maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/Transforms/DeadStoreElimination/inst-limits.ll

262 lines
8.4 KiB
LLVM
Raw Normal View History

Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g. Adding a test case to make sure that the threshold used in the memory dependence analysis is respected. The test case also checks that debug intrinsics are not counted towards this threshold. Differential Revision: http://llvm-reviews.chandlerc.com/D2141 llvm-svn: 194646
2013-11-14 09:10:52 +08:00
; RUN: opt -S -dse < %s | FileCheck %s
; If there are two stores to the same location, DSE should be able to remove
; the first store if the two stores are separated by no more than 98
; instructions. The existence of debug intrinsics between the stores should
; not affect this instruction limit.
@x = global i32 0, align 4
; Function Attrs: nounwind
define i32 @test_within_limit() {
entry:
; The first store; later there is a second store to the same location,
; so this store should be optimized away by DSE.
; CHECK-NOT: store i32 1, i32* @x, align 4
store i32 1, i32* @x, align 4
; Insert 98 dummy instructions between the two stores
%0 = bitcast i32 0 to i32
%1 = bitcast i32 0 to i32
%2 = bitcast i32 0 to i32
%3 = bitcast i32 0 to i32
%4 = bitcast i32 0 to i32
%5 = bitcast i32 0 to i32
%6 = bitcast i32 0 to i32
%7 = bitcast i32 0 to i32
%8 = bitcast i32 0 to i32
%9 = bitcast i32 0 to i32
%10 = bitcast i32 0 to i32
%11 = bitcast i32 0 to i32
%12 = bitcast i32 0 to i32
%13 = bitcast i32 0 to i32
%14 = bitcast i32 0 to i32
%15 = bitcast i32 0 to i32
%16 = bitcast i32 0 to i32
%17 = bitcast i32 0 to i32
%18 = bitcast i32 0 to i32
%19 = bitcast i32 0 to i32
%20 = bitcast i32 0 to i32
%21 = bitcast i32 0 to i32
%22 = bitcast i32 0 to i32
%23 = bitcast i32 0 to i32
%24 = bitcast i32 0 to i32
%25 = bitcast i32 0 to i32
%26 = bitcast i32 0 to i32
%27 = bitcast i32 0 to i32
%28 = bitcast i32 0 to i32
%29 = bitcast i32 0 to i32
%30 = bitcast i32 0 to i32
%31 = bitcast i32 0 to i32
%32 = bitcast i32 0 to i32
%33 = bitcast i32 0 to i32
%34 = bitcast i32 0 to i32
%35 = bitcast i32 0 to i32
%36 = bitcast i32 0 to i32
%37 = bitcast i32 0 to i32
%38 = bitcast i32 0 to i32
%39 = bitcast i32 0 to i32
%40 = bitcast i32 0 to i32
%41 = bitcast i32 0 to i32
%42 = bitcast i32 0 to i32
%43 = bitcast i32 0 to i32
%44 = bitcast i32 0 to i32
%45 = bitcast i32 0 to i32
%46 = bitcast i32 0 to i32
%47 = bitcast i32 0 to i32
%48 = bitcast i32 0 to i32
%49 = bitcast i32 0 to i32
%50 = bitcast i32 0 to i32
%51 = bitcast i32 0 to i32
%52 = bitcast i32 0 to i32
%53 = bitcast i32 0 to i32
%54 = bitcast i32 0 to i32
%55 = bitcast i32 0 to i32
%56 = bitcast i32 0 to i32
%57 = bitcast i32 0 to i32
%58 = bitcast i32 0 to i32
%59 = bitcast i32 0 to i32
%60 = bitcast i32 0 to i32
%61 = bitcast i32 0 to i32
%62 = bitcast i32 0 to i32
%63 = bitcast i32 0 to i32
%64 = bitcast i32 0 to i32
%65 = bitcast i32 0 to i32
%66 = bitcast i32 0 to i32
%67 = bitcast i32 0 to i32
%68 = bitcast i32 0 to i32
%69 = bitcast i32 0 to i32
%70 = bitcast i32 0 to i32
%71 = bitcast i32 0 to i32
%72 = bitcast i32 0 to i32
%73 = bitcast i32 0 to i32
%74 = bitcast i32 0 to i32
%75 = bitcast i32 0 to i32
%76 = bitcast i32 0 to i32
%77 = bitcast i32 0 to i32
%78 = bitcast i32 0 to i32
%79 = bitcast i32 0 to i32
%80 = bitcast i32 0 to i32
%81 = bitcast i32 0 to i32
%82 = bitcast i32 0 to i32
%83 = bitcast i32 0 to i32
%84 = bitcast i32 0 to i32
%85 = bitcast i32 0 to i32
%86 = bitcast i32 0 to i32
%87 = bitcast i32 0 to i32
%88 = bitcast i32 0 to i32
%89 = bitcast i32 0 to i32
%90 = bitcast i32 0 to i32
%91 = bitcast i32 0 to i32
%92 = bitcast i32 0 to i32
%93 = bitcast i32 0 to i32
%94 = bitcast i32 0 to i32
%95 = bitcast i32 0 to i32
%96 = bitcast i32 0 to i32
%97 = bitcast i32 0 to i32
; Insert a meaningless dbg.value intrinsic; it should have no
; effect on the working of DSE in any way.
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. llvm-svn: 218787
2014-10-02 02:55:02 +08:00
call void @llvm.dbg.value(metadata !12, i64 0, metadata !10, metadata !{})
Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g. Adding a test case to make sure that the threshold used in the memory dependence analysis is respected. The test case also checks that debug intrinsics are not counted towards this threshold. Differential Revision: http://llvm-reviews.chandlerc.com/D2141 llvm-svn: 194646
2013-11-14 09:10:52 +08:00
; CHECK: store i32 -1, i32* @x, align 4
store i32 -1, i32* @x, align 4
ret i32 0
}
; Function Attrs: nounwind
define i32 @test_outside_limit() {
entry:
; The first store; later there is a second store to the same location
; CHECK: store i32 1, i32* @x, align 4
store i32 1, i32* @x, align 4
; Insert 99 dummy instructions between the two stores; this is
; one too many instruction for the DSE to take place.
%0 = bitcast i32 0 to i32
%1 = bitcast i32 0 to i32
%2 = bitcast i32 0 to i32
%3 = bitcast i32 0 to i32
%4 = bitcast i32 0 to i32
%5 = bitcast i32 0 to i32
%6 = bitcast i32 0 to i32
%7 = bitcast i32 0 to i32
%8 = bitcast i32 0 to i32
%9 = bitcast i32 0 to i32
%10 = bitcast i32 0 to i32
%11 = bitcast i32 0 to i32
%12 = bitcast i32 0 to i32
%13 = bitcast i32 0 to i32
%14 = bitcast i32 0 to i32
%15 = bitcast i32 0 to i32
%16 = bitcast i32 0 to i32
%17 = bitcast i32 0 to i32
%18 = bitcast i32 0 to i32
%19 = bitcast i32 0 to i32
%20 = bitcast i32 0 to i32
%21 = bitcast i32 0 to i32
%22 = bitcast i32 0 to i32
%23 = bitcast i32 0 to i32
%24 = bitcast i32 0 to i32
%25 = bitcast i32 0 to i32
%26 = bitcast i32 0 to i32
%27 = bitcast i32 0 to i32
%28 = bitcast i32 0 to i32
%29 = bitcast i32 0 to i32
%30 = bitcast i32 0 to i32
%31 = bitcast i32 0 to i32
%32 = bitcast i32 0 to i32
%33 = bitcast i32 0 to i32
%34 = bitcast i32 0 to i32
%35 = bitcast i32 0 to i32
%36 = bitcast i32 0 to i32
%37 = bitcast i32 0 to i32
%38 = bitcast i32 0 to i32
%39 = bitcast i32 0 to i32
%40 = bitcast i32 0 to i32
%41 = bitcast i32 0 to i32
%42 = bitcast i32 0 to i32
%43 = bitcast i32 0 to i32
%44 = bitcast i32 0 to i32
%45 = bitcast i32 0 to i32
%46 = bitcast i32 0 to i32
%47 = bitcast i32 0 to i32
%48 = bitcast i32 0 to i32
%49 = bitcast i32 0 to i32
%50 = bitcast i32 0 to i32
%51 = bitcast i32 0 to i32
%52 = bitcast i32 0 to i32
%53 = bitcast i32 0 to i32
%54 = bitcast i32 0 to i32
%55 = bitcast i32 0 to i32
%56 = bitcast i32 0 to i32
%57 = bitcast i32 0 to i32
%58 = bitcast i32 0 to i32
%59 = bitcast i32 0 to i32
%60 = bitcast i32 0 to i32
%61 = bitcast i32 0 to i32
%62 = bitcast i32 0 to i32
%63 = bitcast i32 0 to i32
%64 = bitcast i32 0 to i32
%65 = bitcast i32 0 to i32
%66 = bitcast i32 0 to i32
%67 = bitcast i32 0 to i32
%68 = bitcast i32 0 to i32
%69 = bitcast i32 0 to i32
%70 = bitcast i32 0 to i32
%71 = bitcast i32 0 to i32
%72 = bitcast i32 0 to i32
%73 = bitcast i32 0 to i32
%74 = bitcast i32 0 to i32
%75 = bitcast i32 0 to i32
%76 = bitcast i32 0 to i32
%77 = bitcast i32 0 to i32
%78 = bitcast i32 0 to i32
%79 = bitcast i32 0 to i32
%80 = bitcast i32 0 to i32
%81 = bitcast i32 0 to i32
%82 = bitcast i32 0 to i32
%83 = bitcast i32 0 to i32
%84 = bitcast i32 0 to i32
%85 = bitcast i32 0 to i32
%86 = bitcast i32 0 to i32
%87 = bitcast i32 0 to i32
%88 = bitcast i32 0 to i32
%89 = bitcast i32 0 to i32
%90 = bitcast i32 0 to i32
%91 = bitcast i32 0 to i32
%92 = bitcast i32 0 to i32
%93 = bitcast i32 0 to i32
%94 = bitcast i32 0 to i32
%95 = bitcast i32 0 to i32
%96 = bitcast i32 0 to i32
%97 = bitcast i32 0 to i32
%98 = bitcast i32 0 to i32
; CHECK: store i32 -1, i32* @x, align 4
store i32 -1, i32* @x, align 4
ret i32 0
}
; Function Attrs: nounwind readnone
Move the complex address expression out of DIVariable and into an extra argument of the llvm.dbg.declare/llvm.dbg.value intrinsics. Previously, DIVariable was a variable-length field that has an optional reference to a Metadata array consisting of a variable number of complex address expressions. In the case of OpPiece expressions this is wasting a lot of storage in IR, because when an aggregate type is, e.g., SROA'd into all of its n individual members, the IR will contain n copies of the DIVariable, all alike, only differing in the complex address reference at the end. By making the complex address into an extra argument of the dbg.value/dbg.declare intrinsics, all of the pieces can reference the same variable and the complex address expressions can be uniqued across the CU, too. Down the road, this will allow us to move other flags, such as "indirection" out of the DIVariable, too. The new intrinsics look like this: declare void @llvm.dbg.declare(metadata %storage, metadata %var, metadata %expr) declare void @llvm.dbg.value(metadata %storage, i64 %offset, metadata %var, metadata %expr) This patch adds a new LLVM-local tag to DIExpressions, so we can detect and pretty-print DIExpression metadata nodes. What this patch doesn't do: This patch does not touch the "Indirect" field in DIVariable; but moving that into the expression would be a natural next step. http://reviews.llvm.org/D4919 rdar://problem/17994491 Thanks to dblaikie and dexonsmith for reviewing this patch! Note: I accidentally committed a bogus older version of this patch previously. llvm-svn: 218787
2014-10-02 02:55:02 +08:00
declare void @llvm.dbg.value(metadata, i64, metadata, metadata)
Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g. Adding a test case to make sure that the threshold used in the memory dependence analysis is respected. The test case also checks that debug intrinsics are not counted towards this threshold. Differential Revision: http://llvm-reviews.chandlerc.com/D2141 llvm-svn: 194646
2013-11-14 09:10:52 +08:00
!llvm.dbg.cu = !{!0}
Debug Info: update testing cases to specify the debug info version number. We are going to drop debug info without a version number or with a different version number, to make sure we don't crash when we see bitcode files with different debug info metadata format. llvm-svn: 195504
2013-11-23 05:49:45 +08:00
!llvm.module.flags = !{!11, !13}
Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g. Adding a test case to make sure that the threshold used in the memory dependence analysis is respected. The test case also checks that debug intrinsics are not counted towards this threshold. Differential Revision: http://llvm-reviews.chandlerc.com/D2141 llvm-svn: 194646
2013-11-14 09:10:52 +08:00
Revert "Revert "DI: Fold constant arguments into a single MDString"" This reverts commit r218918, effectively reapplying r218914 after fixing an Ocaml bindings test and an Asan crash. The root cause of the latter was a tightened-up check in `DILexicalBlock::Verify()`, so I'll file a PR to investigate who requires the loose check (and why). Original commit message follows. -- This patch addresses the first stage of PR17891 by folding constant arguments together into a single MDString. Integers are stringified and a `\0` character is used as a separator. Part of PR17891. Note: I've attached my testcases upgrade scripts to the PR. If I've just broken your out-of-tree testcases, they might help. llvm-svn: 219010
2014-10-04 04:01:09 +08:00
!0 = metadata !{metadata !"0x11\004\00clang version 3.4\001\00\000\00\000", metadata !1, metadata !2, metadata !2, metadata !3, metadata !9, metadata !2} ; [ DW_TAG_compile_unit ] [/home/tmp/test.c] [DW_LANG_C99]
Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g. Adding a test case to make sure that the threshold used in the memory dependence analysis is respected. The test case also checks that debug intrinsics are not counted towards this threshold. Differential Revision: http://llvm-reviews.chandlerc.com/D2141 llvm-svn: 194646
2013-11-14 09:10:52 +08:00
!1 = metadata !{metadata !"test.c", metadata !"/home/tmp"}
!2 = metadata !{i32 0}
!3 = metadata !{metadata !4}
Revert "Revert "DI: Fold constant arguments into a single MDString"" This reverts commit r218918, effectively reapplying r218914 after fixing an Ocaml bindings test and an Asan crash. The root cause of the latter was a tightened-up check in `DILexicalBlock::Verify()`, so I'll file a PR to investigate who requires the loose check (and why). Original commit message follows. -- This patch addresses the first stage of PR17891 by folding constant arguments together into a single MDString. Integers are stringified and a `\0` character is used as a separator. Part of PR17891. Note: I've attached my testcases upgrade scripts to the PR. If I've just broken your out-of-tree testcases, they might help. llvm-svn: 219010
2014-10-04 04:01:09 +08:00
!4 = metadata !{metadata !"0x2e\00test_within_limit\00test_within_limit\00\003\000\001\000\006\00256\000\004", metadata !1, metadata !5, metadata !6, null, i32 ()* @test_within_limit, null, null, metadata !2} ; [ DW_TAG_subprogram ] [line 3] [def] [scope 4] [test]
!5 = metadata !{metadata !"0x29", metadata !1} ; [ DW_TAG_file_type ] [/home/tmp/test.c]
!6 = metadata !{metadata !"0x15\00\000\000\000\000\000\000", i32 0, null, null, metadata !7, null, null, null} ; [ DW_TAG_subroutine_type ] [line 0, size 0, align 0, offset 0] [from ]
Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g. Adding a test case to make sure that the threshold used in the memory dependence analysis is respected. The test case also checks that debug intrinsics are not counted towards this threshold. Differential Revision: http://llvm-reviews.chandlerc.com/D2141 llvm-svn: 194646
2013-11-14 09:10:52 +08:00
!7 = metadata !{metadata !8}
Revert "Revert "DI: Fold constant arguments into a single MDString"" This reverts commit r218918, effectively reapplying r218914 after fixing an Ocaml bindings test and an Asan crash. The root cause of the latter was a tightened-up check in `DILexicalBlock::Verify()`, so I'll file a PR to investigate who requires the loose check (and why). Original commit message follows. -- This patch addresses the first stage of PR17891 by folding constant arguments together into a single MDString. Integers are stringified and a `\0` character is used as a separator. Part of PR17891. Note: I've attached my testcases upgrade scripts to the PR. If I've just broken your out-of-tree testcases, they might help. llvm-svn: 219010
2014-10-04 04:01:09 +08:00
!8 = metadata !{metadata !"0x24\00int\000\0032\0032\000\000\005", null, null} ; [ DW_TAG_base_type ] [int] [line 0, size 32, align 32, offset 0, enc DW_ATE_signed]
Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g. Adding a test case to make sure that the threshold used in the memory dependence analysis is respected. The test case also checks that debug intrinsics are not counted towards this threshold. Differential Revision: http://llvm-reviews.chandlerc.com/D2141 llvm-svn: 194646
2013-11-14 09:10:52 +08:00
!9 = metadata !{metadata !10}
Revert "Revert "DI: Fold constant arguments into a single MDString"" This reverts commit r218918, effectively reapplying r218914 after fixing an Ocaml bindings test and an Asan crash. The root cause of the latter was a tightened-up check in `DILexicalBlock::Verify()`, so I'll file a PR to investigate who requires the loose check (and why). Original commit message follows. -- This patch addresses the first stage of PR17891 by folding constant arguments together into a single MDString. Integers are stringified and a `\0` character is used as a separator. Part of PR17891. Note: I've attached my testcases upgrade scripts to the PR. If I've just broken your out-of-tree testcases, they might help. llvm-svn: 219010
2014-10-04 04:01:09 +08:00
!10 = metadata !{metadata !"0x34\00x\00x\00\001\000\001", null, metadata !5, metadata !8, i32* @x, null} ; [ DW_TAG_variable ] [x] [line 1] [def]
Fixing a heisenbug where the memory dependence analysis behaves differently with and without -g. Adding a test case to make sure that the threshold used in the memory dependence analysis is respected. The test case also checks that debug intrinsics are not counted towards this threshold. Differential Revision: http://llvm-reviews.chandlerc.com/D2141 llvm-svn: 194646
2013-11-14 09:10:52 +08:00
!11 = metadata !{i32 2, metadata !"Dwarf Version", i32 4}
!12 = metadata !{i32* undef}
Revert "Revert "DI: Fold constant arguments into a single MDString"" This reverts commit r218918, effectively reapplying r218914 after fixing an Ocaml bindings test and an Asan crash. The root cause of the latter was a tightened-up check in `DILexicalBlock::Verify()`, so I'll file a PR to investigate who requires the loose check (and why). Original commit message follows. -- This patch addresses the first stage of PR17891 by folding constant arguments together into a single MDString. Integers are stringified and a `\0` character is used as a separator. Part of PR17891. Note: I've attached my testcases upgrade scripts to the PR. If I've just broken your out-of-tree testcases, they might help. llvm-svn: 219010
2014-10-04 04:01:09 +08:00
!13 = metadata !{i32 1, metadata !"Debug Info Version", i32 2}