llvm/test/CodeGen/X86/sse4a.ll
Simon Pilgrim 3aa4772f49 [X86][SSE4A] Regenerated SSE4A intrinsics tests
There are no VEX encoded versions of SSE4A instructions, make sure that AVX targets give the same output

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@272060 91177308-0d34-0410-b5e6-96231b3b80d8
2016-06-07 21:15:45 +00:00

105 lines
3.2 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i386-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=X32
; RUN: llc < %s -mtriple=i386-unknown-unknown -mattr=+sse4a,+avx | FileCheck %s --check-prefix=X32
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4a | FileCheck %s --check-prefix=X64
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4a,+avx | FileCheck %s --check-prefix=X64
define void @test1(i8* %p, <4 x float> %a) nounwind optsize ssp {
; X32-LABEL: test1:
; X32: # BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-NEXT: movntss %xmm0, (%eax)
; X32-NEXT: retl
;
; X64-LABEL: test1:
; X64: # BB#0:
; X64-NEXT: movntss %xmm0, (%rdi)
; X64-NEXT: retq
tail call void @llvm.x86.sse4a.movnt.ss(i8* %p, <4 x float> %a) nounwind
ret void
}
declare void @llvm.x86.sse4a.movnt.ss(i8*, <4 x float>)
define void @test2(i8* %p, <2 x double> %a) nounwind optsize ssp {
; X32-LABEL: test2:
; X32: # BB#0:
; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
; X32-NEXT: movntsd %xmm0, (%eax)
; X32-NEXT: retl
;
; X64-LABEL: test2:
; X64: # BB#0:
; X64-NEXT: movntsd %xmm0, (%rdi)
; X64-NEXT: retq
tail call void @llvm.x86.sse4a.movnt.sd(i8* %p, <2 x double> %a) nounwind
ret void
}
declare void @llvm.x86.sse4a.movnt.sd(i8*, <2 x double>)
define <2 x i64> @test3(<2 x i64> %x) nounwind uwtable ssp {
; X32-LABEL: test3:
; X32: # BB#0:
; X32-NEXT: extrq $2, $3, %xmm0
; X32-NEXT: retl
;
; X64-LABEL: test3:
; X64: # BB#0:
; X64-NEXT: extrq $2, $3, %xmm0
; X64-NEXT: retq
%1 = tail call <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64> %x, i8 3, i8 2)
ret <2 x i64> %1
}
declare <2 x i64> @llvm.x86.sse4a.extrqi(<2 x i64>, i8, i8) nounwind
define <2 x i64> @test4(<2 x i64> %x, <2 x i64> %y) nounwind uwtable ssp {
; X32-LABEL: test4:
; X32: # BB#0:
; X32-NEXT: extrq %xmm1, %xmm0
; X32-NEXT: retl
;
; X64-LABEL: test4:
; X64: # BB#0:
; X64-NEXT: extrq %xmm1, %xmm0
; X64-NEXT: retq
%1 = bitcast <2 x i64> %y to <16 x i8>
%2 = tail call <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64> %x, <16 x i8> %1) nounwind
ret <2 x i64> %2
}
declare <2 x i64> @llvm.x86.sse4a.extrq(<2 x i64>, <16 x i8>) nounwind
define <2 x i64> @test5(<2 x i64> %x, <2 x i64> %y) nounwind uwtable ssp {
; X32-LABEL: test5:
; X32: # BB#0:
; X32-NEXT: insertq $6, $5, %xmm1, %xmm0
; X32-NEXT: retl
;
; X64-LABEL: test5:
; X64: # BB#0:
; X64-NEXT: insertq $6, $5, %xmm1, %xmm0
; X64-NEXT: retq
%1 = tail call <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64> %x, <2 x i64> %y, i8 5, i8 6)
ret <2 x i64> %1
}
declare <2 x i64> @llvm.x86.sse4a.insertqi(<2 x i64>, <2 x i64>, i8, i8) nounwind
define <2 x i64> @test6(<2 x i64> %x, <2 x i64> %y) nounwind uwtable ssp {
; X32-LABEL: test6:
; X32: # BB#0:
; X32-NEXT: insertq %xmm1, %xmm0
; X32-NEXT: retl
;
; X64-LABEL: test6:
; X64: # BB#0:
; X64-NEXT: insertq %xmm1, %xmm0
; X64-NEXT: retq
%1 = tail call <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64> %x, <2 x i64> %y) nounwind
ret <2 x i64> %1
}
declare <2 x i64> @llvm.x86.sse4a.insertq(<2 x i64>, <2 x i64>) nounwind