2208c97c1b
Similar to 806761a7629df268c8aed49657aeccffa6bca449 -mtriple= specifies the full target triple while -march= merely sets the architecture part of the default target triple, leaving a target triple which may not make sense. Therefore, -march= is error-prone and not recommended for tests without a target triple. The issue has been benign as we recognize $unknown-apple-darwin as ELF instead of rejecting it outrightly.
32 lines
860 B
LLVM
32 lines
860 B
LLVM
; RUN: llc -mtriple=hexagon -hexagon-align-loads=0 < %s | FileCheck %s
|
|
|
|
; CHECK-LABEL: aligned_load:
|
|
; CHECK: = vmem({{.*}})
|
|
define <16 x i32> @aligned_load(ptr %p, <16 x i32> %a) #0 {
|
|
%v = load <16 x i32>, ptr %p, align 64
|
|
ret <16 x i32> %v
|
|
}
|
|
|
|
; CHECK-LABEL: aligned_store:
|
|
; CHECK: vmem({{.*}}) =
|
|
define void @aligned_store(ptr %p, <16 x i32> %a) #0 {
|
|
store <16 x i32> %a, ptr %p, align 64
|
|
ret void
|
|
}
|
|
|
|
; CHECK-LABEL: unaligned_load:
|
|
; CHECK: = vmemu({{.*}})
|
|
define <16 x i32> @unaligned_load(ptr %p, <16 x i32> %a) #0 {
|
|
%v = load <16 x i32>, ptr %p, align 32
|
|
ret <16 x i32> %v
|
|
}
|
|
|
|
; CHECK-LABEL: unaligned_store:
|
|
; CHECK: vmemu({{.*}}) =
|
|
define void @unaligned_store(ptr %p, <16 x i32> %a) #0 {
|
|
store <16 x i32> %a, ptr %p, align 32
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { nounwind "target-cpu"="hexagonv60" "target-features"="+hvxv60,+hvx-length64b" }
|