RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 03:33:20 +01:00
Code Issues Projects Releases Wiki Activity
f625c157a1
llvm-mirror/test/DebugInfo/Generic/accel-table-hash-collisions.ll

106 lines
7.1 KiB
LLVM
Raw Normal View History

DwarfAccelTable: Fix handling of hash collisions. It turns out accelerator tables where totally broken if they contained entries with colliding hashes. The failure mode is pretty bad, as it not only impacted the colliding entries, but would basically make all the entries after the first hash collision pointing in the wrong place. The testcase uses the symbol names that where found to collide during a clang build. From a performance point of view, the patch adds a sort and a linear walk over each bucket contents. While it has a measurable impact on the accelerator table emission, it's not showing up significantly in clang profiles (and I'd argue that correctness is priceless :-)). llvm-svn: 231732
2015-03-10 01:46:31 +01:00
; REQUIRES: object-emission
; RUN: %llc_dwarf -dwarf-accel-tables=Enable -filetype=obj -o - < %s | llvm-dwarfdump -debug-dump=apple_names - | FileCheck %s
; Generated from the following C code using
; clang -S -emit-llvm hash-collision.c
;
; The names of the variables have been chosen so that they produce hash collisions.
; There are 12 names here that are hashed to only 6 hashes (each pair of lines
; hashes to the same value, see the CHECK lines below).
;
; int ForceTopDown;
; int _ZNSt3__116allocator_traitsINS_9allocatorINS_11__tree_nodeINS_12__value_typeIPN4llvm10BasicBlockEPNS4_10RegionNodeEEEPvEEEEE11__constructIS9_JNS_4pairIS6_S8_EEEEEvNS_17integral_constantIbLb1EEERSC_PT_DpOT0_;
; int _ZN5clang23DataRecursiveASTVisitorIN12_GLOBAL__N_124UnusedBackingIvarCheckerEE26TraverseCUDAKernelCallExprEPNS_18CUDAKernelCallExprE;
; int _ZN4llvm16DenseMapIteratorIPNS_10MDLocationENS_6detail13DenseSetEmptyENS_10MDNodeInfoIS1_EENS3_12DenseSetPairIS2_EELb0EE23AdvancePastEmptyBucketsEv;
; int _ZNK4llvm12LivePhysRegs5printERNS_11raw_ostreamE;
; int _ZN4llvm15ScalarEvolution14getSignedRangeEPKNS_4SCEVE;
; int k1;
; int is;
; int setStmt;
; int _ZN4llvm5TwineC1Ei;
; int _ZNK5clang12OverrideAttr5cloneERNS_10ASTContextE;
; int _ZN4llvm22MachineModuleInfoMachOD2Ev;
; Check that we have the right amount of hashes.
; CHECK: Bucket count = 6
; CHECK: Hashes count = 6
; Check that all the names are present in the output
; CHECK: Hash = 0x00597841
; CHECK: Name: {{[0-9a-f]*}} "is"
; CHECK: Name: {{[0-9a-f]*}} "k1"
; CHECK: Hash = 0xa4b42a1e
; CHECK: Name: {{[0-9a-f]*}} "_ZN5clang23DataRecursiveASTVisitorIN12_GLOBAL__N_124UnusedBackingIvarCheckerEE26TraverseCUDAKernelCallExprEPNS_18CUDAKernelCallExprE"
; CHECK: Name: {{[0-9a-f]*}} "_ZN4llvm16DenseMapIteratorIPNS_10MDLocationENS_6detail13DenseSetEmptyENS_10MDNodeInfoIS1_EENS3_12DenseSetPairIS2_EELb0EE23AdvancePastEmptyBucketsEv"
; CHECK: Hash = 0xeee7c0b2
; CHECK: Name: {{[0-9a-f]*}} "_ZNK4llvm12LivePhysRegs5printERNS_11raw_ostreamE"
; CHECK: Name: {{[0-9a-f]*}} "_ZN4llvm15ScalarEvolution14getSignedRangeEPKNS_4SCEVE"
; CHECK: Hash = 0xea48ac5f
; CHECK: Name: {{[0-9a-f]*}} "ForceTopDown"
; CHECK: Name: {{[0-9a-f]*}} "_ZNSt3__116allocator_traitsINS_9allocatorINS_11__tree_nodeINS_12__value_typeIPN4llvm10BasicBlockEPNS4_10RegionNodeEEEPvEEEEE11__constructIS9_JNS_4pairIS6_S8_EEEEEvNS_17integral_constantIbLb1EEERSC_PT_DpOT0_"
; CHECK: Hash = 0x6b22f71f
; CHECK: Name: {{[0-9a-f]*}} "_ZNK5clang12OverrideAttr5cloneERNS_10ASTContextE"
; CHECK: Name: {{[0-9a-f]*}} "_ZN4llvm22MachineModuleInfoMachOD2Ev"
; CHECK: Hash = 0x8c248979
; CHECK: Name: {{[0-9a-f]*}} "setStmt"
; CHECK: Name: {{[0-9a-f]*}} "_ZN4llvm5TwineC1Ei"
Renumber testcase metadata nodes after r290153. This patch renumbers the metadata nodes in debug info testcases after https://reviews.llvm.org/D26769. This is a separate patch because it causes so much churn. This was implemented with a python script that pipes the testcases through llvm-as - | llvm-dis - and then goes through the original and new output side-by side to insert all comments at a close-enough location. Differential Revision: https://reviews.llvm.org/D27765 llvm-svn: 290292
2016-12-22 01:45:21 +01:00
source_filename = "test/DebugInfo/Generic/accel-table-hash-collisions.ll"
DwarfAccelTable: Fix handling of hash collisions. It turns out accelerator tables where totally broken if they contained entries with colliding hashes. The failure mode is pretty bad, as it not only impacted the colliding entries, but would basically make all the entries after the first hash collision pointing in the wrong place. The testcase uses the symbol names that where found to collide during a clang build. From a performance point of view, the patch adds a sort and a linear walk over each bucket contents. While it has a measurable impact on the accelerator table emission, it's not showing up significantly in clang profiles (and I'd argue that correctness is priceless :-)). llvm-svn: 231732
2015-03-10 01:46:31 +01:00
Renumber testcase metadata nodes after r290153. This patch renumbers the metadata nodes in debug info testcases after https://reviews.llvm.org/D26769. This is a separate patch because it causes so much churn. This was implemented with a python script that pipes the testcases through llvm-as - | llvm-dis - and then goes through the original and new output side-by side to insert all comments at a close-enough location. Differential Revision: https://reviews.llvm.org/D27765 llvm-svn: 290292
2016-12-22 01:45:21 +01:00
@ForceTopDown = common global i32 0, align 4, !dbg !0
DebugInfo: New metadata representation for global variables. This patch reverses the edge from DIGlobalVariable to GlobalVariable. This will allow us to more easily preserve debug info metadata when manipulating global variables. Fixes PR30362. A program for upgrading test cases is attached to that bug. Differential Revision: http://reviews.llvm.org/D20147 llvm-svn: 281284
2016-09-13 03:12:59 +02:00
@_ZNSt3__116allocator_traitsINS_9allocatorINS_11__tree_nodeINS_12__value_typeIPN4llvm10BasicBlockEPNS4_10RegionNodeEEEPvEEEEE11__constructIS9_JNS_4pairIS6_S8_EEEEEvNS_17integral_constantIbLb1EEERSC_PT_DpOT0_ = common global i32 0, align 4, !dbg !6
Renumber testcase metadata nodes after r290153. This patch renumbers the metadata nodes in debug info testcases after https://reviews.llvm.org/D26769. This is a separate patch because it causes so much churn. This was implemented with a python script that pipes the testcases through llvm-as - | llvm-dis - and then goes through the original and new output side-by side to insert all comments at a close-enough location. Differential Revision: https://reviews.llvm.org/D27765 llvm-svn: 290292
2016-12-22 01:45:21 +01:00
@_ZN5clang23DataRecursiveASTVisitorIN12_GLOBAL__N_124UnusedBackingIvarCheckerEE26TraverseCUDAKernelCallExprEPNS_18CUDAKernelCallExprE = common global i32 0, align 4, !dbg !9
@_ZN4llvm16DenseMapIteratorIPNS_10MDLocationENS_6detail13DenseSetEmptyENS_10MDNodeInfoIS1_EENS3_12DenseSetPairIS2_EELb0EE23AdvancePastEmptyBucketsEv = common global i32 0, align 4, !dbg !11
@_ZNK4llvm12LivePhysRegs5printERNS_11raw_ostreamE = common global i32 0, align 4, !dbg !13
@_ZN4llvm15ScalarEvolution14getSignedRangeEPKNS_4SCEVE = common global i32 0, align 4, !dbg !15
@k1 = common global i32 0, align 4, !dbg !17
@is = common global i32 0, align 4, !dbg !19
@setStmt = common global i32 0, align 4, !dbg !21
@_ZN4llvm5TwineC1Ei = common global i32 0, align 4, !dbg !23
@_ZNK5clang12OverrideAttr5cloneERNS_10ASTContextE = common global i32 0, align 4, !dbg !25
@_ZN4llvm22MachineModuleInfoMachOD2Ev = common global i32 0, align 4, !dbg !27
!llvm.dbg.cu = !{!2}
!llvm.module.flags = !{!29, !30, !31}
!llvm.ident = !{!32}
DwarfAccelTable: Fix handling of hash collisions. It turns out accelerator tables where totally broken if they contained entries with colliding hashes. The failure mode is pretty bad, as it not only impacted the colliding entries, but would basically make all the entries after the first hash collision pointing in the wrong place. The testcase uses the symbol names that where found to collide during a clang build. From a performance point of view, the patch adds a sort and a linear walk over each bucket contents. While it has a measurable impact on the accelerator table emission, it's not showing up significantly in clang profiles (and I'd argue that correctness is priceless :-)). llvm-svn: 231732
2015-03-10 01:46:31 +01:00
Renumber testcase metadata nodes after r290153. This patch renumbers the metadata nodes in debug info testcases after https://reviews.llvm.org/D26769. This is a separate patch because it causes so much churn. This was implemented with a python script that pipes the testcases through llvm-as - | llvm-dis - and then goes through the original and new output side-by side to insert all comments at a close-enough location. Differential Revision: https://reviews.llvm.org/D27765 llvm-svn: 290292
2016-12-22 01:45:21 +01:00
!0 = !DIGlobalVariableExpression(var: !1)
!1 = !DIGlobalVariable(name: "ForceTopDown", scope: !2, file: !3, line: 1, type: !8, isLocal: false, isDefinition: true)
!2 = distinct !DICompileUnit(language: DW_LANG_C99, file: !3, producer: "clang version 3.7.0 (trunk 231548) (llvm/trunk 231547)", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !4, retainedTypes: !4, globals: !5, imports: !4)
!3 = !DIFile(filename: "hash-collisions.c", directory: "/tmp")
!4 = !{}
!5 = !{!0, !6, !9, !11, !13, !15, !17, !19, !21, !23, !25, !27}
!6 = !DIGlobalVariableExpression(var: !7)
!7 = !DIGlobalVariable(name: "_ZNSt3__116allocator_traitsINS_9allocatorINS_11__tree_nodeINS_12__value_typeIPN4llvm10BasicBlockEPNS4_10RegionNodeEEEPvEEEEE11__constructIS9_JNS_4pairIS6_S8_EEEEEvNS_17integral_constantIbLb1EEERSC_PT_DpOT0_", scope: !2, file: !3, line: 2, type: !8, isLocal: false, isDefinition: true)
!8 = !DIBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
!9 = !DIGlobalVariableExpression(var: !10)
!10 = !DIGlobalVariable(name: "_ZN5clang23DataRecursiveASTVisitorIN12_GLOBAL__N_124UnusedBackingIvarCheckerEE26TraverseCUDAKernelCallExprEPNS_18CUDAKernelCallExprE", scope: !2, file: !3, line: 3, type: !8, isLocal: false, isDefinition: true)
!11 = !DIGlobalVariableExpression(var: !12)
!12 = !DIGlobalVariable(name: "_ZN4llvm16DenseMapIteratorIPNS_10MDLocationENS_6detail13DenseSetEmptyENS_10MDNodeInfoIS1_EENS3_12DenseSetPairIS2_EELb0EE23AdvancePastEmptyBucketsEv", scope: !2, file: !3, line: 4, type: !8, isLocal: false, isDefinition: true)
!13 = !DIGlobalVariableExpression(var: !14)
!14 = !DIGlobalVariable(name: "_ZNK4llvm12LivePhysRegs5printERNS_11raw_ostreamE", scope: !2, file: !3, line: 5, type: !8, isLocal: false, isDefinition: true)
!15 = !DIGlobalVariableExpression(var: !16)
!16 = !DIGlobalVariable(name: "_ZN4llvm15ScalarEvolution14getSignedRangeEPKNS_4SCEVE", scope: !2, file: !3, line: 6, type: !8, isLocal: false, isDefinition: true)
!17 = !DIGlobalVariableExpression(var: !18)
!18 = !DIGlobalVariable(name: "k1", scope: !2, file: !3, line: 7, type: !8, isLocal: false, isDefinition: true)
!19 = !DIGlobalVariableExpression(var: !20)
!20 = !DIGlobalVariable(name: "is", scope: !2, file: !3, line: 8, type: !8, isLocal: false, isDefinition: true)
!21 = !DIGlobalVariableExpression(var: !22)
!22 = !DIGlobalVariable(name: "setStmt", scope: !2, file: !3, line: 9, type: !8, isLocal: false, isDefinition: true)
!23 = !DIGlobalVariableExpression(var: !24)
!24 = !DIGlobalVariable(name: "_ZN4llvm5TwineC1Ei", scope: !2, file: !3, line: 10, type: !8, isLocal: false, isDefinition: true)
!25 = !DIGlobalVariableExpression(var: !26)
!26 = !DIGlobalVariable(name: "_ZNK5clang12OverrideAttr5cloneERNS_10ASTContextE", scope: !2, file: !3, line: 11, type: !8, isLocal: false, isDefinition: true)
!27 = !DIGlobalVariableExpression(var: !28)
!28 = !DIGlobalVariable(name: "_ZN4llvm22MachineModuleInfoMachOD2Ev", scope: !2, file: !3, line: 12, type: !8, isLocal: false, isDefinition: true)
!29 = !{i32 2, !"Dwarf Version", i32 2}
!30 = !{i32 2, !"Debug Info Version", i32 3}
!31 = !{i32 1, !"PIC Level", i32 2}
!32 = !{!"clang version 3.7.0 (trunk 231548) (llvm/trunk 231547)"}
DwarfAccelTable: Fix handling of hash collisions. It turns out accelerator tables where totally broken if they contained entries with colliding hashes. The failure mode is pretty bad, as it not only impacted the colliding entries, but would basically make all the entries after the first hash collision pointing in the wrong place. The testcase uses the symbol names that where found to collide during a clang build. From a performance point of view, the patch adds a sort and a linear walk over each bucket contents. While it has a measurable impact on the accelerator table emission, it's not showing up significantly in clang profiles (and I'd argue that correctness is priceless :-)). llvm-svn: 231732
2015-03-10 01:46:31 +01:00
Reference in New Issue Copy Permalink