scalar2:
69 %load = load i64, i64* %ptr
70 %ins = insertelement <2 x i64> undef, i64 %load, i32 0
71 %bc = bitcast <2 x i64> %ins to <8 x i16>
72 %res = tail call <4 x float> @llvm.x86.vcvtph2ps.128(<8 x i16> %bc)
76 define void @test_x86_vcvtps2ph_256_m(<8 x i16>* nocapture %d, <8 x float> %a) nounwind {
78 ; CHECK-LABEL: test_x86_vcvtps2ph_256_m:
80 ; CHECK: vcvtps2ph $3, %ymm0, (%
81 %0 = tail call <8 x i16> @llvm.x86.vcvtps2ph.256(<8 x float> %a, i32 3)
82 store <8 x i16> %0, <8 x i16>* %d, align 16
86 define void @test_x86_vcvtps2ph_128_m(<4 x i16>* nocapture %d, <4 x float> %a) nounwind {
88 ; CHECK-LABEL: test_x86_vcvtps2ph_128_m:
90 ; CHECK: vcvtps2ph $3, %xmm0, (%
91 %0 = tail call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %a, i32 3)
92 %1 = shufflevector <8 x i16> %0, <8 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
93 store <4 x i16> %1, <4 x i16>* %d, align 8
97 define void @test_x86_vcvtps2ph_128_m2(double* nocapture %hf4x16, <4 x float> %f4x32) #0 {
99 ; CHECK-LABEL: test_x86_vcvtps2ph_128_m2:
101 ; CHECK: vcvtps2ph $3, %xmm0, (%
102 %0 = tail call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %f4x32, i32 3)
103 %1 = bitcast <8 x i16> %0 to <2 x double>
104 %vecext = extractelement <2 x double> %1, i32 0
105 store double %vecext, double* %hf4x16, align 8
109 define void @test_x86_vcvtps2ph_128_m3(i64* nocapture %hf4x16, <4 x float> %f4x32) #0 {
111 ; CHECK-LABEL: test_x86_vcvtps2ph_128_m3:
113 ; CHECK: vcvtps2ph $3, %xmm0, (%
114 %0 = tail call <8 x i16> @llvm.x86.vcvtps2ph.128(<4 x float> %f4x32, i32 3)
115 %1 = bitcast <8 x i16> %0 to <2 x i64>
116 %vecext = extractelement <2 x i64> %1, i32 0
117 store i64 %vecext, i64* %hf4x16, align 8