RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-11-30 07:00:57 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Add v16i16 and v32i8 store patterns

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@137166 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Bruno Cardoso Lopes 2011-08-09 22:39:53 +00:00
parent 6ff9aa2aa7
commit 18deb04e9c
2 changed files with 32 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
lib/Target/X86/X86InstrSSE.td
View File

@ -4063,10 +4063,18 @@ let Predicates = [HasAVX] in {
(VMOVAPSYmr addr:$dst, VR256:$src)>;
def : Pat<(alignedstore (v8i32 VR256:$src), addr:$dst),
(VMOVAPSYmr addr:$dst, VR256:$src)>;
def : Pat<(alignedstore (v16i16 VR256:$src), addr:$dst),
(VMOVAPSYmr addr:$dst, VR256:$src)>;
def : Pat<(alignedstore (v32i8 VR256:$src), addr:$dst),
(VMOVAPSYmr addr:$dst, VR256:$src)>;
def : Pat<(store (v4i64 VR256:$src), addr:$dst),
(VMOVUPSYmr addr:$dst, VR256:$src)>;
def : Pat<(store (v8i32 VR256:$src), addr:$dst),
(VMOVUPSYmr addr:$dst, VR256:$src)>;
def : Pat<(store (v16i16 VR256:$src), addr:$dst),
(VMOVUPSYmr addr:$dst, VR256:$src)>;
def : Pat<(store (v32i8 VR256:$src), addr:$dst),
(VMOVUPSYmr addr:$dst, VR256:$src)>;
}
//===----------------------------------------------------------------------===//

24
test/CodeGen/X86/avx-load-store.ll
View File

@ -40,3 +40,27 @@ define <4 x double> @mov01(<4 x double> %v, double * %ptr) nounwind {
ret <4 x double> %i0
}
; CHECK: vmovaps %ymm
define void @storev16i16(<16 x i16> %a) nounwind {
store <16 x i16> %a, <16 x i16>* undef, align 32
unreachable
}
; CHECK: vmovups %ymm
define void @storev16i16_01(<16 x i16> %a) nounwind {
store <16 x i16> %a, <16 x i16>* undef, align 4
unreachable
}
; CHECK: vmovaps %ymm
define void @storev32i8(<32 x i8> %a) nounwind {
store <32 x i8> %a, <32 x i8>* undef, align 32
unreachable
}
; CHECK: vmovups %ymm
define void @storev32i8_01(<32 x i8> %a) nounwind {
store <32 x i8> %a, <32 x i8>* undef, align 4
unreachable
}