Tim Northover [Wed, 2 Jul 2014 12:56:02 +0000 (12:56 +0000)]
ARM: add support for v8 ldaex/stlex builtins.
ARMv8 adds (to both AArch32 and AArch64) acquiring and releasing
variants of the exclusive operations, in line with the C++11 memory
model.
This adds support for two new intrinsics to expose them to C & C++
developers directly: __builtin_arm_ldaex and __builtin_arm_stlex, in
direct analogy with the versions with no implicit barrier.
Tim Northover [Wed, 2 Jul 2014 10:25:45 +0000 (10:25 +0000)]
X86: inline all atomic operations up to 128-bits.
The backend *can* cope with all of these now, so Clang should give it the
chance. On CPUs without cmpxchg16b (e.g. the original athlon64) LLVM can reform
the libcalls.
Richard Trieu [Wed, 2 Jul 2014 04:39:38 +0000 (04:39 +0000)]
Prevent Clang from crashing on template code.
Fixes PR20110, where Clang hits an assertion failure when it expects that the
sub-expression of a bit cast to pointer to also be a pointer, but gets a value
instead.
David Majnemer [Tue, 1 Jul 2014 22:24:56 +0000 (22:24 +0000)]
Driver: Handle /GR- in a compatible way with MSVC
There are slight differences between /GR- and -fno-rtti which made
mapping one to the other inappropriate.
-fno-rtti disables dynamic_cast, typeid, and does not emit RTTI related
information for the v-table.
/GR- does not generate complete object locators and thus will not
reference them in vftables. However, constructs like dynamic_cast and
typeid are permitted.
This should bring our implementation of RTTI up to semantic parity with
MSVC modulo bugs.
Win64: Disable uwtable until we fix .pdata emission in LLVM
LLVM r211399 started emitting .pdata for win64 by default.
Unfortunately, it produces invalid object files. I plan to fix that
Soon. For now, don't request unwind tables. This fixes the clang-cl
self-host on win64.
David Majnemer [Tue, 1 Jul 2014 20:30:31 +0000 (20:30 +0000)]
MS ABI: Reference MSVC RTTI from the VFTable
The pointer for a class's RTTI data comes right before the VFTable but
has no name. To be properly compatible with this, we do the following:
* Create a single GlobalVariable which holds the contents of the VFTable
_and_ the pointer to the RTTI data.
* Create a GlobalAlias, with appropriate linkage/visibility, that points
just after the RTTI data pointer. This ensures that the VFTable
symbol will always refer to VFTable data.
* Create a Comdat with a "Largest" SelectionKind and stick the private
GlobalVariable in it. By transitivity, the GlobalAlias will be a
member of the Comdat group. Using "Largest" ensures that foreign
definitions without an RTTI data pointer will _not_ be chosen in the
final linked image.
Whether or not we emit RTTI data depends on several things:
* The -fno-rtti flag implies that we should never not emit a pointer to
RTTI data before the VFTable.
* __declspec(dllimport) brings in the VFTable from a remote DLL. Use an
available_externally GlobalVariable to provide a local definition of
the VFTable. This means that we won't have any available_externally
definitions of things like complete object locators. This is
acceptable because they are never directly referenced.
To my knowledge, this completes the implementation of MSVC RTTI code
generation.
Further semantic work should be done to properly support /GR-.
This patch removes the dead code, and refines the
getEHResumeBlock() slightly.
The CleanupHackLevel was a hack to the old exception
handling intrinsics, which have several issues with function
inliner.
Since LLVM 3.0, the new landingpad and resume instructions
are added to LLVM IR. With the new exception handling
mechanism, most of the issues are fixed now. We should
always use these instructions to implement the exception
handling code nowadays, and we don't need the hack any more.
Besides, the `CleanupHackLevel` is a compile-time constant,
thus other cases have been considered as dead code for a while.
Simon Atanasyan [Tue, 1 Jul 2014 10:59:09 +0000 (10:59 +0000)]
[Driver][Mips] MIPS ABI names "32" and "64" used as arguments of the "-mabi"
command line option only. Internally we convert them to the "o32" and "n64"
respectively. So we do not need to refer them anywhere after that conversion.
Enable test/CodeGen/indirect-goto.c in 64b for local arrays
In 32b mode the reference count for block addresses
is not zero. This prevents inlining and constant
folding and causes the test to fail. Changing
the triple allows runnning the test in 64b mode.
The array in foo2 is now local instead of static until
at lower optimization levels the interprocedural constant
propagator is invoked before the global optimizer.
Alp Toker [Tue, 1 Jul 2014 03:19:50 +0000 (03:19 +0000)]
clang-interpreter: use LLVM interpreter if JIT is unavailable
Update the strategy in r212083 to try JIT first and otherwise fall back to the
interpreter. This gives the best of both worlds and still builds fine with no
targets enabled.
Alp Toker [Tue, 1 Jul 2014 02:41:55 +0000 (02:41 +0000)]
clang-interpreter: interpret instead of JITing
Fixes the build when no targets are selected, or no native target is built.
This also better matches up with the description/title of the example and
demonstrates how clang can be used to run C++ on constrained environments
without file IO or executable memory permissions (e.g. iOS apps).
A comment is added explaining how to extend the demo with JIT support as
needed.
Alexey Samsonov [Mon, 30 Jun 2014 20:27:16 +0000 (20:27 +0000)]
[UBSan] Don't link UBSan runtime into shared libraries.
It used to be a feature of UBSan (it could sanitize a standalone
shared object instead of the whole program), but now it causes
more problems, like PR20165.
Ben Langmuir [Mon, 30 Jun 2014 20:04:14 +0000 (20:04 +0000)]
Consider module depedencies when checking a preamble in libclang
Add module dependencies (header files, module map files) to the list of
files to check when deciding whether to rebuild a preamble. That fixes
using preambles with module imports so long as they are in
non-overridden files.
My intent is to use to unify the existing dependency collectors to the
new “DependencyCollectory” interface from this commit, starting with the
DependencyFileGenerator.
Andrea Di Biagio [Mon, 30 Jun 2014 18:23:58 +0000 (18:23 +0000)]
[x86] Add Clang support for intrinsic __rdpmc.
This patch adds intrinsic __rdpmc to header file 'ia32intrin.h'.
Intrinsic __rdmpc can be used to read performance monitoring counters. It is
implemented as a direct call to __builtin_ia32_rdpmc.
It takes as input a value representing the index of the performance counter to
read. The value of the performance counter is then returned as a unsigned
64-bit quantity.
Daniel Jasper [Mon, 30 Jun 2014 13:24:54 +0000 (13:24 +0000)]
clang-format: [JS] support free-standing functions again.
This worked initially but was broken by r210887.
Before:
function outer1(a, b) {
function inner1(a, b) { return a; } inner1(a, b);
} function outer2(a, b) { function inner2(a, b) { return a; } inner2(a, b); }
After:
function outer1(a, b) {
function inner1(a, b) { return a; }
inner1(a, b);
}
function outer2(a, b) {
function inner2(a, b) { return a; }
inner2(a, b);
}
Alp Toker [Mon, 30 Jun 2014 01:33:59 +0000 (01:33 +0000)]
CodeGenAction::ExecuteAction(): check for invalid LLVM source locations
Add sign checks to deal with the fact that IR parser line/column pairs are
signed integers and sometimes invalid.
The crash path is potentially triggered by corrupt '.bc' files in practice,
though I don't have a binary input test case that can be checked-in right now.
(Unfortunately the backend itself crashes on various ill-formed '.bc' inputs so
this bandage isn't as helpful as it appears yet.)
Alp Toker [Mon, 30 Jun 2014 01:33:53 +0000 (01:33 +0000)]
Extract an isReservedName() function
We'll want to share the implementation if anything else decides to check
for reserved names in future, so make this little snippet of code more
discoverable.
Also remove the __va_list_tag and __builtin_va_list special-case
checks. They're leftovers from before when the reserved name logic was
added.
Windows on ARM defines va_list as a typedef for char *. Although the semantics
of argument passing for variadic functions matches AAPCS VFP, the wrapped
struct __va_list type is unused. This makes the intrinsic definition for
va_list match that of Visual Studio.
NAKAMURA Takumi [Sun, 29 Jun 2014 16:00:11 +0000 (16:00 +0000)]
Revert r211866, r211895 and r211995, "Driver: use GNU::Link for the Generic_GCC toolchain". It broke users of Generic_GCC, cygwin and mingw32.
It reverts commits as follows:
r211866: "Driver: use GNU::Link for the Generic_GCC toolchain"
r211895: "Replace GetProgramPath("ld") with GetLinkerPath()."
r211995: "Driver: add a cygwin linker tool"
This adds a linker tool for the Windows cygwin environment. This linker
invocation is significantly different from the generic ld invocation. It
requires additional parameters as well as does not accept some normal
parameters. This should fix self-hosting on Cygwin.
Some time ago, I noticed that try would get resolved incorrectly for Windows
Itanium targets. Add an explicit test to exsure that exceptions are handled
correctly for Windows Itanium environments.
Richard Trieu [Sat, 28 Jun 2014 23:25:37 +0000 (23:25 +0000)]
Extend -Wtautological-undefined-compare and -Wundefined-bool-conversion to
trigger on taking the address of a reference that is returned from a function
call.
Summary: This patch introduces ACLE header file, implementing extensions that can be directly mapped to existing Clang intrinsics. It implements for both AArch32 and AArch64.
David Blaikie [Fri, 27 Jun 2014 17:40:03 +0000 (17:40 +0000)]
Remove 'const' from MemoryBuffers used through the SourceManager
This removes a const_cast added in r211884 that occurred due to an
inconsistency in how MemoryBuffers are handled between some parts of
clang and LLVM.
MemoryBuffers are immutable and the general convention in the LLVM
project is to omit const from immutable types as it's simply
redundant/verbose (see llvm::Type, for example). While this change
doesn't remove "const" from /every/ MemoryBuffer, it at least makes this
chain of ownership/usage consistent.
Reid Kleckner [Fri, 27 Jun 2014 17:02:02 +0000 (17:02 +0000)]
clang-cl: Map /EHs- to -fno-exceptions
This isn't 100% compatible with MSVC, but it's close enough. MSVC's /EH
flag doesn't really control exceptions so much as how to clean up after
an exception is thrown. The upshot is that cl.exe /EHs- will compile
try, throw, and catch statements with a warning, but clang-cl will
reject such constructs with a hard error. We can't compile such EH
constructs anyway, but this may matter to consumers of the AST.
Chandler Carruth [Fri, 27 Jun 2014 16:37:27 +0000 (16:37 +0000)]
Fix a bug in my previous patch by restoring the behavior that the fatal
error handler is only registered once.
To avoid the use of std::call_once (the obvious way to do this) I've
wrapped everything up into a managed static and done the work in
a constructor. Silly, but it should be effective.
Some out-of-tree libclang users reported this to me, and I've asked them
to put together a test case which exhibits this behavior, but I wanted
to fix things ASAP since the nature of the fix is straight forward.
Oliver Stannard [Fri, 27 Jun 2014 13:59:27 +0000 (13:59 +0000)]
[ARM] Fix AAPCS non-compliance caused by very large structs
This is a fix to the code in clang which inserts padding arguments to
ensure that the ARM backend can emit AAPCS-VFP compliant code. This code
needs to track the number of registers which have been allocated in order
to do this. When passing a very large struct (>64 bytes) by value, clang
emits IR which takes a pointer to the struct, but the backend converts this
back to passing the struct in registers and on the stack. The bug was that
this was being considered by clang to only use one register, meaning that
there were situations in which padding arguments were incorrectly emitted
by clang.
James Molloy [Fri, 27 Jun 2014 11:53:35 +0000 (11:53 +0000)]
[ARM-BE] Generate correct NEON intrinsics for big endian systems.
The NEON intrinsics in arm_neon.h are designed to work on vectors
"as-if" loaded by (V)LDR. We load vectors "as-if" (V)LD1, so the
intrinsics are currently incorrect.
This patch adds big-endian versions of the intrinsics that does the
"obvious but dumb" thing of reversing all vector inputs and all
vector outputs. This will produce extra REVs, but we trust the
optimizer to remove them.
Driver: use GNU::Link for the Generic_GCC toolchain
This changes the behaviour of the driver for linking to match that of the
Generic_GCC::Assemble. The default link should use "ld" rather than "gcc" for
the linker as gcc does. This avoids the unnecessary round-tripping through gcc.
It also is much more reasonable behaviour from the user's perspective. This
should have been updated with SVN r195554 which changed the behaviour of
Generic_GCC::Assemble.
The gcc_forward test needs to be updated to mark the fact that -march is a flag
for GCC not ld. This was updated as a typo fix, but added a check for a flag
that is not a link flag.
The bindings test covers the change for testing, and thus no new test was added.
Anna Zaks [Fri, 27 Jun 2014 01:03:05 +0000 (01:03 +0000)]
Do not inline methods of C++ containers (coming from headers).
This silences false positives (leaks, use of uninitialized value) in simple
code that uses containers such as std::vector and std::list. The analyzer
cannot reason about the internal invariances of those data structures which
leads to false positives. Until we come up with a better solution to that
problem, let's just not inline the methods of the containers and allow objects
to escape whenever such methods are called.
This just extends an already existing flag "c++-container-inlining" and applies
the heuristic not only to constructors and destructors of the containers, but
to all of their methods.
We have a bunch of distinct user reports all related to this issue
(radar://16058651, radar://16580751, radar://16384286, radar://16795491
[PR19637]).
Reid Kleckner [Thu, 26 Jun 2014 22:42:18 +0000 (22:42 +0000)]
Avoid extra back reference key lookup in msmangler
Avoid a second key lookup when the back reference key is going to be
inserted in the StringMap. The string lookups in the msmangler are the
main responsible for the huge overhead when compared to the itanium
mangler. This patch makes a small but noticeable improvement.
Objective-C ARC. Provide diagnostic and fix-it
when casting a retainable object to a objc_bridge_related
CF type with the suggestion of applying the method
specified in the bridging attribute to the object.
// rdar://15932435
projects / clang / log