31 lines
1.0 KiB
LLVM
31 lines
1.0 KiB
LLVM
; RUN: opt -march=hexagon -hexagon-loop-idiom -S < %s | FileCheck %s
|
|
|
|
; Make sure we don't convert load/store loops into memcpy if the access type
|
|
; is a vector. Using vector instructions is generally better in such cases.
|
|
|
|
; CHECK-NOT: @llvm.memcpy
|
|
|
|
%s.0 = type { i32 }
|
|
|
|
define void @f0(%s.0* noalias %a0, %s.0* noalias %a1) #0 align 2 {
|
|
b0:
|
|
br i1 undef, label %b1, label %b2
|
|
|
|
b1: ; preds = %b1, %b0
|
|
%v0 = phi i32 [ %v7, %b1 ], [ 0, %b0 ]
|
|
%v1 = mul nuw nsw i32 %v0, 64
|
|
%v2 = getelementptr %s.0, %s.0* %a0, i32 %v1
|
|
%v3 = getelementptr %s.0, %s.0* %a1, i32 %v1
|
|
%v4 = bitcast %s.0* %v2 to <64 x i32>*
|
|
%v5 = load <64 x i32>, <64 x i32>* %v4, align 256
|
|
%v6 = bitcast %s.0* %v3 to <64 x i32>*
|
|
store <64 x i32> %v5, <64 x i32>* %v6, align 256
|
|
%v7 = add nuw nsw i32 %v0, 1
|
|
br i1 undef, label %b1, label %b2
|
|
|
|
b2: ; preds = %b1, %b0
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { "target-cpu"="hexagonv60" "target-features"="+hvx,+hvx-length64b" }
|