- krait processor currently modeled with the same features as A9.
- Krait processor additionally has VFP4 (fused multiply add/sub)
and hardware division features enabled.
- krait has currently the same Schedule model as A9
- krait cpu flag is not recognized by the GNU assembler yet,
it is replaced with march=armv7-a to avoid a lower march
from being used.
llvm-svn: 196618
This commit adds the flag '-via-file-asm' to the clang driver. The
purpose of this flag is to have a way to test that clang can consume
the assembly code that it outputs. When passed this flag, clang will
generate a temporary file that contains the assembly output from the
compile step. This assembly file will then be consumed by either the
integrated assembler or the external assembler. To test that the
integrated assembler can consume its own output compile with:
$ clang -integrated-assembler -via-file-asm
Without the '-via-file-asm' flag, clang would directly create the
object file when using the integrated assembler. With the flag it
will first create the temporary assembly file and then read that
file and assemble it with the integrated assembler.
The flow is similar to -save-temps, except that it only effects
the assembly input and the temporary file is not saved.
llvm-svn: 196606
MS-ABI adds padding before *every* vbase if the last field in a record
is a bit-field. This changes clangs behavior to match. I also fix some
windows-style line endings in the test file.
Differential Revision: http://llvm-reviews.chandlerc.com/D2277
llvm-svn: 196605
Adds padding between bases or virtual bases in an attempt to avoid
aliasing of zero-sized sub-objects. The approach used by the ABI adds
two more bits of state. Detailed comments are in the code. Test cases
included.
Differential Revision: http://llvm-reviews.chandlerc.com/D2258
llvm-svn: 196602
This is another regression fixed by reverting r189090.
In this case, the problem is not live variables but the approach that was taken in r189090. This regression was caused by explicitly binding "true" to the condition when we take the true branch. Normally that's okay, but in this case we're planning to reuse that condition as the value of the expression.
llvm-svn: 196599
This reverts commit r189090.
The original patch introduced regressions (see the added live-variables.* tests). The patch depends on the correctness of live variable analyses, which are not computed correctly. I've opened PR18159 to track the proper resolution to this problem.
The patch was a stepping block to r189746. This is why part of the patch reverts temporary destructor tests that started crashing. The temporary destructors feature is disabled by default.
llvm-svn: 196593
Summary:
GCC uses -fauto-profile to enable sample-based PGO. This patch
adds it to Clang as an alias for -fprofile-sample-use.
Differential Revision: http://llvm-reviews.chandlerc.com/D2353
llvm-svn: 196589
In order to make the migration to modules easier, it seems to be helpful
to allow a 1:1 mapping between target names of a current build system
and the corresponding C++ modules. As such targets commonly contain
characters like "-". ":" and "/", allowing arbitrary quote-escaped
strings seems to be a straightforward option.
After several offline discussions, the precise mechanisms for C++
module names especially regarding submodules and import statements has
yet to be determined. Thus, this patch only enables string literals as
names inside the module map files which can be used by automatic module
import (through #include).
Also improve the error message on missing use-declarations.
llvm-svn: 196573
__declspec(align())
This patch implements required alignment in a way that makes
__declspec(align()) and #pragma pack play correctly together. In the
MS-ABI, __declspec(align()) is a hard rule and cannot be overridden by
#pragma pack. This cases each record to have two interesting alignments
"preferred alignment" (which matches Itanium's concept of alignment) and
"required alignment" which is an alignment that must never be violated,
even in the case of #pragma pack. This patch introduces the concept of
Required Alignment to the record builder and tracks/uses it
appropriately. Test cases are included.
Differential Revision: http://llvm-reviews.chandlerc.com/D2283
llvm-svn: 196549
This commit changes -Wassign-enum to compare unqualified types. One could
think that this does not matter much, because who wants a value of enum type
that is const-qualified? But this breaks the intended pattern to silence this
warning with an explicit cast:
static const enum Foo z = (enum Foo) 42;
In this case, source type is 'enum Foo', and destination type is 'const enum
Foo', and if we compare qualified types, they don't match, so we used warn.
llvm-svn: 196548
the following pattern.
If 'case' expression refers to a static const variable of the correct enum
type, then we count this as a sufficient declaration of intent by the user,
so we silence the warning.
llvm-svn: 196546
category is declared in category's primary
class's super class. Because the super class is
expected to implemented the method. // rdar://15580969
llvm-svn: 196531
I happened to notice this while trying to write a test for an iOS simulator
target. I suspect we just missed this when we added separate "macosx" and "ios"
triples instead of the generic "darwin" OS.
llvm-svn: 196527
within their namespace, and such a redeclaration isn't required to be a
definition any more.
Update DR status page to say Clang 3.4 instead of SVN and add new Clang 3.5
category (but keep Clang 3.4 yellow for now).
llvm-svn: 196481
don't assume that it inherits the designated initializers from the super class.
If the assumption was wrong because a new initializer was a designated one that was not marked as such,
we will emit misleading warnings for subclasses of the interface.
llvm-svn: 196476
For an init capture, process the initialization expression
right away. For lambda init-captures such as the following:
const int x = 10;
auto L = [i = x+1](int a) {
return [j = x+2,
&k = x](char b) { };
};
keep in mind that each lambda init-capture has to have:
- its initialization expression executed in the context
of the enclosing/parent decl-context.
- but the variable itself has to be 'injected' into the
decl-context of its lambda's call-operator (which has
not yet been created).
Each init-expression is a full-expression that has to get
Sema-analyzed (for capturing etc.) before its lambda's
call-operator's decl-context, scope & scopeinfo are pushed on their
respective stacks. Thus if any variable is odr-used in the init-capture
it will correctly get captured in the enclosing lambda, if one exists.
The init-variables above are created later once the lambdascope and
call-operators decl-context is pushed onto its respective stack.
Since the lambda init-capture's initializer expression occurs in the
context of the enclosing function or lambda, therefore we can not wait
till a lambda scope has been pushed on before deciding whether the
variable needs to be captured. We also need to process all
lvalue-to-rvalue conversions and discarded-value conversions,
so that we can avoid capturing certain constant variables.
For e.g.,
void test() {
const int x = 10;
auto L = [&z = x](char a) { <-- don't capture by the current lambda
return [y = x](int i) { <-- don't capture by enclosing lambda
return y;
}
};
If x was not const, the second use would require 'L' to capture, and
that would be an error.
Make sure TranformLambdaExpr is also aware of this.
Patch approved by Richard (Thanks!!)
http://llvm-reviews.chandlerc.com/D2092
llvm-svn: 196454
We would skip until the next comma, hoping good things whould lie there,
however this would fail when we have such things as this:
struct A {};
template <typename>
struct D;
template <>
struct D<C> : B, A::D;
Once this happens, we would believe that D with a nested namespace
specifier of A was a variable that was being declared. We would go on
to complain that there was an extraneous 'template <>' on their variable
declaration.
Crashes would happen when 'A' gets defined as 'enum class A {}' as
various asserts would fire.
Instead, we should skip up until the semicolon if we see that we are in
the middle of a definition and the current token is a ':'
This fixes PR17084.
llvm-svn: 196453
Summary:
In general, this type node can be used to represent any type adjustment
that occurs implicitly without losing type sugar. The immediate use of
this is to adjust the calling conventions of member function pointer
types without breaking template instantiation.
Fixes PR17996.
Reviewers: rsmith
Differential Revision: http://llvm-reviews.chandlerc.com/D2332
llvm-svn: 196451
nested-name-specifier, rather than crashing. (In fact, reject all
literal-operator-ids that have a non-namespace nested-name-specifier). The
grammar doesn't allow these in some cases, and in other cases does allow them
but instantiation will always fail.
llvm-svn: 196443
Clang currently croaks on the following:
struct X1 {
struct X2 {
int L = ([] (int i) { return i; })(2);
};
};
asserting that the containing lexical context of the lambda is not Sema's cur context, when pushing the lambda's decl context on.
This occurs because (prior to this patch) getContainingDC always returns the non-nested class for functions at class scope (even for inline member functions of nested classes (to account for delayed parsing of their bodies)). The patch addresses this by having getContainingDC always return the lexical DC for a lambda's call operator.
Link to the bug: http://llvm.org/bugs/show_bug.cgi?id=18052
Link to Richard Smith's feedback on phabricator: http://llvm-reviews.chandlerc.com/D2331
Thanks!
llvm-svn: 196423
which specifies couple of (optional) method selectors
for bridging a CFobject to or from an ObjectiveC
object. This is wip. // rdsr://15499111
llvm-svn: 196408
Summary:
MSVC destroys arguments in the callee from left to right. Because C++
objects have to be destroyed in the reverse order of construction, Clang
has to construct arguments from right to left and destroy arguments from
left to right.
This patch fixes the ordering by reversing the order of evaluation of
all call arguments under the MS C++ ABI.
Fixes PR18035.
Reviewers: rsmith
Differential Revision: http://llvm-reviews.chandlerc.com/D2275
llvm-svn: 196402
I'd misunderstood getIndirect() to mean that the argument should be passed
as a pointer at the ABI level, with the ByVal argument choosing caller-copy
semantics over no-caller-copy (callee-copy-on-write) semantics. But
getIndirect(x) actually means that x is passed by pointer at the IR
level but (at least on all other targets I looked at) directly at the
ABI level. getIndirect(x, false) selects a pointer to a caller-made
copy, which is what SystemZ was aiming for.
This fixes a miscompilation of c-index-test. Structure arguments were being
passed by pointer, but no copy was being made, so a write in the callee
stomped over a caller's local variable.
llvm-svn: 196370
We would lose track of the mangling number assigned to the original
declaration which would cause us to create manglings that didn't match
the Itanium C++ specification.
e.g. Two static fields with the same name inside of a function template
would receive the same mangling with LLVM fixing up the second field so
they wouldn't collide. This would create an incompatibility with other
compilers following the Itanium ABI.
I've confirmed that the new mangling is identical to the ones generated
by icc and gcc.
N.B. This was uncovered while working on Microsoft mangler.
llvm-svn: 196368
In delayed template parsing mode, adjust the template depth counter for each template parameter list associated with an out of line member template specialization.
llvm-svn: 196351
In trunk, we can use features as below:
aarch64-registered-target
hexagon-registered-target
msp430-registered-target
r600-registered-target
systemz-registered-target
xcore-registered-target
Each of them, as below, implies corresponding subtargets:
arm-registered-target -- arm, thumb
mips-registered-target -- mips, mips64, mips64el, mipsel
nvptx-registered-target -- nvptx, nvptx64
sparc-registered-target -- sparc, sparcv9
x86-registered-target -- x86, x86-64
They will be renamed:
cppbackend-registered-target -- was "cpp". Unused in trunk.
powerpc-registered-target -- was "ppc32", "ppc64" and "ppc64le".
The feature "asserts" is also taken from llvm-config.
llvm-svn: 196347
super another initializer and when the implementation does not delegate to
another initializer via a call on 'self'.
A secondary initializer is an initializer method not marked as a designated
initializer within a class that has at least one initializer marked as a
designated initializer.
llvm-svn: 196318
Before, there SourceManager would not return a FileEntry for a
SourceLocation of a macro expansion (if the header name itself is
defined in a macro). We'd then fallback to assume that the module
currently being built is the including module. However, in this case we
are actually interested in the spelling location of the filename loc in
order to derive the including module.
llvm-svn: 196311
clang converts keywords to identifiers for compatibility with various system
headers such as GNU libc.
Implement a -Wkeyword-compat extension warning to diagnose those cases. The
warning is on by default but will generally be ignored in system headers. It
can however be enabled globally to aid standards conformance testing.
This also changes the __uptr keyword avoidance from r195710 to no longer
special-case system headers, bringing it in line with other similar workarounds
in clang.
Implementation returns bool for symmetry with token annotation functions.
Some examples:
warning: keyword '__is_pod' will be treated as an identifier for the remainder of the translation unit [-Wkeyword-compat]
struct __is_pod
warning: keyword '__uptr' will be treated as an identifier here [-Wkeyword-compat]
union w *__uptr;
llvm-svn: 196212
This is a duplicate implementation.
E.g. this patch defines:
float64_t vabd_f64(float64_t a, float64_t b)
But there is already a similar intrinsic "vabdd_f64" with the same types.
Also, this intrinsic will be conflicted to the vector type intrinsic as following(Which is implemented by me and will be committed to trunk):
float64x1_t vabd_f64(float64x1_t a, float64x1_t b).
Two functions shouldn't have a same name in arm_neon.h.
According to ARM ACLE document, such vabd_f64 with float64_t is not existing.
So I revert this commit.
llvm-svn: 196205
lookup, if parsing failed, we did not restore the lexer state properly, and
eventually crashed. This change ensures that we always consume all the tokens
from the new token stream we started to parse the name from inline asm.
llvm-svn: 196182
gcc treats [[gnu:const]], [[gnu::__const]], and [[gnu:__const__]] as all being
equivalent. Add an additional test case to ensure that we do not miss the last
case.
llvm-svn: 195982
'not' on Windows tries and fails to convert the argument to UTF-16 and back for
some reason:
Error: Unable to convert command-line to UTF-16
Let's try replacing it with !.
(There was no obvious way to do these with FileCheck)
llvm-svn: 195907
We wouldn't properly save and restore the pending local instantiations
we had built up prior to instantiation of a variable definition. This
would lead to us instantiating too much causing crashes and other
general badness.
This fixes PR14374.
llvm-svn: 195887
1) Use %clang_cc1 instead of the driver
2) Validate that the input contains a BOM
3) Validate that the BOM has been stripped from the output
llvm-svn: 195886
We would fail to instantiate them when the surrounding function was
instantiated. Instantiate the class and add it's members to the list of
pending instantiations, they should be resolved when we are finished
with the function's body.
This fixes PR9685.
llvm-svn: 195827
The warning from cmake-clang-x64-msc16-R was:
test\SemaCXX\old-style-cast.cpp Line 6: cast to 'void **' from smaller integer type 'long'
llvm-svn: 195813
code for handling triviality, deletedness and constexpr. Fix a few bugs in
these, particularly related to mutable members, and remove some dead code.
llvm-svn: 195809
GNU libc uses '__uptr' as a member name in C mode, conflicting with the
eponymous MSVC pointer modifier keyword.
Detect and mark the token as an identifier when these specific conditions are
met. __uptr will continue to work as a keyword for the remainder of the
translation unit.
Fixes PR17824.
llvm-svn: 195710
The integrated assembler was already the default for win32. It is now able
to handle a clang bootstrap on mingw, so make it the default.
llvm-svn: 195676
It wasn't possible for an anonymous type to show up inside of function arguments.
However, decltype (which MSVC added support for in 2010) makes this
possible. Further, backrefs to these anonymous types can now be formed.
This fixes PR18022.
N.B. We do not, and very likely _will not_, support MSVC's bug where
subsequent typedefs of anonymous types leak into the linkage name; this
is a gross violation of the ABI. A warning should be introduced to
inform our users of this particular shortcoming.
llvm-svn: 195669
MSVC applies these to the following declaration only if present, otherwise
silently ignores them whereas we'll issue a warning.
Handling differs from ordinary attributes appearing in the same place, so add a
Sema test to make sure we get it right.
llvm-svn: 195577
Clang knows how to use the gnu assembler directly from doing so on linux and
hurd. The existing support worked out of the box on cygwin and mingw and I was
able to bootstrap clang with it in both systems (with pending patches for the
new mingw abi, but that is independent of the assembler).
llvm-svn: 195554
module. Use the marker to diagnose cases where we try to transition between
submodules when not at the top level (most likely because a closing brace was
missing at the end of a header file, but is also possible if submodule headers
attempt to do something fundamentally non-modular, like our .def files).
llvm-svn: 195543
This is still an experimental attribute, but I wanted it in tree
for review. It may still get yanked.
This attribute can only be applied to a class @interface, not
a class extension or category. It does not change the type
system rules for Objective-C, but rather the implementation checking
for Objective-C classes that explicitly conform to a protocol.
During protocol conformance checking, clang recursively searches
up the class hierarchy for the set of methods that compose
a protocol. This attribute will cause the compiler to not consider
the methods contributed by a super class, its categories, and those
from its ancestor classes. Thus this attribute is used to force
subclasses to redeclare (and hopefully re-implement) methods if
they decide to explicitly conform to a protocol where some of those
methods may be provided by a super class.
This attribute intentionally leaves out properties, which are associated
with state. This attribute only considers methods (at least right now)
that are non-property accessors. These represent methods that "do something"
as dictated by the protocol. This may be further refined, and this
should be considered a WIP until documentation gets written or this
gets removed.
llvm-svn: 195533
attribute on method declaration and implementation
match. This makes no sense. Most annotations are
meant for declarations only and one is for implementation.
This has been constant source of regresions and hackery to
get around special cases. I am removing this check.
Such checks must be done on a case by case basis and
when it makes sense. For example, it makes sense
for availability/deprecated and I will file a radar
for that. // rdar://15531984
llvm-svn: 195524
available always-inline functions. This breaks libc++'s locale
implementation. Code generation for this case should be fixed, but this
is a stop gap fix for clang 3.4.
llvm-svn: 195501
Not long ago I made the CodeGen of for loops simplify the condition at
-O0 in the same way we do for if and conditionals. Unfortunately this
ties how loops and simple conditions work together too tightly, which
makes features such as instrumentation based PGO awkward.
Ultimately, we should find a more general way to simplify the logic in
a given condition, but for now we'll just avoid using EmitBranchOnBool
for loops, like we already do for while and do loops.
llvm-svn: 195438
Diags aren't usually in the first person, and 'windows' isn't the correct
product spelling to use in prose. Sidestep issues completely by making this
error message platform-neutral.
llvm-svn: 195422
whose semantic is currently identical to objc_bridge,
but their differences may manifest down the road with
further enhancements. // rdar://15498044
llvm-svn: 195376
In OpenCL a vector of 3 elements, acts like a vector of four elements.
So for a vector of size 3 the '.hi' and '.odd' accessors, would access
the elements {2, 3} and {1, 3} respectively.
However, in EmitStoreThroughExtVectorComponentLValue we are still operating on
a vector of size 3, so we should only access {2} and {1}. We do this by checking
the last element to be accessed, and ignore it if it is out-of-bounds.
EmitLoadOfExtVectorElementLValue doesn't have a similar problem, because it does
a direct shufflevector with undef, so an out-of-bounds access just gives an undef
value.
Patch by Anastasia Stulova!
llvm-svn: 195367
After implementing this patch, a few concerns about the language
feature itself emerged in my head that I had previously not considered.
I want to resolve those design concerns first before having
a half-designed language feature in the tree.
llvm-svn: 195328
The idea is to allow a class to stipulate that its methods (and those
of its parents) cannot be used for protocol conformance in a subclass.
A subclass is then explicitly required to re-implement those methods
of they are present in the class marked with this attribute.
Currently the attribute can only be applied to an @interface, and
not a category or class extension. This is by design. Unlike
protocol conformance, where a category can add explicit conformance
of a protocol to class, this anti-conformance really needs to be
observed uniformly by all clients of the class. That's because
the absence of the attribute implies more permissive checking of
protocol conformance.
This unfortunately required changing method lookup in ObjCInterfaceDecl
to take an optional protocol parameter. This should not slow down
method lookup in most cases, and is just used for protocol conformance.
llvm-svn: 195323
data member definitions when the variable has an initializer
in its declaration.
For the following code:
struct S {
static const int x = 42;
};
const int S::x = 42;
This patch changes the diagnostic from:
a.cc:4:14: error: redefinition of 'x'
const int S::x = 42;
^
a.cc:2:20: note: previous definition is here
static const int x = 42;
^
to:
a.cc:4:18: error: static data member 'x' already has an initializer
const int S::x = 42;
^
a.cc:2:24: note: previous initialization is here
static const int x = 42;
^
Differential Revision: http://llvm-reviews.chandlerc.com/D2235
llvm-svn: 195306
This makes Clang emit a linkonce_odr definition for 'val' in the code below,
to be compatible with MSVC-compiled code:
struct Foo {
static const int val = 1;
};
Differential Revision: http://llvm-reviews.chandlerc.com/D2233
llvm-svn: 195283
We are still using Dwarf Version 2 for Darwin systems, make it consistent
with -gline-tables-only.
This should fix an internal buildbot.
llvm-svn: 195267
- If a deprecated class refers to another deprecated class, do not warn.
- @implementations of a deprecated class can refer to other deprecated things.
Fixes <rdar://problem/15407366> and <rdar://problem/15466783>.
llvm-svn: 195259
Summary:
RTTI is not yet implemented for the Microsoft C++ ABI and isn't expected
soon. We could easily add the mangling, but the error is what prevents
us from silently miscompiling code that expects RTTI.
Instead, add a new mangleTypeName entry point that simply forwards to
mangleName or mangleType to produce a string that isn't part of the ABI.
Itanium can continue to use RTTI names to avoid unecessary test
breakage.
This also seems like the right design. The fact that TBAA names happen
to be RTTI names is now an implementation detail of the mangler, rather
than part of TBAA.
Differential Revision: http://llvm-reviews.chandlerc.com/D2153
llvm-svn: 195168
and we see an ill-formed declarator that would probably be well-formed if the
tag definition were just missing a semicolon, use that as the diagnostic
instead of producing some other mysterious error.
llvm-svn: 195163
Microsoft adds an extra byte of padding before laying out zero sized
non-virtual bases if the non-virtual base before it contains a vbptr.
This patch adds the same behavior to clang.
Differential Revision: http://llvm-reviews.chandlerc.com/D2106
llvm-svn: 195158
The previous patches tried to deduce the correct function type. I now realize
this is not possible in general. Consider
class foo {
template <typename T> static void bar(T v);
};
extern template void foo::bar(const void *);
We will only know that bar is static after a lookup, so we have to handle this
in the template instantiation code.
This patch reverts my previous two changes (but not the tests) and instead
handles the issue in DeduceTemplateArguments.
llvm-svn: 195154
should be isolated in the backend (r195123). From the frontend point
of view in case of "-mhard-float -mips16" combination of flags the float
ABI mode should remain unchanged.
The patch reviewed by Reed Kotler.
llvm-svn: 195124
This is failing for me. When I run the command on my own, I get this:
Error reading /usr/local/google/home/morbo/llvm/llvm.obj/tools/clang/test/Format/Output/.clang-format: Invalid argument
void f() {
int* i;
int j;
}
The formatting is like this because I have the Google format version in my
~/.clang-format file. This test should be made independent of that.
llvm-svn: 195080
the -Q flag to the as(1) assembler driver.
We will soon be switching the darwin as(1) assembler driver to call clang(1)
and use the intergated assembler by default. To do this and still support
clang(1)'s -no-integrated-as flag, when clang(1) runs the as(1) assembler
driver and -no-integrated-as is used it needs to pass the -Q flag to as(1)
so it uses its GNU based assembler, and not turn around and call clag(1)'s
integrated assembler.
rdar://15495921
llvm-svn: 195054
logic was not handling typedefs as free functions. This was not
causing problems with the existing tests, but does with the microsoft
abi where they have to get a different calling convention.
I will try to refactor this into a method on Declarator in a second.
llvm-svn: 195050
After r195009, the test would write a .o file to the test dir. Send that to
/dev/null instead. Also fix the typo in test/Frontend/invalid-o-level.c.
llvm-svn: 195047
Before this patch explicit template instatiations of member function templates
were failing with the microsoft abi and 32 bits. This was happening because
the expected and computed function types had different calling conventions.
This patch fixes it by considering the default calling convention in
GetFullTypeForDeclarator.
This fixes pr17973.
llvm-svn: 195032
Summary:
Currently with clang:
$ clang -O20 foo.c
error: invalid value '20' in '-O20'
With the patch:
$ clang -O20 foo.c
warning: optimization level '-O20' is unsupported; using '-O3' instead.
1 warning generated.
This matches the gcc behavior (with a warning added)
Pass all tests:
Testing: 0 .. 10.. 20.. 30.. 40.. 50.. 60.. 70.. 80.. 90..
Testing Time: 94.14s
Expected Passes : 6721
Expected Failures : 20
Unsupported Tests : 17
(which was not the case of http://llvm-reviews.chandlerc.com/D2125)
Differential Revision: http://llvm-reviews.chandlerc.com/D2212
llvm-svn: 195009
Line 559: 'long long' is a C++11 extension
Line 566: 'long long' is a C++11 extension
Line 674: 'long long' is a C++11 extension
See also PR13819.
llvm-svn: 195005
This adds -freroll-loops (and -fno-reroll-loops in the usual way) to enable
loop rerolling as part of the optimization pass manager. This transformation
can enable vectorization, reduce code size (or both).
Briefly, loop rerolling can transform a loop like this:
for (int i = 0; i < 3200; i += 5) {
a[i] += alpha * b[i];
a[i + 1] += alpha * b[i + 1];
a[i + 2] += alpha * b[i + 2];
a[i + 3] += alpha * b[i + 3];
a[i + 4] += alpha * b[i + 4];
}
into this:
for (int i = 0; i < 3200; ++i) {
a[i] += alpha * b[i];
}
Loop rerolling is currently disabled by default at all optimization levels.
llvm-svn: 194967
New rules of invalidation/escape of the source buffer of memcpy: the source buffer contents is invalidated and escape while the source buffer region itself is neither invalidated, nor escape.
In the current modeling of memcpy the information about allocation state of regions, accessible through the source buffer, is not copied to the destination buffer and we can not track the allocation state of those regions anymore. So we invalidate/escape the source buffer indirect regions in anticipation of their being invalidated for real later. This eliminates false-positive leaks reported by the unix.Malloc and alpha.cplusplus.NewDeleteLeaks checkers for the cases like
char *f() {
void *x = malloc(47);
char *a;
memcpy(&a, &x, sizeof a);
return a;
}
llvm-svn: 194953
clang -cc1 skips the driver so it never made sense to include these with the
Driver tests.
Basic type tests and flag tests generally both go in Frontend.
Now that the final -cc1 tests have been moved out of test/Driver, add a
local substitution to enforce and detect future mistakes.
These miscategorized tests were probably the source of confusion in r194817.
llvm-svn: 194919
Teach the '-arch' command line option to enable the compiler-friendly
features of core-avx2 CPUs on Darwin. Pass the information along in the
target triple like Darwin+ARM does.
llvm-svn: 194907
Earlier versions discarded the state too soon, and did not track state changes,
e.g. when passing a temporary to a move constructor. Patch by
chris.wailes@gmail.com; review and minor fixes by delesley.
llvm-svn: 194900
the GNU documentation: the attribute only appertains to the label if it is
followed by a semicolon. Based on a patch by Aaron Ballman!
llvm-svn: 194869
Trying to fix test failures since earlier today.
One of the tests added in this commit is outputting test/Driver/clang_f_opts.s
which the builders that build in-tree (eg. clang-native-arm-cortex-a9) are
trying to run as a test case, causing failures.
clang_f_opts.c:
If -### doesn't emit the warning then this test probably shouldn't be in
here in the first place. Frontend maybe?
invalid-o-level.c:
Running %clang_cc1 in the Driver tests doesn't make sense because -cc1
bypasses the driver. (I'm not reverting the commit that introduced this but
please fix instead of keeping it this way.)
Reverting to fix the build failures and also so that the tests can be thought
out more thoroughly.
This reverts commit r194817.
llvm-svn: 194845
Even if we don't support a flag, we should be able to parse it
to provide a better error message than the current default
"error: no such file or directory: '/foo'" (which we should probably
also tweak, btw).
This also tries to clean up the test file a bit.
llvm-svn: 194837
Instead of storing the vtable offset directly in the function pointer and
doing a branch to check for virtualness at each call site, the MS ABI
generates a thunk for calling the function at a specific vtable offset,
and puts that in the function pointer.
This patch adds support for emitting such thunks. However, it doesn't support
pointers to virtual member functions that are variadic, have an incomplete
aggregate return type or parameter, or are overriding a function in a virtual
base class.
Differential Revision: http://llvm-reviews.chandlerc.com/D2104
llvm-svn: 194827
By adding a default config.excludes pattern we can avoid individual
suppressions in subdirectories.
This matches LLVM's lit.cfg which also excludes a few other common non-test
filenames for consistency.
llvm-svn: 194814
representing the module import rather than making the module immediately
visible. This serves two goals:
* It avoids making declarations in the module visible prematurely, if we
walk past the #include during a tentative parse, for instance, and
* It gives a diagnostic (although, admittedly, not a very nice one) if
a header with a corresponding module is included anywhere other than
at the top level.
llvm-svn: 194782
Ana Pazos