RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-31 20:51:52 +01:00
Code Issues Projects Releases Wiki Activity

[X86][SSE4A] Remove the GCCBuiltins from the movntsd/movntss intrinsic defs so we can emit native IR from clang.

Browse Source 
Clang-side sibling commit to follow.

llvm-svn: 273002
This commit is contained in:
Simon Pilgrim 2016-06-17 14:27:38 +00:00
parent 6b5106aea7
commit 312535b60f
2 changed files with 8 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
include/llvm/IR/IntrinsicsX86.td
View File

@ -942,9 +942,9 @@ let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
def int_x86_sse4a_insertq : GCCBuiltin<"__builtin_ia32_insertq">,
Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty], [IntrNoMem]>;
def int_x86_sse4a_movnt_ss : GCCBuiltin<"__builtin_ia32_movntss">,
def int_x86_sse4a_movnt_ss :
Intrinsic<[], [llvm_ptr_ty, llvm_v4f32_ty], []>;
def int_x86_sse4a_movnt_sd : GCCBuiltin<"__builtin_ia32_movntsd">,
def int_x86_sse4a_movnt_sd :
Intrinsic<[], [llvm_ptr_ty, llvm_v2f64_ty], []>;
}

12
test/CodeGen/X86/sse4a-intrinsics-fast-isel.ll
View File

@ -78,11 +78,10 @@ define void @test_stream_sd(double* %p, <2 x double> %a) {
; X64: # BB#0:
; X64-NEXT: movntsd %xmm0, (%rdi)
; X64-NEXT: retq
%bc = bitcast double* %p to i8*
call void @llvm.x86.sse4a.movnt.sd(i8* %bc, <2 x double> %a)
%1 = extractelement <2 x double> %a, i64 0
store double %1, double* %p, align 1, !nontemporal !1
ret void
}
declare void @llvm.x86.sse4a.movnt.sd(i8*, <2 x double>) nounwind readnone
define void @test_mm_stream_ss(float* %p, <4 x float> %a) {
; X32-LABEL: test_mm_stream_ss:
@ -95,8 +94,9 @@ define void @test_mm_stream_ss(float* %p, <4 x float> %a) {
; X64: # BB#0:
; X64-NEXT: movntss %xmm0, (%rdi)
; X64-NEXT: retq
%bc = bitcast float* %p to i8*
call void @llvm.x86.sse4a.movnt.ss(i8* %bc, <4 x float> %a)
%1 = extractelement <4 x float> %a, i64 0
store float %1, float* %p, align 1, !nontemporal !1
ret void
}
declare void @llvm.x86.sse4a.movnt.ss(i8*, <4 x float>) nounwind readnone
!1 = !{i32 1}