weak linkage. Also, fix a problem where global weak variables
with non-trivial initializers were getting guard variables, or at
least were checking for them and then crashing.
llvm-svn: 129342
for __unknown_anytype resolution to destructively modify the AST. So that's
what it does now, which significantly simplifies some of the implementation.
Normal member calls work pretty cleanly now, and I added support for
propagating unknown-ness through &.
llvm-svn: 129331
__block object copy/dispose helpers for C++ objects with those for
different variables with completely different semantics simply because
they happen to both be no more aligned than a pointer.
Found by inspection.
Also, internalize most of the helper generation logic within CGBlocks.cpp,
and refactor it to fit my peculiar aesthetic sense.
llvm-svn: 128618
Emit them instead with the linkage of the VTT.
I'm actually really ambivalent about this; it's what GCC does, but outside
of improving code size (if the linkage is coalescing), I'm not sure it's
at all relevant. Construction vtables are naturally referenced only by the
VTT, which is itself only referenced by complete-object constructors and
destructors; giving the construction vtables possibly-external linkage is
important if you have an optimization that drills through the VTT to a
reference to a particular construction vtable which it cannot just emit
itself.
llvm-svn: 128374
they don't collide with file-scope extern functions from the same
translation unit. This is basically a matter of applying the same
logic to FunctionDecls as we were previously applying to VarDecls.
llvm-svn: 128072
Issue this as an IR-gen error; it's not really worthwhile doing this
"right", i.e. in Sema, because IR gen knows a lot of tricks beyond
what the constant evaluator knows.
llvm-svn: 127854
allocation and therefore requires a null-check. We were doing that, but
we weren't treating the new-initializer as being conditionally executed,
which means it was possible to get ill-formed IR as in PR9298.
llvm-svn: 127147
computing for a nested decl with explicit visibility. This is all part
of the general philosophy of explicit visibility attributes, where
any information that was obviously available at the attribute site
should probably be ignored. Fixes PR9371.
llvm-svn: 126992
necessarily enclose the innermost normal cleanup depth, because
the top of the jump scope stack might be an EH cleanup or EH scope.
Fixes PR9303.
llvm-svn: 126472
without defining them. This should be an error, but I'm paranoid about
"uses" that end up not actually requiring a definition. I'll revisit later.
Also, teach IR generation to not set internal linkage on variable
declarations, just for safety's sake. Doing so produces an invalid module
if the variable is not ultimately defined.
Also, fix several places in the test suite where we were using internal
functions without definitions.
llvm-svn: 126016
class and to bind the shared value using OpaqueValueExpr. This fixes an
unnoticed problem with deserialization of these expressions where the
deserialized form would lose the vital pointer-equality trait; or rather,
it fixes it because this patch also does the right thing for deserializing
OVEs.
Change OVEs to not be a "temporary object" in the sense that copy elision is
permitted.
This new representation is not totally unawkward to work with, but I think
that's really part and parcel with the semantics we're modelling here. In
particular, it's much easier to fix things like the copy elision bug and to
make the CFG look right.
I've tried to update the analyzer to deal with this in at least some
obvious cases, and I think we get a much better CFG out, but the printing
of OpaqueValueExprs probably needs some work.
llvm-svn: 125744
a zero constant for a complete class. rdar://problem/8424975
To make this happen, track the field indexes for virtual bases
in the complete object. I'm curious whether we might be better
off making CGRecordLayoutBuilder *much* more reliant on
ASTRecordLayout; we're currently duplicating an awful lot of the ABI
layout logic.
llvm-svn: 125555
and we later find the definition, make sure that we add the definition
(not the declaration) to the list of deferred definitions to
emit. Fixes PR8864.
Thanks to Nick Lewycky for testing this patch out
llvm-svn: 125157
is not defined in the current translation unit. Doing so lead to compile errors
such as PR9114.
Instead, when CodeGen is building the vtable, don't try to emit a definition
for functions that aren't marked used in the current translation unit.
Fixes PR9114.
llvm-svn: 124768
extremely rambunctious, both on parsing and on template instantiation.
Calm it down, fixing an internal consistency assert on anonymous enum
instantiation manglings.
llvm-svn: 124653
* llvm-link would complains about mismatched visibility
* If we produce a relocation with an available_externally, it is good to know that
it is hidden.
llvm-svn: 124633
current translation unit as available_externally.
This helps devirtualize the second example in PR3100, comment 18:
struct S { S() {}; virtual void xyzzy(); };
inline void foo(S *s) { s->xyzzy(); }
void bar() { S s; foo(&s); }
This involved four major changes:
1. In DefineUsedVTables, always mark virtual member functions as referenced for
non-template classes and class template specializations.
2. In CodeGenVTables::ShouldEmitVTableInThisTU return true if optimizations are
enabled, even if the key function is not implemented in this translation
unit. We don't ever do this for code compiled with -fapple-kext, because we
don't ever want to devirtualize virtual member function calls in that case.
3. Give the correct linkage for vtables where the key function is not defined.
4. Update the linkage for RTTI structures when necessary.
llvm-svn: 124565
fixing a crash which probably nobody was ever going to see. In doing so,
fix a horrendous number of problems with the conditional-cleanups code.
Also, make conditional cleanups re-use the cleanup's activation variable,
which avoids some unfortunate repetitiveness.
llvm-svn: 124481
non-class prvalues actually require the realization of a
temporary. For everything else, we already have an lvalue (or class
prvalue) in the subexpression.
Note: we're missing some move elision in this case. I'll tackle that next.
llvm-svn: 124453
deallocation function has a two-argument form. Store the result of this
check in new[] and delete[] nodes.
Fixes rdar://problem/8913519
llvm-svn: 124373
using rules that I just made up this morning. This encoding has now
been proposed to the Itanium C++ ABI group for inclusion, but of
course it's still possible that the mangling will change.
llvm-svn: 124296
I'm separately committing this because it incidentally changes some
block orderings and minor IR issues, like using a phi instead of
an unnecessary alloca.
llvm-svn: 124277
NRVO candidate for a return statement, to
Sema::getCopyElisionCandidate(), and teach it enough to also determine
the NRVO candidate for a throw expression. We still don't use the
latter information, however.
Along the way, implement core issue 1148, which eliminates copy
elision from catch parameters and clarifies that copy elision cannot
occur from function parameters (which we already implemented).
llvm-svn: 123982
For example:
class A{
public:
A& operator=(const A& that) {
if (this != &that) {
this->A::~A();
this->A::A(that); // <=== explicit constructor call.
}
return *this;
}
};
More work will be needed to support an explicit call to a template constructor.
llvm-svn: 123735
The problem was that we were asserting the we never added an empty class
to the same offset twice. This is not true for unions, where two members, empty
or not, can have the some offset.
llvm-svn: 122633
inconsistent with the type that the builtin *should* have, forget
about the builtin altogether: we don't want subsequence analyses,
CodeGen, etc., to think that we have a proper builtin function.
C is protected from errors here because it allows one to use a
library builtin without having a declaration, and detects inconsistent
(re-)declarations of builtins during declaration merging. C++ was
unprotected, and therefore would crash.
Fixes PR8839.
llvm-svn: 122351
within the class. Teach IR gen to look for function definitions in record
lexical contexts when deciding whether to emit a function whose address
was taken. Fixes PR8789.
llvm-svn: 121833
visibility. Fixes PR8713.
I've disabled a test which was testing that you can #pragma pop visibility
to get out of a namespace's visibility attribute. We should probably just
diagnose that as an error unless it's instrumental to someone's system
headers.
llvm-svn: 121459
"inline", we weren't giving the definition weak linkage because the
"inline" bit wasn't propagated. This was a longstanding FIXME that,
somehow, hadn't triggered a bug in the wild. Fix this problem by
tracking whether any declaration was marked "inline", and clean up the
semantics of GNU's "extern inline" semantics calculation based on this
change.
Fixes <rdar://problem/8740363>.
llvm-svn: 121373
the LHS, or else the pointer might be invalid. This is kindof dumb, but
go ahead and make sure we're doing that for l-value scalar assignment,
which fixes a miscompile of obj-c++.dg/block-seq.mm.
Leave a FIXME for how to solve this problem for agg __blocks.
llvm-svn: 120992
Fix a bug in the emission of complex compound assignment l-values.
Introduce a method to emit an expression whose value isn't relevant.
Make that method evaluate its operand as an l-value if it is one.
Fixes our volatile compliance in C++.
llvm-svn: 120931
mangler. Now member functions and pointers thereof have their calling
convention mangled as __thiscall if they have the default CC (even though,
they technically still have the __cdecl CC).
llvm-svn: 118598
there's no return adjustment from the overridden to the overrider doesn't
mean there isn't a return adjustment from the overrider to the final
overrider. This matters if we're emitting a virtual this-adjustment thunk
because the overrider virtually inherits from the class providing the
nearest overridden method. Do the appropriate return adjustment in this case.
Fixes PR7611.
llvm-svn: 118466
data members by delaying the emission of the initializer until after
linkage and visibility have been set on the global. Also, don't
emit a guard unless the variable actually ends up with vague linkage,
and don't use thread-safe statics in any case.
llvm-svn: 118336
e.g. for:
template <int i> class A {
class B *g;
};
'class B' has the template as lexical context but semantically it is
introduced in namespace scope.
Fixes rdar://8611125 & http://llvm.org/PR8505
llvm-svn: 118235
with their own explicit visibility attributes. Basically we only want to
apply a single visibility attribute from any particular ancestry.
llvm-svn: 117998
independently of whether they're definitions, then teach IR generation to
ignore non-explicit visibility when emitting declarations. Use this to
make sure that RTTI, vtables, and VTTs get the right visibility.
More of rdar://problem/8613093
llvm-svn: 117781
whether it's a declaration or not, then ignores that information for
declarations unless it was explicitly given. It's not totally clear
how that should be mapped into a sane system, but make an effort.
llvm-svn: 117780
for namespace-scope variable declarations.
Apply visibility in IR gen to variables that are merely declared
and never defined. We were previously emitting these with default
visibility unless they were declared with private_extern.
Ignore global visibility settings when computing visibility for
a declaration's context, and key several conditions on whether a
visibility attribute exists anywhere in the hierarchy as opposed
to whether it exists at the current level.
llvm-svn: 117729
and never defined. We were previously emitting these with default
visibility unless they were declared with private_extern.
Ignore global visibility settings when computing visibility for
a declaration's context, and key several conditions on whether a
visibility attribute exists anywhere in the hierarchy as opposed
to whether it exists at the current level.
llvm-svn: 117644
- tags with C linkage should ignore visibility=hidden
- functions and variables with explicit visibility attributes should
ignore the linkage of their types
Either of these should be sufficient to fix PR8457.
Also, FileCheck-ize a test case.
llvm-svn: 117351
more closely parallel the computation of linkage. This gets us to a state
much closer to what gcc emits, modulo bugs, which will undoubtedly arise in
abundance.
llvm-svn: 117147
by marking the decl invalid isn't. Make some steps towards supporting these
and then hastily shut them down at the last second by marking them as
unsupported.
llvm-svn: 116661
arguments in either the placement or constructor arguments. This is
important if the default arguments refer to a declaration or create a
temporary.
llvm-svn: 115700
-Wpadded warns when undesired padding is introduced in a struct. (rdar://7469556)
-Wpacked warns if a struct is given the packed attribute, but the packed attribute has no effect
on the layout or the size of the struct. Such structs may be mis-aligned for little benefit.
The warnings are emitted at the point where layout is calculated, that is at RecordLayoutBuilder.
To avoid calculating the layouts of all structs regardless of whether they are needed or not,
I let the layouts be lazily constructed when needed. This has the disadvantage that the above warnings
will be emitted only when they are used for IR gen, and not e.g with -fsyntax-only:
$ cat t.c
struct S {
char c;
int i;
};
void f(struct S* s) {}
$ clang -fsyntax-only -Wpadded t.c
$ clang -c -Wpadded t.c -o t.o
t.c:3:7: warning: padding struct 'struct S' with 3 bytes to align 'i' [-Wpadded]
int i;
^
1 warning generated.
This is a good tradeoff between providing the warnings and not calculating layouts for all
structs in case the user has enabled a couple of rarely used warnings.
llvm-svn: 114544
LHS and when conditional expression is an array. Since
it will be decayed, saved expression must be saved with
decayed expression. This is necessary to preserve semantics
of this extension (and prevent an IRGen crash which expects
an array to always be decayed). I am sure there will be other
cases in c++ (aggregate conditionals for example) when saving of the
expression must happen after some transformation on conditional
expression has happened.
Doug, please review. Fixes // rdar://8446940
llvm-svn: 114296
the bases are completely initialized. This won't work --- base
initializer expressions can rely on the vtables having been set up.
Check for uses of 'this' in the initializers and force a vtable
initialization if found.
This might not be good enough; we might need to extend this to handle
the possibility of arbitrary code finding an external reference to this
(not yet completely-constructed!) object and accessing through it,
in which case we'll probably find ourselves doing a lot more unnecessary
stores.
llvm-svn: 114153
the cleanup might not be dominated by the allocation code.
In this case, we have to store aside all the delete arguments
in case we need them later. There's room for optimization here
in cases where we end up not actually needing the cleanup in
different branches (or being able to pop it after the
initialization code).
Also make sure we only call this operator delete along the path
where we actually allocated something.
Fixes rdar://problem/8439196.
llvm-svn: 114145
slot. The easiest way to do that was to bundle up the information
we care about for aggregate slots into a new structure which demands
that its creators at least consider the question.
I could probably be convinced that the ObjC 'needs GC' bit should
be rolled into this structure.
Implement generalized copy elision. The main obstacle here is that
IR-generation must be much more careful about making sure that exactly
llvm-svn: 113962
complains when the element type of a C++ "delete" expression is
different from what we would expect from the pointer type. When
deleting a bool*, we end up with an i1 on one side (where we compute
the LLVM type from the Clang bool type) and i8 on the other (where we
grab the LLVM type from the LLVM pointer type). I've weakened the
assertion appropriately, and the Boost Parallel Graph Library now
passes its regression tests.
llvm-svn: 112821
constructing an LLVM PointerType directly from the "bool"'s LLVM type
(i1), which resulted in unfortunate pointer type i1*. The fix is to
build the LLVM PointerType from the corresponding Clang PointerType,
so that we get i8* in the case of a bool.
John, please review. I also left a FIXME there because we seem to be
dropping "volatile", which would be rather unfortunate.
llvm-svn: 112819
implement ARM array cookies. Also fix a few unfortunate bugs:
- throwing dtors in deletes prevented the allocation from being deleted
- adding the cookie to the new[] size was not being considered for
overflow (and, more seriously, was screwing up the earlier checks)
- deleting an array via a pointer to array of class type was not
causing any destructors to be run and was passing the unadjusted
pointer to the deallocator
- lots of address-space problems, in case anyone wants to support
free store in a variant address space :)
llvm-svn: 112814
caused by my ABI work. Passing:
struct outer {
int x;
struct epsilon_matcher {} e;
int f;
};
as {i32,i32} isn't safe, because the offset of the second element
needs to be at 8 when it is interpreted as a memory value.
llvm-svn: 112686
templates when only the declaration is in scope. This requires deferring the
instantiation to be lazy, and ensuring the definition is required for that
translation unit. We re-use the existing pending instantiation queue,
previously only used to track implicit instantiations which were required to be
lazy. Fixes PR7979.
A subsequent change will rename *PendingImplicitInstantiations to
*PendingInstatiations for clarity given its broader role.
llvm-svn: 112037
but not in C++, so don't emit aggregate loads of volatile references
in null context in C++. Happens to have been caught by an assertion.
We do not get the scalar case right. Volatiles are really broken.
llvm-svn: 112019
That revision started classifying truly empty structs like "Y" and "X"
as being NoClass/NoClass and turning them into 'ignore'. The call code
turns around and allocates space for the ignored argument with
GetUndefRValue. The bug is that GetUndefRValue would return the address
as undef, instead of returning an object with a defined address but
undefined contents.
llvm-svn: 111794
class; they should just be completely opaque throughout IR gen now,
although I haven't really audited that.
Fix a bug apparently inherited from gcc-4.2 where we failed to null-check
member data pointers when performing derived-to-base or base-to-derived
conversions on them.
llvm-svn: 111789
implicitly-defined default constructor, zero-initialize the memory
before calling the default constructor. Previously, we would only
zero-initialize in the case of a trivial default constructor.
Also, simplify the hideous logic that determines when we have a
trivial default constructor and, therefore, don't need to emit any
call at all.
llvm-svn: 111779
pointers. I find the resulting code to be substantially cleaner, and it
makes it very easy to use the same APIs for data member pointers (which I have
conscientiously avoided here), and it avoids a plethora of potential
inefficiencies due to excessive memory copying, but we'll have to see if it
actually works.
llvm-svn: 111776
duplication between the constant and non-constant paths in all of this.
Implement ARM ABI semantics for member pointer constants and conversion.
llvm-svn: 111772
trivial default constructors. We're weren't zero-initializing them,
which manifested as <rdar://problem/8320532> (a regression in the GCC
test suite) and is likely to have caused significant other breakage.
llvm-svn: 111650
mangleCallExpression. Also, operator names with unknown arity should
be mangled as binary operators; this is actually covered by an oddly-
positioned sentence in the ABI document. Fixes PR7891.
llvm-svn: 111395
a -cc1 option. The Darwin linker complains about mixed visibility when linking
gcc-built objects with clang-built objects, and the optimization isn't really
that valuable. Platforms with less ornery linkers can feel free to enable this.
llvm-svn: 110979
to avoid the awesome-but-wrong-in-this-case assertion in the canon EAC.
Fixes PR7834.
Also fix a subtle address-space bug in the memset path.
llvm-svn: 110511
do the right thing with mixed-visibility symbols, so disable the visibility
optimization where that's possible, i.e. with template classes (since it's
possible that an arbitrary template might be subject to an explicit
instantiation elsewhere). 447.dealII actually does this.
I've put the code under an option that's currently not hooked up to anything.
llvm-svn: 110374
functions with in-line definitions, since such thunks will be emitted at any
use of the function.
Completes the feature work for rdar://problem/7523229.
llvm-svn: 110285
Apply hidden visibility to most RTTI; libstdc++ does not rely on exact
pointer equality for the type info (just the type info names). Apply
the same optimization to RTTI that we do to vtables.
Fixes PR5962.
llvm-svn: 110192
The X86-64 ABI code didn't handle the case when a struct
would get classified and turn up as "NoClass INTEGER" for
example. This is perfectly possible when the first slot
is all padding (e.g. due to empty base classes). In this
situation, the first 8-byte doesn't take a register at all,
only the second 8-byte does.
This fixes this by enhancing the x86-64 abi stuff to allow
and handle this case, reverts the broken fix for PR5831,
and enhances the target independent stuff to be able to
handle an argument value in registers being accessed at an
offset from the memory value.
This is the last x86-64 calling convention related miscompile
that I'm aware of.
llvm-svn: 109848
return where the struct has a base but no fields. This
was because the x86-64 abi logic was checking the wrong
predicate in one place.
This was introduced in r91874, which was a fix for PR5831,
which lacked a CHECK line, so I verified and added it.
llvm-svn: 109759
have a "coerce to" type which often matches the default lowering of Clang
type to LLVM IR type, but the coerce case can be handled by making them
not be the same.
This simplifies things and fixes issues where X86-64 abi lowering would
return coerce after making preferred types exactly match up. This caused
us to compile:
typedef float v4f32 __attribute__((__vector_size__(16)));
v4f32 foo(v4f32 X) {
return X+X;
}
into this code at -O0:
define <4 x float> @foo(<4 x float> %X.coerce) nounwind {
entry:
%retval = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=2]
%coerce = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=2]
%X.addr = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=3]
store <4 x float> %X.coerce, <4 x float>* %coerce
%X = load <4 x float>* %coerce ; <<4 x float>> [#uses=1]
store <4 x float> %X, <4 x float>* %X.addr
%tmp = load <4 x float>* %X.addr ; <<4 x float>> [#uses=1]
%tmp1 = load <4 x float>* %X.addr ; <<4 x float>> [#uses=1]
%add = fadd <4 x float> %tmp, %tmp1 ; <<4 x float>> [#uses=1]
store <4 x float> %add, <4 x float>* %retval
%0 = load <4 x float>* %retval ; <<4 x float>> [#uses=1]
ret <4 x float> %0
}
Now we get:
define <4 x float> @foo(<4 x float> %X) nounwind {
entry:
%X.addr = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=3]
store <4 x float> %X, <4 x float>* %X.addr
%tmp = load <4 x float>* %X.addr ; <<4 x float>> [#uses=1]
%tmp1 = load <4 x float>* %X.addr ; <<4 x float>> [#uses=1]
%add = fadd <4 x float> %tmp, %tmp1 ; <<4 x float>> [#uses=1]
ret <4 x float> %add
}
This implements rdar://8248065
llvm-svn: 109733
enclosing normal cleanup, not the top of the EH stack. I'm *really*
surprised this hasn't been causing more problems.
Fixes rdar://problem/8231514.
llvm-svn: 109569
when the RHS of the ||/&& is ever 0 or 1. This handles a variety of
creative idioms for "true" used in C programs and fixes many false
positives at the expense of a few false negatives. This fixes
rdar://8230351.
llvm-svn: 109314
initializer of (). Make sure to use a simple memset() when we can, or
fall back to generating a loop when a simple memset will not
suffice. Fixes <rdar://problem/8212208>, a regression due to my work
in r107857.
llvm-svn: 108977
which generates more efficient and more obviously conformant
code. We now test for overflow of the multiply then force
the result to -1 if so. On X86, this generates nice code
like this:
__Z4testl: ## @_Z4testl
## BB#0: ## %entry
subl $12, %esp
movl $4, %eax
mull 16(%esp)
testl %edx, %edx
movl $-1, %ecx
cmovel %eax, %ecx
movl %ecx, (%esp)
call __Znam
addl $12, %esp
ret
llvm-svn: 108927
causing clang to compile this code into something that correctly throws a
length error, fixing a potential integer overflow security attack:
void *test(long N) {
return new int[N];
}
int main() {
test(1L << 62);
}
We do this even when exceptions are disabled, because it is better for the
code to abort than for the attack to succeed.
This is heavily based on a patch that Fariborz wrote.
llvm-svn: 108915
that involve binding a reference to a pure rvalue temporary (e.g., not
a class temporary), by creating a new temporary and copying the result
there. Fixes PR6024.
llvm-svn: 108431
mostly in avoiding unnecessary work at compile time but also in producing more
sensible block orderings.
Move the destructor cleanups for local variables over to use lazy cleanups.
Eventually all cleanups will do this; for now we have some awkward code
duplication.
Tell IR generation just to never produce landing pads in -fno-exceptions.
This is a much more comprehensive solution to a problem which previously was
half-solved by checks in most cleanup-generation spots.
llvm-svn: 108270
t2.c:2:12: warning: use of logical && with constant operand; switch to bitwise &
or remove constant [-Wlogical-bitwise-confusion]
return x && 4;
^ ~
wording improvement suggestions are welcome.
llvm-svn: 108260
at -O0. The only change from the previous patch is that we don't try
to generate virtual method thunks for an available_externally
function.
llvm-svn: 108230
-O0, since we won't be using the definitions for anything anyway. For
lib/System/Path.o when built in Debug+Asserts mode, this leads to a 4%
improvement in compile time (and suppresses 440 function bodies).
<rdar://problem/7987644>
llvm-svn: 108156
CXXConstructExpr/CXXTemporaryObjectExpr/CXXNewExpr as
appropriate. Fixes PR7556, and provides a slide codegen improvement
when copy-initializing a POD class type from a value-initialized
temporary. Previously, we weren't eliding the copy.
llvm-svn: 107827
self-host. Hopefully these results hold up on different platforms.
I tried to keep the GNU ObjC runtime happy, but it's hard for me to test.
Reimplement how clang generates IR for exceptions. Instead of creating new
invoke destinations which sequentially chain to the previous destination,
push a more semantic representation of *why* we need the cleanup/catch/filter
behavior, then collect that information into a single landing pad upon request.
Also reorganizes how normal cleanups (i.e. cleanups triggered by non-exceptional
control flow) are generated, since it's actually fairly closely tied in with
the former. Remove the need to track which cleanup scope a block is associated
with.
Document a lot of previously poorly-understood (by me, at least) behavior.
The new framework implements the Horrible Hack (tm), which requires every
landing pad to have a catch-all so that inlining will work. Clang no longer
requires the Horrible Hack just to make exceptions flow correctly within
a function, however. The HH is an unfortunate requirement of LLVM's EH IR.
llvm-svn: 107631
ObjC pointers were easy enough (as far as the ABI is concerned, they're
just pointers to structs), but I had to invent a new mangling for block
pointers. This is particularly worrying with the Microsoft ABI, because
it is a vendor-specific ABI; extending it could come back to bite us
later when MS extends it on their own (and you know they will).
llvm-svn: 107572
Only actual functions get mangled correctly; I don't know how to fix it for
function pointers yet. Thanks to John McCall for the hint.
Also, mangle anonymous tag types. I don't have a suitable testcase yet; I have
a feeling that that's going to need support for static locals, and I haven't
figured out exactly how MSVC's scheme for mangling those works.
llvm-svn: 107561
aren't dropping all exception specifications on destructors, the
exception specifications on implicitly-declared destructors were
detected as being wrong (which they were).
Introduce logic to provide a proper exception-specification for
implicitly-declared destructors. This also fixes PR6972.
Note that the other implicitly-declared special member functions also
need to get exception-specifications. I'll deal with that in a
subsequent commit.
llvm-svn: 107385
r107173, "fix PR7519: after thrashing around and remembering how all this stuff"
r107216, "fix PR7523, which was caused by the ABI code calling ConvertType instead"
This includes a fix to make ConvertTypeForMem handle the "recursive" case, and call
it as such when lowering function types which have an indirect result.
llvm-svn: 107310
doesn't mangle array parameters right), but I think that should be fixed
in Sema (Doug, John, what do you think?).
Also, stub out the remaining mangleType() routines.
llvm-svn: 107264
would trigger an extra method call).
- While in the area, I also changed Clang to not emit an unnecessary load from
'x' in cases like 'y = (x = 1)'.
llvm-svn: 107210
load/store nonsense in the epilog. For example, for:
int foo(int X) {
int A[100];
return A[X];
}
we used to generate:
%arrayidx = getelementptr inbounds [100 x i32]* %A, i32 0, i64 %idxprom ; <i32*> [#uses=1]
%tmp1 = load i32* %arrayidx ; <i32> [#uses=1]
store i32 %tmp1, i32* %retval
%0 = load i32* %retval ; <i32> [#uses=1]
ret i32 %0
}
which codegen'd to this code:
_foo: ## @foo
## BB#0: ## %entry
subq $408, %rsp ## imm = 0x198
movl %edi, 400(%rsp)
movl 400(%rsp), %edi
movslq %edi, %rax
movl (%rsp,%rax,4), %edi
movl %edi, 404(%rsp)
movl 404(%rsp), %eax
addq $408, %rsp ## imm = 0x198
ret
Now we generate:
%arrayidx = getelementptr inbounds [100 x i32]* %A, i32 0, i64 %idxprom ; <i32*> [#uses=1]
%tmp1 = load i32* %arrayidx ; <i32> [#uses=1]
ret i32 %tmp1
}
and:
_foo: ## @foo
## BB#0: ## %entry
subq $408, %rsp ## imm = 0x198
movl %edi, 404(%rsp)
movl 404(%rsp), %edi
movslq %edi, %rax
movl (%rsp,%rax,4), %eax
addq $408, %rsp ## imm = 0x198
ret
This actually does matter, cutting out 2000 lines of IR from CGStmt.ll
for example.
Another interesting effect is that altivec.h functions which are dead
now get dce'd by the inliner. Hence all the changes to
builtins-ppc-altivec.c to ensure the calls aren't dead.
llvm-svn: 106970
objective-c++ class objects which have GC'able objc object
pointers and need to use ObjC's objc_memmove_collectable
API (radar 8070772).
llvm-svn: 106061
- Mangle qualifiers.
- Start mangling variables' types into the name. A variable declared with a
builtin type should now mangle properly.
llvm-svn: 105931
(but not their types; that's later).
NOTE: Right now, variables in the global namespace don't get mangled, even
though they're supposed to be. This is because the default mangler
implements the shouldMangleDeclName() method that tells clang not to mangle
them. This will be fixed in a later patch.
llvm-svn: 105805
isn't possible to compute.
This patch is mostly refactoring; the key change is the addition of the code
starting with the comment, "Check whether the function has a computable LLVM
signature." The solution here is essentially the same as the way the
vtable code handles such functions.
llvm-svn: 105151
variables should have that linkage. Otherwise, its static local
variables should have internal linkage. To avoid computing this excessively,
set a function's linkage before we emit code for it.
Previously we were assigning weak linkage to the static variables of
static inline functions in C++, with predictably terrible results. This
fixes that and also gives better linkage than 'weak' when merging is required.
llvm-svn: 104581
- I think this can be cleaned up, since this means we may notify the consumer about the vtable twice, but I didn't see an easy fix for this without more substantial refactoring.
- Doug, please review!
llvm-svn: 104577
This works around a crash where malloc reused the memory of an erased BB for a
new BB leaving old cleanup information pointing at the new block.
llvm-svn: 104472
in several important ways:
- VLAs of non-POD types are not permitted.
- VLAs cannot be used in conjunction with C++ templates.
These restrictions are intended to keep VLAs out of the parts of the
C++ type system where they cause the most trouble. Fixes PR5678 and
<rdar://problem/8013618>.
llvm-svn: 104443
temporaries. There are actually several interrelated fixes here:
- When converting an object to a base class, it's only an lvalue
cast when the original object was an lvalue and we aren't casting
pointer-to-derived to pointer-to-base. Previously, we were
misclassifying derived-to-base casts of class rvalues as lvalues,
causing various oddities (including problems with reference binding
not extending the lifetimes of some temporaries).
- Teach the code for emitting a reference binding how to look
through no-op casts and parentheses directly, since
Expr::IgnoreParenNoOpCasts is just plain wrong for this. Also, make
sure that we properly look through multiple levels of indirection
from the temporary object, but destroy the actual temporary object;
this fixes the reference-binding issue mentioned above.
- Teach Objective-C message sends to bind the result as a temporary
when needed. This is actually John's change, but it triggered the
reference-binding problem above, so it's included here. Now John
can actually test his return-slot improvements.
llvm-svn: 104434
not make copies non-POD arguments or arguments passed by reference:
just copy the pointers directly. This eliminates another source of the
dreaded memcpy-of-non-PODs. Fixes PR7188.
llvm-svn: 104327
recursively, e.g. so that members of anonymous unions inside anonymous structs
still get initialized. Also generate default constructor calls for anonymous
struct members when necessary.
llvm-svn: 104292
subobject. Previously, we could only properly bind to a base class
subobject while extending the lifetime of the complete object (of a
derived type); for non-static data member subobjects, we could memcpy
(!) the result and bind to that, which is rather broken.
Now, we pull apart the expression that we're binding to, to figure out
which subobject we're accessing, then construct the temporary object
(adding a destruction if needed) and, finally, dig out the subobject
we actually meant to access.
This fixes yet another instance where we were memcpy'ing rather than
doing the right thing. However, note the FIXME in references.cpp:
there's more work to be done for binding to subobjects, since the AST
is incorrectly modeling some member accesses in base classes as
lvalues when they are really rvalues.
llvm-svn: 104219
class type (that uses a return slot), pass the return slot to the
callee directly rather than allocating new storage and trying to copy
the object. This appears to have been the cause of the remaining two
Boost.Interprocess failures.
llvm-svn: 104215
initializer, don't fold paramters. Their initializers are just default
arguments which can be overridden. This fixes some spectacular regressions due
to more things making it into the constant folding.
llvm-svn: 103904
__cxa_guard_abort along the exceptional edge into (in effect) a nested
"try" that rethrows after aborting. Fixes PR7144 and the remaining
Boost.ProgramOptions failures, along with the regressions that r103880
caused.
The crucial difference between this and r103880 is that we now follow
LLVM's little dance with the llvm.eh.exception and llvm.eh.selector
calls, then use _Unwind_Resume_or_Rethrow to rethrow.
llvm-svn: 103892
__cxa_guard_abort along the exceptional edge into (in effect) a nested
"try" that rethrows after aborting. Fixes PR7144 and the remaining
Boost.ProgramOptions failures.
llvm-svn: 103880
return statements. We perform NRVO only when all of the return
statements in the function return the same variable. Fixes some link
failures in Boost.Interprocess (which is relying on NRVO), and
probably improves performance for some C++ applications.
llvm-svn: 103867
throw, it should use invoke when needed. The fixes the
Boost.Statechrt failures that motivated PR7132, but there are a few
side issues to tackle as well.
llvm-svn: 103803
methods for which the key function is guaranteed to be in another
translation unit. Unfortunately, this guarantee isn't the case when
dealing with shared libraries that fail to export these virtual method
definitions.
I'm reopening PR6747 so we can consider this again at a later point in
time.
llvm-svn: 103741
"used" (e.g., we will refer to the vtable in the generated code) and
when they are defined (i.e., because we've seen the key function
definition). Previously, we were effectively tracking "potential
definitions" rather than uses, so we were a bit too eager about emitting
vtables for classes without key functions.
The new scheme:
- For every use of a vtable, Sema calls MarkVTableUsed() to indicate
the use. For example, this occurs when calling a virtual member
function of the class, defining a constructor of that class type,
dynamic_cast'ing from that type to a derived class, casting
to/through a virtual base class, etc.
- For every definition of a vtable, Sema calls MarkVTableUsed() to
indicate the definition. This happens at the end of the translation
unit for classes whose key function has been defined (so we can
delay computation of the key function; see PR6564), and will also
occur with explicit template instantiation definitions.
- For every vtable defined/used, we mark all of the virtual member
functions of that vtable as defined/used, unless we know that the key
function is in another translation unit. This instantiates virtual
member functions when needed.
- At the end of the translation unit, Sema tells CodeGen (via the
ASTConsumer) which vtables must be defined (CodeGen will define
them) and which may be used (for which CodeGen will define the
vtables lazily).
From a language perspective, both the old and the new schemes are
permissible: we're allowed to instantiate virtual member functions
whenever we want per the standard. However, all other C++ compilers
were more lazy than we were, and our eagerness was both a performance
issue (we instantiated too much) and a portability problem (we broke
Boost test cases, which now pass).
Notes:
(1) There's a ton of churn in the tests, because the order in which
vtables get emitted to IR has changed. I've tried to isolate some of
the larger tests from these issues.
(2) Some diagnostics related to
implicitly-instantiated/implicitly-defined virtual member functions
have moved to the point of first use/definition. It's better this
way.
(3) I could use a review of the places where we MarkVTableUsed, to
see if I missed any place where the language effectively requires a
vtable.
Fixes PR7114 and PR6564.
llvm-svn: 103718
for, and switch), be careful to construct the full expressions as soon
as we perform template instantation, so we don't either forget to call
temporary destructors or destroy temporaries at the wrong time. This
is the template-instantiation analogue to r103187, during which I
hadn't realized that the issue would affect the handling of these
constructs differently inside and outside of templates.
Fixes a regression in Boost.Function.
llvm-svn: 103357
Richard Smith