1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -fast-isel -mtriple=i386-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=X32
3 ; RUN: llc < %s -fast-isel -mtriple=i386-unknown-unknown -mattr=+sse4a,+avx | FileCheck %s --check-prefix=X32
4 ; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=X64
5 ; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+sse4a,+avx | FileCheck %s --check-prefix=X64
7 ; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/sse4a-builtins.c
9 define <2 x i64> @test_mm_extracti_si64(<2 x i64> %x) {
10 ; X32-LABEL: test_mm_extracti_si64:
12 ; X32-NEXT: extrq $2, $3, %xmm0
15 ; X64-LABEL: test_mm_extracti_si64:
17 ; X64-NEXT: extrq $2, $3, %xmm0
19 %res = call <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64> %x, i8 3, i8 2)
22 declare <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64>, i8, i8) nounwind readnone
24 define <2 x i64> @test_mm_extract_si64(<2 x i64> %x, <2 x i64> %y) {
25 ; X32-LABEL: test_mm_extract_si64:
27 ; X32-NEXT: extrq %xmm1, %xmm0
30 ; X64-LABEL: test_mm_extract_si64:
32 ; X64-NEXT: extrq %xmm1, %xmm0
34 %bc = bitcast <2 x i64> %y to <16 x i8>
35 %res = call <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64> %x, <16 x i8> %bc)
38 declare <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64>, <16 x i8>) nounwind readnone
40 define <2 x i64> @test_mm_inserti_si64(<2 x i64> %x, <2 x i64> %y) {
41 ; X32-LABEL: test_mm_inserti_si64:
43 ; X32-NEXT: insertq $6, $5, %xmm1, %xmm0
46 ; X64-LABEL: test_mm_inserti_si64:
48 ; X64-NEXT: insertq $6, $5, %xmm1, %xmm0
50 %res = call <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64> %x, <2 x i64> %y, i8 5, i8 6)
53 declare <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64>, <2 x i64>, i8, i8) nounwind readnone
55 define <2 x i64> @test_mm_insert_si64(<2 x i64> %x, <2 x i64> %y) {
56 ; X32-LABEL: test_mm_insert_si64:
58 ; X32-NEXT: insertq %xmm1, %xmm0
61 ; X64-LABEL: test_mm_insert_si64:
63 ; X64-NEXT: insertq %xmm1, %xmm0
65 %res = call <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64> %x, <2 x i64> %y)
68 declare <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64>, <2 x i64>) nounwind readnone
70 define void @test_stream_sd(double* %p, <2 x double> %a) {
71 ; X32-LABEL: test_stream_sd:
73 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
74 ; X32-NEXT: movntsd %xmm0, (%eax)
77 ; X64-LABEL: test_stream_sd:
79 ; X64-NEXT: movntsd %xmm0, (%rdi)
81 %1 = extractelement <2 x double> %a, i64 0
82 store double %1, double* %p, align 1, !nontemporal !1
86 define void @test_mm_stream_ss(float* %p, <4 x float> %a) {
87 ; X32-LABEL: test_mm_stream_ss:
89 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
90 ; X32-NEXT: movntss %xmm0, (%eax)
93 ; X64-LABEL: test_mm_stream_ss:
95 ; X64-NEXT: movntss %xmm0, (%rdi)
97 %1 = extractelement <4 x float> %a, i64 0
98 store float %1, float* %p, align 1, !nontemporal !1