This revision was automatically updated to reflect the committed changes.
Closed by commit rL275231: [CUDA] Don't assume that destructors can't be
overloaded. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21912?vs=62444&id=63749#toc
Repository:
rL LLVM
http://revie
Author: jlebar
Date: Tue Jul 12 18:23:12 2016
New Revision: 275232
URL: http://llvm.org/viewvc/llvm-project?rev=275232&view=rev
Log:
[CUDA] Add additional testcases for EraseUnwantedCUDAMatches.
Summary:
Specifically, this patch adds testcases for all three calls to
EraseUnwantedCUDAMatches. The
Author: jlebar
Date: Tue Jul 12 18:23:01 2016
New Revision: 275231
URL: http://llvm.org/viewvc/llvm-project?rev=275231&view=rev
Log:
[CUDA] Don't assume that destructors can't be overloaded.
Summary:
You can overload a destructor in CUDA, and SemaOverload needs to be
tweaked not to crash when it
Author: jlebar
Date: Fri Jul 8 12:59:24 2016
New Revision: 274897
URL: http://llvm.org/viewvc/llvm-project?rev=274897&view=rev
Log:
Fix flag name in comment in cuda-version-check.cu.
Modified:
cfe/trunk/test/Driver/cuda-version-check.cu
Modified: cfe/trunk/test/Driver/cuda-version-check.cu
Author: jlebar
Date: Thu Jul 7 13:24:28 2016
New Revision: 274782
URL: http://llvm.org/viewvc/llvm-project?rev=274782&view=rev
Log:
[CUDA] s/OPT_nocuda_version_chec/OPT_no_cuda_version_check/.
Fix build breakage.
Modified:
cfe/trunk/lib/Driver/ToolChains.cpp
cfe/trunk/lib/Driver/Tools.c
jlebar added a comment.
Friendly ping
http://reviews.llvm.org/D21912
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
This revision was automatically updated to reflect the committed changes.
Closed by commit rL274781: [CUDA] Check that our CUDA install supports the
requested architectures. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21869?vs=63094&id=63100#toc
Repository:
rL LLV
Author: jlebar
Date: Thu Jul 7 13:17:52 2016
New Revision: 274781
URL: http://llvm.org/viewvc/llvm-project?rev=274781&view=rev
Log:
[CUDA] Check that our CUDA install supports the requested architectures.
Summary:
Raise an error if you're using a CUDA installation that's too old for
the requeste
Author: jlebar
Date: Thu Jul 7 13:15:03 2016
New Revision: 274780
URL: http://llvm.org/viewvc/llvm-project?rev=274780&view=rev
Log:
[CUDA] Rename the __nvvm_bar0 builtin back to __syncthreads.
The builtin was renamed in r274770. But __syncthreads is part of our
user-facing API, so we need to ke
jlebar added inline comments.
Comment at: include/clang/Driver/Options.td:1722-1724
@@ -1721,2 +1721,5 @@
def nocudalib : Flag<["-"], "nocudalib">;
+def nocuda_version_check : Flag<["-"], "nocuda-version-check">,
+ HelpText<"Don't error out if the detected version of the CUDA in
jlebar updated this revision to Diff 63094.
jlebar marked 3 inline comments as done.
jlebar added a comment.
Address review comments.
http://reviews.llvm.org/D21869
Files:
include/clang/Basic/DiagnosticDriverKinds.td
include/clang/Driver/Options.td
lib/Driver/ToolChains.cpp
lib/Driver/T
Author: jlebar
Date: Wed Jul 6 20:06:59 2016
New Revision: 274713
URL: http://llvm.org/viewvc/llvm-project?rev=274713&view=rev
Log:
[CUDA] Fix "control reaches end of non-void function" warnings in Cuda.cpp.
Some compilers are too dumb to realize that the switch statement covers
all cases.
(Don
jlebar added a comment.
Friendly ping.
http://reviews.llvm.org/D21869
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
Author: jlebar
Date: Wed Jul 6 16:45:44 2016
New Revision: 274689
URL: http://llvm.org/viewvc/llvm-project?rev=274689&view=rev
Log:
[CUDA] Add missing namespace qualification on CudaArch in Action.cpp.
Fix build breakage with MSVC.
Modified:
cfe/trunk/lib/Driver/Action.cpp
Modified: cfe/tr
This revision was automatically updated to reflect the committed changes.
Closed by commit rL274681: [CUDA] Add utility functions for dealing with CUDA
versions / architectures. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21867?vs=62409&id=62975#toc
Repository:
rL
This revision was automatically updated to reflect the committed changes.
Closed by commit rL274682: [CUDA] Rename member variables in
CudaInstallationDetector. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21868?vs=62300&id=62976#toc
Repository:
rL LLVM
http://rev
Author: jlebar
Date: Wed Jul 6 16:21:43 2016
New Revision: 274682
URL: http://llvm.org/viewvc/llvm-project?rev=274682&view=rev
Log:
[CUDA] Rename member variables in CudaInstallationDetector.
Summary:
Remove the "Cuda" prefix from these variables -- it's clear that they
related to CUDA given the
Author: jlebar
Date: Wed Jul 6 16:21:39 2016
New Revision: 274681
URL: http://llvm.org/viewvc/llvm-project?rev=274681&view=rev
Log:
[CUDA] Add utility functions for dealing with CUDA versions / architectures.
Summary:
Currently our handling of CUDA architectures is scattered all around
clang. T
This revision was automatically updated to reflect the committed changes.
Closed by commit rL274680: [CUDA] Add support for CUDA 8 and sm_60-62.
(authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21778?vs=62908&id=62974#toc
Repository:
rL LLVM
http://reviews.llvm.org/D2
Author: jlebar
Date: Wed Jul 6 16:21:14 2016
New Revision: 274680
URL: http://llvm.org/viewvc/llvm-project?rev=274680&view=rev
Log:
[CUDA] Add support for CUDA 8 and sm_60-62.
Summary: Also add sm_32, which was missing.
Reviewers: tra
Subscribers: cfe-commits
Differential Revision: http://rev
jlebar added a comment.
> They will need to wait for corresponding patch on LLVM side to deal with new
> SM variants, though.
This is in now, http://reviews.llvm.org/D22068.
Comment at: lib/Driver/ToolChains.cpp:1715
@@ -1714,2 +1714,3 @@
CudaPathCandidates.push_back(D.S
jlebar updated this revision to Diff 62908.
jlebar marked an inline comment as done.
jlebar added a comment.
Don't pull in the cuda 8 headers by default.
http://reviews.llvm.org/D21778
Files:
lib/Basic/Targets.cpp
lib/Driver/Action.cpp
lib/Driver/ToolChains.cpp
Index: lib/Driver/ToolChai
jlebar added a comment.
Hi, Alexy. Would you mind not asking for 'final' in additional reviews until
we've resolved this thread elsewhere? Feel free to find me on IRC if you want
to talk about it synchronously.
Thanks!
http://reviews.llvm.org/D21845
__
jlebar added a subscriber: jlebar.
jlebar added a comment.
Yeah, I'd say that in the absence of a rule, consistency with surrounding
code is king. Otherwise we're sending a message when we don't mean to be.
I'm not at my machine, but my recollection is that most of the driver uses
final sparingl
Yeah, I'd say that in the absence of a rule, consistency with surrounding
code is king. Otherwise we're sending a message when we don't mean to be.
I'm not at my machine, but my recollection is that most of the driver uses
final sparingly. But whatever the convention is we should do that, I thin
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: bkramer, cfe-commits.
Bug pointed out by Benjamin Kramer in r264008. I think the bug is
benign because by the time this is called, we should only have at most
two overloads to consider (either a host and a devic
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
Specifically, this patch adds testcases for all three calls to
EraseUnwantedCUDAMatches. The addr-of-overloaded-fn test I accidentally
neutered in r264207, which moved much of
CodeGenCUDA/function-
jlebar created this revision.
jlebar added a reviewer: rsmith.
jlebar added subscribers: tra, cfe-commits.
You can overload a destructor in CUDA, and SemaOverload needs to be
tweaked not to crash when it sees an explicit call to an overloaded
destructor.
http://reviews.llvm.org/D21912
Files:
l
jlebar added a comment.
Alexey, it seems that you're asking for "final" on all classes that are not
inherited from. Forgive my ignorance, but would you mind pointing me to the
document that talks about our position on "final" in LLVM source? I don't see
it in the style guide, but I may be mis
t 5:08 AM, Benjamin Kramer wrote:
>> On Tue, Mar 22, 2016 at 1:09 AM, Justin Lebar via cfe-commits
>> wrote:
>>> Author: jlebar
>>> Date: Mon Mar 21 19:09:25 2016
>>> New Revision: 264008
>>>
>>> URL: http://llvm.or
1 test from 1 test case ran. (20 ms total)
> [ PASSED ] 0 tests.
> [ FAILED ] 1 test, listed below:
> [ FAILED ] DeclarationMatcher.MatchClass
>
> 1 FAILED TEST
>
>
>
>
>
> 2016-06-30 21:12 GMT+03:00 Justin Lebar via cfe-commits
> :
>>
jlebar marked 4 inline comments as done.
jlebar added a comment.
http://reviews.llvm.org/D21867
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
jlebar updated this revision to Diff 62409.
jlebar added a comment.
Address Art's review.
http://reviews.llvm.org/D21867
Files:
include/clang/Basic/Cuda.h
include/clang/Driver/Action.h
lib/Basic/CMakeLists.txt
lib/Basic/Cuda.cpp
lib/Basic/Targets.cpp
lib/Driver/Action.cpp
lib/Driv
Interestingly all the clang tests pass with that whole line commented
out. So something *really* seems missing here.
Thank you for finding this.
On Thu, Jun 30, 2016 at 5:08 AM, Benjamin Kramer wrote:
> On Tue, Mar 22, 2016 at 1:09 AM, Justin Lebar via cfe-commits
> wrote:
>> Au
Author: jlebar
Date: Thu Jun 30 15:29:29 2016
New Revision: 274269
URL: http://llvm.org/viewvc/llvm-project?rev=274269&view=rev
Log:
Fix ASTMatchersNodeTest to work on Windows.
It was failing because it had an explicit check for whether we're on
Windows.
There are a few other similar explicit ch
jlebar marked an inline comment as done.
Comment at: lib/Driver/Driver.cpp:1026-1028
@@ -1024,4 +1025,5 @@
} else if (CudaDeviceAction *CDA = dyn_cast(A)) {
-os << '"'
- << (CDA->getGpuArchName() ? CDA->getGpuArchName() : "(multiple archs)")
+os << '"' << (CDA->ge
jlebar marked an inline comment as done.
Comment at: lib/Basic/Cuda.cpp:8-19
@@ +7,14 @@
+
+const char *CudaVersionToString(CudaVersion V) {
+ switch (V) {
+ case CudaVersion::UNKNOWN:
+return "unknown";
+ case CudaVersion::CUDA_70:
+return "7.0";
+ case CudaVersion::C
Author: jlebar
Date: Thu Jun 30 13:41:33 2016
New Revision: 274261
URL: http://llvm.org/viewvc/llvm-project?rev=274261&view=rev
Log:
[CUDA] Give templated device functions internal linkage, templated kernels
external linkage.
Summary:
This lets LLVM perform IPO over these functions. In particul
This revision was automatically updated to reflect the committed changes.
Closed by commit rL274261: [CUDA] Give templated device functions internal
linkage, templated kernels… (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21337?vs=60728&id=62391#toc
Repository:
rL
This revision was automatically updated to reflect the committed changes.
Closed by commit rL274257: Don't instantiate a full host toolchain in
ASTMatchersTest. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21810?vs=62132&id=62385#toc
Repository:
rL LLVM
http://rev
Author: jlebar
Date: Thu Jun 30 13:12:25 2016
New Revision: 274257
URL: http://llvm.org/viewvc/llvm-project?rev=274257&view=rev
Log:
Don't instantiate a full host toolchain in ASTMatchersTest.
Summary:
This test was stat()'ing large swaths of /usr/lib hundreds of times, as
every invocation of mat
jlebar added a comment.
> But I think this is a reasonable workaround until such an API can be provided.
Should I take that as an LG, or are we waiting for someone else to approve this?
http://reviews.llvm.org/D21810
___
cfe-commits mailing list
c
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
Remove the "Cuda" prefix from these variables -- it's clear that they
related to CUDA given their containing type.
http://reviews.llvm.org/D21868
Files:
lib/Driver/ToolChains.cpp
lib/Driver/To
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
Raise an error if you're using a CUDA installation that's too old for
the requested architectures. In practice, this means that you need a
CUDA 8 install to compile for sm_6*.
http://reviews.llvm.
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
Currently our handling of CUDA architectures is scattered all around
clang. This patch centralizes it.
A key advantage of this centralization is that you can now write a C++
switch on e.g. CudaArc
jlebar updated this revision to Diff 62132.
jlebar marked 3 inline comments as done.
jlebar added a comment.
Fix typo in comment.
http://reviews.llvm.org/D21810
Files:
unittests/ASTMatchers/ASTMatchersTest.h
Index: unittests/ASTMatchers/ASTMatchersTest.h
=
jlebar added inline comments.
Comment at: unittests/ASTMatchers/ASTMatchersTest.h:81-83
@@ +80,5 @@
+ //
+ // FIXME: This is a hack to work around the fact that there's no way to do
the
+ // equivalent of runToolOnCodeWithArgs without instantiating a full Driver.
+ // We shou
jlebar created this revision.
jlebar added a reviewer: chandlerc.
jlebar added a subscriber: cfe-commits.
Herald added a subscriber: klimek.
This test was stat()'ing large swaths of /usr/lib hundreds of times, as
every invocation of matchesConditionally*() created a new Linux
toolchain.
In additi
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
Also add sm_32, which was missing.
http://reviews.llvm.org/D21778
Files:
lib/Basic/Targets.cpp
lib/Driver/Action.cpp
lib/Driver/ToolChains.cpp
Index: lib/Driver/ToolChains.cpp
=
jlebar added a comment.
Friendly ping.
http://reviews.llvm.org/D21337
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
jlebar added a subscriber: jlebar.
jlebar added a comment.
There seem to be many nontrivial whitespace errors introduced by this patch.
For example,
-Attrs.push_back(HTMLStartTagComment::Attribute(Ident.getLocation(),
-
Ident.ge
This revision was automatically updated to reflect the committed changes.
Closed by commit rL272857: [CUDA] Don't pass top-level -march down to device
cc1 or ptxas. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21419?vs=60932&id=60935#toc
Repository:
rL LLVM
http:/
Author: jlebar
Date: Wed Jun 15 18:46:11 2016
New Revision: 272857
URL: http://llvm.org/viewvc/llvm-project?rev=272857&view=rev
Log:
[CUDA] Don't pass top-level -march down to device cc1 or ptxas.
Summary:
Previously if you did e.g.
$ clang -march=haswell -x cuda foo.cu
we would pass "-march=
jlebar updated this revision to Diff 60932.
jlebar added a comment.
Fix tests for real this time.
http://reviews.llvm.org/D21419
Files:
lib/Driver/ToolChains.cpp
test/Driver/cuda-march.cu
Index: test/Driver/cuda-march.cu
===
-
jlebar added inline comments.
Comment at: test/Driver/cuda-march.cu:15-16
@@ +14,4 @@
+
+// RUN: %clang -### -target x86_64-linux-gnu -c -march=skylake
--cuda-gpu-arch=sm_30 %s 2>&1 | \
+// RUN: FileCheck -check-prefix SKYLAKE -check-prefix SM30 %s
+
tra wrote:
>
jlebar updated this revision to Diff 60931.
jlebar added a comment.
Remove redundant test.
http://reviews.llvm.org/D21419
Files:
lib/Driver/ToolChains.cpp
test/Driver/cuda-march.cu
Index: test/Driver/cuda-march.cu
===
--- /dev
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: echristo, cfe-commits.
Previously if you did e.g.
$ clang -march=haswell -x cuda foo.cu
we would pass "-march=haswell -march=sm_20" down to the ptxas tool.
This causes it to assert, and rightly so!
http://re
jlebar added a comment.
tra makes the good point that maybe this should be done in ASTContext, where we
already have a special case for __global__. (I think I gravitated to doing it
this way because the GVA* enums have zero documentation -- at least I have a
vague idea of what the LLVM attribu
jlebar created this revision.
jlebar added a reviewer: rsmith.
jlebar added subscribers: tra, cfe-commits.
This lets LLVM perform IPO over these functions. In particular, it
allows LLVM to emit ld.global.nc for loads to __restrict pointers in
kernels that are never written to.
http://reviews.llv
This revision was automatically updated to reflect the committed changes.
Closed by commit rL272299: [CUDA] Implement __shfl* intrinsics in clang
headers. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21162?vs=60223&id=60230#toc
Repository:
rL LLVM
http://reviews.l
Author: jlebar
Date: Thu Jun 9 15:04:57 2016
New Revision: 272299
URL: http://llvm.org/viewvc/llvm-project?rev=272299&view=rev
Log:
[CUDA] Implement __shfl* intrinsics in clang headers.
Summary: Clang changes to make use of the LLVM intrinsics added in D21160.
Reviewers: tra
Subscribers: jhole
jlebar added inline comments.
Comment at: lib/Headers/__clang_cuda_intrinsics.h:77-80
@@ +76,6 @@
+_Static_assert(sizeof(__tmp) == sizeof(__in));
\
+memcpy(&__tmp, &__in, sizeof(__in));
\
+__tmp = ::__
jlebar updated this revision to Diff 60223.
jlebar marked 2 inline comments as done.
jlebar added a comment.
Update after tra's review.
http://reviews.llvm.org/D21162
Files:
include/clang/Basic/BuiltinsNVPTX.def
lib/Headers/__clang_cuda_intrinsics.h
lib/Headers/__clang_cuda_runtime_wrappe
jlebar added a comment.
Thank you for the reviews, Justin!
http://reviews.llvm.org/D21162
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
jlebar added a comment.
(Art, I would appreciate a second set of eyes on this one, as the last time I
did this -- with ldg -- I messed up pretty badly.)
http://reviews.llvm.org/D21162
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://l
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: cfe-commits, jholewinski.
Clang changes to make use of the LLVM intrinsics added in D21160.
http://reviews.llvm.org/D21162
Files:
include/clang/Basic/BuiltinsNVPTX.def
lib/Headers/__clang_cuda_intrinsics.h
jlebar added inline comments.
Comment at: lib/Sema/SemaDeclAttr.cpp:4079
@@ +4078,3 @@
+ if (ValArg.isInvalid())
+return nullptr;
+
OK, so then we want an assert, not an if?
http://reviews.llvm.org/D20985
___
c
jlebar accepted this revision.
This revision is now accepted and ready to land.
Comment at: lib/Sema/SemaDeclAttr.cpp:4044
@@ +4043,3 @@
+// Checks whether an argument of launch_bounds attribute is
+// acceptable, performs implicit conversion to Rvalue and returns
+// non-nullptr
jlebar added a comment.
In http://reviews.llvm.org/D20985#448836, @tra wrote:
> In http://reviews.llvm.org/D20985#448822, @jlebar wrote:
>
> > How is this different from test/SemaCUDA/launch_bounds.cu:27-28? It does
> >
> > const int constint = 512;
> > __launch_bounds__(constint) void TestC
jlebar added a comment.
How is this different from test/SemaCUDA/launch_bounds.cu:27-28? It does
const int constint = 512;
__launch_bounds__(constint) void TestConstInt(void);
which looks verbatim the same as this testcase.
http://reviews.llvm.org/D20985
___
Thank you, Tom. I will have a look.
On Wed, Jun 1, 2016 at 11:22 AM, Tom Stellard wrote:
> On Tue, May 31, 2016 at 09:27:13PM -0000, Justin Lebar via cfe-commits wrote:
>> Author: jlebar
>> Date: Tue May 31 16:27:13 2016
>> New Revision: 271336
>>
>> URL: http:/
jlebar added a comment.
In http://reviews.llvm.org/D20836#444911, @tra wrote:
> I guess we would not be able to remove convergent from inline asm
> automatically. Do we need a way to explicitly remove convergent from inline
> asm?
We can think about it. I'm not sure it will make a big differ
This revision was automatically updated to reflect the committed changes.
Closed by commit rL271336: [CUDA] Conservatively mark inline asm as convergent.
(authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D20836?vs=59130&id=59133#toc
Repository:
rL LLVM
http://reviews.ll
Author: jlebar
Date: Tue May 31 16:27:13 2016
New Revision: 271336
URL: http://llvm.org/viewvc/llvm-project?rev=271336&view=rev
Log:
[CUDA] Conservatively mark inline asm as convergent.
Summary:
This is particularly important because a some convergent CUDA intrinsics
(e.g. __shfl_down) are imple
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
This is particularly important because a some convergent CUDA intrinsics
(e.g. __shfl_down) are implemented in terms of inline asm.
http://reviews.llvm.org/D20836
Files:
lib/CodeGen/CGStmt.cpp
This revision was automatically updated to reflect the committed changes.
Closed by commit rL271215: [CUDA] Fix order of vectorized ldg intrinsics'
elements. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D20794?vs=58972&id=58976#toc
Repository:
rL LLVM
http://review
Author: jlebar
Date: Mon May 30 12:12:55 2016
New Revision: 271215
URL: http://llvm.org/viewvc/llvm-project?rev=271215&view=rev
Log:
[CUDA] Fix order of vectorized ldg intrinsics' elements.
Summary: The order is [x, y, z, w], not [w, x, y, z].
Subscribers: cfe-commits, tra
Differential Revision
jlebar created this revision.
jlebar added subscribers: tra, cfe-commits.
The order is [x, y, z, w], not [w, x, y, z].
http://reviews.llvm.org/D20794
Files:
lib/Headers/__clang_cuda_intrinsics.h
Index: lib/Headers/__clang_cuda_intrinsics.h
=
This revision was automatically updated to reflect the committed changes.
Closed by commit rL270484: [CUDA] Add -fcuda-approx-transcendentals flag.
(authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D20493?vs=58123&id=58145#toc
Repository:
rL LLVM
http://reviews.llvm.org
Author: jlebar
Date: Mon May 23 15:19:56 2016
New Revision: 270484
URL: http://llvm.org/viewvc/llvm-project?rev=270484&view=rev
Log:
[CUDA] Add -fcuda-approx-transcendentals flag.
Summary:
This lets us emit e.g. sin.approx.f32. See
http://docs.nvidia.com/cuda/parallel-thread-execution/#floating-
jlebar updated this revision to Diff 58123.
jlebar added a comment.
More tightly scope the __USE_FAST_MATH__ macro.
tra pointed out that device_functions.hpp uses __USE_FAST_MATH__ for its own
purposes. For this CL, we only want to define __USE_FAST_MATH__ around
math_functions.hpp.
http://rev
jlebar marked 2 inline comments as done.
jlebar added a comment.
Repository:
rL LLVM
http://reviews.llvm.org/D20457
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
jlebar added a comment.
Thank you for the review!
http://reviews.llvm.org/D20457
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
Author: jlebar
Date: Fri May 20 16:33:01 2016
New Revision: 270279
URL: http://llvm.org/viewvc/llvm-project?rev=270279&view=rev
Log:
Update -ffast-math documentation to match reality.
Reviewers: rsmith
Subscribers: cfe-commits
Differential Revision: http://reviews.llvm.org/D20457
Modified:
This revision was automatically updated to reflect the committed changes.
Closed by commit rL270279: Update -ffast-math documentation to match reality.
(authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D20457?vs=57958&id=57995#toc
Repository:
rL LLVM
http://reviews.llvm
jlebar created this revision.
jlebar added a reviewer: rnk.
jlebar added subscribers: cfe-commits, tra.
This lets us emit e.g. sin.approx.f32. See
http://docs.nvidia.com/cuda/parallel-thread-execution/#floating-point-instructions-sin
http://reviews.llvm.org/D20493
Files:
include/clang/Basic/L
jlebar abandoned this revision.
jlebar added a comment.
Actually, after talking offline with Chandler, I need something more
complicated than this. I will send a new patch.
Sorry for the noise.
http://reviews.llvm.org/D20481
___
cfe-commits maili
jlebar created this revision.
jlebar added a reviewer: rsmith.
jlebar added a subscriber: cfe-commits.
The CUDA headers look for __USE_FAST_MATH__.
http://reviews.llvm.org/D20481
Files:
lib/Headers/__clang_cuda_runtime_wrapper.h
Index: lib/Headers/__clang_cuda_runtime_wrapper.h
==
jlebar updated this revision to Diff 57958.
jlebar added a comment.
Update per Richard's review.
http://reviews.llvm.org/D20457
Files:
docs/UsersManual.rst
include/clang/Basic/LangOptions.def
include/clang/Driver/Options.td
Index: include/clang/Driver/Options.td
=
jlebar created this revision.
jlebar added a reviewer: rsmith.
jlebar added a subscriber: cfe-commits.
http://reviews.llvm.org/D20457
Files:
include/clang/Basic/LangOptions.def
include/clang/Driver/Options.td
Index: include/clang/Driver/Options.td
This revision was automatically updated to reflect the committed changes.
Closed by commit rL270150: [CUDA] Implement __ldg using intrinsics. (authored
by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D19990?vs=56603&id=57873#toc
Repository:
rL LLVM
http://reviews.llvm.org/D1999
Author: jlebar
Date: Thu May 19 17:49:13 2016
New Revision: 270150
URL: http://llvm.org/viewvc/llvm-project?rev=270150&view=rev
Log:
[CUDA] Implement __ldg using intrinsics.
Summary:
Previously it was implemented as inline asm in the CUDA headers.
This change allows us to use the [addr+imm] addr
jlebar accepted this revision.
jlebar added a comment.
This revision is now accepted and ready to land.
Well, if the CUDA documentation says so...let's do it. :) Thanks for your
patience, everyone.
http://reviews.llvm.org/D20341
___
cfe-commits m
jlebar added a comment.
> But people also don't expect IEEE compliance on GPUs
Is that true? You have a lot more experience with this than I do, but my
observation of nvidia's hardware is that it's moved to add *more* IEEE
compliance as it's matured. For example, older hardware didn't suppor
jlebar added a comment.
I am not sure we want this? Although it matches nvcc, it does not match our
floating-point behavior for C++ in general -- it makes us non-IEEE-whatever
compliant by default.
Although I agree that if we don't do this, lots of people are not going to pass
-fp-contract=fa
jlebar added a comment.
Friendly ping. This is a big help with some Tensorflow benchmarks.
http://reviews.llvm.org/D19990
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
jlebar added a comment.
This patch regresses Eigen, because it raises an error even on host+device
functions.
Repository:
rL LLVM
http://reviews.llvm.org/D20034
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bi
jlebar accepted this revision.
jlebar added a comment.
This revision is now accepted and ready to land.
OK, if the function explicitly says it accepts null values and if we check
elsewhere in the function, I'm personally OK adding the checks.
http://reviews.llvm.org/D20141
__
jlebar added a comment.
Can we have a test?
http://reviews.llvm.org/D20141
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
jlebar accepted this revision.
jlebar added a comment.
This revision is now accepted and ready to land.
lgtm, but I'd like Richard to sign off on this too.
Comment at: lib/Sema/SemaDecl.cpp:10438
@@ -10437,1 +10437,3 @@
+ // Also make sure that destructor, ifthere is one,
401 - 500 of 874 matches
Mail list logo
