Compiler Explorer

Source code

source_filename = "/app/example.cpp"
target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-p7:160:256:256:32-p8:128:128:128:48-p9:192:256:256:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5-G1-ni:7:8:9"
target triple = "amdgcn-amd-amdhsa"

define amdgpu_kernel void @kernel_sum(ptr addrspace(1) noundef %partial_sums.coerce, i32 noundef %Nparticles) #0 {
entry:
  %partial_sums = alloca ptr, align 8, addrspace(5)
  %partial_sums.addr = alloca ptr, align 8, addrspace(5)
  %Nparticles.addr = alloca i32, align 4, addrspace(5)
  %x = alloca i32, align 4, addrspace(5)
  %sum = alloca float, align 4, addrspace(5)
  %num_blocks = alloca i32, align 4, addrspace(5)
  %partial_sums.ascast = addrspacecast ptr addrspace(5) %partial_sums to ptr
  %partial_sums.addr.ascast = addrspacecast ptr addrspace(5) %partial_sums.addr to ptr
  %Nparticles.addr.ascast = addrspacecast ptr addrspace(5) %Nparticles.addr to ptr
  %x.ascast = addrspacecast ptr addrspace(5) %x to ptr
  %sum.ascast = addrspacecast ptr addrspace(5) %sum to ptr
  %num_blocks.ascast = addrspacecast ptr addrspace(5) %num_blocks to ptr
  store ptr addrspace(1) %partial_sums.coerce, ptr %partial_sums.ascast, align 8
  %partial_sums1 = load ptr, ptr %partial_sums.ascast, align 8
  store ptr %partial_sums1, ptr %partial_sums.addr.ascast, align 8
  store i32 %Nparticles, ptr %Nparticles.addr.ascast, align 4
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %x)
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %sum)
  store float 0.000000e+00, ptr %sum.ascast, align 4
  call void @llvm.lifetime.start.p5(ptr addrspace(5) %num_blocks)
  %0 = load i32, ptr %Nparticles.addr.ascast, align 4
  %add = add nsw i32 %0, 256
  %sub = sub nsw i32 %add, 1
  %div = sdiv i32 %sub, 256
  store i32 %div, ptr %num_blocks.ascast, align 4
  store i32 0, ptr %x.ascast, align 4
  br label %for.cond

for.cond:                                         ; preds = %for.inc, %entry
  %1 = load i32, ptr %x.ascast, align 4
  %2 = load i32, ptr %num_blocks.ascast, align 4
  %cmp = icmp slt i32 %1, %2
  br i1 %cmp, label %for.body, label %for.end

for.body:                                         ; preds = %for.cond
  %3 = load ptr, ptr %partial_sums.addr.ascast, align 8
  %4 = load i32, ptr %x.ascast, align 4
  %idxprom = sext i32 %4 to i64
  %arrayidx = getelementptr inbounds float, ptr %3, i64 %idxprom
  %5 = load float, ptr %arrayidx, align 4
  %6 = load float, ptr %sum.ascast, align 4
  %add2 = fadd contract float %6, %5
  store float %add2, ptr %sum.ascast, align 4
  br label %for.inc

for.inc:                                          ; preds = %for.body
  %7 = load i32, ptr %x.ascast, align 4
  %inc = add nsw i32 %7, 1
  store i32 %inc, ptr %x.ascast, align 4
  br label %for.cond

for.end:                                          ; preds = %for.cond
  %8 = load float, ptr %sum.ascast, align 4
  %9 = load ptr, ptr %partial_sums.addr.ascast, align 8
  %arrayidx3 = getelementptr inbounds float, ptr %9, i64 0
  store float %8, ptr %arrayidx3, align 4
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %num_blocks)
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %sum)
  call void @llvm.lifetime.end.p5(ptr addrspace(5) %x)
  ret void
}

define hidden spir_kernel void @kernel_sum_with_spirv_opts_allowed(ptr addrspace(1) noundef %partial_sums.coerce, i32 noundef %Nparticles) #0 {
entry:
  %0 = ptrtoint ptr addrspace(1) %partial_sums.coerce to i64
  %1 = inttoptr i64 %0 to ptr
  %cmp8 = icmp sgt i32 %Nparticles, 0
  br i1 %cmp8, label %for.body.preheader, label %for.end

for.body.preheader:
  %sub = add nuw nsw i32 %Nparticles, 255
  %div1112 = lshr i32 %sub, 8
  %wide.trip.count = zext nneg i32 %div1112 to i64
  br label %for.body

for.body:
  %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]
  %sum.09 = phi float [ 0.000000e+00, %for.body.preheader ], [ %add2, %for.body ]
  %arrayidx = getelementptr inbounds nuw float, ptr %1, i64 %indvars.iv
  %2 = load float, ptr %arrayidx, align 4
  %add2 = fadd contract float %sum.09, %2
  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
  %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count
  br i1 %exitcond.not, label %for.end, label %for.body

for.end:
  %sum.0.lcssa = phi float [ 0.000000e+00, %entry ], [ %add2, %for.body ]
  store float %sum.0.lcssa, ptr %1, align 4
  ret void
}

attributes #0 = { mustprogress nofree norecurse nosync nounwind memory(argmem: readwrite) "amdgpu-agpr-alloc"="0" "amdgpu-flat-work-group-size"="1,1024" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx906" "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64" "uniform-work-group-size"="true" }

!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !1, producer: "clang version 22.0.0git (https://github.com/llvm/llvm-project.git e665cf397686e881cfb0728d896d2f375c0aead5)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, splitDebugInlining: false, nameTableKind: None)
!1 = !DIFile(filename: "/app/example.cpp", directory: "/app", checksumkind: CSK_MD5, checksum: "ef522d5ba4f4e8751ebcefa37a16d95a")
!2 = !{i32 1, !"amdhsa_code_object_version", i32 600}
!3 = !{i32 1, !"amdgpu_printf_kind", !"hostcall"}
!4 = !{i32 7, !"Dwarf Version", i32 5}
!5 = !{i32 2, !"Debug Info Version", i32 3}
!6 = !{i32 1, !"wchar_size", i32 4}
!7 = !{i32 8, !"PIC Level", i32 2}
!8 = !{i32 7, !"frame-pointer", i32 2}
!9 = !{i32 0, i32 0}
!10 = !{!"clang version 22.0.0git (https://github.com/llvm/llvm-project.git e665cf397686e881cfb0728d896d2f375c0aead5)"}
!11 = distinct !DISubprogram(name: "kernel_sum", linkageName: "kernel_sum(float*, int)", scope: !12, file: !12, line: 4, type: !13, scopeLine: 5, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !19)
!12 = !DIFile(filename: "example.cpp", directory: "/app", checksumkind: CSK_MD5, checksum: "ef522d5ba4f4e8751ebcefa37a16d95a")
!13 = !DISubroutineType(cc: DW_CC_LLVM_SpirFunction, types: !14)
!14 = !{null, !15, !17}
!15 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !16, size: 64, dwarfAddressSpace: 4)
!16 = !DIBasicType(name: "float", size: 32, encoding: DW_ATE_float)
!17 = !DIDerivedType(tag: DW_TAG_const_type, baseType: !18)
!18 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
!19 = !{!20, !21, !22, !23, !24}
!20 = !DILocalVariable(name: "partial_sums", arg: 1, scope: !11, file: !12, line: 4, type: !15)
!21 = !DILocalVariable(name: "Nparticles", arg: 2, scope: !11, file: !12, line: 4, type: !17)
!22 = !DILocalVariable(name: "x", scope: !11, file: !12, line: 6, type: !18)
!23 = !DILocalVariable(name: "sum", scope: !11, file: !12, line: 7, type: !16)
!24 = !DILocalVariable(name: "num_blocks", scope: !11, file: !12, line: 8, type: !18)
!25 = !{i32 1024, i32 1, i32 1}
!26 = !DILocation(line: 0, scope: !11)
!27 = !DILocation(line: 9, column: 17, scope: !28)
!28 = distinct !DILexicalBlock(scope: !29, file: !12, line: 9, column: 3)
!29 = distinct !DILexicalBlock(scope: !11, file: !12, line: 9, column: 3)
!30 = !DILocation(line: 9, column: 3, scope: !29)
!31 = !DILocation(line: 8, column: 45, scope: !11)
!32 = !DILocation(line: 8, column: 50, scope: !11)
!33 = !DILocation(line: 10, column: 12, scope: !34)
!34 = distinct !DILexicalBlock(scope: !28, file: !12, line: 9, column: 36)
!35 = !{!36, !36, i64 0}
!36 = !{!"float", !37, i64 0}
!37 = !{!"omnipotent char", !38, i64 0}
!38 = !{!"Simple C++ TBAA"}
!39 = !DILocation(line: 10, column: 9, scope: !34)
!40 = !DILocation(line: 9, column: 32, scope: !28)
!41 = distinct !{!41, !30, !42, !43}
!42 = !DILocation(line: 11, column: 3, scope: !29)
!43 = !{!"llvm.loop.mustprogress"}
!44 = !DILocation(line: 12, column: 19, scope: !11)
!45 = !DILocation(line: 13, column: 1, scope: !11)