[llvm] r235140 - DebugInfo: Fix UserValue::match() in LiveDebugVariables after r235050
Duncan P. N. Exon Smith
dexonsmith at apple.com
Thu Apr 16 17:23:19 PDT 2015
> On 2015-Apr-16, at 17:07, David Blaikie <dblaikie at gmail.com> wrote:
>
> On Thu, Apr 16, 2015 at 4:55 PM, Duncan P. N. Exon Smith
> <dexonsmith at apple.com> wrote:
>>
>>> On 2015-Apr-16, at 15:44, David Blaikie <dblaikie at gmail.com> wrote:
>>>
>>>
>>>
>>> On Thu, Apr 16, 2015 at 3:27 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
>>> Author: dexonsmith
>>> Date: Thu Apr 16 17:27:54 2015
>>> New Revision: 235140
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=235140&view=rev
>>> Log:
>>> DebugInfo: Fix UserValue::match() in LiveDebugVariables after r235050
>>>
>>> r235050 dropped the inlined-at field from `MDLocalVariable`, deferring
>>> to the `!dbg` attachments. Fix `UserValue` to take the `!dbg` into
>>> account when differentiating between variables.
>>>
>>> Added:
>>> llvm/trunk/test/DebugInfo/X86/inlined-formal-parameter.ll
>>> Modified:
>>> llvm/trunk/lib/CodeGen/LiveDebugVariables.cpp
>>>
>>> Modified: llvm/trunk/lib/CodeGen/LiveDebugVariables.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LiveDebugVariables.cpp?rev=235140&r1=235139&r2=235140&view=diff
>>> ==============================================================================
>>> --- llvm/trunk/lib/CodeGen/LiveDebugVariables.cpp (original)
>>> +++ llvm/trunk/lib/CodeGen/LiveDebugVariables.cpp Thu Apr 16 17:27:54 2015
>>> @@ -158,10 +158,10 @@ public:
>>> UserValue *getNext() const { return next; }
>>>
>>> /// match - Does this UserValue match the parameters?
>>> - bool match(const MDNode *Var, const MDNode *Expr, unsigned Offset,
>>> - bool indirect) const {
>>> - return Var == Variable && Expr == Expression && Offset == offset &&
>>> - indirect == IsIndirect;
>>> + bool match(const MDNode *Var, const MDNode *Expr, const MDLocation *IA,
>>> + unsigned Offset, bool indirect) const {
>>> + return Var == Variable && Expr == Expression && dl->getInlinedAt() == IA &&
>>> + Offset == offset && indirect == IsIndirect;
>>> }
>>>
>>> /// merge - Merge equivalence classes.
>>> @@ -464,7 +464,7 @@ UserValue *LDVImpl::getUserValue(const M
>>> UserValue *UV = Leader->getLeader();
>>> Leader = UV;
>>> for (; UV; UV = UV->getNext())
>>> - if (UV->match(Var, Expr, Offset, IsIndirect))
>>> + if (UV->match(Var, Expr, DL->getInlinedAt(), Offset, IsIndirect))
>>> return UV;
>>> }
>>>
>>>
>>> Added: llvm/trunk/test/DebugInfo/X86/inlined-formal-parameter.ll
>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/DebugInfo/X86/inlined-formal-parameter.ll?rev=235140&view=auto
>>> ==============================================================================
>>> --- llvm/trunk/test/DebugInfo/X86/inlined-formal-parameter.ll (added)
>>> +++ llvm/trunk/test/DebugInfo/X86/inlined-formal-parameter.ll Thu Apr 16 17:27:54 2015
>>> @@ -0,0 +1,105 @@
>>> +; RUN: llc -filetype=obj -o %t.o %s
>>> +; RUN: llvm-dwarfdump -debug-dump=info %t.o | FileCheck %s
>>> +
>>> +; Testcase generated using 'clang -O2 -S -emit-llvm' from the following:
>>>
>>> Reckon it might be worth making sure both the function calls don't get optimized away? (maybe you can replace g and l with a call to some external function - then they can't get collapsed together or otherwise optimized away - but I'm not sure if that'll preserve the bug?) Just so this test case doesn't get confused with other behavior?
>>>
>>> Yep, this is a bit simpler & doesn't tickle the optimized-away case:
>>>
>>> void sink();
>>>
>>> static __attribute__((always_inline)) void f(int a) {
>>> sink();
>>> }
>>>
>>> void call() {
>>> f(0);
>>> f(0);
>>> }
>>
>> Happy to add this testcase, or do you think it should replace the one
>> I have already?
>
> I was meaning to suggest replacing (well, simplifying a bit) the one
> already committed.
>
>> Any value in also having the case where one of the
>> inlined subroutines is optimized away?
>
> I don't think so - it doesn't seem relevant to the bug & I think it
> clouds the purpose of the test/bug/issue at hand. The version I've
> suggested is consistent with, at least, my own habits of using a call
> to an external funciton as the implementation of an inline function to
> ensure no interesting optimizations occur on the resulting inlined
> instructions - it's a trivial/robust case of inlining.
>
> (maybe rename 'sink'/f to just f1/f2, possibly rename 'call' to
> 'calls' in my example - dunno, whatever names seem simple enough &
> don't have weird/extra/unknown connotations)
>
>
Makes sense to me; thanks for reviewing. r235149.
I left the names mostly as is; I think "sink" has some descriptive value.
>>
>>>
>>> +;; int *g;
>>> +;;
>>> +;; static __attribute__((always_inline)) int f(int a) {
>>> +;; int l;
>>> +;; g = &l;
>>> +;; return a;
>>> +;; }
>>> +;;
>>> +;; int main(void) {
>>> +;; f(0);
>>> +;; f(0);
>>> +;; return 0;
>>> +;; }
>>> +
>>> +; Check that we the first call to f(0) has no inlined subroutine (since the
>>> +; function is optimized out), and the second call correctly describes the
>>> +; formal parameter 'a'.
>>> +
>>> +; CHECK: DW_TAG_inlined_subroutine
>>> +; CHECK-NOT: DW_TAG_inlined_subroutine
>>> +; CHECK: DW_AT_low_pc
>>> +; CHECK-NEXT: DW_AT_high_pc
>>> +; CHECK-NOT: DW_TAG_inlined_subroutine
>>> +; CHECK: DW_TAG_formal_parameter
>>> +; CHECK-NEXT: DW_AT_const_value
>>> +; CHECK-NEXT: DW_AT_abstract_origin {{.*}} "a"
>>> +; CHECK-NOT: DW_TAG_inlined_subroutine
>>> +; CHECK: DW_TAG_variable
>>> +; CHECK-NEXT: DW_AT_location
>>> +; CHECK-NEXT: DW_AT_abstract_origin {{.*}} "l"
>>> +
>>> +target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
>>> +target triple = "x86_64-apple-darwin"
>>> +
>>> + at g = common global i32* null, align 8
>>> +
>>> +; Function Attrs: nounwind ssp uwtable
>>> +define i32 @main() #0 {
>>> +entry:
>>> + %l.i2 = alloca i32, align 4
>>> + tail call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !12, metadata !21), !dbg !22
>>> + %0 = bitcast i32* %l.i2 to i8*, !dbg !24
>>> + call void @llvm.lifetime.start(i64 4, i8* %0), !dbg !24
>>> + tail call void @llvm.dbg.value(metadata i32 0, i64 0, metadata !12, metadata !21), !dbg !24
>>> + tail call void @llvm.dbg.value(metadata i32* %l.i2, i64 0, metadata !13, metadata !21), !dbg !26
>>> + store i32* %l.i2, i32** @g, align 8, !dbg !27, !tbaa !28
>>> + call void @llvm.lifetime.end(i64 4, i8* %0), !dbg !32
>>> + ret i32 0, !dbg !33
>>> +}
>>> +
>>> +; Function Attrs: nounwind readnone
>>> +declare void @llvm.dbg.value(metadata, i64, metadata, metadata) #1
>>> +
>>> +; Function Attrs: nounwind
>>> +declare void @llvm.lifetime.start(i64, i8* nocapture) #2
>>> +
>>> +; Function Attrs: nounwind
>>> +declare void @llvm.lifetime.end(i64, i8* nocapture) #2
>>> +
>>> +attributes #0 = { nounwind ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="core2" "unsafe-fp-math"="false" "use-soft-float"="false" }
>>> +attributes #1 = { nounwind readnone }
>>> +attributes #2 = { nounwind }
>>> +
>>> +!llvm.dbg.cu = !{!0}
>>> +!llvm.module.flags = !{!17, !18, !19}
>>> +!llvm.ident = !{!20}
>>> +
>>> +!0 = !MDCompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 3.7.0 (trunk 235110) (llvm/trunk 235108)", isOptimized: true, runtimeVersion: 0, emissionKind: 1, enums: !2, retainedTypes: !2, subprograms: !3, globals: !14, imports: !2)
>>> +!1 = !MDFile(filename: "t.c", directory: "/path/to/dir")
>>> +!2 = !{}
>>> +!3 = !{!4, !8}
>>> +!4 = !MDSubprogram(name: "main", scope: !1, file: !1, line: 9, type: !5, isLocal: false, isDefinition: true, scopeLine: 9, flags: DIFlagPrototyped, isOptimized: true, function: i32 ()* @main, variables: !2)
>>> +!5 = !MDSubroutineType(types: !6)
>>> +!6 = !{!7}
>>> +!7 = !MDBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
>>> +!8 = !MDSubprogram(name: "f", scope: !1, file: !1, line: 3, type: !9, isLocal: true, isDefinition: true, scopeLine: 3, flags: DIFlagPrototyped, isOptimized: true, variables: !11)
>>> +!9 = !MDSubroutineType(types: !10)
>>> +!10 = !{!7, !7}
>>> +!11 = !{!12, !13}
>>> +!12 = !MDLocalVariable(tag: DW_TAG_arg_variable, name: "a", arg: 1, scope: !8, file: !1, line: 3, type: !7)
>>> +!13 = !MDLocalVariable(tag: DW_TAG_auto_variable, name: "l", scope: !8, file: !1, line: 4, type: !7)
>>> +!14 = !{!15}
>>> +!15 = !MDGlobalVariable(name: "g", scope: !0, file: !1, line: 1, type: !16, isLocal: false, isDefinition: true, variable: i32** @g)
>>> +!16 = !MDDerivedType(tag: DW_TAG_pointer_type, baseType: !7, size: 64, align: 64)
>>> +!17 = !{i32 2, !"Dwarf Version", i32 2}
>>> +!18 = !{i32 2, !"Debug Info Version", i32 3}
>>> +!19 = !{i32 1, !"PIC Level", i32 2}
>>> +!20 = !{!"clang version 3.7.0 (trunk 235110) (llvm/trunk 235108)"}
>>> +!21 = !MDExpression()
>>> +!22 = !MDLocation(line: 3, column: 49, scope: !8, inlinedAt: !23)
>>> +!23 = distinct !MDLocation(line: 10, column: 3, scope: !4)
>>> +!24 = !MDLocation(line: 3, column: 49, scope: !8, inlinedAt: !25)
>>> +!25 = distinct !MDLocation(line: 11, column: 3, scope: !4)
>>> +!26 = !MDLocation(line: 4, column: 7, scope: !8, inlinedAt: !25)
>>> +!27 = !MDLocation(line: 5, column: 5, scope: !8, inlinedAt: !25)
>>> +!28 = !{!29, !29, i64 0}
>>> +!29 = !{!"any pointer", !30, i64 0}
>>> +!30 = !{!"omnipotent char", !31, i64 0}
>>> +!31 = !{!"Simple C/C++ TBAA"}
>>> +!32 = !MDLocation(line: 11, column: 3, scope: !4)
>>> +!33 = !MDLocation(line: 12, column: 3, scope: !4)
>>>
>>>
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>
More information about the llvm-commits
mailing list