RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-21 20:12:56 +02:00
Code Issues Projects Releases Wiki Activity
1f3fc31b75
llvm-mirror/test/DebugInfo/Generic/recursive_inlining.ll

275 lines
12 KiB
LLVM
Raw Normal View History

DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
; REQUIRES: object-emission
Fix llvm/test/DebugInfo/X86/recursive_inlining.ll to use %llc_dwarf. llvm-svn: 215181
2014-08-08 04:24:05 +02:00
; RUN: %llc_dwarf -filetype=obj -O0 < %s | llvm-dwarfdump -debug-dump=info - | FileCheck %s
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
; This isn't a very pretty test case - I imagine there might be other ways to
; tickle the optimizers into producing the desired code, but I haven't found
; them.
; The issue is when a function is inlined into itself, the inlined argument
; accidentally overwrote the concrete argument and was lost.
; IR generated from the following source compiled with clang -g:
; void fn1(void *);
; void fn2(int, int, int, int);
; void fn3();
; void fn8();
; struct C {
; int b;
; void m_fn2() {
; fn8();
; if (b) fn2(0, 0, 0, 0);
; fn3();
; }
; };
; C *x;
; inline void fn7() {}
; void fn6() {
; fn8();
; x->m_fn2();
; fn7();
; }
; void fn3() { fn6(); }
; void fn4() { x->m_fn2(); }
; void fn5() { x->m_fn2(); }
; The definition of C and declaration of C::m_fn2
; CHECK: DW_TAG_structure_type
; CHECK-NOT: {{DW_TAG|NULL}}
; CHECK: DW_TAG_member
; CHECK-NOT: {{DW_TAG|NULL}}
Make DW_AT_[MIPS_]linkage_name optional, and off by default for SCE. Mangled "linkage" names can be huge, and if the debugger (or other tools) have no use for them, the size savings can be very impressive (on the order of 40%). Add one test for controlling behavior, and modify a number of tests to either stop using linkage names, or make llc emit them (so these tests will still run when the default triple is for PS4). Differential Revision: http://reviews.llvm.org/D11374 llvm-svn: 244678
2015-08-11 23:36:45 +02:00
; CHECK: [[M_FN2_DECL:.*]]: DW_TAG_subprogram
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
; CHECK-NOT: DW_TAG
; CHECK: DW_AT_name {{.*}} "m_fn2"
; CHECK-NOT: {{DW_TAG|NULL}}
Make DW_AT_[MIPS_]linkage_name optional, and off by default for SCE. Mangled "linkage" names can be huge, and if the debugger (or other tools) have no use for them, the size savings can be very impressive (on the order of 40%). Add one test for controlling behavior, and modify a number of tests to either stop using linkage names, or make llc emit them (so these tests will still run when the default triple is for PS4). Differential Revision: http://reviews.llvm.org/D11374 llvm-svn: 244678
2015-08-11 23:36:45 +02:00
; CHECK: DW_TAG_formal_parameter
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
; The abstract definition of C::m_fn2
; CHECK: [[M_FN2_ABS_DEF:.*]]: DW_TAG_subprogram
; CHECK-NOT: DW_TAG
Make DW_AT_[MIPS_]linkage_name optional, and off by default for SCE. Mangled "linkage" names can be huge, and if the debugger (or other tools) have no use for them, the size savings can be very impressive (on the order of 40%). Add one test for controlling behavior, and modify a number of tests to either stop using linkage names, or make llc emit them (so these tests will still run when the default triple is for PS4). Differential Revision: http://reviews.llvm.org/D11374 llvm-svn: 244678
2015-08-11 23:36:45 +02:00
; CHECK: DW_AT_specification {{.*}} {[[M_FN2_DECL]]}
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
; CHECK-NOT: DW_TAG
; CHECK: DW_AT_inline
; CHECK-NOT: {{DW_TAG|NULL}}
; CHECK: [[M_FN2_THIS_ABS_DEF:.*]]: DW_TAG_formal_parameter
; CHECK-NOT: DW_TAG
; CHECK: DW_AT_name {{.*}} "this"
; Skip some other functions
; CHECK: DW_TAG_subprogram
; CHECK: DW_TAG_subprogram
; CHECK: DW_TAG_subprogram
; The concrete definition of C::m_fn2
; CHECK: DW_TAG_subprogram
; CHECK-NOT: DW_TAG
Make DW_AT_[MIPS_]linkage_name optional, and off by default for SCE. Mangled "linkage" names can be huge, and if the debugger (or other tools) have no use for them, the size savings can be very impressive (on the order of 40%). Add one test for controlling behavior, and modify a number of tests to either stop using linkage names, or make llc emit them (so these tests will still run when the default triple is for PS4). Differential Revision: http://reviews.llvm.org/D11374 llvm-svn: 244678
2015-08-11 23:36:45 +02:00
; CHECK: DW_AT_abstract_origin {{.*}} {[[M_FN2_ABS_DEF]]}
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
; CHECK-NOT: {{DW_TAG|NULL}}
; CHECK: DW_TAG_formal_parameter
; CHECK-NOT: DW_TAG
; CHECK: DW_AT_abstract_origin {{.*}} {[[M_FN2_THIS_ABS_DEF]]}
; CHECK-NOT: {{DW_TAG|NULL}}
; Inlined fn3:
; CHECK: DW_TAG_inlined_subroutine
; CHECK-NOT: {{DW_TAG|NULL}}
; Inlined fn6:
; CHECK: DW_TAG_inlined_subroutine
; CHECK-NOT: {{DW_TAG|NULL}}
; Inlined C::m_fn2:
; CHECK: DW_TAG_inlined_subroutine
; CHECK-NOT: DW_TAG
Make DW_AT_[MIPS_]linkage_name optional, and off by default for SCE. Mangled "linkage" names can be huge, and if the debugger (or other tools) have no use for them, the size savings can be very impressive (on the order of 40%). Add one test for controlling behavior, and modify a number of tests to either stop using linkage names, or make llc emit them (so these tests will still run when the default triple is for PS4). Differential Revision: http://reviews.llvm.org/D11374 llvm-svn: 244678
2015-08-11 23:36:45 +02:00
; CHECK: DW_AT_abstract_origin {{.*}} {[[M_FN2_ABS_DEF]]}
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
; CHECK-NOT: {{DW_TAG|NULL}}
; CHECK: DW_TAG_formal_parameter
; CHECK-NOT: DW_TAG
; CHECK: DW_AT_abstract_origin {{.*}} {[[M_FN2_THIS_ABS_DEF]]}
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/recursive_inlining.ll"
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
%struct.C = type { i32 }
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
@x = global %struct.C* null, align 8, !dbg !0
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
; Function Attrs: nounwind
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
define void @_Z3fn6v() #0 !dbg !20 {
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
entry:
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
tail call void @_Z3fn8v() #3, !dbg !23
%0 = load %struct.C*, %struct.C** @x, align 8, !dbg !24, !tbaa !25
tail call void @llvm.dbg.value(metadata %struct.C* %0, i64 0, metadata !29, metadata !32) #3, !dbg !33
tail call void @_Z3fn8v() #3, !dbg !34
%b.i = getelementptr inbounds %struct.C, %struct.C* %0, i64 0, i32 0, !dbg !35
%1 = load i32, i32* %b.i, align 4, !dbg !35, !tbaa !37
%tobool.i = icmp eq i32 %1, 0, !dbg !35
br i1 %tobool.i, label %_ZN1C5m_fn2Ev.exit, label %if.then.i, !dbg !35
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
if.then.i: ; preds = %entry
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
tail call void @_Z3fn2iiii(i32 0, i32 0, i32 0, i32 0) #3, !dbg !40
br label %_ZN1C5m_fn2Ev.exit, !dbg !40
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
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
_ZN1C5m_fn2Ev.exit: ; preds = %if.then.i, %entry
tail call void @_Z3fn3v() #3, !dbg !42
ret void, !dbg !43
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
}
declare void @_Z3fn8v() #1
; Function Attrs: nounwind
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
define linkonce_odr void @_ZN1C5m_fn2Ev(%struct.C* nocapture readonly %this) #0 align 2 !dbg !30 {
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
entry:
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
tail call void @llvm.dbg.value(metadata %struct.C* %this, i64 0, metadata !29, metadata !32), !dbg !44
tail call void @_Z3fn8v() #3, !dbg !45
%b = getelementptr inbounds %struct.C, %struct.C* %this, i64 0, i32 0, !dbg !46
%0 = load i32, i32* %b, align 4, !dbg !46, !tbaa !37
%tobool = icmp eq i32 %0, 0, !dbg !46
br i1 %tobool, label %if.end, label %if.then, !dbg !46
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
if.then: ; preds = %entry
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
tail call void @_Z3fn2iiii(i32 0, i32 0, i32 0, i32 0) #3, !dbg !47
br label %if.end, !dbg !47
if.end: ; preds = %if.then, %entry
tail call void @_Z3fn8v() #3, !dbg !48
%1 = load %struct.C*, %struct.C** @x, align 8, !dbg !52, !tbaa !25
tail call void @llvm.dbg.value(metadata %struct.C* %1, i64 0, metadata !29, metadata !32) #3, !dbg !53
tail call void @_Z3fn8v() #3, !dbg !54
%b.i.i = getelementptr inbounds %struct.C, %struct.C* %1, i64 0, i32 0, !dbg !55
%2 = load i32, i32* %b.i.i, align 4, !dbg !55, !tbaa !37
%tobool.i.i = icmp eq i32 %2, 0, !dbg !55
br i1 %tobool.i.i, label %_Z3fn6v.exit, label %if.then.i.i, !dbg !55
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
if.then.i.i: ; preds = %if.end
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
tail call void @_Z3fn2iiii(i32 0, i32 0, i32 0, i32 0) #3, !dbg !56
br label %_Z3fn6v.exit, !dbg !56
_Z3fn6v.exit: ; preds = %if.then.i.i, %if.end
tail call void @_Z3fn3v() #3, !dbg !57
ret void, !dbg !58
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
}
; Function Attrs: nounwind
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
define void @_Z3fn3v() #0 !dbg !50 {
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
entry:
br label %tailrecurse
tailrecurse: ; preds = %tailrecurse.backedge, %entry
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
tail call void @_Z3fn8v() #3, !dbg !59
%0 = load %struct.C*, %struct.C** @x, align 8, !dbg !61, !tbaa !25
tail call void @llvm.dbg.value(metadata %struct.C* %0, i64 0, metadata !29, metadata !32) #3, !dbg !62
tail call void @_Z3fn8v() #3, !dbg !63
%b.i.i = getelementptr inbounds %struct.C, %struct.C* %0, i64 0, i32 0, !dbg !64
%1 = load i32, i32* %b.i.i, align 4, !dbg !64, !tbaa !37
%tobool.i.i = icmp eq i32 %1, 0, !dbg !64
br i1 %tobool.i.i, label %tailrecurse.backedge, label %if.then.i.i, !dbg !64
tailrecurse.backedge: ; preds = %if.then.i.i, %tailrecurse
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
br label %tailrecurse
if.then.i.i: ; preds = %tailrecurse
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
tail call void @_Z3fn2iiii(i32 0, i32 0, i32 0, i32 0) #3, !dbg !65
br label %tailrecurse.backedge, !dbg !65
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
}
; Function Attrs: nounwind
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
define void @_Z3fn4v() #0 !dbg !66 {
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
entry:
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 = load %struct.C*, %struct.C** @x, align 8, !dbg !67, !tbaa !25
tail call void @_ZN1C5m_fn2Ev(%struct.C* %0), !dbg !67
ret void, !dbg !67
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
}
; Function Attrs: nounwind
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
define void @_Z3fn5v() #0 !dbg !68 {
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
entry:
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 = load %struct.C*, %struct.C** @x, align 8, !dbg !69, !tbaa !25
tail call void @_ZN1C5m_fn2Ev(%struct.C* %0), !dbg !69
ret void, !dbg !69
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
}
declare void @_Z3fn2iiii(i32, i32, i32, i32) #1
; 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
2014-10-01 20:55:02 +02:00
declare void @llvm.dbg.value(metadata, i64, metadata, metadata) #2
DebugInfo: Fix overwriting/loss of inlined arguments to recursively inlined functions. Due to an unnecessary special case, inlined arguments that happened to be from the same function as they were inlined into were misclassified as non-inline arguments and would overwrite the non-inlined arguments. Assert that we never overwrite a function's arguments, and stop misclassifying inlined arguments as non-inline arguments to fix this issue. Excuse the rather crappy test case - handcrafted IR might do better, or someone who understands better how to tickle the inliner to create a recursive inlining situation like this (though it may also be necessary to tickle the variable in a particular way to cause it to be recorded in the MMI side table and go down this particular path for location information). llvm-svn: 215157
2014-08-08 00:22:49 +02:00
attributes #0 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-realign-stack" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #1 = { "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-realign-stack" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #2 = { nounwind readnone }
attributes #3 = { nounwind }
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
!llvm.dbg.cu = !{!12}
!llvm.module.flags = !{!17, !18}
!llvm.ident = !{!19}
!0 = !DIGlobalVariableExpression(var: !1)
!1 = !DIGlobalVariable(name: "x", scope: null, file: !2, line: 13, type: !3, isLocal: false, isDefinition: true)
!2 = !DIFile(filename: "recursive_inlining.cpp", directory: "/usr/local/google/home/blaikie/dev/scratch/missing_concrete_variable_on_darwin/reduce")
!3 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !4, size: 64, align: 64)
!4 = !DICompositeType(tag: DW_TAG_structure_type, name: "C", file: !2, line: 5, size: 32, align: 32, elements: !5, identifier: "_ZTS1C")
!5 = !{!6, !8}
!6 = !DIDerivedType(tag: DW_TAG_member, name: "b", scope: !4, file: !2, line: 6, baseType: !7, size: 32, align: 32)
!7 = !DIBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
!8 = !DISubprogram(name: "m_fn2", linkageName: "_ZN1C5m_fn2Ev", scope: !4, file: !2, line: 7, type: !9, isLocal: false, isDefinition: false, scopeLine: 7, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: true)
!9 = !DISubroutineType(types: !10)
!10 = !{null, !11}
!11 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !4, size: 64, align: 64, flags: DIFlagArtificial | DIFlagObjectPointer)
!12 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !13, producer: "clang version 3.6.0 ", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !14, retainedTypes: !15, globals: !16, imports: !14)
!13 = !DIFile(filename: "<stdin>", directory: "/usr/local/google/home/blaikie/dev/scratch/missing_concrete_variable_on_darwin/reduce")
!14 = !{}
!15 = !{!4}
!16 = !{!0}
!17 = !{i32 2, !"Dwarf Version", i32 4}
!18 = !{i32 2, !"Debug Info Version", i32 3}
!19 = !{!"clang version 3.6.0 "}
!20 = distinct !DISubprogram(name: "fn6", linkageName: "_Z3fn6v", scope: !2, file: !2, line: 15, type: !21, isLocal: false, isDefinition: true, scopeLine: 15, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: true, unit: !12, variables: !14)
!21 = !DISubroutineType(types: !22)
!22 = !{null}
!23 = !DILocation(line: 16, scope: !20)
!24 = !DILocation(line: 17, scope: !20)
!25 = !{!26, !26, i64 0}
!26 = !{!"any pointer", !27, i64 0}
!27 = !{!"omnipotent char", !28, i64 0}
!28 = !{!"Simple C/C++ TBAA"}
!29 = !DILocalVariable(name: "this", arg: 1, scope: !30, type: !3, flags: DIFlagArtificial | DIFlagObjectPointer)
!30 = distinct !DISubprogram(name: "m_fn2", linkageName: "_ZN1C5m_fn2Ev", scope: !4, file: !2, line: 7, type: !9, isLocal: false, isDefinition: true, scopeLine: 7, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: true, unit: !12, declaration: !8, variables: !31)
!31 = !{!29}
!32 = !DIExpression()
!33 = !DILocation(line: 0, scope: !30, inlinedAt: !24)
!34 = !DILocation(line: 8, scope: !30, inlinedAt: !24)
!35 = !DILocation(line: 9, scope: !36, inlinedAt: !24)
!36 = distinct !DILexicalBlock(scope: !30, file: !2, line: 9)
!37 = !{!38, !39, i64 0}
!38 = !{!4, !39, i64 0}
!39 = !{!"int", !27, i64 0}
!40 = !DILocation(line: 9, scope: !41, inlinedAt: !24)
!41 = distinct !DILexicalBlock(scope: !36, file: !2, line: 9)
!42 = !DILocation(line: 10, scope: !30, inlinedAt: !24)
!43 = !DILocation(line: 19, scope: !20)
!44 = !DILocation(line: 0, scope: !30)
!45 = !DILocation(line: 8, scope: !30)
!46 = !DILocation(line: 9, scope: !36)
!47 = !DILocation(line: 9, scope: !41)
!48 = !DILocation(line: 16, scope: !20, inlinedAt: !49)
!49 = !DILocation(line: 20, scope: !50, inlinedAt: !51)
!50 = distinct !DISubprogram(name: "fn3", linkageName: "_Z3fn3v", scope: !2, file: !2, line: 20, type: !21, isLocal: false, isDefinition: true, scopeLine: 20, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: true, unit: !12, variables: !14)
!51 = !DILocation(line: 10, scope: !30)
!52 = !DILocation(line: 17, scope: !20, inlinedAt: !49)
!53 = !DILocation(line: 0, scope: !30, inlinedAt: !52)
!54 = !DILocation(line: 8, scope: !30, inlinedAt: !52)
!55 = !DILocation(line: 9, scope: !36, inlinedAt: !52)
!56 = !DILocation(line: 9, scope: !41, inlinedAt: !52)
!57 = !DILocation(line: 10, scope: !30, inlinedAt: !52)
!58 = !DILocation(line: 11, scope: !30)
!59 = !DILocation(line: 16, scope: !20, inlinedAt: !60)
!60 = !DILocation(line: 20, scope: !50)
!61 = !DILocation(line: 17, scope: !20, inlinedAt: !60)
!62 = !DILocation(line: 0, scope: !30, inlinedAt: !61)
!63 = !DILocation(line: 8, scope: !30, inlinedAt: !61)
!64 = !DILocation(line: 9, scope: !36, inlinedAt: !61)
!65 = !DILocation(line: 9, scope: !41, inlinedAt: !61)
!66 = distinct !DISubprogram(name: "fn4", linkageName: "_Z3fn4v", scope: !2, file: !2, line: 21, type: !21, isLocal: false, isDefinition: true, scopeLine: 21, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: true, unit: !12, variables: !14)
!67 = !DILocation(line: 21, scope: !66)
!68 = distinct !DISubprogram(name: "fn5", linkageName: "_Z3fn5v", scope: !2, file: !2, line: 22, type: !21, isLocal: false, isDefinition: true, scopeLine: 22, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: true, unit: !12, variables: !14)
!69 = !DILocation(line: 22, scope: !68)
Reference in New Issue Copy Permalink