mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-22 20:43:44 +02:00
262321d1ff
This patch lets the llvm tools handle the new HVX target features that
are added by frontend (clang). The target-features are of the form
"hvx-length64b" for 64 Byte HVX mode, "hvx-length128b" for 128 Byte mode HVX.
"hvx-double" is an alias to "hvx-length128b" and is soon will be deprecated.
The hvx version target feature is upgated form "+hvx" to "+hvxv{version_number}.
Eg: "+hvxv62"
For the correct HVX code generation, the user must use the following
target features.
For 64B mode: "+hvxv62" "+hvx-length64b"
For 128B mode: "+hvxv62" "+hvx-length128b"
Clang picks a default length if none is specified. If for some reason,
no hvx-length is specified to llvm, the compilation will bail out.
There is a corresponding clang patch.
Differential Revision: https://reviews.llvm.org/D38851
llvm-svn: 316101
29 lines
1.0 KiB
LLVM
29 lines
1.0 KiB
LLVM
; RUN: llc -march=hexagon < %s | FileCheck %s
|
|
target triple = "hexagon"
|
|
|
|
; Function Attrs: norecurse nounwind
|
|
define void @test(<32 x i32>* nocapture readonly %x, <32 x i32>* nocapture readnone %y, <32 x i32>* nocapture %a, <32 x i32>* nocapture %b) #0 {
|
|
entry:
|
|
; CHECK: v0 = vmem(r0+#7):nt
|
|
%add.ptr = getelementptr inbounds <32 x i32>, <32 x i32>* %x, i32 7
|
|
%0 = load <32 x i32>, <32 x i32>* %add.ptr, align 128, !tbaa !1, !nontemporal !4
|
|
|
|
; CHECK: v1.cur = vmem(r2+#0):nt
|
|
%1 = load <32 x i32>, <32 x i32>* %a, align 128, !tbaa !1, !nontemporal !4
|
|
|
|
; CHECK: vmem(r3+#3):nt = v1
|
|
%add.ptr2 = getelementptr inbounds <32 x i32>, <32 x i32>* %b, i32 3
|
|
store <32 x i32> %1, <32 x i32>* %add.ptr2, align 128, !tbaa !1, !nontemporal !4
|
|
|
|
; CHECK: vmem(r2+#0):nt = v0
|
|
store <32 x i32> %0, <32 x i32>* %a, align 128, !tbaa !1, !nontemporal !4
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { norecurse nounwind "target-cpu"="hexagonv60" "target-features"="+hvxv60,+hvx-length128b" }
|
|
|
|
!1 = !{!2, !2, i64 0}
|
|
!2 = !{!"omnipotent char", !3, i64 0}
|
|
!3 = !{!"Simple C/C++ TBAA"}
|
|
!4 = !{i32 1}
|