Summary:
Different devices may in some cases require different code generation schemes in order to implement OpenMP. This is required not only for performance reasons, but also because it may not be possible to have the current (default) implementation working for these devices. E.g. GPU's cannot implement the same scheme a target such as powerpc or x86b would use, in the sense that it does not have the ability to fork threads, instead all the threads are always executing and need to be managed by the implementation.
This patch proposes a reorganization of the code in the OpenMP code generation to pave the way to have specialized implementation of OpenMP support. More than a "real" patch this is more a request for comments in order to understand if what is proposed is acceptable or if there are better/easier ways to do it.
In this patch part of the common OpenMP codegen infrastructure is moved to a new file under a new namespace (CGOpenMPCommon) so it can be shared between the default implementation and the specialized one. When CGOpenMPRuntime is created, an attempt to select a specialized implementation is done.
In the patch a specialization for nvptx targets is done which currently checks if the target is an OpenMP device and trap if it is not.
Let me know comments suggestions you may have.
Reviewers: hfinkel, carlo.bertolli, arpith-jacob, kkwli0, ABataev
Subscribers: Hahnfeld, cfe-commits, fraggamuffin, caomhin, jholewinski
Differential Revision: http://reviews.llvm.org/D16784
llvm-svn: 259977
Avoid crashing when printing diagnostics for vtable-related CFI
errors. In diagnostic mode, the frontend does an additional check of
the vtable pointer against the set of all known vtable addresses and
lets the runtime handler know if it is safe to inspect the vtable.
http://reviews.llvm.org/D16823
llvm-svn: 259716
In general CUDA does not allow dynamic initialization of
global device-side variables. One exception is that CUDA allows
records with empty constructors as described in section E2.2.1 of
CUDA 7.5 Programming guide.
This patch applies initializer checks for all device-side variables.
Empty constructors are accepted, but no code is generated for them.
Differential Revision: http://reviews.llvm.org/D15305
llvm-svn: 259592
Member pointers in the MS ABI are tricky for a variety of reasons.
The size of a member pointer is indeterminate until the program reaches
a point where the representation is required to be known. However,
*pointers* to member pointers may exist without knowing the pointee
type's representation. In these cases, we synthesize an opaque LLVM
type for the pointee type.
However, we can be in a situation where the underlying member pointer's
representation became known mid-way through the program. To account for
this, we attempted to manicure CodeGen's type-cache so that we can
replace the opaque member pointer type with the real deal while leaving
the pointer types unperturbed. This, unfortunately, is a problematic
approach to take as we will violate CodeGen's invariants.
These violations are mostly harmless but let's do the right thing
instead: invalidate the type-cache if a member pointer's LLVM
representation changes.
This fixes PR26313.
llvm-svn: 258839
* Runtime diagnostic data for cfi-icall changed to match the rest of
cfi checks
* Layout of all CFI diagnostic data changed to put Kind at the
beginning. There is no ABI stability promise yet.
* Call cfi_slowpath_diag instead of cfi_slowpath when needed.
* Emit __cfi_check_fail function, which dispatches a CFI check
faliure according to trap/recover settings of the current module.
* A tiny driver change to match the way the new handlers are done in
compiler-rt.
llvm-svn: 258745
Summary:
These aliases are done to support inline asm, but there's nothing we can
do: NVPTX doesn't support aliases.
Reviewers: tra
Subscribers: cfe-commits, jhen, echristo
Differential Revision: http://reviews.llvm.org/D16501
llvm-svn: 258734
This is part of a new statistics gathering feature for the sanitizers.
See clang/docs/SanitizerStats.rst for further info and docs.
Differential Revision: http://reviews.llvm.org/D16175
llvm-svn: 257971
Proper diagnostic and resolution of mangled names' conflicts in variables.
When there is a declaration and a definition using the same name but different
types, we emit what is in the definition. When there are two conflicting
definitions, we issue an error.
Differential Revision: http://reviews.llvm.org/D15686
llvm-svn: 257754
Clang got itself into the situation where we mangled the same
constructor twice with two different constructor types. After one of
the constructors were utilized, the tag used for one of the types
changed from class to struct because a class template became complete.
This resulted in one of the constructor types varying from the other
constructor.
Instead, force "base" constructor types to "complete" if the ABI doesn't
have constructor variants. This will ensure that GlobalDecls for both
variants will get the same mangled name.
This fixes PR26029.
llvm-svn: 257205
This patch attempts to fix the regressions identified when the patch was committed initially.
Thanks to Michael Liao for identifying the fix in the offloading metadata generation
related with side effects in evaluation of function arguments.
llvm-svn: 256933
Summary:
In order to offloading work properly two things need to be in place:
- a descriptor with all the offloading information (device entry functions, and global variable) has to be created by the host and registered in the OpenMP offloading runtime library.
- all the device functions need to be emitted for the device and a convention has to be in place so that the runtime library can easily map the host ID of an entry point with the actual function in the device.
This patch adds support for these two things. However, only entry functions are being registered given that 'declare target' directive is not yet implemented.
About offloading descriptor:
The details of the descriptor are explained with more detail in http://goo.gl/L1rnKJ. Basically the descriptor will have fields that specify the number of devices, the pointers to where the device images begin and end (that will be defined by the linker), and also pointers to a the begin and end of table whose entries contain information about a specific entry point. Each entry has the type:
```
struct __tgt_offload_entry{
void *addr;
char *name;
int64_t size;
};
```
and will be implemented in a pre determined (ELF) section `.omp_offloading.entries` with 1-byte alignment, so that when all the objects are linked, the table is in that section with no padding in between entries (will be like a C array). The code generation ensures that all `__tgt_offload_entry` entries are emitted in the same order for both host and device so that the runtime can have the corresponding entries in both host and device in same index of the table, and efficiently implement the mapping.
The resulting descriptor is registered/unregistered with the runtime library using the calls `__tgt_register_lib` and `__tgt_unregister_lib`. The registration is implemented in a high priority global initializer so that the registration happens always before any initializer (that can potentially include target regions) is run.
The driver flag -omptargets= was created to specify a comma separated list of devices the user wants to support so that the new functionality can be exercised. Each device is specified with its triple.
About target codegen:
The target codegen is pretty much straightforward as it reuses completely the logic of the host version for the same target region. The tricky part is to identify the meaningful target regions in the device side. Unlike other programming models, like CUDA, there are no already outlined functions with attributes that mark what should be emitted or not. So, the information on what to emit is passed in the form of metadata in host bc file. This requires a new option to pass the host bc to the device frontend. Then everything is similar to what happens in CUDA: the global declarations emission is intercepted to check to see if it is an "interesting" declaration. The difference is that instead of checking an attribute, the metadata information in checked. Right now, there is only a form of metadata to pass information about the device entry points (target regions). A class `OffloadEntriesInfoManagerTy` was created to manage all the information and queries related with the metadata. The metadata looks like this:
```
!omp_offload.info = !{!0, !1, !2, !3, !4, !5, !6}
!0 = !{i32 0, i32 52, i32 77426347, !"_ZN2S12r1Ei", i32 479, i32 13, i32 4}
!1 = !{i32 0, i32 52, i32 77426347, !"_ZL7fstatici", i32 461, i32 11, i32 5}
!2 = !{i32 0, i32 52, i32 77426347, !"_Z9ftemplateIiET_i", i32 444, i32 11, i32 6}
!3 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 99, i32 11, i32 0}
!4 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 272, i32 11, i32 3}
!5 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 127, i32 11, i32 1}
!6 = !{i32 0, i32 52, i32 77426347, !"_Z3fooi", i32 159, i32 11, i32 2}
```
The fields in each metadata entry are (in sequence):
Entry 1) an ID of the type of metadata - right now only zero is used meaning "OpenMP target region".
Entry 2) a unique ID of the device where the input source file that contain the target region lives.
Entry 3) a unique ID of the file where the input source file that contain the target region lives.
Entry 4) a mangled name of the function that encloses the target region.
Entries 5) and 6) line and column number where the target region was found.
Entry 7) is the order the entry was emitted.
Entry 2) and 3) are required to distinguish files that have the same function name.
Entry 4) is required to distinguish different instances of the same declaration (usually templated ones)
Entries 5) and 6) are required to distinguish the particular target region in body of the function (it is possible that a given target region is not an entry point - if clause can evaluate always to zero - and therefore we need to identify the "interesting" target regions. )
This patch replaces http://reviews.llvm.org/D12306.
Reviewers: ABataev, hfinkel, tra, rjmccall, sfantao
Subscribers: FBrygidyn, piotr.rak, Hahnfeld, cfe-commits
Differential Revision: http://reviews.llvm.org/D12614
llvm-svn: 256842
This sets the maximum entry count among all functions in the program to the module using module flags. This allows the optimizer to use this information.
Differential Revision: http://reviews.llvm.org/D15163
llvm-svn: 255918
Clang-side cross-DSO CFI.
* Adds a command line flag -f[no-]sanitize-cfi-cross-dso.
* Links a runtime library when enabled.
* Emits __cfi_slowpath calls is bitset test fails.
* Emits extra hash-based bitsets for external CFI checks.
* Sets a module flag to enable __cfi_check generation during LTO.
This mode does not yet support diagnostics.
llvm-svn: 255694
This updates clang to use bundle operands to associate an invoke with
the funclet which it is contained within.
Depends on D15517.
Differential Revision: http://reviews.llvm.org/D15518
llvm-svn: 255675
This sets the maximum entry count among all functions in the program to the
module using module flags. This allows the optimizer to use this information.
Differential Revision: http://reviews.llvm.org/D15163
llvm-svn: 255397
This reverts commit r254195.
From the description, I suspect that the wrong patch was committed here,
and this is causing assertion failures in EmitDeferred() when the global
value ends up being a bitcast of a global.
llvm-svn: 254823
`pass_object_size` is our way of enabling `__builtin_object_size` to
produce high quality results without requiring inlining to happen
everywhere.
A link to the design doc for this attribute is available at the
Differential review link below.
Differential Revision: http://reviews.llvm.org/D13263
llvm-svn: 254554
This is about how we handle static member of a template. Before this commit,
we use internal linkage for the IR thread-local variable, which is inefficient.
With this commit, we will start to follow Itanium C++ ABI.
rdar://problem/23415206
Reviewed by John McCall.
llvm-svn: 252814
Summary: This fixes a bug that's easily encountered in LLDB
(https://llvm.org/bugs/show_bug.cgi?id=22875). The problem here is that we
mangle a name during debug info emission, but never actually emit the actual
Decl, so we run into problems in EmitDeclMetadata (which assumes such a Decl
exists). Fix that by just skipping metadata emissions for mangled names that
don't have associated Decls.
Reviewers: rjmccall
Subscribers: labath, cfe-commits
Differential Revision: http://reviews.llvm.org/D13959
llvm-svn: 252229
This sets the mostly expected Darwin default ABI options for these two
platforms. Active changes from these defaults for watchOS are in a later patch.
llvm-svn: 251708
CGBlocks.cpp.
This commit fixes a bug in clang's code-gen where it creates the
following functions but doesn't attach function attributes to them:
__copy_helper_block_
__destroy_helper_block_
__Block_byref_object_copy_
__Block_byref_object_dispose_
rdar://problem/20828324
Differential Revision: http://reviews.llvm.org/D13525
llvm-svn: 249735
early.
This is needed in a patch I plan to commit later, in which a null Decl
pointer is passed to SetLLVMFunctionAttributesForDefinition.
Relevant discussion is in http://reviews.llvm.org/D13525.
llvm-svn: 249722
No ABI for C++ currently makes it possible to implement the standard
100% perfectly. We wrongly hid some of our compatible behavior behind
-fms-compatibility instead of tying it to the compiler ABI.
llvm-svn: 249656
This commit fixes an assert that is triggered when optnone is being
added to an IR function that is already marked with minsize and optsize.
rdar://problem/22723716
Differential Revision: http://reviews.llvm.org/D13004
llvm-svn: 248191
This avoids building a fake LLVM IR global variable just to ferry an i32
down into LLVM codegen. It also puts a nail in the coffin of using MS
ABI C++ EH with landingpads, since now we'll assert in the lpad code
when flags are present.
llvm-svn: 247843
It is dangerous to do LTO on code with strict-vtable-pointers, because
one module has invariant.group.barriers, and the other one not.
In the future I want to just strip all invariant.group metadata from
vptrs loads/stores and get rid of invariant.group.barrier calls.
http://reviews.llvm.org/D12580
llvm-svn: 247724
These are a few cleanups I happened to have from trying to go in a
different direction recently, so just flushing them out while I have
them.
llvm-svn: 247593
This was the wrong direction to take anyway (because ultimately the
GlobalValue needed the pointee type again and /it/ used
PointerType::getElementType eventually anyway)... let's go a different way.
This reverts commit r236161.
llvm-svn: 247586
Current implementation may end up emitting an undefined reference for
an "inline __attribute__((always_inline))" function by generating an
"available_externally alwaysinline" IR function for it and then failing to
inline all the calls. This happens when a call to such function is in dead
code. As the inliner is an SCC pass, it does not process dead code.
Libc++ relies on the compiler never emitting such undefined reference.
With this patch, we emit a pair of
1. internal alwaysinline definition (called F.alwaysinline)
2a. A stub F() { musttail call F.alwaysinline }
-- or, depending on the linkage --
2b. A declaration of F.
The frontend ensures that F.inlinefunction is only used for direct
calls, and the stub is used for everything else (taking the address of
the function, really). Declaration (2b) is emitted in the case when
"inline" is meant for inlining only (like __gnu_inline__ and some
other cases).
This approach, among other nice properties, ensures that alwaysinline
functions are always internal, making it impossible for a direct call
to such function to produce an undefined symbol reference.
This patch is based on ideas by Chandler Carruth and Richard Smith.
llvm-svn: 247494
Current implementation may end up emitting an undefined reference for
an "inline __attribute__((always_inline))" function by generating an
"available_externally alwaysinline" IR function for it and then failing to
inline all the calls. This happens when a call to such function is in dead
code. As the inliner is an SCC pass, it does not process dead code.
Libc++ relies on the compiler never emitting such undefined reference.
With this patch, we emit a pair of
1. internal alwaysinline definition (called F.alwaysinline)
2a. A stub F() { musttail call F.alwaysinline }
-- or, depending on the linkage --
2b. A declaration of F.
The frontend ensures that F.inlinefunction is only used for direct
calls, and the stub is used for everything else (taking the address of
the function, really). Declaration (2b) is emitted in the case when
"inline" is meant for inlining only (like __gnu_inline__ and some
other cases).
This approach, among other nice properties, ensures that alwaysinline
functions are always internal, making it impossible for a direct call
to such function to produce an undefined symbol reference.
This patch is based on ideas by Chandler Carruth and Richard Smith.
llvm-svn: 247465
This flag causes the compiler to emit bit set entries for functions as well
as runtime bitset checks at indirect call sites. Depends on the new function
bitset mechanism.
Differential Revision: http://reviews.llvm.org/D11857
llvm-svn: 247238
This function can be used to create a metadata identifier for a specific
type. No functionality change, but this will be used by D11857 and D12026.
Differential Revision: http://reviews.llvm.org/D12038
llvm-svn: 247098
Introduce an Address type to bundle a pointer value with an
alignment. Introduce APIs on CGBuilderTy to work with Address
values. Change core APIs on CGF/CGM to traffic in Address where
appropriate. Require alignments to be non-zero. Update a ton
of code to compute and propagate alignment information.
As part of this, I've promoted CGBuiltin's EmitPointerWithAlignment
helper function to CGF and made use of it in a number of places in
the expression emitter.
The end result is that we should now be significantly more correct
when performing operations on objects that are locally known to
be under-aligned. Since alignment is not reliably tracked in the
type system, there are inherent limits to this, but at least we
are no longer confused by standard operations like derived-to-base
conversions and array-to-pointer decay. I've also fixed a large
number of bugs where we were applying the complete-object alignment
to a pointer instead of the non-virtual alignment, although most of
these were hidden by the very conservative approach we took with
member alignment.
Also, because IRGen now reliably asserts on zero alignments, we
should no longer be subject to an absurd but frustrating recurring
bug where an incomplete type would report a zero alignment and then
we'd naively do a alignmentAtOffset on it and emit code using an
alignment equal to the largest power-of-two factor of the offset.
We should also now be emitting much more aggressive alignment
attributes in the presence of over-alignment. In particular,
field access now uses alignmentAtOffset instead of min.
Several times in this patch, I had to change the existing
code-generation pattern in order to more effectively use
the Address APIs. For the most part, this seems to be a strict
improvement, like doing pointer arithmetic with GEPs instead of
ptrtoint. That said, I've tried very hard to not change semantics,
but it is likely that I've failed in a few places, for which I
apologize.
ABIArgInfo now always carries the assumed alignment of indirect and
indirect byval arguments. In order to cut down on what was already
a dauntingly large patch, I changed the code to never set align
attributes in the IR on non-byval indirect arguments. That is,
we still generate code which assumes that indirect arguments have
the given alignment, but we don't express this information to the
backend except where it's semantically required (i.e. on byvals).
This is likely a minor regression for those targets that did provide
this information, but it'll be trivial to add it back in a later
patch.
I partially punted on applying this work to CGBuiltin. Please
do not add more uses of the CreateDefaultAligned{Load,Store}
APIs; they will be going away eventually.
llvm-svn: 246985
We were crashing in CodeGen given input like this:
int self_alias(void) __attribute__((weak, alias("self_alias")));
such a self-alias is invalid, but instead of diagnosing the situation, we'd
proceed to produce IR for both the function declaration and the alias. Because
we already had a function named 'self_alias', the alias could not be named the
same thing, and so LLVM would pick a different name ('self_alias1' for example)
for that value. When we later called CodeGenModule::checkAliases, we'd look up
the IR value corresponding to the alias name, find the function declaration
instead, and then assert in a cast to llvm::GlobalAlias. The easiest way to prevent
this is simply to avoid creating the wrongly-named alias value in the first
place and issue the diagnostic there (instead of in checkAliases). We detect a
related cycle case in CodeGenModule::EmitAliasDefinition already, so this just
adds a second such check.
Even though the other test cases for this 'alias definition is part of a cycle'
diagnostic are in test/Sema/attr-alias-elf.c, I've added a separate regression
test for this case. This is because I can't add this check to
test/Sema/attr-alias-elf.c without disturbing the other test cases in that
file. In order to avoid construction of the bad IR values, this diagnostic
is emitted from within CodeGenModule::EmitAliasDefinition (and the relevant
declaration is not added to the Aliases vector). The other cycle checks are
done within the CodeGenModule::checkAliases function based on the Aliases
vector, called from CodeGenModule::Release. However, if there have been errors
earlier, HandleTranslationUnit does not call Release, and so checkAliases is
never called, and so none of the other diagnostics would be produced.
Fixes PR23509.
llvm-svn: 246882
This implements basic support for compiling (though not yet assembling
or linking) for a WebAssembly target. Note that ABI details are not yet
finalized, and may change.
Differential Revision: http://reviews.llvm.org/D12002
llvm-svn: 246814
Proper diagnostic and resolution of mangled names conflicts between C++ methods
and C functions. This patch implements support for functions/methods only;
support for variables is coming separately.
Differential Revision: http://reviews.llvm.org/D11297
llvm-svn: 246438
This patch does two things:
1) Don't error about dllimport/export on thread-local static local variables.
We put those attributes on static locals in dllimport/export functions
implicitly in case the function gets inlined. Now, for TLS variables this
is a problem because we can't import such variables, but it's a benign
problem becase:
2) Make sure we never inline a dllimport function TLS static locals. In fact,
never inline a dllimport function that references a non-imported function
or variable (because these are not defined in the importing library). This
seems to match MSVC's behaviour.
Differential Revision: http://reviews.llvm.org/D12422
llvm-svn: 246338
Summary:
According to CUDA documentation, global variables declared with __device__,
__constant__ can be initialized from host code, so mark them as
externally initialized. Because __shared__ variables cannot have an
initialization as part of their declaration and since the value maybe kept
across different kernel invocation, the value of __shared__ is effectively
undefined instead of zero initialized.
Wrongly using zero initializer may cause illegitimate optimization, e.g.
removing unused __constant__ variable because it's not updated in the device
code and the value is initialized with zero.
Test Plan: test/CodeGenCUDA/address-spaces.cu
Patch by Xuetian Weng
Reviewers: jholewinski, eliben, tra, jingyue
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D12241
llvm-svn: 245786
We risk iterator invalidation issues if we use a DenseMap to hold the
backing storage for an APValue. Instead, BumpPtrAllocate them and
use APValue * as our DenseMap value.
Also, don't assume that MaterializedGlobalTemporaryMap won't regrow
between when we initially perform a lookup and later on when we actually
try to insert into it.
This fixes PR24289.
Differential Revision: http://reviews.llvm.org/D11629
llvm-svn: 244989
After r244870 flush() will only compare two null pointers and return,
doing nothing but wasting run time. The call is not required any more
as the stream and its SmallString are always in sync.
Thanks to David Blaikie for reviewing.
llvm-svn: 244928
Summary:
By default, 'clang' emits dwarf and 'clang-cl' emits codeview. You can
force emission of one or both by passing -gcodeview and -gdwarf to
either driver.
Reviewers: dblaikie, hans
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D11742
llvm-svn: 244097
This was calling FD->hasBody(), meaning "Does the function that this
decl refers to have a body?", rather than
FD->doesThisDeclarationHaveABody(), meaning "Is this decl a
non-deleted definition?".
We might want to consider renaming these APIs :/
llvm-svn: 243360
This allows a module-aware debugger such as LLDB to import the currently
visible modules before dropping into the expression evaluator.
rdar://problem/20965932
llvm-svn: 241084
isTriviallyRecursive is a hack used to bridge a gap between the
expectations that source code assumes and the semantics that LLVM IR can
provide. Specifically, asm labels on functions are treated as an
explicit name for a GlobalObject in Clang but treated like an
output-processing step in GCC. Tweak this hack a little further to emit
calls to library functions instead of emitting an incorrect definition.
The definition in question would have available_externally linkage (this
is OK) but result in a call to itself which will either result in an
infinite loop or stack overflow.
This fixes PR23964.
llvm-svn: 241043
When a profile file cannot be opened, we used to display just the error
message but not the name of the profile the compiler was trying to open.
This will become useful in the next set of patches that introduce
GCC-compatible flags to specify profiles.
llvm-svn: 240715
The patch is generated using this command:
$ tools/extra/clang-tidy/tool/run-clang-tidy.py -fix \
-checks=-*,llvm-namespace-comment -header-filter='llvm/.*|clang/.*' \
work/llvm/tools/clang
To reduce churn, not touching namespaces spanning less than 10 lines.
llvm-svn: 240270
This patch adds initial support for the -fsanitize=kernel-address flag to Clang.
Right now it's quite restricted: only out-of-line instrumentation is supported, globals are not instrumented, some GCC kasan flags are not supported.
Using this patch I am able to build and boot the KASan tree with LLVMLinux patches from github.com/ramosian-glider/kasan/tree/kasan_llvmlinux.
To disable KASan instrumentation for a certain function attribute((no_sanitize("kernel-address"))) can be used.
llvm-svn: 240131
If the type isn't trivially moveable emplace can skip a potentially
expensive move. It also saves a couple of characters.
Call sites were found with the ASTMatcher + some semi-automated cleanup.
memberCallExpr(
argumentCountIs(1), callee(methodDecl(hasName("push_back"))),
on(hasType(recordDecl(has(namedDecl(hasName("emplace_back")))))),
hasArgument(0, bindTemporaryExpr(
hasType(recordDecl(hasNonTrivialDestructor())),
has(constructExpr()))),
unless(isInTemplateInstantiation()))
No functional change intended.
llvm-svn: 238601
This is a follow-up to r238266. It turned out structors are codegened through a different path,
and didn't get the storage class set in EmitGlobalFunctionDefinition.
llvm-svn: 238443
With this change, enabling -fmodules-local-submodule-visibility results in name
visibility rules being applied to submodules of the current module in addition
to imported modules (that is, names no longer "leak" between submodules of the
same top-level module). This also makes it much safer to textually include a
non-modular library into a module: each submodule that textually includes that
library will get its own "copy" of that library, and so the library becomes
visible no matter which including submodule you import.
llvm-svn: 237473
Functions with available_externally linkage will not be emitted to object
files (they will just be undefined symbols), so it does not make sense to
put them in comdats.
Creates a second overload of maybeSetTrivialComdat that uses the GlobalObject
instead of the Decl, and uses that in several places that had the faulty
logic.
Differential Revision: http://reviews.llvm.org/D9580
llvm-svn: 236879
- added -fcuda-include-gpubinary option to incorporate results of
device-side compilation into host-side one.
- generate code to register GPU binaries and associated kernels
with CUDA runtime and clean-up on exit.
- added test case for init/deinit code generation.
Differential Revision: http://reviews.llvm.org/D9507
llvm-svn: 236765
This issue was fixed elsewhere in r235396 in a more general way, hence these
changes no longer do anything. Keep the testcase however, to ensure that we
don't regress this for ARM.
llvm-svn: 236104
When creating a global variable with a type of a struct with bitfields, we must
forcibly set the alignment of the global from the RecordDecl. We must do this so
that the proper bitfield alignment makes its way down to LLVM, since clang will
mangle the bitfields into one large type.
llvm-svn: 235976
Currently clang emits file-scope asm during *both* host and device
compilation modes which is usually a wrong thing to do.
There's no way to attach any attribute to an __asm statement, so
there's no way to differentiate between host-side and device-side
file-scope asm. This patch makes clang to match nvcc behavior and
emit file-scope-asm only during host-side compilation.
Differential Revision: http://reviews.llvm.org/D9270
llvm-svn: 235905
Code in CodeGenModule::GetOrCreateLLVMGlobal that sets up GlobalValue
object for LLVM external symbols has this comment:
// FIXME: This code is overly simple and should be merged with other global
// handling.
One part does seems to be "overly simple" currently is that this code
never sets any alignment info on the GlobalValue, so that the emitted
IR does not have any align attribute on external globals. This can
lead to unnecessarily inefficient code generation.
This patch adds a GV->setAlignment call to set alignment info.
llvm-svn: 235396
Things can't both be in comdats and have common linkage, so never give things
in comdats common linkage. Common linkage is only used in .c files, and the
only thing that can trigger a comdat in c is selectany from what I can tell.
Fixes PR23243.
Also address an over-the-shoulder review comment from rnk by moving the
hasAttr<SelectAnyAttr>() in Decl.cpp around a bit. It only makes a minor
difference for selectany on global variables, so it goes well with the rest of
this patch.
http://reviews.llvm.org/D9042
llvm-svn: 235053
This reverts commit r234767, as it was breaking all ARM buildbots for two days and the
assert is not in the code, making it difficult to spot the error, which would keep the
bots red for a few more days. New errors were silently introduced because of this bug,
and we don't want this to escalate.
llvm-svn: 234983
Now the GEP constant utility functions require the type to be explicitly
passed (since eventually the pointer type will be opaque and not convey
the required type information). For now callers can still pass nullptr
(though none were needed here in Clang, which is nice) if
convenienc/necessary, but eventually that will be disallowed as well.
llvm-svn: 233937
