PR 14529 was opened because neither Clang or LLVM was expanding
calls to creal* or cimag* into instructions that just load the
respective complex field. After some discussion, it was not
considered realistic to do this in LLVM because of the platform
specific way complex types are expanded. Thus a way to solve
this in Clang was pursued. GCC does a similar expansion.
This patch adds the feature to Clang by making the creal* and
cimag* functions library builtins and modifying the builtin code
generator to look for the new builtin types.
llvm-svn: 170455
This fixes the storage class of extern decls that are merged with file level
statics. The patch also fixes the linkage computation so that they are
considered internal.
llvm-svn: 170406
This also requires adding support to -cc1as for passing the detecting
PWD down through LLVM's debug info (which in turn required the LLVM
change in r170371).
The test case is weak (we only test the driver behavior) because there
is currently to infrastructure for running cc1as in the test suite. So
those four lines are untested (much like all other lines in that file),
but we have a test for the same pattern using llvm-mc in the LLVM
repository.
llvm-svn: 170373
performance heuristic
After inlining a function with more than 13 basic blocks 32 times, we
are not going to inline it anymore. The idea is that inlining large
functions leads to drastic performance implications. Since the function
has already been inlined, we know that we've analyzed it in many
contexts.
The following metrics are used:
- Large function is a function with more than 13 basic blocks (we
should switch to another metric, like cyclomatic complexity)
- We consider that we've inlined a function many times if it's been
inlined 32 times. This number is configurable with -analyzer-config
max-times-inline-large=xx
This heuristic addresses a performance regression introduced with
inlining on one benchmark. The analyzer on this benchmark became 60
times slower with inlining turned on. The heuristic allows us to analyze
it in 24% of the time. The performance improvements on the other
benchmarks I've tested with are much lower - under 10%, which is
expected.
llvm-svn: 170361
incompatibility with how complex values are returned. It is sufficient
to flag all complex types as direct rather than indirect.
A new test case is provided that checks correct IR generation for the
various supported flavors of _Complex.
llvm-svn: 170302
of a member function with parenthesized declarator.
Like this test case:
class Foo {
const char *(baz)() {
return __PRETTY_FUNCTION__;
}
};
llvm-svn: 170233
C++11 allowed writing "vector<vector<int>>" without a space between the two ">".
This change allows this for protocols in template lists too in -std=c++11 mode,
and improves the diagnostic in c++98 mode.
llvm-svn: 170223
This fixes the missing warning here:
struct S {
template <typename T>
void meth() {
char arr[3];
arr[4] = 0; // warning: array index 4 is past the end of the array
}
};
template <typename T>
void func() {
char arr[3];
arr[4] = 0; // no warning
}
llvm-svn: 170180
I wasn't sure where to put the test case for this, but this seemed like as good
a place as any. I had to reorder the tests here to make them legible while
still matching the order of metadata output in the IR file (for some reason
making it virtual changed the ordering).
Relevant commit to fix up LLVM to actually respect 'artificial' member
variables is coming once I write up a test case for it.
llvm-svn: 170154
This is a Band-Aid fix to a false positive, where we complain about not
initializing self to [super init], where self is not coming from the
init method, but is coming from the caller to init.
The proper solution would be to associate the self and it's state with
the enclosing init.
llvm-svn: 170059
This still isn't quite right, but it fixes a crash.
I factored out findCommonParent because we need it on the result of
getImmediateExpansionRange: for a function macro, the beginning
and end of an expansion range can come out of different
macros/macro arguments, which means the resulting range is a complete
mess to handle consistently.
I also made some changes to how findCommonParent works; it works somewhat
better in some cases, and somewhat worse in others, but I think overall
it's a better balance. I'm coming to the conclusion that mapDiagnosticRanges
isn't using the right algorithm, though: chasing the caret is fundamentally
more complicated than any algorithm which only considers one FileID for the
caret can handle because each SourceLocation doesn't really have a single parent.
We need to follow the same path of choosing expansion locations and spelling
locations which the caret used to come up with the correct range
in the general case.
Fixes <rdar://problem/12847524>.
llvm-svn: 170049
has inconsistent ownership with the backing ivar, point the error location to the
ivar.
Pointing to the ivar (instead of the @synthesize) is better since this is where a fix is needed.
Also provide the location of @synthesize via a note.
This also fixes the problem where an auto-synthesized property would emit an error without
any location.
llvm-svn: 170039
My variadics patch, r169588, changed these calls to typically be
bitcasts rather than calls to a supposedly variadic function.
This totally subverted a hack where we intentionally dropped
excess arguments from such calls in order to appease the inliner
and a "warning" from the optimizer. This patch extends the hack
to also work with bitcasts, as well as teaching it to rewrite
invokes.
llvm-svn: 170034
We don't handle array destructors correctly yet, but we now apply the same
hack (explicitly destroy the first element, implicitly invalidate the rest)
for multidimensional arrays that we already use for linear arrays.
<rdar://problem/12858542>
llvm-svn: 170000
call sites as tail calls unconditionally. While it's theoretically true that
this is just an optimization, it's an optimization that we very much want to
happen even at -O0, or else ARC applications become substantially harder to
debug. See r169796 for the llvm/fast-isel side of things.
rdar://12553082
llvm-svn: 169996
Add -fslp-vectorize (with -ftree-slp-vectorize as an alias for gcc compatibility)
to provide a way to enable the basic-block vectorization pass. This uses the same
acronym as gcc, superword-level parallelism (SLP), also common in the literature,
to refer to basic-block vectorization.
Nadav suggested this as a follow-up to the adding of -fvectorize.
llvm-svn: 169909
We don't want to relax all instructions in
$ clang -c test.s
since most users don't pass -O when using the driver to assemble.
On the other hand, -save-temps should not change the output unnecessary, so in
$ clang -c test.c -save-temps
we should relax all instructions.
llvm-svn: 169815
definition, rather than at the end of the definition of the set of nested
classes. We still defer checking of the user-specified exception specification
to the end of the nesting -- we can't check that until we've parsed the
in-class initializers for non-static data members.
llvm-svn: 169805
both LE and BE targets.
AFAICT, Clang get's this correct for PPC64. I've compared it to GCC 4.8
output for PPC64 (thanks Roman!) and to my limited ability to read power
assembly, it looks functionally equivalent. It would be really good to
fill in the assertions on this test case for x86-32, PPC32, ARM, etc.,
but I've reached the limit of my time and energy... Hopefully other
folks can chip in as it would be good to have this in place to test any
subsequent changes.
To those who care about PPC64 performance, a side note: there is some
*obnoxiously* bad code generated for these test cases. It would be worth
someone's time to sit down and teach the PPC backend to pattern match
these IR constructs better. It appears that things like '(shr %foo,
<imm>)' turn into 'rldicl R, R, 64-<imm>, <imm>' or some such. They
don't even get combined with other 'rldicl' instructions *immediately
adjacent*. I'll add a couple of these patterns to the README, but
I think it would be better to look at all the patterns produced by this
and other bitfield access code, and systematically build up a collection
of patterns that efficiently reduce them to the minimal code.
llvm-svn: 169693
This was an egregious bug due to the several iterations of refactorings
that took place. Size no longer meant what it original did by the time
I finished, but this line of code never got updated. Unfortunately we
had essentially zero tests for this in the regression test suite. =[
I've added a PPC64 run over the bitfield test case I've been primarily
using. I'm still looking at adding more tests and making sure this is
the *correct* bitfield access code on PPC64 linux, but it looks pretty
close to me, and it is *worlds* better than before this patch as it no
longer asserts! =] More commits to follow with at least additional tests
and maybe more fixes.
Sorry for the long breakage due to this....
llvm-svn: 169691
array from a braced-init-list. There seems to be a core wording wart
here (it suggests we should be testing whether the elements of the init
list are implicitly convertible to the array element type, not whether
there is an implicit conversion sequence) but our prior behavior appears
to be a bug, not a deliberate effort to implement the standard as written.
llvm-svn: 169690
Linux too, as I think we inherited it from there. The ABI spec says 128-bit,
although I think SGI's compiler on IRIX may be the only thing ever to support
this.
llvm-svn: 169674
the cases where we can't determine whether special members would be trivial
while building the class, we eagerly declare those special members. The impact
of this is bounded, since it does not trigger implicit declarations of special
members in classes which merely *use* those classes.
In order to determine whether we need to apply this rule, we also need to
eagerly declare move operations and destructors in cases where they might be
deleted. If a move operation were supposed to be deleted, it would instead
be suppressed, and we could need overload resolution to determine if we fall
back to a trivial copy operation. If a destructor were implicitly deleted,
it would cause the move constructor of any derived classes to be suppressed.
As discussed on cxx-abi-dev, C++11's selected constructor rules are also
retroactively applied as a defect resolution in C++03 mode, in order to
identify that class B has a non-trivial copy constructor (since it calls
A's constructor template, not A's copy constructor):
struct A { template<typename T> A(T &); };
struct B { mutable A a; };
llvm-svn: 169673
Remove pre-standard restriction on explicitly-defaulted copy constructors with
'incorrect' parameter types, and instead just make those special members
non-trivial as the standard requires.
This required making CXXRecordDecl correctly handle classes which have both a
trivial and a non-trivial special member of the same kind.
This also fixes PR13217 by reimplementing DiagnoseNontrivial in terms of the
new triviality computation technology.
llvm-svn: 169667
directive as a macro expansion.
This is more of a "macro reference" than a macro expansion but it's close enough
for libclang's purposes. If it causes issues we can revisit and introduce a new
kind of cursor.
llvm-svn: 169666
properly, rather than faking it up by pretending that a reference member makes
the default constructor non-trivial. That leads to rejects-valids when putting
such types inside unions.
llvm-svn: 169662
their USR should contain a location.
This uniques them from other declarations with the same name but in different translation units.
rdar://10546541
llvm-svn: 169647
top level.
This heuristic is already turned on for non-ObjC methods
(inlining-mode=noredundancy). If a method has been previously analyzed,
while being inlined inside of another method, do not reanalyze it as top
level.
This commit applies it to ObjCMethods as well. The main caveat here is
that to catch the retain release errors, we are still going to reanalyze
all the ObjC methods but without inlining turned on.
Gives 21% performance increase on one heavy ObjC benchmark, which
suffered large performance regressions due to ObjC inlining.
llvm-svn: 169639
This is the case where the analyzer tries to print out source locations
for code within a synthesized function body, which of course does not have
a valid source location. The previous fix attempted to do this during
diagnostic path pruning, but some diagnostics have pruning disabled, and
so any diagnostic with a path that goes through a synthesized body will
either hit an assertion or emit invalid output.
<rdar://problem/12657843> (again)
llvm-svn: 169631
We were emitting calls to blocks as if all arguments were
required --- i.e. with signature (A,B,C,D,...) rather than
(A,B,...). This patch fixes that and accounts for the
implicit block-context argument as a required argument.
In addition, this patch changes the function type under which
we call unprototyped functions on platforms like x86-64 that
guarantee compatibility of variadic functions with unprototyped
function types; previously we would always call such functions
under the LLVM type T (...)*, but now we will call them under
the type T (A,B,C,D,...)*. This last change should have no
material effect except for making the type conventions more
explicit; it was a side-effect of the most convenient implementation.
llvm-svn: 169588
* Look for i686-linux-android under <sysroot>/lib/gcc.
* This patch also slightly enhance the test suite for
Android GCC toolchain detection.
llvm-svn: 169557
that were already parsed in the same "indexing session".
An indexing session is defined as using the same CXIndexAction object
for multiple clang_indexSourceFile calls.
Passing CXIndexOpt_SkipParsedBodiesInSession as an indexing option will
enable the mode where we try to skip bodies that were already parsed in
another translation unit.
If a function's body was skipped, the "flags" field in the CXIdxDeclInfo
structure will have "CXIdxDeclFlag_Skipped" bit was set.
llvm-svn: 169539
Meador Inge