tra added a comment.
In https://reviews.llvm.org/D25845#576819, @jlebar wrote:
> > This patch changes the way we match target attributes of base template vs
> > attributes used in explicit instantiation or specialization so that only
> > explicitly specified attributes are considered.
>
> Anoth
tra updated this revision to Diff 75511.
tra marked 5 inline comments as done.
tra added a comment.
addressed jlebar's comments.
https://reviews.llvm.org/D25845
Files:
include/clang/Sema/Sema.h
lib/Sema/SemaCUDA.cpp
lib/Sema/SemaDeclAttr.cpp
lib/Sema/SemaTemplate.cpp
test/SemaCUDA/fun
tra added inline comments.
Comment at: include/clang/Sema/Sema.h:9396
+ CUDAFunctionTarget IdentifyCUDATarget(const FunctionDecl *D,
+bool IgnoreImplicitHDAttr = false);
CUDAFunctionTarget IdentifyCUDATarget(const AttributeList *Attr);
tra added inline comments.
Comment at: lib/Sema/SemaCUDA.cpp:87
+
+ if ((HasHostAttr && HasDeviceAttr) || ForceCUDAHostDeviceDepth > 0)
+return CFT_HostDevice;
jlebar wrote:
> Checking ForceCUDAHostDeviceDepth here is...yeah. Especially because the
> other
tra updated this revision to Diff 75515.
tra marked 5 inline comments as done.
tra added a comment.
addressed jlebar's comments.
https://reviews.llvm.org/D25809
Files:
include/clang/Basic/DiagnosticSemaKinds.td
include/clang/Sema/Sema.h
lib/Sema/SemaCUDA.cpp
lib/Sema/SemaDecl.cpp
lib/
tra updated this revision to Diff 75516.
tra added a comment.
removed pragma check from IdentifyCUDATarget for real.
https://reviews.llvm.org/D25809
Files:
include/clang/Basic/DiagnosticSemaKinds.td
include/clang/Sema/Sema.h
lib/Sema/SemaCUDA.cpp
lib/Sema/SemaDecl.cpp
lib/Sema/SemaOve
tra updated this revision to Diff 75652.
tra marked an inline comment as done.
tra added a comment.
Addressed remaining nits.
https://reviews.llvm.org/D25809
Files:
include/clang/Basic/DiagnosticSemaKinds.td
include/clang/Sema/Sema.h
lib/Sema/SemaCUDA.cpp
lib/Sema/SemaDecl.cpp
lib/Sem
tra added inline comments.
Comment at: lib/Sema/SemaCUDA.cpp:791
+ CUDAFunctionTarget NewTarget = IdentifyCUDATarget(NewFD);
+ for (auto OldND : Previous) {
+FunctionDecl *OldFD = OldND->getAsFunction();
jlebar wrote:
> tra wrote:
> > jlebar wrote:
> > > If
Author: tra
Date: Mon Oct 24 17:52:39 2016
New Revision: 285037
URL: http://llvm.org/viewvc/llvm-project?rev=285037&view=rev
Log:
Link clangCodeGen with clangAnalysis required after r284990.
Fixes build break for configurations that use shared libraries.
Modified:
cfe/trunk/lib/CodeGen/CMake
tra added a comment.
CodeGen's CMakeLists.txt needs to be updated to link with clangAnalysis now,
otherwise it breaks builds that use shared libraries.
Fixed in r285037.
diff --git a/lib/CodeGen/CMakeLists.txt b/lib/CodeGen/CMakeLists.txt
index f5d5d69..9cf34f6 100644
--- a/lib/CodeGen/CMa
tra updated this revision to Diff 75812.
tra added a comment.
- Instead of relying on the first attribute we find, check all matching ones.
- Specializations inherit their target attributes from their base template
only. Their effective target always matches that of the template and is no
longer
tra updated this revision to Diff 75816.
tra added a comment.
- handle using declarations found in the overload set we check.
https://reviews.llvm.org/D25809
Files:
include/clang/Basic/DiagnosticSemaKinds.td
include/clang/Sema/Sema.h
lib/Sema/SemaCUDA.cpp
lib/Sema/SemaDecl.cpp
lib/Sem
tra added a comment.
I'm OK with the change, but the comments suggest that things may be more
complicated.
How about disabling assert for CUDA only?
https://reviews.llvm.org/D25640
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://list
tra accepted this revision.
tra added a comment.
This revision is now accepted and ready to land.
LGTM.
Comment at: clang/test/SemaCUDA/add-inline-in-definition.cu:13-14
+//
+// The trickiness here comes from the fact that the FunctionDecl bar() sees for
+// foo() does not have
tra updated this revision to Diff 34230.
tra marked 2 inline comments as done.
tra added a comment.
Implementes Richard Smiths' suggestions:
Fixed attribute order in name mangling.
Replaced std::function with template argument.
http://reviews.llvm.org/D12453
Files:
include/clang/Basic/LangOpt
tra marked 2 inline comments as done.
tra added a comment.
http://reviews.llvm.org/D12453
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
tra abandoned this revision.
tra added a comment.
Emitting IR is not sufficient to ensure that the kernels survive GDCE, so the
patch does not work with optimizations on.
http://reviews.llvm.org/D11666 would have to do for now.
http://reviews.llvm.org/D11993
_
tra updated the summary for this revision.
tra updated this revision to Diff 34251.
tra added a comment.
Assume (and test) that NVPTX back-end includes NVVMReflect by default.
http://reviews.llvm.org/D11664
Files:
include/clang/Basic/LangOptions.def
include/clang/Driver/CC1Options.td
lib/
tra updated this revision to Diff 34368.
tra added a comment.
Implemented pcc@'s suggestion to disallow mixing HD and H/D functions.
Removed name mangling as it's no longer needed. Restriction on HD overloading
guarantees that we'll emit only one viable function during particular side of
compila
tra created this revision.
tra added reviewers: jingyue, eliben, wengxt, jholewinski.
tra added a subscriber: cfe-commits.
r245786 (D12241) disabled all initializers for __shared__ variables. It also
prevents placement of records in __shared__ space as they have constructor as
their initializer.
Author: tra
Date: Thu Sep 10 12:26:58 2015
New Revision: 247307
URL: http://llvm.org/viewvc/llvm-project?rev=247307&view=rev
Log:
[CUDA] Allow trivial constructors as initializer for __shared__ variables.
Differential Revision: http://reviews.llvm.org/D12739
Modified:
cfe/trunk/lib/CodeGen/C
This revision was automatically updated to reflect the committed changes.
Closed by commit rL247307: [CUDA] Allow trivial constructors as initializer for
__shared__ variables. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D12739?vs=34372&id=34460#toc
Repository:
rL LLVM
tra updated this revision to Diff 34467.
tra added a comment.
Removed unneeded #includes.
http://reviews.llvm.org/D11664
Files:
include/clang/Basic/LangOptions.def
include/clang/Driver/CC1Options.td
lib/CodeGen/CodeGenAction.cpp
lib/Frontend/CompilerInvocation.cpp
test/CodeGenCUDA/Inp
This revision was automatically updated to reflect the committed changes.
Closed by commit rL247317: [CUDA] Postprocess bitcode linked in during
device-side CUDA compilation. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D11664?vs=34467&id=34470#toc
Repository:
rL LLVM
Author: tra
Date: Thu Sep 10 13:24:23 2015
New Revision: 247317
URL: http://llvm.org/viewvc/llvm-project?rev=247317&view=rev
Log:
[CUDA] Postprocess bitcode linked in during device-side CUDA compilation.
Link in and internalize the symbols we need from supplied bitcode library.
Differential Revi
tra created this revision.
tra added a reviewer: echristo.
tra added a subscriber: cfe-commits.
* Only the last of the --cuda-host-only/--cuda-device-only options has effect.
* Added test cases.
* Fixed printing of empty action lists for -ccc-print-phases
http://reviews.llvm.org/D12892
Files:
tra created this revision.
tra added a reviewer: echristo.
tra added a subscriber: cfe-commits.
Paves the way for the upcoming patch where host-side compilation will need to
know device-side triple.
http://reviews.llvm.org/D12893
Files:
include/clang/Driver/Action.h
lib/Driver/Action.cpp
tra added a comment.
In http://reviews.llvm.org/D12892#246699, @echristo wrote:
> Yuck. No better ideas off the top of my head other than diagnosing and
> erroring though.
>
> (Is that a possibility? :)
I need to be able to override preceding options, so no error is intentional.
http://revie
tra created this revision.
tra added reviewers: echristo, eliben, jpienaar.
tra added a subscriber: cfe-commits.
Herald added subscribers: dschuff, jfb.
The patch:
- adds -aux-triple option to specify target triple
- propagates aux target info to AST context and Preprocessor
- pulls in targe
tra updated this revision to Diff 35015.
tra marked an inline comment as done.
tra added a comment.
cosmetic fixes.
http://reviews.llvm.org/D12917
Files:
include/clang/AST/ASTContext.h
include/clang/Basic/Builtins.h
include/clang/Driver/CC1Options.td
include/clang/Frontend/CompilerInsta
tra marked an inline comment as done.
tra added a comment.
In http://reviews.llvm.org/D12917#247868, @jpienaar wrote:
> Nice, so this will allow parsing/AST construction with builtins from 2
> architectures but will fail to compile if a builtin for the host/device is
> called from device/host.
Author: tra
Date: Thu Sep 17 15:19:33 2015
New Revision: 247930
URL: http://llvm.org/viewvc/llvm-project?rev=247930&view=rev
Log:
Fixed the comment to match reality.
Modified:
cfe/trunk/include/clang/Frontend/CompilerInstance.h
Modified: cfe/trunk/include/clang/Frontend/CompilerInstance.h
UR
tra updated this revision to Diff 35031.
tra marked an inline comment as done.
tra added a comment.
Updated to address Eric's comments.
http://reviews.llvm.org/D12917
Files:
include/clang/AST/ASTContext.h
include/clang/Basic/Builtins.h
include/clang/Driver/CC1Options.td
include/clang/Fr
tra marked 2 inline comments as done.
Comment at: lib/Frontend/CompilerInstance.cpp:82-87
@@ -84,6 +81,8 @@
+void CompilerInstance::setTarget(TargetInfo *Value) { Target = Value; }
+void CompilerInstance::setAuxTarget(TargetInfo *Value) { AuxTarget = Value; }
void CompilerInsta
tra created this revision.
tra added a reviewer: echristo.
tra added a subscriber: cfe-commits.
The class searches for a CUDA installation directory and provides accessors for
relevant bits of info about it.
Added new option --cuda-path= which allows overriding default search
paths. If it's not
tra added a comment.
Any suggestions how to test this?
http://reviews.llvm.org/D12989
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
tra added a comment.
I'll add tests for install dir detection. As for detection of bitcode files,
perhaps I should remove it from this patch and commit it along with the code
that's going to use it and where I can test it.
http://reviews.llvm.org/D12989
_
tra updated the summary for this revision.
tra updated this revision to Diff 35333.
tra added a comment.
Added test case for CUDA detection.
Removed libdevice file detection for now.
http://reviews.llvm.org/D12989
Files:
include/clang/Driver/Options.td
lib/Driver/ToolChains.cpp
lib/Driver
Author: tra
Date: Tue Sep 22 12:22:51 2015
New Revision: 248293
URL: http://llvm.org/viewvc/llvm-project?rev=248293&view=rev
Log:
[CUDA] Add implicit __attribute__((used)) to all __global__ functions.
This makes sure that we emit kernels that were instantiated from the
host code and which would n
Author: tra
Date: Tue Sep 22 12:22:59 2015
New Revision: 248295
URL: http://llvm.org/viewvc/llvm-project?rev=248295&view=rev
Log:
[CUDA] Allow function overloads in CUDA based on host/device attributes.
The patch makes it possible to parse CUDA files that contain host/device
functions with identi
Author: tra
Date: Tue Sep 22 12:23:05 2015
New Revision: 248296
URL: http://llvm.org/viewvc/llvm-project?rev=248296&view=rev
Log:
[CUDA] Add appropriate host/device attribute to builtins.
The changes are part of attribute-based CUDA function overloading (D12453)
and as such are only enabled when
Author: tra
Date: Tue Sep 22 12:23:09 2015
New Revision: 248297
URL: http://llvm.org/viewvc/llvm-project?rev=248297&view=rev
Log:
[CUDA] Fixes minor cuda-related issues in the driver
* Only the last of the --cuda-host-only/--cuda-device-only options has effect.
* CudaHostAction always wraps host-
Author: tra
Date: Tue Sep 22 12:23:13 2015
New Revision: 248298
URL: http://llvm.org/viewvc/llvm-project?rev=248298&view=rev
Log:
Augmented CudaHostAction to carry device-side triple.
Differential Revision: http://reviews.llvm.org/D12893
Modified:
cfe/trunk/include/clang/Driver/Action.h
Author: tra
Date: Tue Sep 22 12:23:22 2015
New Revision: 248299
URL: http://llvm.org/viewvc/llvm-project?rev=248299&view=rev
Log:
[CUDA] Allow parsing of host and device code simultaneously.
* adds -aux-triple option to specify target triple
* propagates aux target info to AST context and Prepr
This revision was automatically updated to reflect the committed changes.
Closed by commit rL248295: [CUDA] Allow function overloads in CUDA based on
host/device attributes. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D12453?vs=34368&id=35387#toc
Repository:
rL LLVM
This revision was automatically updated to reflect the committed changes.
Closed by commit rL248298: Augmented CudaHostAction to carry device-side
triple. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D12893?vs=34857&id=35388#toc
Repository:
rL LLVM
http://reviews.llvm
This revision was automatically updated to reflect the committed changes.
Closed by commit rL248297: [CUDA] Fixes minor cuda-related issues in the driver
(authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D12892?vs=34855&id=35386#toc
Repository:
rL LLVM
http://reviews.llvm.
This revision was automatically updated to reflect the committed changes.
tra marked an inline comment as done.
Closed by commit rL248299: [CUDA] Allow parsing of host and device code
simultaneously. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D12917?vs=35031&id=35389#to
tra created this revision.
tra added reviewers: rsmith, echristo.
tra added a subscriber: cfe-commits.
Adjusted __global__ functions with DiscardableODR linkage to use StrongODR
linkage instead so they are visible externally.
Replaces D11666 / r248293
http://reviews.llvm.org/D13067
Files:
l
works much better. I've just sent http://reviews.llvm.org/D13067
--Artem
> On Sep 22, 2015 10:24 AM, "Artem Belevich via cfe-commits" <
> cfe-commits@lists.llvm.org> wrote:
>
>> Author: tra
>> Date: Tue Sep 22 12:22:51 2015
>> New Revision: 248293
&
Author: tra
Date: Wed Sep 23 12:44:53 2015
New Revision: 248400
URL: http://llvm.org/viewvc/llvm-project?rev=248400&view=rev
Log:
[CUDA] __global__ functions should always be visible externally.
Adjust __global__ functions with DiscardableODR linkage to use
StrongODR linkage instead, so they are
This revision was automatically updated to reflect the committed changes.
Closed by commit rL248400: [CUDA] __global__ functions should always be visible
externally. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D13067?vs=35412&id=35529#toc
Repository:
rL LLVM
http://r
tra updated this revision to Diff 35531.
tra added a comment.
Removed unneeded temporary variable.
http://reviews.llvm.org/D12989
Files:
include/clang/Driver/Options.td
lib/Driver/ToolChains.cpp
lib/Driver/ToolChains.h
test/Driver/Inputs/CUDA/usr/local/cuda/include/.keep
test/Driver/I
Author: tra
Date: Wed Sep 23 13:13:25 2015
New Revision: 248408
URL: http://llvm.org/viewvc/llvm-project?rev=248408&view=rev
Log:
[CUDA] Added CUDA installation detector class.
Added new option --cuda-path= which allows
overriding default search paths.
If it's not specified we look for CUDA insta
This revision was automatically updated to reflect the committed changes.
Closed by commit rL248408: [CUDA] Added CUDA installation detector class.
(authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D12989?vs=35531&id=35533#toc
Repository:
rL LLVM
http://reviews.llvm.org/D1
Author: tra
Date: Wed Sep 23 13:28:30 2015
New Revision: 248413
URL: http://llvm.org/viewvc/llvm-project?rev=248413&view=rev
Log:
Reverted r248408 "[CUDA] Added CUDA installation detector class."
because included test fails on some platforms.
Removed:
cfe/trunk/test/Driver/Inputs/CUDA/usr/lo
Author: tra
Date: Wed Sep 23 16:49:39 2015
New Revision: 248433
URL: http://llvm.org/viewvc/llvm-project?rev=248433&view=rev
Log:
[CUDA] Added CUDA installation detector class.
Added new option --cuda-path= which allows
overriding default search paths.
If it's not specified we look for CUDA insta
I'm looking into it. Can you reproduce it on linux?
On Wed, Sep 23, 2015 at 4:43 PM, Rui Ueyama wrote:
> On Wed, Sep 23, 2015 at 2:49 PM, Artem Belevich via cfe-commits <
> cfe-commits@lists.llvm.org> wrote:
>
>> Author: tra
>> Date: Wed Sep 23 16:49:39 2015
&
gt; On Wed, Sep 23, 2015 at 4:47 PM, Artem Belevich wrote:
>
>> I'm looking into it. Can you reproduce it on linux?
>>
>>
>> On Wed, Sep 23, 2015 at 4:43 PM, Rui Ueyama wrote:
>>
>>> On Wed, Sep 23, 2015 at 2:49 PM, Artem Belevich via cfe-commits <
;>> your Inputs/CUDA, and run "clang -v --target=i386-unknown-linux
>>> --sysroot=/Inputs/CUDA".
>>>
>>> On Wed, Sep 23, 2015 at 4:47 PM, Artem Belevich wrote:
>>>
>>>> I'm looking into it. Can you reproduce it on linux?
>>&g
Author: tra
Date: Wed Sep 23 19:15:47 2015
New Revision: 248448
URL: http://llvm.org/viewvc/llvm-project?rev=248448&view=rev
Log:
Disabled one of the test runs in cuda-detect.cu
--sysroot apparently can't handle long paths which triggers test
failure on some platforms.
Modified:
cfe/trunk/te
Thank you for the fix.
--Artem
On Wed, Sep 23, 2015 at 9:32 PM, NAKAMURA Takumi
wrote:
> chapuni added a subscriber: chapuni.
>
>
> Comment at: cfe/trunk/lib/Driver/ToolChains.cpp:1492
> @@ +1491,3 @@
> +const llvm::opt::ArgList
> &Ar
tra created this revision.
tra added a reviewer: echristo.
tra added a subscriber: cfe-commits.
Propagates AuxTriple throughout job pipeline construction in driver and on to
Tool::ConstructJob().
It in turn passes target triple of opposite side of CUDA compilation as
-aux-triple option to sub-co
tra created this revision.
tra added a reviewer: echristo.
tra added a subscriber: cfe-commits.
- added detection of libdevice bitcode file and API to find one appropriate
for the GPU we're compiling for.
- added include paths to detected CUDA installation
- added include paths for -aux-tri
tra created this revision.
tra added reviewers: echristo, eliben, jholewinski.
tra added a subscriber: cfe-commits.
Header files that come with CUDA are assuming split host/device compilation and
are not usable by clang out of the box.
With a bit of preprocessor magic it's possible to twist them
tra updated this revision to Diff 35747.
tra added a comment.
Fixed typos and whitespace nits.
use #pragma push_macro for __CUDACC_RTC__, too.
http://reviews.llvm.org/D13171
Files:
lib/Headers/CMakeLists.txt
lib/Headers/clang_cuda_support.h
Index: lib/Headers/clang_cuda_support.h
=
tra updated this revision to Diff 35760.
tra added a comment.
Include cuda.h and #error if we see wrong CUDA_VERSION.
http://reviews.llvm.org/D13171
Files:
lib/Headers/CMakeLists.txt
lib/Headers/clang_cuda_support.h
Index: lib/Headers/clang_cuda_support.h
==
tra marked an inline comment as done.
tra added a comment.
http://reviews.llvm.org/D13171
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
tra created this revision.
tra added reviewers: eliben, jholewinski, echristo.
tra added a subscriber: cfe-commits.
Herald added a subscriber: jholewinski.
Currently 32-bit NVPTX happens to have 64-bit long which does not match host's
long in 32-bit mode.
http://reviews.llvm.org/D13181
Files:
The (vague) idea was to make clear that the header is *not* part of cuda
distribution.
That said, the file could use a better name.
Do any of these sound better?
* fix_cuda_headers.h
* adapt_cuda_headers.h
* cuda_shim.h
--Artem
On Sat, Sep 26, 2015 at 12:20 PM, Eric Christopher
wrote:
> echri
Author: tra
Date: Mon Sep 28 17:54:08 2015
New Revision: 248753
URL: http://llvm.org/viewvc/llvm-project?rev=248753&view=rev
Log:
[CUDA] 32-bit NVPTX should have 32-bit long type.
Currently it's 64-bit which will lead to mismatch between host and
device code if we compile for i386.
Differential
This revision was automatically updated to reflect the committed changes.
Closed by commit rL248753: [CUDA] 32-bit NVPTX should have 32-bit long type.
(authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D13181?vs=35772&id=35918#toc
Repository:
rL LLVM
http://reviews.llvm.org
cuda_runtime.h may be a better choice. nvcc -includes it during both host
and device compilation and this wrapper file is intended to serve a similar
purpose and will probably end up being -included by cc1 in the end.
--Artem
On Mon, Sep 28, 2015 at 6:26 PM, Eric Christopher
wrote:
> echristo
tra updated this revision to Diff 36048.
tra added a comment.
Renamed wrapper to cuda_runtime.h
Similarly to nvcc, automatically add "-include cuda_runtime.h" to CC1
invocations unless -nocudainc is specified.
http://reviews.llvm.org/D13171
Files:
lib/Driver/ToolChains.cpp
lib/Headers/CMak
tra added a comment.
LGTM'ed by Aaron Ballman via email that didn't make it to Phabricator.
http://reviews.llvm.org/D11690
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
Author: tra
Date: Mon Aug 10 15:33:56 2015
New Revision: 244497
URL: http://llvm.org/viewvc/llvm-project?rev=244497&view=rev
Log:
[CUDA] Added stubs for new attributes used by CUDA headers.
The main purpose is to avoid errors and warnings while parsing CUDA
header files. The attributes are curren
Author: tra
Date: Mon Aug 10 15:57:02 2015
New Revision: 244501
URL: http://llvm.org/viewvc/llvm-project?rev=244501&view=rev
Log:
[CUDA] Add implicit __attribute__((used)) to all __global__ functions.
This allows emitting kernels that were instantiated from the host code
and which would never be
This revision was automatically updated to reflect the committed changes.
Closed by commit rL244501: [CUDA] Add implicit __attribute__((used)) to all
__global__ functions. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D11666?vs=31128&id=31718#toc
Repository:
rL LLVM
ht
tra created this revision.
tra added reviewers: eliben, echristo.
tra added a subscriber: cfe-commits.
The patch makes sure that register names in named register variables and inline
assembly are checked only on appropriate side of CUDA compilation.
http://reviews.llvm.org/D11950
Files:
lib/S
tra created this revision.
tra added reviewers: echristo, rsmith, eliben.
tra added a subscriber: cfe-commits.
This is a somewhat different way to do it than D11666 which got rolled back.
Codegen postpones emitting instantiated kernel function template until it's
used.
If kernel is used only fro
tra created this revision.
tra added reviewers: eliben, echristo.
tra added a subscriber: cfe-commits.
The patch adds appropriate __host__ or __device__ attributes to target-specific
builtins
so we can properly check whether they may or may not be called from particular
context.
http://review
tra updated this revision to Diff 32478.
tra marked an inline comment as done.
tra added a comment.
used isTSBuiltin in SemaChecking.cpp
http://reviews.llvm.org/D12122
Files:
include/clang/Basic/Builtins.h
lib/Sema/SemaChecking.cpp
lib/Sema/SemaDecl.cpp
test/SemaCUDA/builtins.cu
test/
tra updated this revision to Diff 32481.
tra added a comment.
Added a comment explaining reasoning behind attribute choice for
target-specific builtins.
http://reviews.llvm.org/D12122
Files:
include/clang/Basic/Builtins.h
lib/Sema/SemaChecking.cpp
lib/Sema/SemaDecl.cpp
test/SemaCUDA/bu
tra marked an inline comment as done.
Comment at: lib/Sema/SemaDecl.cpp:11166
@@ +11165,3 @@
+!FD->hasAttr() && !FD->hasAttr()) {
+ if (getLangOpts().CUDAIsDevice)
+FD->addAttr(CUDADeviceAttr::CreateImplicit(Context,
FD->getLocation()));
elib
This revision was automatically updated to reflect the committed changes.
Closed by commit rL245496: [CUDA] Add appropriate host/device attribute to
builtins. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D12122?vs=32481&id=32607#toc
Repository:
rL LLVM
http://reviews.
Author: tra
Date: Wed Aug 19 15:48:20 2015
New Revision: 245496
URL: http://llvm.org/viewvc/llvm-project?rev=245496&view=rev
Log:
[CUDA] Add appropriate host/device attribute to builtins.
Differential Revision: http://reviews.llvm.org/D12122
Added:
cfe/trunk/test/SemaCUDA/builtins.cu
Modifie
tra added a comment.
Ping.
http://reviews.llvm.org/D11694
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
tra updated this revision to Diff 32611.
tra added a comment.
Removed rarely used temp var.
http://reviews.llvm.org/D11950
Files:
lib/Sema/SemaDecl.cpp
lib/Sema/SemaStmtAsm.cpp
test/SemaCUDA/asm-constraints-mixed.cu
Index: test/SemaCUDA/asm-constraints-mixed.cu
==
tra added inline comments.
Comment at: lib/Sema/SemaDecl.cpp:5944
@@ -5943,3 +5943,3 @@
ProcessDeclAttributes(S, NewVD, D);
-
+ bool ShouldHandleTargetErrors = DeclAttrsMatchCUDAMode(getLangOpts(), NewVD);
if (getLangOpts().CUDA) {
eliben wrote:
> Since thi
Author: tra
Date: Wed Aug 19 16:11:37 2015
New Revision: 245502
URL: http://llvm.org/viewvc/llvm-project?rev=245502&view=rev
Log:
[CUDA] Added stubs for __nvvm_atom_add_*_d() builtins.
They show up in CUDA headers but are not currently supported by
NVPTX back-end.
Differential Revision: http://r
This revision was automatically updated to reflect the committed changes.
Closed by commit rL245502: [CUDA] Added stubs for __nvvm_atom_add_*_d()
builtins. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D11694?vs=31157&id=32616#toc
Repository:
rL LLVM
http://reviews.llv
Author: tra
Date: Thu Aug 20 13:28:56 2015
New Revision: 245592
URL: http://llvm.org/viewvc/llvm-project?rev=245592&view=rev
Log:
Revert r245496 "[CUDA] Add appropriate host/device attribute to builtins."
It's breaking internal test.
Removed:
cfe/trunk/test/SemaCUDA/builtins.cu
Modified:
tra added a comment.
Reverted in r245592 due to breaking internal tests.
Repository:
rL LLVM
http://reviews.llvm.org/D12122
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
tra abandoned this revision.
tra added a comment.
There is no way to close revision. Abandoning it to get it off my list of
pending changes.
http://reviews.llvm.org/D11690
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.or
tra added inline comments.
Comment at: lib/CodeGen/CodeGenAction.cpp:166-170
@@ +165,7 @@
+std::vector ModuleFuncNames;
+// We need to internalize contents of the linked module but it
+// has to be done *after* the linking because internalized
+// s
Author: tra
Date: Thu Aug 27 13:10:41 2015
New Revision: 246174
URL: http://llvm.org/viewvc/llvm-project?rev=246174&view=rev
Log:
[CUDA] Improve CUDA compilation pipeline creation.
Current implementation tries to guess which Action will result in a
job which needs to incorporate device-side GPU b
This revision was automatically updated to reflect the committed changes.
Closed by commit rL246174: [CUDA] Improve CUDA compilation pipeline creation.
(authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D11280?vs=29947&id=33341#toc
Repository:
rL LLVM
http://reviews.llvm.or
This revision was automatically updated to reflect the committed changes.
Closed by commit rL246193: [CUDA] Check register names on appropriate side of
cuda compilation only. (authored by tra).
Changed prior to commit:
http://reviews.llvm.org/D11950?vs=32611&id=33348#toc
Repository:
rL LLVM
Author: tra
Date: Thu Aug 27 14:54:21 2015
New Revision: 246193
URL: http://llvm.org/viewvc/llvm-project?rev=246193&view=rev
Log:
[CUDA] Check register names on appropriate side of cuda compilation only.
Differential Revision: http://reviews.llvm.org/D11950
Modified:
cfe/trunk/lib/Sema/SemaD
tra created this revision.
tra added reviewers: eliben, rsmith, jholewinski, echristo, jingyue.
tra added a subscriber: cfe-commits.
The patch makes it possible to parse CUDA files that contain host/device
functions with identical signatures, but different attributes without having
to physicall
401 - 500 of 1265 matches
Mail list logo
