1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE
3 ; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
4 ; RUN: llc < %s -fast-isel -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512dq,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX,AVX512
6 ; NOTE: This should use IR equivalent to what is generated by clang/test/CodeGen/sse2-builtins.c
8 define i64 @test_mm_cvtsd_si64(<2 x double> %a0) nounwind {
9 ; SSE-LABEL: test_mm_cvtsd_si64:
11 ; SSE-NEXT: cvtsd2si %xmm0, %rax
14 ; AVX-LABEL: test_mm_cvtsd_si64:
16 ; AVX-NEXT: vcvtsd2si %xmm0, %rax
18 %res = call i64 @llvm.x86.sse2.cvtsd2si64(<2 x double> %a0)
21 declare i64 @llvm.x86.sse2.cvtsd2si64(<2 x double>) nounwind readnone
23 define i64 @test_mm_cvtsi128_si64(<2 x i64> %a0) nounwind {
24 ; SSE-LABEL: test_mm_cvtsi128_si64:
26 ; SSE-NEXT: movq %xmm0, %rax
29 ; AVX-LABEL: test_mm_cvtsi128_si64:
31 ; AVX-NEXT: vmovq %xmm0, %rax
33 %res = extractelement <2 x i64> %a0, i32 0
37 define <2 x double> @test_mm_cvtsi64_sd(<2 x double> %a0, i64 %a1) nounwind {
38 ; SSE-LABEL: test_mm_cvtsi64_sd:
40 ; SSE-NEXT: cvtsi2sd %rdi, %xmm0
43 ; AVX-LABEL: test_mm_cvtsi64_sd:
45 ; AVX-NEXT: vcvtsi2sd %rdi, %xmm0, %xmm0
47 %res = call <2 x double> @llvm.x86.sse2.cvtsi642sd(<2 x double> %a0, i64 %a1)
50 declare <2 x double> @llvm.x86.sse2.cvtsi642sd(<2 x double>, i64) nounwind readnone
52 define <2 x i64> @test_mm_cvtsi64_si128(i64 %a0) nounwind {
53 ; SSE-LABEL: test_mm_cvtsi64_si128:
55 ; SSE-NEXT: movq %rdi, %xmm0
58 ; AVX-LABEL: test_mm_cvtsi64_si128:
60 ; AVX-NEXT: vmovq %rdi, %xmm0
62 %res0 = insertelement <2 x i64> undef, i64 %a0, i32 0
63 %res1 = insertelement <2 x i64> %res0, i64 0, i32 1
67 define i64 @test_mm_cvttsd_si64(<2 x double> %a0) nounwind {
68 ; SSE-LABEL: test_mm_cvttsd_si64:
70 ; SSE-NEXT: cvttsd2si %xmm0, %rax
73 ; AVX-LABEL: test_mm_cvttsd_si64:
75 ; AVX-NEXT: vcvttsd2si %xmm0, %rax
77 %res = call i64 @llvm.x86.sse2.cvttsd2si64(<2 x double> %a0)
80 declare i64 @llvm.x86.sse2.cvttsd2si64(<2 x double>) nounwind readnone
82 define <2 x i64> @test_mm_loadu_si64(i64* %a0) nounwind {
83 ; SSE-LABEL: test_mm_loadu_si64:
85 ; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
88 ; AVX-LABEL: test_mm_loadu_si64:
90 ; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
92 %ld = load i64, i64* %a0, align 1
93 %res0 = insertelement <2 x i64> undef, i64 %ld, i32 0
94 %res1 = insertelement <2 x i64> %res0, i64 0, i32 1
98 define void @test_mm_stream_si64(i64 *%a0, i64 %a1) {
99 ; CHECK-LABEL: test_mm_stream_si64:
101 ; CHECK-NEXT: movntiq %rsi, (%rdi)
103 store i64 %a1, i64* %a0, align 1, !nontemporal !0