[PATCH] D29358: [OpenMP] Report an error for -faltivec on anything other than PowerPC.

gtbercea created this revision. Herald added a subscriber: nemanjai. When performing an offload, throw an error if the host tool chain uses the altivec flag on an architecture other than PowerPC. Repository: rL LLVM https://reviews.llvm.org/D29358 Files: lib/Driver/Tools.cpp

[PATCH] D29339: [OpenMP] Add support for auxiliary triple specification

gtbercea updated this revision to Diff 86635. gtbercea added a comment. Condition changed to only check triple once. Repository: rL LLVM https://reviews.llvm.org/D29339 Files: lib/Driver/Tools.cpp lib/Frontend/CompilerInstance.cpp lib/Frontend/CompilerInvocation.cpp

[PATCH] D29339: [OpenMP] Add support for auxiliary triple specification

gtbercea created this revision. Device offloading requires the specification of an additional flag containing the triple of the //other// architecture the code is being compiled on if such an architecture exists. If compiling for the host, the auxiliary triple flag will contain the triple

[PATCH] D29659: [OpenMP] Add flag for disabling the default generation of relocatable OpenMP target code for NVIDIA GPUs.

gtbercea created this revision. Previously we have added the "-c" flag which gets passed to PTXAS by default to generate relocatable OpenMP target code by default. This set of flags exposes control over this behaviour. Repository: rL LLVM https://reviews.llvm.org/D29659 Files:

[PATCH] D29659: [OpenMP] Add flag for disabling the default generation of relocatable OpenMP target code for NVIDIA GPUs.

gtbercea updated this revision to Diff 87463. gtbercea added a comment. Additional test added to check "-c" is passed in appropriately. Repository: rL LLVM https://reviews.llvm.org/D29659 Files: include/clang/Driver/Options.td lib/Driver/Tools.cpp test/Driver/openmp-offload.c Index:

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea created this revision. This flag "--fopenmp-ptx=" enables the overwriting of the default PTX version used for GPU offloaded OpenMP target regions: "+ptx42". Repository: rL LLVM https://reviews.llvm.org/D29660 Files: include/clang/Driver/Options.td lib/Driver/ToolChains.cpp

[PATCH] D29904: [OpenMP] Prevent emission of exception handling code when using OpenMP to offload to NVIDIA devices.

gtbercea created this revision. For the OpenMP toolchain which offloads to NVIDIA GPUs make sure that no exception handling code is emitted. Repository: rL LLVM https://reviews.llvm.org/D29904 Files: lib/Frontend/CompilerInvocation.cpp test/OpenMP/target_parallel_no_exceptions.cpp

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea created this revision. When a scalar argument is explicitly mapped, the device kernel needs to be passed that argument by reference. This is a partial fix to this bug . The full fix requires data sharing support which will land in future

[PATCH] D29659: [OpenMP] Add flag for disabling the default generation of relocatable OpenMP target code for NVIDIA GPUs.

gtbercea updated this revision to Diff 93240. gtbercea added a comment. Herald added a subscriber: rengolin. Update patch to reflect latest source code changes. Repository: rL LLVM https://reviews.llvm.org/D29659 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea updated this revision to Diff 93244. gtbercea added a comment. Herald added a subscriber: rengolin. Update patch to reflect latest source code changes. Repository: rL LLVM https://reviews.llvm.org/D29660 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp

[PATCH] D29339: [OpenMP] Add support for auxiliary triple specification

gtbercea updated this revision to Diff 93171. gtbercea added a comment. Herald added a subscriber: rengolin. Update patch to reflect latest source code changes. Repository: rL LLVM https://reviews.llvm.org/D29339 Files: lib/Driver/ToolChains/Clang.cpp lib/Frontend/CompilerInstance.cpp

[PATCH] D29904: [OpenMP] Prevent emission of exception handling code when using OpenMP to offload to NVIDIA devices.

gtbercea updated this revision to Diff 93513. gtbercea added a comment. Herald added a subscriber: rengolin. Clean-up test. Repository: rL LLVM https://reviews.llvm.org/D29904 Files: lib/Frontend/CompilerInvocation.cpp test/OpenMP/target_parallel_no_exceptions.cpp Index:

[PATCH] D29644: [OpenMP] Pass -v to PTXAS if it was passed to the driver.

gtbercea updated this revision to Diff 93502. gtbercea added a comment. Remove redundant check. Repository: rL LLVM https://reviews.llvm.org/D29644 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29642: [OpenMP] Make OpenMP generated code for the NVIDIA device relocatable by default

gtbercea updated this revision to Diff 93501. gtbercea added a comment. Remove redundant check. Repository: rL LLVM https://reviews.llvm.org/D29642 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29644: [OpenMP] Pass -v to PTXAS if it was passed to the driver.

gtbercea updated this revision to Diff 93559. gtbercea added a comment. Change prefix name in test. Repository: rL LLVM https://reviews.llvm.org/D29644 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29904: [OpenMP] Prevent emission of exception handling code when using OpenMP to offload to NVIDIA devices.

gtbercea added inline comments. Comment at: lib/Frontend/CompilerInvocation.cpp:2167-2172 +// Set the flag to prevent the implementation from emitting device exception +// handling code for those requiring so. +if (Opts.OpenMPIsDevice && T.isNVPTX()) { +

[PATCH] D29642: [OpenMP] Make OpenMP generated code for the NVIDIA device relocatable by default

gtbercea updated this revision to Diff 93557. gtbercea added a comment. Change prefix name in test. Repository: rL LLVM https://reviews.llvm.org/D29642 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea updated this revision to Diff 93181. gtbercea marked 5 inline comments as done. gtbercea added a comment. Herald added subscribers: sbc100, dschuff, jfb, rengolin. Update patch to reflect latest source code changes. Repository: rL LLVM https://reviews.llvm.org/D29647 Files:

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea added inline comments. Comment at: lib/Driver/ToolChains.cpp:4902 + DeviceOffloadingKind == Action::OFK_Cuda) && + "The offloading kind is not OpenMP or CUDA."); jlebar wrote: > Not sure this assertion message helps us much beyond

[PATCH] D29642: [OpenMP] Make OpenMP generated code for the NVIDIA device relocatable by default

gtbercea updated this revision to Diff 93172. gtbercea added a comment. Herald added a subscriber: rengolin. Update patch to reflect latest source code changes. Repository: rL LLVM https://reviews.llvm.org/D29642 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea added inline comments. Comment at: lib/Driver/Tools.cpp:12136 // Obtain architecture from the action. - CudaArch gpu_arch = StringToCudaArch(JA.getOffloadingArch()); assert(gpu_arch != CudaArch::UNKNOWN && jlebar wrote: > Why does

[PATCH] D29644: [OpenMP] Pass -v to PTXAS if it was passed to the driver.

gtbercea updated this revision to Diff 93176. gtbercea added a comment. Herald added a subscriber: rengolin. Update patch to reflect latest source code changes. Repository: rL LLVM https://reviews.llvm.org/D29644 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c

[PATCH] D29654: [OpenMP] Integrate OpenMP target region cubin into host binary

gtbercea updated this revision to Diff 93192. gtbercea added a comment. Herald added a subscriber: rengolin. Update patch to reflect latest source code changes. Repository: rL LLVM https://reviews.llvm.org/D29654 Files: lib/Driver/ToolChains/CommonArgs.cpp

[PATCH] D29651: [OpenMP] Consider LIBRARY_PATH when selecting library paths for NVPTX targets in OpenMP mode.

gtbercea updated this revision to Diff 93188. gtbercea added a comment. Herald added a subscriber: rengolin. Update patch to reflect latest source code changes. Repository: rL LLVM https://reviews.llvm.org/D29651 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c

[PATCH] D29339: [OpenMP] Add support for auxiliary triple specification

gtbercea updated this revision to Diff 93643. gtbercea added a comment. Re-order conditions and run clang-format. Repository: rL LLVM https://reviews.llvm.org/D29339 Files: lib/Driver/ToolChains/Clang.cpp lib/Frontend/CompilerInstance.cpp lib/Frontend/CompilerInvocation.cpp

[PATCH] D29644: [OpenMP] Pass -v to PTXAS if it was passed to the driver.

gtbercea updated this revision to Diff 93676. gtbercea added a comment. Move test. Repository: rL LLVM https://reviews.llvm.org/D29644 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/cuda-external-tools.cu Index: test/Driver/cuda-external-tools.cu

[PATCH] D29642: [OpenMP] Make OpenMP generated code for the NVIDIA device relocatable by default

gtbercea updated this revision to Diff 93672. gtbercea marked an inline comment as done. gtbercea added a comment. Fix prefix. Repository: rL LLVM https://reviews.llvm.org/D29642 Files: lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29658: [OpenMP] Customize CUDA-based tool chain selection

gtbercea updated this revision to Diff 93679. gtbercea added a comment. Herald added a subscriber: rengolin. Split patch. Repository: rL LLVM https://reviews.llvm.org/D29658 Files: include/clang/Basic/DiagnosticDriverKinds.td lib/Driver/Driver.cpp test/OpenMP/target_messages.cpp

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea marked an inline comment as done. gtbercea added inline comments. Comment at: lib/Sema/SemaOpenMP.cpp:358-360 + /// Do the check specified in \a Check to all component lists at a given level + /// and return true if any issue is found. + bool

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea updated this revision to Diff 94824. gtbercea added a comment. Integrate review. Repository: rL LLVM https://reviews.llvm.org/D29660 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea updated this revision to Diff 94841. gtbercea added a comment. Refactor. Repository: rL LLVM https://reviews.llvm.org/D29660 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea updated this revision to Diff 94842. gtbercea added a comment. Run clang format. Repository: rL LLVM https://reviews.llvm.org/D29660 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea updated this revision to Diff 94867. gtbercea added a comment. Refactor code. Repository: rL LLVM https://reviews.llvm.org/D29905 Files: lib/Sema/SemaOpenMP.cpp test/OpenMP/target_map_codegen.cpp Index: test/OpenMP/target_map_codegen.cpp

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea updated this revision to Diff 93859. gtbercea added a comment. Herald added a subscriber: rengolin. Update test. Repository: rL LLVM https://reviews.llvm.org/D29905 Files: lib/Sema/SemaOpenMP.cpp test/OpenMP/target_map_codegen.cpp Index: test/OpenMP/target_map_codegen.cpp

[PATCH] D29659: [OpenMP] Add flag for disabling the default generation of relocatable OpenMP target code for NVIDIA GPUs.

gtbercea updated this revision to Diff 93883. gtbercea added a comment. Update test. Repository: rL LLVM https://reviews.llvm.org/D29659 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea updated this revision to Diff 93901. gtbercea added a comment. Update test. Repository: rL LLVM https://reviews.llvm.org/D29660 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea updated this revision to Diff 94972. gtbercea added a comment. Pass correct levels. Refactor code. Repository: rL LLVM https://reviews.llvm.org/D29905 Files: lib/Sema/SemaOpenMP.cpp test/OpenMP/target_map_codegen.cpp Index: test/OpenMP/target_map_codegen.cpp

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea updated this revision to Diff 94985. gtbercea added a comment. Refactor code. Repository: rL LLVM https://reviews.llvm.org/D29905 Files: lib/Sema/SemaOpenMP.cpp test/OpenMP/target_map_codegen.cpp Index: test/OpenMP/target_map_codegen.cpp

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea updated this revision to Diff 94983. gtbercea added a comment. Clean-up. Repository: rL LLVM https://reviews.llvm.org/D29905 Files: lib/Sema/SemaOpenMP.cpp test/OpenMP/target_map_codegen.cpp Index: test/OpenMP/target_map_codegen.cpp

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea updated this revision to Diff 94990. gtbercea added a comment. Fix for loop range. Repository: rL LLVM https://reviews.llvm.org/D29905 Files: lib/Sema/SemaOpenMP.cpp test/OpenMP/target_map_codegen.cpp Index: test/OpenMP/target_map_codegen.cpp

[PATCH] D32035: [OpenMP] Error when trying to offload to an unsupported architecture

gtbercea created this revision. Herald added a subscriber: rengolin. Throw an error when offloading is unsupported for a particular target architecture. Repository: rL LLVM https://reviews.llvm.org/D32035 Files: include/clang/Basic/DiagnosticDriverKinds.td

[PATCH] D29654: [OpenMP] Integrate OpenMP target region cubin into host binary

gtbercea updated this revision to Diff 95175. gtbercea added a comment. Use the rename() utility function of LLVM for renaming the PTXAS output before invoking NVLINK. Repository: rL LLVM https://reviews.llvm.org/D29654 Files: lib/Driver/Driver.cpp lib/Driver/ToolChains/CommonArgs.cpp

[PATCH] D29654: [OpenMP] Integrate OpenMP target region cubin into host binary

gtbercea marked 4 inline comments as done. gtbercea added inline comments. Comment at: test/Driver/openmp-offload.c:594 +/// Check cubin file generation and usage by nvlink +// RUN: %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda -save-temps

[PATCH] D32035: [OpenMP] Error when trying to offload to an unsupported architecture

gtbercea updated this revision to Diff 95198. gtbercea added a comment. Re-use an already existing flag rather than creating a new one. Repository: rL LLVM https://reviews.llvm.org/D32035 Files: lib/Frontend/CompilerInvocation.cpp test/OpenMP/target_messages.cpp Index:

[PATCH] D29654: [OpenMP] Integrate OpenMP target region cubin into host binary

gtbercea updated this revision to Diff 95737. gtbercea marked an inline comment as done. gtbercea added a comment. Avoid renaming by enabling PTXAS to generate an output file with the appropriate extension, in this case a cubin extension. Repository: rL LLVM https://reviews.llvm.org/D29654

[PATCH] D29659: [OpenMP] Add flag for disabling the default generation of relocatable OpenMP target code for NVIDIA GPUs.

gtbercea updated this revision to Diff 96144. gtbercea added a comment. Refactor if condition. Repository: rL LLVM https://reviews.llvm.org/D29659 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29658: [OpenMP] Customize CUDA-based tool chain selection

gtbercea updated this revision to Diff 95171. gtbercea added a comment. Remove tests which belong into a different patch. Repository: rL LLVM https://reviews.llvm.org/D29658 Files: lib/Driver/Driver.cpp Index: lib/Driver/Driver.cpp

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea updated this revision to Diff 95126. gtbercea added a comment. Fix. Repository: rL LLVM https://reviews.llvm.org/D29660 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D32035: [OpenMP] Error when trying to offload to an unsupported architecture

gtbercea updated this revision to Diff 95461. gtbercea added a comment. Merge IF statements. Repository: rL LLVM https://reviews.llvm.org/D32035 Files: lib/Frontend/CompilerInvocation.cpp test/OpenMP/target_messages.cpp Index: test/OpenMP/target_messages.cpp

[PATCH] D29654: [OpenMP] Integrate OpenMP target region cubin into host binary

gtbercea updated this revision to Diff 94475. gtbercea added a comment. Address some of the reviews. Repository: rL LLVM https://reviews.llvm.org/D29654 Files: lib/Driver/Driver.cpp lib/Driver/ToolChains/CommonArgs.cpp lib/Driver/ToolChains/CommonArgs.h

[PATCH] D29904: [OpenMP] Prevent emission of exception handling code when using OpenMP to offload to NVIDIA devices.

gtbercea updated this revision to Diff 93665. gtbercea added a comment. Redo regression test. Repository: rL LLVM https://reviews.llvm.org/D29904 Files: lib/Frontend/CompilerInvocation.cpp test/OpenMP/target_parallel_no_exceptions.cpp Index:

[PATCH] D29904: [OpenMP] Prevent emission of exception handling code when using OpenMP to offload to NVIDIA devices.

gtbercea marked 2 inline comments as done. gtbercea added inline comments. Comment at: lib/Frontend/CompilerInvocation.cpp:2167-2172 +// Set the flag to prevent the implementation from emitting device exception +// handling code for those requiring so. +if

[PATCH] D29904: [OpenMP] Prevent emission of exception handling code when using OpenMP to offload to NVIDIA devices.

gtbercea updated this revision to Diff 93670. gtbercea added a comment. run clang-format on test. Repository: rL LLVM https://reviews.llvm.org/D29904 Files: lib/Frontend/CompilerInvocation.cpp test/OpenMP/target_parallel_no_exceptions.cpp Index:

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. I have re-enabled the previous offloading tests and moved the new GPU offloading tests to a new file which is disabled for linux (for now). 310718 Alex thanks so much for the logs, they have been very useful to understand what's going on. Aleksey, I have since tried

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. First of all, I apologize if I've upset you with my previous post. I am actively working on understanding what is causing these issues. It is not my intention to write tests that work on local configurations only. I am upset to see that these tests keep failing for

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. In https://reviews.llvm.org/D29660#836927, @arphaman wrote: > Looks like this test is failing on macOS again after this change: > > http://lab.llvm.org:8080/green/job/clang-stage1-cmake-RA-incremental_check/39231/testReport/Clang/Driver/openmp_offload_c/ > > Can you

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. Revision 310505 fixes the tests for this patch. Repository: rL LLVM https://reviews.llvm.org/D29660 ___ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[PATCH] D36537: [OpenMP] Enable executable lookup into driver directory.

gtbercea created this revision. Invoking the compiler inside a script causes the clang-offload-bundler executable to not be found. This patch fixes this error. Repository: rL LLVM https://reviews.llvm.org/D36537 Files: lib/Driver/ToolChains/Cuda.cpp Index:

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. 310549 should solve this problem by using a default architecture that is supported by the current device version. Repository: rL LLVM https://reviews.llvm.org/D29660 ___ cfe-commits mailing list

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. In https://reviews.llvm.org/D29660#838194, @arphaman wrote: > The failures were very widespread, e.g. there's a linux buildbot that was red > until the revert: http://bb.pgr.jp/builders/test-clang-i686-linux-RA. If you > have access to a linux machine you should be

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. In https://reviews.llvm.org/D29660#839275, @alekseyshl wrote: > In https://reviews.llvm.org/D29660#839209, @gtbercea wrote: > > > I have re-enabled the previous offloading tests and moved the new GPU > > offloading tests to a new file which is disabled for linux (for

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. In https://reviews.llvm.org/D29660#839550, @alekseyshl wrote: > http://lab.llvm.org:8011/builders/sanitizer-x86_64-linux-fast/builds/7109 > failed on r310718, please fix. Couldn't fix/find the actual error so for now, just moving the flag patch tests to

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. Hi Alex, I have no idea what the errors you mention are. I worked until very late last night to fix the previous error with the compute capability. Maybe fixing that unlocked some other errors. I would need a lot of help to figure out what's going on since I probably

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. In https://reviews.llvm.org/D29660#839736, @alekseyshl wrote: > In https://reviews.llvm.org/D29660#839728, @alekseyshl wrote: > > > In https://reviews.llvm.org/D29660#839550, @alekseyshl wrote: > > > > >

[PATCH] D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets

gtbercea added a comment. In https://reviews.llvm.org/D29660#838185, @alekseyshl wrote: > In https://reviews.llvm.org/D29660#838150, @gtbercea wrote: > > > First of all, I apologize if I've upset you with my previous post. I am > > actively working on understanding what is causing these issues.

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea updated this revision to Diff 105932. gtbercea added a comment. Address comments. https://reviews.llvm.org/D34784 Files: include/clang/Basic/DiagnosticDriverKinds.td include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index:

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added a comment. @hfinkel I think I have something that works which is similar to what you were requesting. Please let me know your thoughts! Thanks, --Doru https://reviews.llvm.org/D34784 ___ cfe-commits mailing list

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added inline comments. Comment at: lib/Driver/ToolChains/Cuda.cpp:474 +for (StringRef Opt : OptList) { + AddMArchOption(DAL, Opts, Opt); +} hfinkel wrote: > gtbercea wrote: > > hfinkel wrote: > > > Shouldn't you be adding all of the

[PATCH] D29905: [OpenMP] Pass argument to device kernel by reference when map is used.

gtbercea added a comment. In https://reviews.llvm.org/D29905#800416, @Hahnfeld wrote: > In https://reviews.llvm.org/D34888#799576, @gtbercea wrote: > > > Does this also include the fixes in the following revision? > > > > https://reviews.llvm.org/D29905 > > > Sorry, I wasn't aware of this

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea added inline comments. Comment at: test/Driver/openmp-offload.c:614 +/// Check -march propagates compute capability to device offloading toolchain. +// RUN: %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda -save-temps -no-canonical-prefixes

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea added inline comments. Comment at: test/Driver/openmp-offload.c:614 +/// Check -march propagates compute capability to device offloading toolchain. +// RUN: %clang -### -fopenmp=libomp -fopenmp-targets=nvptx64-nvidia-cuda -save-temps -no-canonical-prefixes

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added a comment. In https://reviews.llvm.org/D34784#795367, @hfinkel wrote: > In https://reviews.llvm.org/D34784#795353, @gtbercea wrote: > > > In https://reviews.llvm.org/D34784#795287, @hfinkel wrote: > > > > > What happens if you have multiple targets? Maybe this should be > > >

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added a comment. In https://reviews.llvm.org/D34784#795934, @hfinkel wrote: > In https://reviews.llvm.org/D34784#795871, @gtbercea wrote: > > > In https://reviews.llvm.org/D34784#795367, @hfinkel wrote: > > > > > In https://reviews.llvm.org/D34784#795353, @gtbercea wrote: > > > > > > >

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea updated this revision to Diff 104527. gtbercea added a comment. Split previous diff into a "device offloading kind" patch (show here) and a **new** patch which relies on a new compiler flag. A TODO has been added to signal that the compute capability is to be handled in the **new**

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea updated this revision to Diff 104532. Repository: rL LLVM https://reviews.llvm.org/D34784 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index: test/Driver/openmp-offload.c

[PATCH] D29654: [OpenMP] Integrate OpenMP target region cubin into host binary

gtbercea updated this revision to Diff 104535. gtbercea added a comment. [Update regression tests] Add a test for propagating the compute capability to the OpenMP device offloading toolchain which targets NVIDIA GPUs. This is a test for patch D34784 which is

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea created this revision. OpenMP has the ability to offload target regions to devices which may have different architectures. A new -fopenmp-target-arch flag is introduced to specify the device architecture. In this patch I use the new flag to specify the compute capability of the

[PATCH] D29339: [OpenMP] Add support for auxiliary triple specification

gtbercea updated this revision to Diff 104656. gtbercea added a comment. Rebase https://reviews.llvm.org/D29339 Files: lib/Driver/ToolChains/Clang.cpp lib/Frontend/CompilerInstance.cpp lib/Frontend/CompilerInvocation.cpp lib/Frontend/InitPreprocessor.cpp test/Driver/openmp-offload.c

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added a comment. In https://reviews.llvm.org/D34784#795287, @hfinkel wrote: > What happens if you have multiple targets? Maybe this should be > -fopenmp-targets-arch=foo,bar,whatever? > > Once this all lands, please make sure that you add additional test cases > here. Make sure that

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea updated this revision to Diff 104962. gtbercea added a comment. Check -fopenmp-targets has one entry when using default toolchain in -Xopenmp-target. https://reviews.llvm.org/D34784 Files: include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added a comment. @hfinkel I've add the flag as suggested. There is one minor change, I used "=" instead of ":" when specifying the toolchain/triple. I also support the triple being omitted when there is only one offloading toolchain specified with -fopenmp-targets.

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea updated this revision to Diff 104960. gtbercea retitled this revision from "[OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading " to "[OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading". gtbercea added a

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added a comment. In https://reviews.llvm.org/D34784#795988, @hfinkel wrote: > In https://reviews.llvm.org/D34784#795980, @gtbercea wrote: > > > In https://reviews.llvm.org/D34784#795934, @hfinkel wrote: > > > > > In https://reviews.llvm.org/D34784#795871, @gtbercea wrote: > > > > > > >

[PATCH] D29658: [OpenMP] Customize CUDA-based tool chain selection

gtbercea updated this revision to Diff 105446. gtbercea added a comment. Use str() https://reviews.llvm.org/D29658 Files: lib/Driver/Driver.cpp Index: lib/Driver/Driver.cpp === --- lib/Driver/Driver.cpp +++

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea updated this revision to Diff 105453. gtbercea added a comment. . https://reviews.llvm.org/D29647 Files: include/clang/Driver/ToolChain.h lib/Driver/ToolChain.cpp lib/Driver/ToolChains/BareMetal.cpp lib/Driver/ToolChains/BareMetal.h lib/Driver/ToolChains/Clang.cpp

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea updated this revision to Diff 105280. gtbercea marked 3 inline comments as done. gtbercea added a comment. Address comments. https://reviews.llvm.org/D34784 Files: include/clang/Basic/DiagnosticDriverKinds.td include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added inline comments. Comment at: lib/Driver/ToolChains/Cuda.cpp:443 + +// Get the compute capability from the -fopenmp-targets flag. +// The default compute capability is sm_20 since this is a CUDA hfinkel wrote: > Is this first sentence

[PATCH] D34888: [OpenMP] Fix mapping of scalars for combined directives

gtbercea added a comment. Does this also include the fixes in the following revision? https://reviews.llvm.org/D29905 Repository: rL LLVM https://reviews.llvm.org/D34888 ___ cfe-commits mailing list cfe-commits@lists.llvm.org

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added inline comments. Comment at: test/Driver/openmp-offload.c:607 + +// CHK-FOPENMP-EQ-TARGET: clang{{.*}} argument unused during compilation: '-Xopenmp-target=powerpc64le-ibm-linux-gnu -march=pwr8' + hfinkel wrote: > I don't see why you'd check that

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea updated this revision to Diff 105355. gtbercea added a comment. Address Comments. https://reviews.llvm.org/D34784 Files: include/clang/Basic/DiagnosticDriverKinds.td include/clang/Driver/Options.td lib/Driver/ToolChains/Cuda.cpp test/Driver/openmp-offload.c Index:

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea added inline comments. Comment at: lib/Driver/ToolChains/Cuda.cpp:478 +auto MArchList = DAL->getAllArgValues(options::OPT_march_EQ); +assert(MArchList.size() < 2 && "At most one GPU arch allowed."); +if (MArchList.empty()) hfinkel wrote: >

[PATCH] D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading

gtbercea updated this revision to Diff 105354. gtbercea added a comment. Address comments. https://reviews.llvm.org/D34784 Files: include/clang/Basic/DiagnosticDriverKinds.td include/clang/Driver/Options.td lib/Driver/Driver.cpp lib/Driver/ToolChains/Clang.cpp

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea added inline comments. Comment at: lib/Driver/ToolChains/Cuda.cpp:435 + +// TODO: get the compute capability from offloading arguments when not +// using the default compute capability of sm_20. hfinkel wrote: > gtbercea wrote: > > hfinkel

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea marked an inline comment as done. gtbercea added inline comments. Comment at: lib/Driver/ToolChains/Cuda.cpp:435 + +// TODO: get the compute capability from offloading arguments when not +// using the default compute capability of sm_20. hfinkel

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea marked an inline comment as done. gtbercea added inline comments. Comment at: lib/Driver/ToolChains/Cuda.cpp:435 + +// TODO: get the compute capability from offloading arguments when not +// using the default compute capability of sm_20. hfinkel

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea marked an inline comment as done. gtbercea added inline comments. Comment at: lib/Driver/ToolChains/Cuda.cpp:217 + std::vector GPUArchNames; + // If this is an OpenMP action we need to extract the device architecture from + // the -march option.

[PATCH] D29647: [OpenMP] Extend CLANG target options with device offloading kind.

gtbercea updated this revision to Diff 104427. gtbercea added a comment. Herald added subscribers: aheejin, jgravelle-google. Updated diff to address comments. Repository: rL LLVM https://reviews.llvm.org/D29647 Files: include/clang/Driver/ToolChain.h lib/Driver/ToolChain.cpp

[PATCH] D29651: [OpenMP] Consider LIBRARY_PATH when selecting library paths for NVPTX targets in OpenMP mode.

gtbercea abandoned this revision. gtbercea added a comment. Not needed. These changes are related to looking up the .bc library for inlining purposes. I believe @arpith-jacob has already handled this in trunk. Therefore this is obsolete code. Repository: rL LLVM

[PATCH] D29658: [OpenMP] Customize CUDA-based tool chain selection

gtbercea updated this revision to Diff 105320. gtbercea edited the summary of this revision. gtbercea added a comment. Rebase on latest master. https://reviews.llvm.org/D29658 Files: lib/Driver/Driver.cpp Index: lib/Driver/Driver.cpp

[PATCH] D29658: [OpenMP] Customize CUDA-based tool chain selection

gtbercea added inline comments. Comment at: lib/Driver/Driver.cpp:564 + auto = + ToolChains[TT.str() + "/" + HostTC->getTriple().normalize()]; + if (!CudaTC) Hahnfeld wrote: > The code above uses `HostTriple.str()`,

  1   2   3   4   5   6   7   >