jlebar marked an inline comment as done.
Comment at: clang/lib/Sema/SemaExprCXX.cpp:688
@@ +687,3 @@
+ if (getLangOpts().CUDA)
+CheckCUDAExceptionExpr(OpLoc, "throw");
+
tra wrote:
> Do you need/want to check returned result?
We could, and we could return Exp
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: jhen, cfe-commits.
https://reviews.llvm.org/D25036
Files:
clang/include/clang/Basic/DiagnosticSemaKinds.td
clang/include/clang/Sema/Sema.h
clang/lib/Sema/SemaCUDA.cpp
clang/lib/Sema/SemaExprCXX.cpp
cla
jlebar added a comment.
> What happens if there are trailing tokens after the pragma?
Added code to make this an error.
https://reviews.llvm.org/D24975
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/l
jlebar updated this revision to Diff 72734.
jlebar marked 2 inline comments as done.
jlebar added a comment.
Address Richard's comments.
I'm fairly neutral on whether we want to make it an error not to match all of
your "begin" pragmas with "end"s. I checked pragma push_macro, and it looks
like
jlebar updated this revision to Diff 72719.
jlebar added a comment.
Herald added a subscriber: mehdi_amini.
s/libgcc/runtime/
https://reviews.llvm.org/D24979
Files:
clang/lib/Driver/ToolChains.cpp
clang/lib/Headers/CMakeLists.txt
clang/lib/Headers/__clang_cuda_complex_builtins.h
clang/l
jlebar updated this revision to Diff 72717.
jlebar marked 2 inline comments as done.
jlebar added a comment.
Address Richard Smith's review comments:
- Change macro format.
- Add tests (these Just Worked).
https://reviews.llvm.org/D24975
Files:
clang/include/clang/Basic/DiagnosticParseKinds.
jlebar added inline comments.
Comment at: lib/Headers/__clang_cuda_runtime_wrapper.h:139
@@ -137,1 +138,3 @@
+// CUDA 8.0.41 relies on __USE_FAST_MATH__ and __CUDA_PREC_DIV's values
+// Previous versions used to check thether they are defined or not.
Nit, missin
jlebar added a comment.
> I' personally would prefer to force-include these files. I suspect it will
> not change things much as we already include a lot.
We have already had bugs filed by users whose root cause was that we #included
more things than nvcc #includes. I know exact compatibility
jlebar added a comment.
> That is way too much knowledge about details of standard library
> implementation.
Honestly I think this looks a lot scarier than it is. Or, to be specific, I
think we are already relying on implementation details much more implicit and
fragile than what is explicit
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: cfe-commits, jhen.
Herald added subscribers: mgorny, beanz.
We do this by wrapping and .
Tests are in the test-suite. support to come separately.
https://reviews.llvm.org/D24979
Files:
clang/lib/Driver/To
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: jhen, cfe-commits.
Currently we declare our inline __device__ math functions in namespace
std. But libstdc++ and libc++ declare these functions in an inline
namespace inside namespace std. We need to match this
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
Herald added subscribers: mgorny, beanz.
This matches the idiom we use for our other CUDA wrapper headers.
https://reviews.llvm.org/D24978
Files:
clang/lib/Frontend/CompilerInvocation.cpp
clan
jlebar created this revision.
jlebar added a reviewer: rsmith.
jlebar added subscribers: cfe-commits, jhen, tra.
These cause us to consider all functions in-between to be __host__
__device__.
You can nest these pragmas; you just can't have more 'end's than
'begin's.
https://reviews.llvm.org/D249
jlebar added a comment.
> I was not able to figure out how to comandeer a revision, so i just went
> ahead and pushed it.
Under "leap into action", one of the options is to commandeer the revision.
https://reviews.llvm.org/D9168
___
cfe-commits m
jlebar added a subscriber: jlebar.
jlebar added a comment.
FWIW I have run into this in the past and just not managed to muster up the
energy to fix it. So, thank you!
https://reviews.llvm.org/D9168
___
cfe-commits mailing list
cfe-commits@lists.l
This revision was automatically updated to reflect the committed changes.
Closed by commit rL281680: [CUDA] Don't try to run sanitizers on NVPTX.
(authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D24640?vs=71572&id=71573#toc
Repository:
rL LLVM
https://reviews.llvm.org
Author: jlebar
Date: Thu Sep 15 18:44:13 2016
New Revision: 281680
URL: http://llvm.org/viewvc/llvm-project?rev=281680&view=rev
Log:
[CUDA] Don't try to run sanitizers on NVPTX.
Summary:
Sanitizers aren't supported on NVPTX -- don't try to run them.
This lets you e.g. pass -fsanitize=address and
jlebar created this revision.
jlebar added a reviewer: kcc.
jlebar added subscribers: jhen, tra, cfe-commits.
Sanitizers aren't supported on NVPTX -- don't try to run them.
This lets you e.g. pass -fsanitize=address and get asan on your host
code.
https://reviews.llvm.org/D24640
Files:
clang/
This revision was automatically updated to reflect the committed changes.
Closed by commit rL281552: [test-suite] [CUDA] Update README. (authored by
jlebar).
Changed prior to commit:
https://reviews.llvm.org/D24590?vs=71445&id=71457#toc
Repository:
rL LLVM
https://reviews.llvm.org/D24590
F
jlebar marked an inline comment as done.
jlebar added a comment.
Will add. Thank you for the reviews!
https://reviews.llvm.org/D24590
___
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 rL281544: [CUDA] Make __clang_cuda_cmath.h compatible with
libc++. (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D24588?vs=71443&id=71447#toc
Repository:
rL LLVM
https://revi
This revision was automatically updated to reflect the committed changes.
Closed by commit rL281543: [CUDA] Add test checking our ability to take a
function pointer to a… (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D24581?vs=71423&id=71446#toc
Repository:
rL LLVM
Author: jlebar
Date: Wed Sep 14 16:50:11 2016
New Revision: 281543
URL: http://llvm.org/viewvc/llvm-project?rev=281543&view=rev
Log:
[CUDA] Add test checking our ability to take a function pointer to a __global__
function on the host side.
Summary: This functionality is used by Thrust.
Reviewer
Author: jlebar
Date: Wed Sep 14 16:50:14 2016
New Revision: 281544
URL: http://llvm.org/viewvc/llvm-project?rev=281544&view=rev
Log:
[CUDA] Make __clang_cuda_cmath.h compatible with libc++.
Summary:
We need to add a bunch more "using"s, which weren't necessary with
libstdc++.
Once this is in I c
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
Fix typo (s/Externals/External/), and mention cuda-tests-simple.
https://reviews.llvm.org/D24590
Files:
External/CUDA/README
Index: External/CUDA/README
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
We need to add a bunch more "using"s, which weren't necessary with
libstdc++.
Once this is in I can check in a test to the test-suite.
https://reviews.llvm.org/D24588
Files:
clang/lib/Headers/_
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
This functionality is used by Thrust.
https://reviews.llvm.org/D24581
Files:
clang/test/SemaCUDA/reference-to-kernel-fn.cu
Index: clang/test/SemaCUDA/reference-to-kernel-fn.cu
=
jlebar created this revision.
jlebar added a reviewer: rsmith.
jlebar added subscribers: tra, cfe-commits.
Move CheckCUDACall from ActOnCallExpr and BuildDeclRefExpr to
DiagnoseUseOfDecl. This lets us catch some edge cases we were missing,
specifically around class operators.
This necessitates a
jlebar created this revision.
jlebar added a reviewer: rsmith.
jlebar added subscribers: tra, cfe-commits.
We'd attempted to allow this, but turns out we were doing a very bad
job. :)
Making this work properly would be a giant change in clang. For
example, we'd need to make CXXRecordDecl::getDe
jlebar accepted this revision.
jlebar added a comment.
This revision is now accepted and ready to land.
Yay, this is great.
Comment at: test/SemaCUDA/target_attr_inheritance.cu:4
@@ +3,3 @@
+// REQUIRES: x86-registered-target
+// REQUIRES: nvptx-registered-target
+
-
This revision was automatically updated to reflect the committed changes.
Closed by commit rL281089: [CUDA] Make __GCC_ATOMIC_XXX_LOCK_FREE macros the
same on host/device. (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D24407?vs=70857&id=70895#toc
Repository:
rL LLVM
Author: jlebar
Date: Fri Sep 9 15:35:43 2016
New Revision: 281089
URL: http://llvm.org/viewvc/llvm-project?rev=281089&view=rev
Log:
[CUDA] Make __GCC_ATOMIC_XXX_LOCK_FREE macros the same on host/device.
Summary:
This fixes a bug where we were unable to compile the following CUDA
file with libstd
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: jhen, cfe-commits.
This fixes a bug where we were unable to compile the following CUDA
file with libstdc++ (didn't try libc++):
#include
void foo() { std::shared_future x; }
The problem is that only defin
This revision was automatically updated to reflect the committed changes.
Closed by commit rL279140: [CUDA] Improve handling of math functions. (authored
by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23627?vs=68427&id=68600#toc
Repository:
rL LLVM
https://reviews.llvm.org/D
Author: jlebar
Date: Thu Aug 18 15:43:13 2016
New Revision: 279140
URL: http://llvm.org/viewvc/llvm-project?rev=279140&view=rev
Log:
[CUDA] Improve handling of math functions.
Summary:
A bunch of related changes here to our CUDA math headers.
- The second arg to nexttoward is a double (well, tec
jlebar added a comment.
These changes have always been kind of scary. tra tested this against Thrust
all combinations of CUDA 7.0/7.5, c++98/11,
libc++/libstdc++{4.8.5/4.9.3,5.3.0}. So we should be good here. I hope.
https://reviews.llvm.org/D23627
___
jlebar added inline comments.
Comment at: clang/lib/Headers/__clang_cuda_cmath.h:125-133
@@ -122,8 +124,11 @@
__DEVICE__ float modf(float __x, float *__iptr) { return ::modff(__x, __iptr);
}
-__DEVICE__ float nexttoward(float __from, float __to) {
+__DEVICE__ float nexttoward(fl
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
A bunch of related changes here to our CUDA math headers.
- The second arg to nexttoward is a double (well, technically, long
double, but we don't have that), not a float.
- Add a forward-declar
Author: jlebar
Date: Mon Aug 15 19:48:21 2016
New Revision: 278767
URL: http://llvm.org/viewvc/llvm-project?rev=278767&view=rev
Log:
[CUDA] Fix "declared here" note on deferred wrong-side errors.
Previously we weren't deferring these "declared here" notes, which is
obviously wrong.
Modified:
This revision was automatically updated to reflect the committed changes.
Closed by commit rL278759: [CUDA] Raise an error if a wrong-side call is
codegen'ed. (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23242?vs=67378&id=68103#toc
Repository:
rL LLVM
https://rev
Author: jlebar
Date: Mon Aug 15 18:00:49 2016
New Revision: 278759
URL: http://llvm.org/viewvc/llvm-project?rev=278759&view=rev
Log:
[CUDA] Raise an error if a wrong-side call is codegen'ed.
Summary:
Some function calls in CUDA are allowed to appear in
semantically-correct programs but are an err
jlebar added a comment.
Wow, I have no idea if this is right. I can try to figure it out, but if
sfantao can review, that's easier.
Comment at: test/Driver/cuda-bindings.cu:10
@@ +9,3 @@
+// way. Instead we check whether we've generated a permanent name on
+// device side whic
This revision was automatically updated to reflect the committed changes.
Closed by commit rL278735: Add the notion of deferred diagnostics. (authored by
jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23241?vs=68063&id=68073#toc
Repository:
rL LLVM
https://reviews.llvm.org/D232
This revision was automatically updated to reflect the committed changes.
Closed by commit rL278734: [CUDA] Include CUDA headers before anything else.
(authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23341?vs=67439&id=68072#toc
Repository:
rL LLVM
https://reviews.llv
This revision was automatically updated to reflect the committed changes.
Closed by commit rL278733: [CUDA] Fix CUDA install version parsing. (authored
by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23340?vs=67438&id=68071#toc
Repository:
rL LLVM
https://reviews.llvm.org/D23
Author: jlebar
Date: Mon Aug 15 15:38:52 2016
New Revision: 278734
URL: http://llvm.org/viewvc/llvm-project?rev=278734&view=rev
Log:
[CUDA] Include CUDA headers before anything else.
Summary:
There's no point to --cuda-path if we then go and include /usr/include
first. And if you install the rig
Author: jlebar
Date: Mon Aug 15 15:38:56 2016
New Revision: 278735
URL: http://llvm.org/viewvc/llvm-project?rev=278735&view=rev
Log:
Add the notion of deferred diagnostics.
Summary:
This patch lets you create diagnostics that are emitted if and only if a
particular FunctionDecl is codegen'ed.
Th
Author: jlebar
Date: Mon Aug 15 15:38:48 2016
New Revision: 278733
URL: http://llvm.org/viewvc/llvm-project?rev=278733&view=rev
Log:
[CUDA] Fix CUDA install version parsing.
Summary:
getAsInteger returns true on error. Oops.
No test because the behavior at the moment is identical with or withou
jlebar updated this revision to Diff 68063.
jlebar added a comment.
Move deferred diags storage into ASTContext.
https://reviews.llvm.org/D23241
Files:
clang/include/clang/AST/ASTContext.h
clang/include/clang/AST/Decl.h
clang/lib/AST/Decl.cpp
clang/lib/CodeGen/CodeGenModule.cpp
clang/
jlebar added inline comments.
Comment at: lib/CodeGen/CGCUDANV.cpp:62
@@ -60,2 +61,3 @@
llvm::ConstantInt::get(SizeTy, 0)};
-auto ConstStr = CGM.GetAddrOfConstantCString(Str, Name.c_str());
+auto ConstStr = CGM.GetAddrOfConstantCString(Str,
This revision was automatically updated to reflect the committed changes.
Closed by commit rL278194: [CUDA] Add __device__ overloads for placement new
and delete. (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23239?vs=67282&id=67447#toc
Repository:
rL LLVM
https:/
This revision was automatically updated to reflect the committed changes.
Closed by commit rL278195: [CUDA] Print a "previous-decl" note when calling an
illegal member fn. (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23240?vs=67081&id=67448#toc
Repository:
rL LLVM
This revision was automatically updated to reflect the committed changes.
Closed by commit rL278196: [CUDA] Reject calls to __device__ functions from
host variable global… (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23335?vs=67435&id=67449#toc
Repository:
rL LLVM
Author: jlebar
Date: Tue Aug 9 20:09:21 2016
New Revision: 278196
URL: http://llvm.org/viewvc/llvm-project?rev=278196&view=rev
Log:
[CUDA] Reject calls to __device__ functions from host variable global
initializers.
Reviewers: tra
Subscribers: cfe-commits
Differential Revision: https://review
Author: jlebar
Date: Tue Aug 9 20:09:11 2016
New Revision: 278193
URL: http://llvm.org/viewvc/llvm-project?rev=278193&view=rev
Log:
[CUDA] Rename CheckCUDATarget to IsAllowedCUDACall. NFC
Summary:
I want to reuse "CheckCUDAFoo" in a later patch. Also, I think
IsAllowedCUDACall gets the point ac
Author: jlebar
Date: Tue Aug 9 20:09:07 2016
New Revision: 278192
URL: http://llvm.org/viewvc/llvm-project?rev=278192&view=rev
Log:
[Diag] Fix idiom in comment: "on the lam", not "on the lamb".
Modified:
cfe/trunk/lib/Basic/Diagnostic.cpp
Modified: cfe/trunk/lib/Basic/Diagnostic.cpp
URL:
h
Author: jlebar
Date: Tue Aug 9 20:09:14 2016
New Revision: 278194
URL: http://llvm.org/viewvc/llvm-project?rev=278194&view=rev
Log:
[CUDA] Add __device__ overloads for placement new and delete.
Summary:
Previously these sort of worked because they didn't end up resulting in
calls at the ptx laye
Author: jlebar
Date: Tue Aug 9 20:09:18 2016
New Revision: 278195
URL: http://llvm.org/viewvc/llvm-project?rev=278195&view=rev
Log:
[CUDA] Print a "previous-decl" note when calling an illegal member fn.
Summary:
When we emit err_ref_bad_target, we should emit a "'method' declared
here" note. We
This revision was automatically updated to reflect the committed changes.
Closed by commit rL278193: [CUDA] Rename CheckCUDATarget to IsAllowedCUDACall.
NFC (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D23238?vs=67079&id=67446#toc
Repository:
rL LLVM
https://revie
Author: jlebar
Date: Tue Aug 9 19:40:43 2016
New Revision: 278183
URL: http://llvm.org/viewvc/llvm-project?rev=278183&view=rev
Log:
[CUDA] Minor comment nits.
Modified:
cfe/trunk/lib/Sema/SemaOverload.cpp
Modified: cfe/trunk/lib/Sema/SemaOverload.cpp
URL:
http://llvm.org/viewvc/llvm-projec
Author: jlebar
Date: Tue Aug 9 19:40:46 2016
New Revision: 278184
URL: http://llvm.org/viewvc/llvm-project?rev=278184&view=rev
Log:
[CUDA] Fix out-of-date comment.
LangOpts.CUDADisableTargetCallChecks no longer exists (it's effectively
always true now).
Modified:
cfe/trunk/include/clang/Sem
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: Prazek, cfe-commits.
There's no point to --cuda-path if we then go and include /usr/include
first. And if you install the right packages, Ubuntu will install (very
old) CUDA headers there.
https://reviews.llvm.
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added subscribers: Prazek, cfe-commits.
getAsInteger returns true on error. Oops.
No test because the behavior at the moment is identical with or without
this change.
https://reviews.llvm.org/D23340
Files:
clang/lib/Driver/To
jlebar updated this revision to Diff 67435.
jlebar marked an inline comment as done.
jlebar added a comment.
Test tweak per review.
https://reviews.llvm.org/D23335
Files:
clang/include/clang/Basic/DiagnosticSemaKinds.td
clang/lib/Sema/SemaDecl.cpp
clang/test/SemaCUDA/global-initializers-h
jlebar added inline comments.
Comment at: clang/test/SemaCUDA/global-initializers-host.cu:2
@@ +1,3 @@
+// RUN: %clang_cc1 %s --std=c++11 -triple x86_64-linux-unknown -emit-llvm -o -
-verify
+
+#include "Inputs/cuda.h"
tra wrote:
> I'd add the same comment as in
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
https://reviews.llvm.org/D23335
Files:
clang/include/clang/Basic/DiagnosticSemaKinds.td
clang/lib/Sema/SemaDecl.cpp
clang/test/SemaCUDA/call-device-fn-from-host.cu
clang/test/SemaCUDA/globa
jlebar added a comment.
> Hm, these seem completely broken, before and after this patch. I'll handle it
> separately if that's OK.
https://reviews.llvm.org/D23335. The device side equivalent works afaict (and
is very well-tested).
https://reviews.llvm.org/D23242
_
jlebar updated this revision to Diff 67378.
jlebar added a comment.
Address review comments.
Also swap the two tests -- I had the wrong names on them.
https://reviews.llvm.org/D23242
Files:
clang/include/clang/Sema/Sema.h
clang/lib/Sema/SemaCUDA.cpp
clang/lib/Sema/SemaDeclCXX.cpp
clang
jlebar marked 3 inline comments as done.
jlebar added a comment.
> Should be add few tests for calling device functions from host-side global
> initializers? Perhaps for device->host, too, as there may be unexpected
> interplay with constructor emptiness checks.
Hm, these seem completely broke
jlebar added a comment.
Reid, I'd still like you to have a look at this one if you don't mind, since
it's outside my and Art's core competencies.
Comment at: clang/lib/CodeGen/CodeGenModule.cpp:2886
@@ +2885,3 @@
+ // Check if this function has diagnostics that should be emitt
jlebar updated this revision to Diff 67367.
jlebar marked an inline comment as done.
jlebar added a comment.
Address review comments, and don't abort codegen'ing a function if it has only
deferred warnings -- check specifically for errors.
https://reviews.llvm.org/D23241
Files:
clang/include/
jlebar added a comment.
In https://reviews.llvm.org/D23239#508863, @tra wrote:
> I think we need to add `noexcept` for these in c++11.
Argh, you are completely right.
Thank you for the review.
https://reviews.llvm.org/D23239
___
cfe-commits mail
jlebar updated this revision to Diff 67282.
jlebar added a comment.
Add noexcept.
https://reviews.llvm.org/D23239
Files:
clang/lib/Headers/__clang_cuda_runtime_wrapper.h
Index: clang/lib/Headers/__clang_cuda_runtime_wrapper.h
==
jlebar created this revision.
jlebar added reviewers: tra, rnk.
jlebar added a subscriber: cfe-commits.
Some function calls in CUDA are allowed to appear in
semantically-correct programs but are an error if they're ever
codegen'ed. Specifically, a host+device function may call a host
function, bu
jlebar created this revision.
jlebar added a reviewer: rnk.
jlebar added subscribers: tra, cfe-commits.
This patch lets you create diagnostics that are emitted if and only if a
particular FunctionDecl is codegen'ed.
This is necessary for CUDA, where some constructs -- e.g. calls from
host+device
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
When we emit err_ref_bad_target, we should emit a "'method' declared
here" note. We already do so in most places, just not in
BuildCallToMemberFunction.
https://reviews.llvm.org/D23240
Files:
c
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
I want to reuse "CheckCUDAFoo" in a later patch. Also, I think
IsAllowedCUDACall gets the point across more clearly.
https://reviews.llvm.org/D23238
Files:
clang/include/clang/Sema/Sema.h
cla
jlebar created this revision.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
Previously these sort of worked because they didn't end up resulting in
calls at the ptx layer. But I'm adding stricter checks that break
placement new without these changes.
https://reviews.llvm.
jlebar accepted this revision.
jlebar added a comment.
This revision is now accepted and ready to land.
> Restore assertions for presence of -march flag.
We don't need an explicit assertion in TranslateArgs?
Comment at: lib/Driver/Driver.cpp:1412
@@ +1411,3 @@
+ assert(HostTC
jlebar added inline comments.
Comment at: lib/Driver/ToolChains.cpp:4834
@@ -4831,2 +4833,3 @@
+getDriver().Diag(diag::err_drv_cuda_nvptx_host);
}
return DAL;
IRL we talked about putting an assert() here and bailing out earlier. Does
that not work?
My
jlebar added inline comments.
Comment at: lib/Driver/ToolChains.cpp:4773
@@ +4772,3 @@
+ StringRef GpuArch =
+ DriverArgs.getLastArgValue(options::OPT_march_EQ, "sm_20");
+ std::string LibDeviceFile = CudaInstallation.getLibDeviceFile(GpuArch);
Hm. I total
jlebar added inline comments.
Comment at: lib/Driver/ToolChains.cpp:4773
@@ +4772,3 @@
+ StringRef GpuArch =
+ DriverArgs.getLastArgValue(options::OPT_march_EQ, "sm_20");
+ std::string LibDeviceFile = CudaInstallation.getLibDeviceFile(GpuArch);
I *think* ma
Author: jlebar
Date: Fri Jul 29 00:26:58 2016
New Revision: 277113
URL: http://llvm.org/viewvc/llvm-project?rev=277113&view=rev
Log:
[Typo police] s/proccess/process/, s/auxiliary/auxilliary/.
Modified:
cfe/trunk/include/clang/AST/Decl.h
Modified: cfe/trunk/include/clang/AST/Decl.h
URL:
htt
Author: jlebar
Date: Fri Jul 29 00:26:58 2016
New Revision: 277113
URL: http://llvm.org/viewvc/llvm-project?rev=277113&view=rev
Log:
[Typo police] s/proccess/process/, s/auxiliary/auxilliary/.
Modified:
clang-tools-extra/trunk/clang-tidy/ClangTidyDiagnosticConsumer.cpp
Modified: clang-tools-
This revision was automatically updated to reflect the committed changes.
Closed by commit rL276927: [CUDA] Align kernel launch args correctly when the
LLVM type's alignment is… (authored by jlebar).
Changed prior to commit:
https://reviews.llvm.org/D22879?vs=65800&id=65824#toc
Repository:
r
Author: jlebar
Date: Wed Jul 27 17:36:21 2016
New Revision: 276927
URL: http://llvm.org/viewvc/llvm-project?rev=276927&view=rev
Log:
[CUDA] Align kernel launch args correctly when the LLVM type's alignment is
different from the clang type's alignment.
Summary:
Before this patch, we computed the
Author: jlebar
Date: Wed Jul 27 17:04:24 2016
New Revision: 276925
URL: http://llvm.org/viewvc/llvm-project?rev=276925&view=rev
Log:
Don't crash when generating code for __attribute__((naked)) member functions.
Summary:
Previously this crashed inside EmitThisParam(). There should be no
prelude f
jlebar added inline comments.
Comment at: test/CodeGenCUDA/kernel-args-alignment.cu:1-2
@@ +1,3 @@
+// REQUIRES: x86-registered-target
+// REQUIRES: nvptx-registered-target
+
rnk wrote:
> Typically clang doesn't need a registered backend for a target to generate I
jlebar updated this revision to Diff 65800.
jlebar added a comment.
Remove REQUIRES lines.
https://reviews.llvm.org/D22879
Files:
lib/CodeGen/CGCUDABuiltin.cpp
lib/CodeGen/CGCUDANV.cpp
test/CodeGenCUDA/kernel-args-alignment.cu
Index: test/CodeGenCUDA/kernel-args-alignment.cu
jlebar created this revision.
jlebar added a reviewer: rnk.
jlebar added subscribers: tra, cfe-commits.
Before this patch, we computed the offsets in memory of args passed to
GPU kernel functions by throwing all of the args into an LLVM struct.
clang emits packed llvm structs basically whenever i
jlebar added a comment.
In https://reviews.llvm.org/D22463#495211, @vladisld wrote:
> In https://reviews.llvm.org/D22463#494828, @jlebar wrote:
>
> > I think the general feeling is that most of us (myself included) would
> > rather not learn a new tool if there's a simpler >alternative, such as
jlebar added a comment.
In https://reviews.llvm.org/D22463#494568, @vladisld wrote:
> Have the alternatives to sub-modules and monolithic repository been discussed
> ?
Hi, Vlad.
Please see the ongoing thread in llvm-dev, entitled "[RFC] One or many git
repositories?".
Tools such as git-repo
jlebar added a comment.
Hi, Renato.
Just to explain why I'm going to go forward with this RFC about a monolithic
repository: From speaking with some top contributors on IRC, I have heard that
they feel that the discussion of whether to move to git has been conflated with
the discussion of how
jlebar added a comment.
In https://reviews.llvm.org/D22463#489461, @rengolin wrote:
> You will not be required to use submodules at all, as we'll all use the
> individual projects, like we have always been. I don't understand why people
> keep going back to it.
There is a key use case that is
jlebar added a comment.
FYI after talking to Chandler, I'm going to write up a separate proposal for
the one-repository thing and send it to the list tomorrow. The suggestion was
that this phabricator thread isn't the right place to have this discussion.
https://reviews.llvm.org/D22463
___
jlebar added a subscriber: jlebar.
jlebar added a comment.
I'm sure you all have thought about this more than I have, and I apologize if
this has been brought up before because I haven't been following the thread
closely. But I am not convinced by this document that using subrepositories
beats
This revision was automatically updated to reflect the committed changes.
Closed by commit rL275232: [CUDA] Add additional testcases for
EraseUnwantedCUDAMatches. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21913?vs=62445&id=63750#toc
Repository:
rL LLVM
http://r
This revision was automatically updated to reflect the committed changes.
Closed by commit rL275233: [CUDA] Use the multi-element remove function in
EraseUnwantedCUDAMatches. (authored by jlebar).
Changed prior to commit:
http://reviews.llvm.org/D21914?vs=62446&id=63751#toc
Repository:
rL LL
Author: jlebar
Date: Tue Jul 12 18:23:13 2016
New Revision: 275233
URL: http://llvm.org/viewvc/llvm-project?rev=275233&view=rev
Log:
[CUDA] Use the multi-element remove function in EraseUnwantedCUDAMatches.
Summary:
Bug pointed out by Benjamin Kramer in r264008. I think the bug is
benign because
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
301 - 400 of 873 matches
Mail list logo