1; RUN: llc < %s | FileCheck %s 2; RUN: %if ptxas %{ llc < %s | %ptxas-verify %} 3 4target triple = "nvptx64-nvidia-cuda" 5 6declare void @llvm.nvvm.barrier0() 7 8; Load a value, then syncthreads. Branch, and use the loaded value only on one 9; side of the branch. The load shouldn't be sunk beneath the call, because 10; syncthreads is modeled as maystore. 11define i32 @f(i32 %x, ptr %ptr, i1 %cond) { 12Start: 13 ; CHECK: ld.u32 14 %ptr_val = load i32, ptr %ptr 15 ; CHECK: bar.sync 16 call void @llvm.nvvm.barrier0() 17 br i1 %cond, label %L1, label %L2 18L1: 19 %ptr_val2 = add i32 %ptr_val, 100 20 br label %L2 21L2: 22 %v4 = phi i32 [ %x, %Start ], [ %ptr_val2, %L1 ] 23 %v5 = add i32 %v4, 1000 24 ret i32 %v5 25} 26