30 lines · plain
1; RUN: opt %s -passes='print<uniformity>' -disable-output 2>&1 | FileCheck %s2 3target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64"4target triple = "nvptx64-nvidia-cuda"5 6define ptx_kernel i32 @hidden_diverge(i32 %n, i32 %a, i32 %b) {7; CHECK-LABEL: for function 'hidden_diverge'8entry:9 %tid = call i32 @llvm.nvvm.read.ptx.sreg.tid.x()10 %cond.var = icmp slt i32 %tid, 011 br i1 %cond.var, label %B, label %C ; divergent12; CHECK: DIVERGENT: %cond.var =13; CHECK: DIVERGENT: br i1 %cond.var,14B:15 %cond.uni = icmp slt i32 %n, 016 br i1 %cond.uni, label %C, label %merge ; uniform17; CHECK-NOT: DIVERGENT: %cond.uni =18; CHECK-NOT: DIVERGENT: br i1 %cond.uni,19C:20 %phi.var.hidden = phi i32 [ 1, %entry ], [ 2, %B ]21; CHECK: DIVERGENT: %phi.var.hidden = phi i3222 br label %merge23merge:24 %phi.ipd = phi i32 [ %a, %B ], [ %b, %C ]25; CHECK: DIVERGENT: %phi.ipd = phi i3226 ret i32 %phi.ipd27}28 29declare i32 @llvm.nvvm.read.ptx.sreg.tid.x()30