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
Previously we made three passes over the set of dead symbols, and removed
them from the state /twice/. Now we combine the autorelease pass and the
symbol death pass, and only have to remove the bindings for the symbols
that leaked.
llvm-svn: 169527
ToolChains.cpp
This is in anticipation of forthcoming library path changes.
Also ...
- Fixes some inconsistencies in how the arch is passed to tools.
- Add test cases for various forms of arch flags
llvm-svn: 169505
paths
- Inherit from Linux rather than ToolChain
- Override AddClangSystemIncludeArgs and AddClangCXXStdlibIncludeArgs
to properly set include paths.
llvm-svn: 169495
generally support the C++11 memory model requirements for bitfield
accesses by relying more heavily on LLVM's memory model.
The primary change this introduces is to move from a manually aligned
and strided access pattern across the bits of the bitfield to a much
simpler lump access of all bits in the bitfield followed by math to
extract the bits relevant for the particular field.
This simplifies the code significantly, but relies on LLVM to
intelligently lowering these integers.
I have tested LLVM's lowering both synthetically and in benchmarks. The
lowering appears to be functional, and there are no really significant
performance regressions. Different code patterns accessing bitfields
will vary in how this impacts them. The only real regressions I'm seeing
are a few patterns where the LLVM code generation for loads that feed
directly into a mask operation don't take advantage of the x86 ability
to do a smaller load and a cheap zero-extension. This doesn't regress
any benchmark in the nightly test suite on my box past the noise
threshold, but my box is quite noisy. I'll be watching the LNT numbers,
and will look into further improvements to the LLVM lowering as needed.
llvm-svn: 169489
We still need to do a recursive walk to determine all static/global variables
referenced by a block, which is needed for region invalidation.
llvm-svn: 169481
Don't require that, during template deduction, a template specialization type
as a function parameter has at least as many template arguments as one used in
a function argument (not even if the argument has been resolved to an exact
type); the additional parameters might be provided by default template
arguments in the template. We don't need this check, since we now implement
[temp.deduct.call]p4 with an additional check after deduction.
llvm-svn: 169475
RUN: a
RUN: b || true
lit expands it to a && b || true, and the || true applies to both commands (thus ignoring failures in 'a')! This is PR10867 again.
llvm-svn: 169434
For most cases where a conversion specifier doesn't match an argument,
we usually guess that the conversion specifier is wrong. However, if
the argument is an integer type and the specifier is %C, it's likely
the user really did mean to print the integer as a character.
(This is more common than %c because there is no way to specify a unichar
literal -- you have to write an integer literal, such as '0x2603',
and then cast it to unichar.)
This does not change the behavior of %S, since there are fewer cases
where printing a literal Unicode *string* is necessary, but this could
easily be changed in the future.
<rdar://problem/11982013>
llvm-svn: 169400
The type of a character literal is 'int' in C, but if the user writes a
character /as/ a literal, we should assume they meant it to be a
character and not a numeric value, and thus offer %c as a correction
rather than %d.
There's a special case for multi-character literals (like 'MooV'), which
have implementation-defined value and usually cannot be printed with %c.
These still use %d as the suggestion.
In C++, the type of a character literal is 'char', and so this problem
doesn't exist.
<rdar://problem/12282316>
llvm-svn: 169398
We tried to account for 'uint8_t' by saying that /typedefs/ of 'char'
should be corrected as %hhd rather than %c, but the condition was wrong.
llvm-svn: 169397
Our error recovery path may have made the class anonymous, and that has a pretty
disastrous impact on any attempt to parse a class body containing constructors.
llvm-svn: 169374
This ensures that even though it comes first, we pick up its .o files.
Note that if we can use this (or something similar / equivalent) on
other platforms, we could potentially remove
ReplaceOperatorsNewAndDelete from the ASan runtimes.
We should probably do something similar for TSan and MSan as well.
llvm-svn: 169328
The count attribute is more accurate with regards to the size of an array. It
also obviates the upper bound attribute in the subrange. We can also better
handle an unbound array by setting the count to -1 instead of the lower bound to
1 and upper bound to 0.
llvm-svn: 169311
the link command. This all works fine when the driver is also responsible for
adding -lstdc++ to the link command. But, if -lstdc++ (or libstdc++.a, etc) is
passed explicitly to the driver, the ASan runtime will appear in the link
command after the standard library, leading to multiple-definition errors for
the global 'operator new' and 'operator delete'. Fix this in a painfully
simple way, by inserting libclang_rt.asan.a at the start of the link command
instead of the end.
If we need to do something more clever, we can walk the link command looking
for something that resembles libstdc++ and insert libclang_rt.asan.a as late
as possible, but the simple solution works for now.
llvm-svn: 169310
The count field is necessary because there isn't a difference between the 'lo'
and 'hi' attributes for a one-element array and a zero-element array. When the
count is '0', we know that this is a zero-element array. When it's >=1, then
it's a normal constant sized array. When it's -1, then the array is unbounded.
llvm-svn: 169219
<declaration> tag of Comment XML and fixed a
missing declaration of ivars private to @implementation
as result of the testing. // rdar://12378714
llvm-svn: 169193
Fixes a crash printing diagnostics on the gcc testsuite, and also makes
diagnostic range printing print nicer results for token pastes.
llvm-svn: 169068
state so that all of the various clones end up rendering their
diagnostics into the same serialized-diagnostics file. This is
important when we actually want failures during module build to be
reported back to the translation unit that tried to import the
not-yet-built or out-of-date module. <rdar://problem/12565727>
llvm-svn: 169057
the output size is greater than the register size. No truncation occurs with
those. Reword warning to make it clearer what's the problem is.
llvm-svn: 169054
module, provide a module import stack similar to what we would get for
an include stack, e.g.,
In module 'DependsOnModule' imported from build-fail-notes.m:4:
In module 'Module' imported from DependsOnModule.framework/Headers/DependsOnModule.h:1:
Inputs/Module.framework/Headers/Module.h:15:12: note: previous definition is here
@interface Module
<rdar://problem/12696425>
llvm-svn: 169042
building module 'Foo' imported from..." notes (the same we we provide
"In file included from..." notes) in the diagnostic, so that we know
how this module got included in the first place. This is part of
<rdar://problem/12696425>.
llvm-svn: 169021
import of that module elsewhere, don't try to build the module again:
it won't work, and the experience is quite dreadful. We track this
information somewhat globally, shared among all of the related
CompilerInvocations used to build modules on-the-fly, so that a
particular Clang instance will only try to build a given module once.
Fixes <rdar://problem/12552849>.
llvm-svn: 168961
1) init-order sanitizer: initialization-order checker.
Status: usable, but may produce false positives w/o proper blacklisting.
2) use-after-return sanitizer
Status: implemented, but heavily understed.
Should be optional, as it significanlty slows program down.
3) use-after-scope sanitizer
Status: in progress.
llvm-svn: 168950
Among other differences, GCC accepts
typedef int IA[];
typedef int A10[10];
static A10 *f(void);
static IA *f(void);
void g(void) {
(void)sizeof(*f());
}
but clang used to reject it with:
invalid application of 'sizeof' to an incomplete type 'IA' (aka 'int []')
The intention of c99's 6.2.7 seems to be that we should use the composite type
and accept as gcc does.
Doing the type merging required some extra fixes:
* Use the type from the function type in initializations, even if an parameter
is available.
* Fix the merging of the noreturn attribute in function types.
* Make CodeGen handle the fact that an parameter type can be different from
the corresponding type in the function type.
llvm-svn: 168895
performed, to determine whether that special member is deleted or constexpr.
That overload resolution process can in turn trigger the instantiation of a
template, which can do anything, including triggering the declaration of that
very same special member function. When this happens, do not try to recursively
declare the special member -- that's impossible. Instead, only try to realise
the truth. There is no special member.
llvm-svn: 168847
in deciding a copy/dispose field is needed in a byref structure
and when generating the copy/dispose helpers. In certain
cases, these fields were being added but no copy/dispose was
being generated. This was uncovered in ARC, but not in MRR.
// rdar://12759433
llvm-svn: 168825
the original parameter or return type.
Since we do not accurately represent the data fields of a union, we should not
directly load or store a union type.
As an exmple, if we have i8,i8, i32, i32 as one field type and i32,i32 as
another field type, the first field type will be chosen to represent the union.
If we load with the union's type, the 3rd byte and the 4th byte will be skipped.
rdar://12723368
llvm-svn: 168820
determine which member function would be the callee from within the template
definition, don't pass that function as a "non-member function" to
CreateOverloadedBinOp. Instead, just rely on it to select the member function
for itself.
llvm-svn: 168818
string literal needs cleaning (because it contains line-splicing in the
encoding prefix or in the ud-suffix), do not clean the section between the
double-quotes -- that's the "raw" bit!
llvm-svn: 168776
constructor/assignment operator with a const-qualified parameter type. The
prior method for determining this incorrectly used overload resolution.
llvm-svn: 168775
a special member" diagnostic from warning to error, and fix the cases where it
produced diagnostics with incorrect wording.
We don't support this as an extension, and we ban it even in C++98 mode. This
breaks too much (for instance, the ABI-specified calling convention for a type
can change if it acquires a copy constructor through the addition of a default
argument).
llvm-svn: 168769
The stop-gap here is to just drop such objects when processing the InitListExpr.
We still need a better solution.
Fixes <rdar://problem/12755044>.
llvm-svn: 168757
This was also covered by <rdar://problem/12753384>. The static analyzer
evaluates a CXXConstructExpr within an initializer expression and
RegionStore doesn't know how to handle the resulting CXXTempObjectRegion
that gets created. We need a better solution than just dropping the
value, but we need to better understand how to implement the right
semantics here.
Thanks to Jordan for his help diagnosing the behavior here.
llvm-svn: 168741
objc_loadWeak. This retains and autorelease the weakly-refereced
object. This hidden autorelease sometimes makes __weak variable alive even
after the weak reference is erased, because the object is still referenced
by an autorelease pool. This patch overcomes this behavior by loading a
weak object via call to objc_loadWeakRetained(), followng it by objc_release
at appropriate place, thereby removing the hidden autorelease. // rdar://10849570
llvm-svn: 168740
Douglas Gregor