1; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_20 | FileCheck %s 2; RUN: %if ptxas %{ llc < %s -mtriple=nvptx64 -mcpu=sm_20 | %ptxas-verify %} 3 4target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64" 5target triple = "nvptx64-nvidia-cuda" 6 7%class.float3 = type { float, float, float } 8 9; Function Attrs: nounwind 10; CHECK-LABEL: some_kernel 11define ptx_kernel void @some_kernel(ptr nocapture %dst) #0 { 12_ZL11compute_vecRK6float3jb.exit: 13 %ret_vec.sroa.8.i = alloca float, align 4 14 %0 = tail call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x() 15 %1 = tail call i32 @llvm.nvvm.read.ptx.sreg.ntid.x() 16 %2 = mul nsw i32 %1, %0 17 %3 = tail call i32 @llvm.nvvm.read.ptx.sreg.tid.x() 18 %4 = add nsw i32 %2, %3 19 %5 = zext i32 %4 to i64 20 call void @llvm.lifetime.start.p0(i64 4, ptr %ret_vec.sroa.8.i) 21 %6 = and i32 %4, 15 22 %7 = icmp eq i32 %6, 0 23 %8 = select i1 %7, float 0.000000e+00, float -1.000000e+00 24 store float %8, ptr %ret_vec.sroa.8.i, align 4 25; CHECK: max.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, 0f00000000 26 %9 = fcmp olt float %8, 0.000000e+00 27 %ret_vec.sroa.8.i.val = load float, ptr %ret_vec.sroa.8.i, align 4 28 %10 = select i1 %9, float 0.000000e+00, float %ret_vec.sroa.8.i.val 29 call void @llvm.lifetime.end.p0(i64 4, ptr %ret_vec.sroa.8.i) 30 %11 = getelementptr inbounds %class.float3, ptr %dst, i64 %5, i32 0 31 store float 0.000000e+00, ptr %11, align 4 32 %12 = getelementptr inbounds %class.float3, ptr %dst, i64 %5, i32 1 33 store float %10, ptr %12, align 4 34 %13 = getelementptr inbounds %class.float3, ptr %dst, i64 %5, i32 2 35 store float 0.000000e+00, ptr %13, align 4 36 ret void 37} 38 39; Function Attrs: nounwind readnone 40declare i32 @llvm.nvvm.read.ptx.sreg.ctaid.x() #1 41 42; Function Attrs: nounwind readnone 43declare i32 @llvm.nvvm.read.ptx.sreg.ntid.x() #1 44 45; Function Attrs: nounwind readnone 46declare i32 @llvm.nvvm.read.ptx.sreg.tid.x() #1 47 48; Function Attrs: nounwind 49declare void @llvm.lifetime.start.p0(i64, ptr nocapture) #2 50 51; Function Attrs: nounwind 52declare void @llvm.lifetime.end.p0(i64, ptr nocapture) #2 53 54attributes #0 = { nounwind "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="true" "no-nans-fp-math"="true" "stack-protector-buffer-size"="8" "no-signed-zeros-fp-math"="true" "use-soft-float"="false" } 55attributes #1 = { nounwind readnone } 56attributes #2 = { nounwind } 57 58!llvm.ident = !{!1} 59!1 = !{!"clang version 3.5.1 (tags/RELEASE_351/final)"} 60