[llvm] r282506 - Propagate DBG_VALUE entries when there are unvisited predecessors
Mehdi Amini via llvm-commits
llvm-commits at lists.llvm.org
Wed Sep 28 23:01:12 PDT 2016
Hi,
After this commit, an LTO link of llvm-tblgen with debug info is trigger many more calls to strlen().
It is not calling it 21928214 (~22M) times, versus 12585102 (12.5M) before. And the livedebugvalues pass gets >3x slower (it is still only 3% of the total link time though).
Is it expected?
The object file produced is slightly larger (31052564 after vs 30953876 before), which is due to the expected “better” debug info I guess.
—
Mehdi
> On Sep 27, 2016, at 9:46 AM, Keith Walker via llvm-commits <llvm-commits at lists.llvm.org> wrote:
>
> Author: kwalker
> Date: Tue Sep 27 11:46:07 2016
> New Revision: 282506
>
> URL: http://llvm.org/viewvc/llvm-project?rev=282506&view=rev
> Log:
> Propagate DBG_VALUE entries when there are unvisited predecessors
>
> Variables are sometimes missing their debug location information in
> blocks in which the variables should be available. This would occur
> when one or more predecessor blocks had not yet been visited by the
> routine which propagated the information from predecessor blocks.
>
> This is addressed by only considering predecessor blocks which have
> already been visited.
>
> The solution to this problem was suggested by Daniel Berlin on the
> LLVM developer mailing list.
>
> Differential Revision: https://reviews.llvm.org/D24927
>
> Added:
> llvm/trunk/test/CodeGen/ARM/dbg-range-extension.mir
> Modified:
> llvm/trunk/lib/CodeGen/LiveDebugValues.cpp
>
> Modified: llvm/trunk/lib/CodeGen/LiveDebugValues.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LiveDebugValues.cpp?rev=282506&r1=282505&r2=282506&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/LiveDebugValues.cpp (original)
> +++ llvm/trunk/lib/CodeGen/LiveDebugValues.cpp Tue Sep 27 11:46:07 2016
> @@ -201,7 +201,8 @@ private:
> VarLocInMBB &OutLocs, VarLocMap &VarLocIDs);
>
> bool join(MachineBasicBlock &MBB, VarLocInMBB &OutLocs, VarLocInMBB &InLocs,
> - const VarLocMap &VarLocIDs);
> + const VarLocMap &VarLocIDs,
> + SmallPtrSet<const MachineBasicBlock *, 16> &Visited);
>
> bool ExtendRanges(MachineFunction &MF);
>
> @@ -368,7 +369,8 @@ bool LiveDebugValues::transfer(MachineIn
> /// inserting a new DBG_VALUE instruction at the start of the @MBB - if the same
> /// source variable in all the predecessors of @MBB reside in the same location.
> bool LiveDebugValues::join(MachineBasicBlock &MBB, VarLocInMBB &OutLocs,
> - VarLocInMBB &InLocs, const VarLocMap &VarLocIDs) {
> + VarLocInMBB &InLocs, const VarLocMap &VarLocIDs,
> + SmallPtrSet<const MachineBasicBlock *, 16> &Visited) {
> DEBUG(dbgs() << "join MBB: " << MBB.getName() << "\n");
> bool Changed = false;
>
> @@ -376,21 +378,32 @@ bool LiveDebugValues::join(MachineBasicB
>
> // For all predecessors of this MBB, find the set of VarLocs that
> // can be joined.
> + int NumVisited = 0;
> for (auto p : MBB.predecessors()) {
> + // Ignore unvisited predecessor blocks. As we are processing
> + // the blocks in reverse post-order any unvisited block can
> + // be considered to not remove any incoming values.
> + if (!Visited.count(p))
> + continue;
> auto OL = OutLocs.find(p);
> // Join is null in case of empty OutLocs from any of the pred.
> if (OL == OutLocs.end())
> return false;
>
> - // Just copy over the Out locs to incoming locs for the first predecessor.
> - if (p == *MBB.pred_begin()) {
> + // Just copy over the Out locs to incoming locs for the first visited
> + // predecessor, and for all other predecessors join the Out locs.
> + if (!NumVisited)
> InLocsT = OL->second;
> - continue;
> - }
> - // Join with this predecessor.
> - InLocsT &= OL->second;
> + else
> + InLocsT &= OL->second;
> + NumVisited++;
> }
>
> + // As we are processing blocks in reverse post-order we
> + // should have processed at least one predecessor, unless it
> + // is the entry block which has no predecessor.
> + assert((NumVisited || MBB.pred_empty()) &&
> + "Should have processed at least one predecessor");
> if (InLocsT.empty())
> return false;
>
> @@ -463,6 +476,7 @@ bool LiveDebugValues::ExtendRanges(Machi
> // To solve it, we perform join() and transfer() using the two worklist method
> // until the ranges converge.
> // Ranges have converged when both worklists are empty.
> + SmallPtrSet<const MachineBasicBlock *, 16> Visited;
> while (!Worklist.empty() || !Pending.empty()) {
> // We track what is on the pending worklist to avoid inserting the same
> // thing twice. We could avoid this with a custom priority queue, but this
> @@ -472,8 +486,8 @@ bool LiveDebugValues::ExtendRanges(Machi
> while (!Worklist.empty()) {
> MachineBasicBlock *MBB = OrderToBB[Worklist.top()];
> Worklist.pop();
> - MBBJoined = join(*MBB, OutLocs, InLocs, VarLocIDs);
> -
> + MBBJoined = join(*MBB, OutLocs, InLocs, VarLocIDs, Visited);
> + Visited.insert(MBB);
> if (MBBJoined) {
> MBBJoined = false;
> Changed = true;
>
> Added: llvm/trunk/test/CodeGen/ARM/dbg-range-extension.mir
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/dbg-range-extension.mir?rev=282506&view=auto
> ==============================================================================
> --- llvm/trunk/test/CodeGen/ARM/dbg-range-extension.mir (added)
> +++ llvm/trunk/test/CodeGen/ARM/dbg-range-extension.mir Tue Sep 27 11:46:07 2016
> @@ -0,0 +1,282 @@
> +# RUN: llc -mtriple=arm-eabi -run-pass=livedebugvalues %s -o - | FileCheck %s
> +#
> +# Check that the debug information for variables are propagated into the correct blocks.
> +#
> +# Generated from the C source:
> +#
> +# int func2(int, int);
> +# void func(int a) {
> +# int b = func2(10, 11);
> +# if (a) {
> +# int c = func2(12, 13);
> +# for(int i = 1; i < a; i++) {
> +# func2(i, i+b);
> +# }
> +# func2(b,c);
> +# }
> +# func2(b,a);
> +# }
> +
> +# CHECK: [[VAR_A:![0-9]+]] = !DILocalVariable(name: "a",
> +# CHECK: [[VAR_B:![0-9]+]] = !DILocalVariable(name: "b",
> +# CHECK: [[VAR_C:![0-9]+]] = !DILocalVariable(name: "c",
> +# CHECK: [[VAR_I:![0-9]+]] = !DILocalVariable(name: "i",
> +
> +# CHECK: bb.0.entry
> +# CHECK: DBG_VALUE debug-use %r0, debug-use _, [[VAR_A]]
> +# CHECK: DBG_VALUE debug-use [[REG_A:%r[0-9]+]], debug-use _, [[VAR_A]]
> +# CHECK: DBG_VALUE debug-use [[REG_B:%r[0-9]+]], debug-use _, [[VAR_B]]
> +
> +# CHECK: bb.1.if.then
> +# CHECK: DBG_VALUE debug-use [[REG_B]], debug-use _, [[VAR_B]]
> +# CHECK: DBG_VALUE debug-use [[REG_A]], debug-use _, [[VAR_A]]
> +# CHECK: DBG_VALUE debug-use [[REG_C:%r[0-9]+]], debug-use _, [[VAR_C]]
> +# CHECK: DBG_VALUE 1, 0, [[VAR_I]]
> +
> +# CHECK: bb.2.for.body
> +# CHECK: DBG_VALUE debug-use [[REG_I:%r[0-9]+]], debug-use _, [[VAR_I]]
> +# CHECK: DBG_VALUE debug-use [[REG_C]], debug-use _, [[VAR_C]]
> +# CHECK: DBG_VALUE debug-use [[REG_B]], debug-use _, [[VAR_B]]
> +# CHECK: DBG_VALUE debug-use [[REG_A]], debug-use _, [[VAR_A]]
> +# CHECK: DBG_VALUE debug-use [[REG_I]], debug-use _, [[VAR_I]]
> +
> +# CHECK: bb.3.for.cond
> +# CHECK: DBG_VALUE debug-use [[REG_C]], debug-use _, [[VAR_C]]
> +# CHECK: DBG_VALUE debug-use [[REG_B]], debug-use _, [[VAR_B]]
> +# CHECK: DBG_VALUE debug-use [[REG_A]], debug-use _, [[VAR_A]]
> +# CHECK: DBG_VALUE debug-use [[REG_I]], debug-use _, [[VAR_I]]
> +
> +# CHECK: bb.4.for.cond.cleanup
> +# CHECK: DBG_VALUE debug-use [[REG_I]], debug-use _, [[VAR_I]]
> +# CHECK: DBG_VALUE debug-use [[REG_C]], debug-use _, [[VAR_C]]
> +# CHECK: DBG_VALUE debug-use [[REG_B]], debug-use _, [[VAR_B]]
> +# CHECK: DBG_VALUE debug-use [[REG_A]], debug-use _, [[VAR_A]]
> +
> +# CHECK: bb.5.if.end
> +# CHECK: DBG_VALUE debug-use [[REG_B]], debug-use _, [[VAR_B]]
> +# CHECK: DBG_VALUE debug-use [[REG_A]], debug-use _, [[VAR_A]]
> +--- |
> + ; ModuleID = '/data/kwalker/work/OpenSource-llvm/llvm/test/CodeGen/ARM/dbg-range-extension.ll'
> + source_filename = "/data/kwalker/work/OpenSource-llvm/llvm/test/CodeGen/ARM/dbg-range-extension.ll"
> + target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
> + target triple = "arm---eabi"
> +
> + ; Function Attrs: minsize nounwind optsize
> + define void @func(i32 %a) local_unnamed_addr #0 !dbg !8 {
> + entry:
> + tail call void @llvm.dbg.value(metadata i32 %a, i64 0, metadata !13, metadata !20), !dbg !21
> + %call = tail call i32 @func2(i32 10, i32 11) #0, !dbg !22
> + tail call void @llvm.dbg.value(metadata i32 %call, i64 0, metadata !14, metadata !20), !dbg !23
> + %tobool = icmp eq i32 %a, 0, !dbg !24
> + br i1 %tobool, label %if.end, label %if.then, !dbg !25
> +
> + if.then: ; preds = %entry
> + %call1 = tail call i32 @func2(i32 12, i32 13) #0, !dbg !26
> + tail call void @llvm.dbg.value(metadata i32 %call1, i64 0, metadata !15, metadata !20), !dbg !27
> + tail call void @llvm.dbg.value(metadata i32 1, i64 0, metadata !18, metadata !20), !dbg !28
> + br label %for.cond, !dbg !29
> +
> + for.cond: ; preds = %for.body, %if.then
> + %i.0 = phi i32 [ 1, %if.then ], [ %inc, %for.body ]
> + tail call void @llvm.dbg.value(metadata i32 %i.0, i64 0, metadata !18, metadata !20), !dbg !28
> + %cmp = icmp slt i32 %i.0, %a, !dbg !30
> + br i1 %cmp, label %for.body, label %for.cond.cleanup, !dbg !33
> +
> + for.cond.cleanup: ; preds = %for.cond
> + %call3 = tail call i32 @func2(i32 %call, i32 %call1) #0, !dbg !34
> + br label %if.end, !dbg !35
> +
> + for.body: ; preds = %for.cond
> + %0 = add i32 %call, %i.0, !dbg !36
> + %call2 = tail call i32 @func2(i32 %i.0, i32 %0) #0, !dbg !36
> + %inc = add nuw nsw i32 %i.0, 1, !dbg !38
> + tail call void @llvm.dbg.value(metadata i32 %inc, i64 0, metadata !18, metadata !20), !dbg !28
> + br label %for.cond, !dbg !40, !llvm.loop !41
> +
> + if.end: ; preds = %for.cond.cleanup, %entry
> + %call4 = tail call i32 @func2(i32 %call, i32 %a) #0, !dbg !43
> + ret void, !dbg !44
> + }
> +
> + ; Function Attrs: minsize optsize
> + declare i32 @func2(i32, i32) local_unnamed_addr #1
> +
> + ; Function Attrs: nounwind readnone
> + declare void @llvm.dbg.value(metadata, i64, metadata, metadata) #2
> +
> + ; Function Attrs: nounwind
> + declare void @llvm.stackprotector(i8*, i8**) #3
> +
> + attributes #0 = { minsize nounwind optsize }
> + attributes #1 = { minsize optsize }
> + attributes #2 = { nounwind readnone }
> + attributes #3 = { nounwind }
> +
> + !llvm.dbg.cu = !{!0}
> + !llvm.module.flags = !{!3, !4, !5, !6}
> + !llvm.ident = !{!7}
> +
> + !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2)
> + !1 = !DIFile(filename: "loop.c", directory: "/tmp")
> + !2 = !{}
> + !3 = !{i32 2, !"Dwarf Version", i32 4}
> + !4 = !{i32 2, !"Debug Info Version", i32 3}
> + !5 = !{i32 1, !"wchar_size", i32 4}
> + !6 = !{i32 1, !"min_enum_size", i32 4}
> + !7 = !{!"clang version 4.0.0 (http://llvm.org/git/clang.git b8f10df3679b36f51e1de7c4351b82d297825089) (http://llvm.org/git/llvm.git c2a5d16d1e3b8c49f5bbb1ff87a76ac4f88edb89)"}
> + !8 = distinct !DISubprogram(name: "func", scope: !1, file: !1, line: 2, type: !9, isLocal: false, isDefinition: true, scopeLine: 2, flags: DIFlagPrototyped, isOptimized: true, unit: !0, variables: !12)
> + !9 = !DISubroutineType(types: !10)
> + !10 = !{null, !11}
> + !11 = !DIBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
> + !12 = !{!13, !14, !15, !18}
> + !13 = !DILocalVariable(name: "a", arg: 1, scope: !8, file: !1, line: 2, type: !11)
> + !14 = !DILocalVariable(name: "b", scope: !8, file: !1, line: 3, type: !11)
> + !15 = !DILocalVariable(name: "c", scope: !16, file: !1, line: 5, type: !11)
> + !16 = distinct !DILexicalBlock(scope: !17, file: !1, line: 4, column: 9)
> + !17 = distinct !DILexicalBlock(scope: !8, file: !1, line: 4, column: 6)
> + !18 = !DILocalVariable(name: "i", scope: !19, file: !1, line: 6, type: !11)
> + !19 = distinct !DILexicalBlock(scope: !16, file: !1, line: 6, column: 3)
> + !20 = !DIExpression()
> + !21 = !DILocation(line: 2, column: 15, scope: !8)
> + !22 = !DILocation(line: 3, column: 17, scope: !8)
> + !23 = !DILocation(line: 3, column: 13, scope: !8)
> + !24 = !DILocation(line: 4, column: 6, scope: !17)
> + !25 = !DILocation(line: 4, column: 6, scope: !8)
> + !26 = !DILocation(line: 5, column: 11, scope: !16)
> + !27 = !DILocation(line: 5, column: 7, scope: !16)
> + !28 = !DILocation(line: 6, column: 11, scope: !19)
> + !29 = !DILocation(line: 6, column: 7, scope: !19)
> + !30 = !DILocation(line: 6, column: 20, scope: !31)
> + !31 = !DILexicalBlockFile(scope: !32, file: !1, discriminator: 1)
> + !32 = distinct !DILexicalBlock(scope: !19, file: !1, line: 6, column: 3)
> + !33 = !DILocation(line: 6, column: 3, scope: !31)
> + !34 = !DILocation(line: 9, column: 3, scope: !16)
> + !35 = !DILocation(line: 10, column: 2, scope: !16)
> + !36 = !DILocation(line: 7, column: 4, scope: !37)
> + !37 = distinct !DILexicalBlock(scope: !32, file: !1, line: 6, column: 30)
> + !38 = !DILocation(line: 6, column: 26, scope: !39)
> + !39 = !DILexicalBlockFile(scope: !32, file: !1, discriminator: 3)
> + !40 = !DILocation(line: 6, column: 3, scope: !39)
> + !41 = distinct !{!41, !42}
> + !42 = !DILocation(line: 6, column: 3, scope: !16)
> + !43 = !DILocation(line: 11, column: 2, scope: !8)
> + !44 = !DILocation(line: 12, column: 1, scope: !8)
> +
> +...
> +---
> +name: func
> +alignment: 2
> +exposesReturnsTwice: false
> +legalized: false
> +regBankSelected: false
> +selected: false
> +tracksRegLiveness: false
> +liveins:
> + - { reg: '%r0' }
> +calleeSavedRegisters: [ '%lr', '%d8', '%d9', '%d10', '%d11', '%d12', '%d13',
> + '%d14', '%d15', '%q4', '%q5', '%q6', '%q7', '%r4',
> + '%r5', '%r6', '%r7', '%r8', '%r9', '%r10', '%r11',
> + '%s16', '%s17', '%s18', '%s19', '%s20', '%s21',
> + '%s22', '%s23', '%s24', '%s25', '%s26', '%s27',
> + '%s28', '%s29', '%s30', '%s31', '%d8_d10', '%d9_d11',
> + '%d10_d12', '%d11_d13', '%d12_d14', '%d13_d15',
> + '%q4_q5', '%q5_q6', '%q6_q7', '%q4_q5_q6_q7', '%r4_r5',
> + '%r6_r7', '%r8_r9', '%r10_r11', '%d8_d9_d10', '%d9_d10_d11',
> + '%d10_d11_d12', '%d11_d12_d13', '%d12_d13_d14',
> + '%d13_d14_d15', '%d8_d10_d12', '%d9_d11_d13', '%d10_d12_d14',
> + '%d11_d13_d15', '%d8_d10_d12_d14', '%d9_d11_d13_d15',
> + '%d9_d10', '%d11_d12', '%d13_d14', '%d9_d10_d11_d12',
> + '%d11_d12_d13_d14' ]
> +frameInfo:
> + isFrameAddressTaken: false
> + isReturnAddressTaken: false
> + hasStackMap: false
> + hasPatchPoint: false
> + stackSize: 24
> + offsetAdjustment: 0
> + maxAlignment: 4
> + adjustsStack: true
> + hasCalls: true
> + maxCallFrameSize: 0
> + hasOpaqueSPAdjustment: false
> + hasVAStart: false
> + hasMustTailInVarArgFunc: false
> +stack:
> + - { id: 0, type: spill-slot, offset: -4, size: 4, alignment: 4, callee-saved-register: '%lr' }
> + - { id: 1, type: spill-slot, offset: -8, size: 4, alignment: 4, callee-saved-register: '%r11' }
> + - { id: 2, type: spill-slot, offset: -12, size: 4, alignment: 4, callee-saved-register: '%r7' }
> + - { id: 3, type: spill-slot, offset: -16, size: 4, alignment: 4, callee-saved-register: '%r6' }
> + - { id: 4, type: spill-slot, offset: -20, size: 4, alignment: 4, callee-saved-register: '%r5' }
> + - { id: 5, type: spill-slot, offset: -24, size: 4, alignment: 4, callee-saved-register: '%r4' }
> +body: |
> + bb.0.entry:
> + successors: %bb.5.if.end, %bb.1.if.then
> + liveins: %r0, %r4, %r5, %r6, %r7, %r11, %lr
> +
> + %sp = frame-setup STMDB_UPD %sp, 14, _, killed %r4, killed %r5, killed %r6, killed %r7, killed %r11, killed %lr
> + frame-setup CFI_INSTRUCTION def_cfa_offset 24
> + frame-setup CFI_INSTRUCTION offset %lr, -4
> + frame-setup CFI_INSTRUCTION offset %r11, -8
> + frame-setup CFI_INSTRUCTION offset %r7, -12
> + frame-setup CFI_INSTRUCTION offset %r6, -16
> + frame-setup CFI_INSTRUCTION offset %r5, -20
> + frame-setup CFI_INSTRUCTION offset %r4, -24
> + DBG_VALUE debug-use %r0, debug-use _, !13, !20, debug-location !21
> + %r4 = MOVr killed %r0, 14, _, _
> + DBG_VALUE debug-use %r4, debug-use _, !13, !20, debug-location !21
> + %r0 = MOVi 10, 14, _, _, debug-location !22
> + %r1 = MOVi 11, 14, _, _, debug-location !22
> + BL @func2, csr_aapcs, implicit-def dead %lr, implicit %sp, implicit killed %r0, implicit killed %r1, implicit-def %sp, implicit-def %r0, debug-location !22
> + %r5 = MOVr killed %r0, 14, _, _, debug-location !22
> + DBG_VALUE debug-use %r5, debug-use _, !14, !20, debug-location !23
> + CMPri %r4, 0, 14, _, implicit-def %cpsr, debug-location !25
> + Bcc %bb.5.if.end, 0, killed %cpsr
> +
> + bb.1.if.then:
> + successors: %bb.3.for.cond
> + liveins: %r4, %r5
> +
> + %r0 = MOVi 12, 14, _, _, debug-location !26
> + %r1 = MOVi 13, 14, _, _, debug-location !26
> + BL @func2, csr_aapcs, implicit-def dead %lr, implicit %sp, implicit killed %r0, implicit killed %r1, implicit-def %sp, implicit-def %r0, debug-location !26
> + %r6 = MOVr killed %r0, 14, _, _, debug-location !26
> + DBG_VALUE debug-use %r6, debug-use _, !15, !20, debug-location !27
> + %r7 = MOVi 1, 14, _, _
> + DBG_VALUE 1, 0, !18, !20, debug-location !28
> + B %bb.3.for.cond
> +
> + bb.2.for.body:
> + successors: %bb.3.for.cond
> + liveins: %r4, %r5, %r6, %r7
> +
> + %r1 = ADDrr %r5, %r7, 14, _, _, debug-location !36
> + %r0 = MOVr %r7, 14, _, _, debug-location !36
> + BL @func2, csr_aapcs, implicit-def dead %lr, implicit %sp, implicit killed %r0, implicit killed %r1, implicit-def %sp, implicit-def dead %r0, debug-location !36
> + %r7 = ADDri killed %r7, 1, 14, _, _, debug-location !38
> + DBG_VALUE debug-use %r7, debug-use _, !18, !20, debug-location !28
> +
> + bb.3.for.cond:
> + successors: %bb.2.for.body, %bb.4.for.cond.cleanup
> + liveins: %r4, %r5, %r6, %r7
> +
> + DBG_VALUE debug-use %r7, debug-use _, !18, !20, debug-location !28
> + CMPrr %r7, %r4, 14, _, implicit-def %cpsr, debug-location !33
> + Bcc %bb.2.for.body, 11, killed %cpsr, debug-location !33
> +
> + bb.4.for.cond.cleanup:
> + successors: %bb.5.if.end
> + liveins: %r4, %r5, %r6
> +
> + %r0 = MOVr %r5, 14, _, _, debug-location !34
> + %r1 = MOVr killed %r6, 14, _, _, debug-location !34
> + BL @func2, csr_aapcs, implicit-def dead %lr, implicit %sp, implicit killed %r0, implicit killed %r1, implicit-def %sp, implicit-def dead %r0, debug-location !34
> +
> + bb.5.if.end:
> + liveins: %r4, %r5
> +
> + %r0 = MOVr killed %r5, 14, _, _, debug-location !43
> + %r1 = MOVr killed %r4, 14, _, _, debug-location !43
> + %sp = LDMIA_UPD %sp, 14, _, def %r4, def %r5, def %r6, def %r7, def %r11, def %lr, debug-location !43
> + TAILJMPd @func2, implicit %sp, implicit %sp, implicit killed %r0, implicit killed %r1, debug-location !43
> +
> +...
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
More information about the llvm-commits
mailing list