llvm/test/CodeGen/AMDGPU/mesa_regression.ll

; RUN: llc -O2 -mtriple=amdgpu8.03--amdhsa -amdgpu-scalarize-global-loads=false  < %s | FileCheck %s

; CHECK-LABEL: %entry
; CHECK: flat_load_dwordx4

define amdgpu_kernel void @store_global(ptr addrspace(1) nocapture %out, ptr addrspace(1) nocapture readonly %in) {
entry:
  %tmp = load <16 x double>, ptr addrspace(1) %in
  store <16 x double> %tmp, ptr addrspace(1) %out
  ret void
}