Hahnfeld (Jonas Hahnfeld)
User

Projects

User does not belong to any projects.

User Details

User Since
Apr 2 2015, 4:52 AM (133 w, 4 d)
Intern @ IBM Research

Recent Activity

Today

Hahnfeld added a reviewer for D39070: [zorg] Refactor Libiomp5Builder to OpenMPBuilder: vleschuk.
Mon, Oct 23, 6:27 AM
Hahnfeld added a comment to D39052: [CMake] Re-enable libomptarget and restrict tests to Clang 6.0.0.
In D39052#900899, @hans wrote:

When we ran into the test failures last time, in PR33859, it turned out there were no upstream buildbots running the tests, so we didn't find out they were failing until we started the release process. Are the tests run by any buildbots now?

Good point, let me find that out and come back to you.

Mon, Oct 23, 6:24 AM
Hahnfeld added a dependent revision for D39070: [zorg] Refactor Libiomp5Builder to OpenMPBuilder: D39052: [CMake] Re-enable libomptarget and restrict tests to Clang 6.0.0.
Mon, Oct 23, 6:24 AM
Hahnfeld added a dependency for D39052: [CMake] Re-enable libomptarget and restrict tests to Clang 6.0.0: D39070: [zorg] Refactor Libiomp5Builder to OpenMPBuilder.
Mon, Oct 23, 6:24 AM

Fri, Oct 20

Hahnfeld committed rL316235: Revert "[OpenMP] Avoid VLAs for some reductions on array sections".
Revert "[OpenMP] Avoid VLAs for some reductions on array sections"
Fri, Oct 20, 1:19 PM
Hahnfeld reopened D39136: [OpenMP] Avoid VLAs for some reductions on array sections.

At least two buildbots failing:
http://lab.llvm.org:8011/builders/clang-cmake-x86_64-avx2-linux/builds/1175
http://lab.llvm.org:8011/builders/clang-atom-d525-fedora-rel/builds/10478

Fri, Oct 20, 1:18 PM
Hahnfeld closed D38920: Updated .clang-format file and applied reformating.

https://reviews.llvm.org/rL316227

Fri, Oct 20, 12:53 PM
Hahnfeld accepted D38920: Updated .clang-format file and applied reformating.

Accepting to close

Fri, Oct 20, 12:52 PM
Hahnfeld committed rL316229: [OpenMP] Avoid VLAs for some reductions on array sections.
[OpenMP] Avoid VLAs for some reductions on array sections
Fri, Oct 20, 12:42 PM
Hahnfeld closed D39136: [OpenMP] Avoid VLAs for some reductions on array sections by committing rL316229: [OpenMP] Avoid VLAs for some reductions on array sections.
Fri, Oct 20, 12:42 PM
Hahnfeld created D39136: [OpenMP] Avoid VLAs for some reductions on array sections.
Fri, Oct 20, 12:27 PM

Thu, Oct 19

Hahnfeld abandoned D26244: [Driver] Add CLANG_PREFER_GCC_LIBRARIES which can be disabled.

Abandoning as I lost interest in this.

Thu, Oct 19, 6:25 PM

Wed, Oct 18

Hahnfeld added inline comments to D38185: Implementation of OMPT as specified in OpenMP 5.0 Preview 1.
Wed, Oct 18, 6:02 PM
Hahnfeld created D39070: [zorg] Refactor Libiomp5Builder to OpenMPBuilder.
Wed, Oct 18, 2:18 PM
Hahnfeld added inline comments to D38185: Implementation of OMPT as specified in OpenMP 5.0 Preview 1.
Wed, Oct 18, 12:23 PM
Hahnfeld added inline comments to D38185: Implementation of OMPT as specified in OpenMP 5.0 Preview 1.
Wed, Oct 18, 8:26 AM
Hahnfeld added a comment to D39052: [CMake] Re-enable libomptarget and restrict tests to Clang 6.0.0.
In D39052#900899, @hans wrote:

When we ran into the test failures last time, in PR33859, it turned out there were no upstream buildbots running the tests, so we didn't find out they were failing until we started the release process. Are the tests run by any buildbots now?

Wed, Oct 18, 8:00 AM
Hahnfeld added inline comments to D38185: Implementation of OMPT as specified in OpenMP 5.0 Preview 1.
Wed, Oct 18, 7:19 AM
Hahnfeld created D39052: [CMake] Re-enable libomptarget and restrict tests to Clang 6.0.0.
Wed, Oct 18, 6:46 AM
Hahnfeld accepted D38878: Add C++ support for testcases.

LGTM

Wed, Oct 18, 6:12 AM

Tue, Oct 17

Hahnfeld added a comment to D38185: Implementation of OMPT as specified in OpenMP 5.0 Preview 1.

Some more comments, there is still a lot of commented code that should be removed.
Another, higher level question: When should frame addresses be set? Currently, that's inconsistently guarded with OMPT_OPTIONAL. However, if I remember correctly this is mandatory?

Tue, Oct 17, 5:10 PM
Hahnfeld committed rL316001: [OpenMP] Implement omp_is_initial_device() as builtin.
[OpenMP] Implement omp_is_initial_device() as builtin
Tue, Oct 17, 7:28 AM
Hahnfeld closed D38968: [OpenMP] Implement omp_is_initial_device() as builtin by committing rL316001: [OpenMP] Implement omp_is_initial_device() as builtin.
Tue, Oct 17, 7:28 AM
Hahnfeld committed rL315996: [CMake][OpenMP] Customize default offloading arch.
[CMake][OpenMP] Customize default offloading arch
Tue, Oct 17, 6:37 AM
Hahnfeld closed D38883: [CMake][OpenMP] Customize default offloading arch by committing rL315996: [CMake][OpenMP] Customize default offloading arch.
Tue, Oct 17, 6:37 AM

Mon, Oct 16

Hahnfeld added a comment to D38968: [OpenMP] Implement omp_is_initial_device() as builtin.

Now that this issue has been addressed and regressions tests pass, should we re-enable Cmake to build libomptarget by default?

Mon, Oct 16, 1:44 PM
Hahnfeld created D38968: [OpenMP] Implement omp_is_initial_device() as builtin.
Mon, Oct 16, 11:59 AM
Hahnfeld committed rL315902: [CUDA] Require libdevice only if needed.
[CUDA] Require libdevice only if needed
Mon, Oct 16, 6:31 AM
Hahnfeld closed D38901: [CUDA] Require libdevice only if needed by committing rL315902: [CUDA] Require libdevice only if needed.
Mon, Oct 16, 6:31 AM

Fri, Oct 13

Hahnfeld updated the diff for D38901: [CUDA] Require libdevice only if needed.

Fix one more condition that checks for nvvm/libdevice and add a test.

Fri, Oct 13, 1:33 PM
Hahnfeld created D38901: [CUDA] Require libdevice only if needed.
Fri, Oct 13, 12:57 PM
Hahnfeld updated the diff for D38883: [CMake][OpenMP] Customize default offloading arch.

Check that the user didn't specify a value lower than sm_30 and re-add some code as discussed.

Fri, Oct 13, 12:55 PM
Hahnfeld added inline comments to D38883: [CMake][OpenMP] Customize default offloading arch.
Fri, Oct 13, 10:58 AM
Hahnfeld added inline comments to D38883: [CMake][OpenMP] Customize default offloading arch.
Fri, Oct 13, 10:15 AM
Hahnfeld created D38883: [CMake][OpenMP] Customize default offloading arch.
Fri, Oct 13, 8:07 AM
Hahnfeld added a comment to D38880: Detect available features of compiler in lit.cfg.

GCC has none of these features?! I can't believe that, maybe only some older versions that don't implement OpenMP 4.x?

Fri, Oct 13, 6:21 AM
Hahnfeld accepted D38881: Load Intel License environment variable in tests if available.

LGTM

Fri, Oct 13, 6:21 AM
Hahnfeld added a comment to D38878: Add C++ support for testcases.

The changes look good, but title and summary don't describe then change?

Fri, Oct 13, 6:19 AM
Hahnfeld accepted D38876: Restrict OMPT to OpenMP version 5.0 and remove old header files.

LGTM with a comment about the error message

Fri, Oct 13, 6:17 AM

Thu, Oct 12

Hahnfeld added a comment to D38837: Add explicit values to .clang-format.

This might work for one patch but still doesn't solve the future problem, i.e. we don't want to be bound to 3.8.1 forever.

Thu, Oct 12, 2:04 PM
Hahnfeld added a comment to D38837: Add explicit values to .clang-format.

Right. We found that each (we haven't tried all versions of cause:) version of the clang-format produces different formatting, so we decided to stick with the version 3.8.1:
$ clang-format --version
clang-format version 3.8.1 (http://llvm.org/git/clang.git 9fd77bd68130d9b2fbc56a3138b6f981d560480a) (http://llvm.org/git/llvm.git 8b47c17a53d683f313eaaa93c4a53de26d8fcba5)

Thu, Oct 12, 12:08 PM
Hahnfeld added a comment to D38837: Add explicit values to .clang-format.

So maybe the reason for that is using different versions of clang-format? @jlpeyton can you share which version you have been using to initially format the source code?

Thu, Oct 12, 12:00 PM

Wed, Oct 4

Hahnfeld committed rL314904: [test] Pass in fixed triple for openmp-offload.c.
[test] Pass in fixed triple for openmp-offload.c
Wed, Oct 4, 6:55 AM
Hahnfeld committed rL314902: [OpenMP] Fix passing of -m arguments correctly.
[OpenMP] Fix passing of -m arguments correctly
Wed, Oct 4, 6:34 AM
Hahnfeld closed D38372: [OpenMP] Fix passing of -m arguments correctly by committing rL314902: [OpenMP] Fix passing of -m arguments correctly.
Wed, Oct 4, 6:34 AM

Mon, Oct 2

Hahnfeld committed rL314691: [CUDA] Fix name of __activemask().
[CUDA] Fix name of __activemask()
Mon, Oct 2, 10:51 AM
Hahnfeld closed D38468: [CUDA] Fix name of __activemask() by committing rL314691: [CUDA] Fix name of __activemask().
Mon, Oct 2, 10:51 AM
Hahnfeld created D38468: [CUDA] Fix name of __activemask().
Mon, Oct 2, 10:32 AM

Fri, Sep 29

Hahnfeld added inline comments to D38277: [compiler-rt][CMake] Fix configuration on PowerPC with sanitizers.
Fri, Sep 29, 9:13 AM
Hahnfeld committed rL314513: [test] Fix uninitialized memory in omp_taskloop_grainsize.c.
[test] Fix uninitialized memory in omp_taskloop_grainsize.c
Fri, Sep 29, 6:54 AM
Hahnfeld committed rL314512: [CMake] Fix configuration on PowerPC with sanitizers.
[CMake] Fix configuration on PowerPC with sanitizers
Fri, Sep 29, 6:34 AM
Hahnfeld closed D38277: [compiler-rt][CMake] Fix configuration on PowerPC with sanitizers by committing rL314512: [CMake] Fix configuration on PowerPC with sanitizers.
Fri, Sep 29, 6:34 AM

Thu, Sep 28

Hahnfeld created D38372: [OpenMP] Fix passing of -m arguments correctly.
Thu, Sep 28, 11:06 AM
Hahnfeld updated the diff for D38277: [compiler-rt][CMake] Fix configuration on PowerPC with sanitizers.
Thu, Sep 28, 6:51 AM
Hahnfeld added a comment to D38277: [compiler-rt][CMake] Fix configuration on PowerPC with sanitizers.

The error with -DLLVM_USE_SANITIZER=Address is

-- Check if the system is big endian
-- Searching 16 bit integer
-- Looking for stddef.h
-- Looking for stddef.h - not found
-- Check size of unsigned short
-- Check size of unsigned short - failed
-- Check size of unsigned int
-- Check size of unsigned int - failed
-- Check size of unsigned long
-- Check size of unsigned long - failed
CMake Error at <...>/cmake/share/cmake-3.5/Modules/TestBigEndian.cmake:51 (message):
  no suitable type found
Call Stack (most recent call first):
  projects/compiler-rt/cmake/base-config-ix.cmake:151 (TEST_BIG_ENDIAN)
  projects/compiler-rt/cmake/config-ix.cmake:138 (test_targets)
  projects/compiler-rt/CMakeLists.txt:99 (include)
Thu, Sep 28, 6:50 AM

Wed, Sep 27

Hahnfeld resigned from D37810: [test] Enable LeakSanitizer on 64-bit Darwin ASan clang builds.

Resigning as this isn't related to OpenMP anymore

Wed, Sep 27, 1:36 PM
Hahnfeld committed rL314329: [OpenMP] Fix passing of -m arguments to device toolchain.
[OpenMP] Fix passing of -m arguments to device toolchain
Wed, Sep 27, 11:14 AM
Hahnfeld committed rL314330: [OpenMP] Fix translation of target args.
[OpenMP] Fix translation of target args
Wed, Sep 27, 11:14 AM
Hahnfeld closed D38259: [OpenMP] Fix translation of target args by committing rL314330: [OpenMP] Fix translation of target args.
Wed, Sep 27, 11:14 AM
Hahnfeld closed D38258: [OpenMP] Fix passing of -m arguments to device toolchain by committing rL314329: [OpenMP] Fix passing of -m arguments to device toolchain.
Wed, Sep 27, 11:14 AM
Hahnfeld committed rL314328: [OpenMP] Fix memory leak when translating arguments.
[OpenMP] Fix memory leak when translating arguments
Wed, Sep 27, 11:14 AM
Hahnfeld closed D38257: [OpenMP] Fix memory leak when translating arguments by committing rL314328: [OpenMP] Fix memory leak when translating arguments.
Wed, Sep 27, 11:14 AM
Hahnfeld added inline comments to D38258: [OpenMP] Fix passing of -m arguments to device toolchain.
Wed, Sep 27, 7:45 AM

Tue, Sep 26

Hahnfeld accepted D38040: [OpenMP] Add an additional test for D34888.

LGTM

Tue, Sep 26, 3:30 PM
Hahnfeld created D38277: [compiler-rt][CMake] Fix configuration on PowerPC with sanitizers.
Tue, Sep 26, 3:30 PM

Mon, Sep 25

Hahnfeld added a dependency for D38259: [OpenMP] Fix translation of target args: D38258: [OpenMP] Fix passing of -m arguments to device toolchain.
Mon, Sep 25, 2:11 PM
Hahnfeld added a dependent revision for D38258: [OpenMP] Fix passing of -m arguments to device toolchain: D38259: [OpenMP] Fix translation of target args.
Mon, Sep 25, 2:11 PM
Hahnfeld added a dependency for D38258: [OpenMP] Fix passing of -m arguments to device toolchain: D38257: [OpenMP] Fix memory leak when translating arguments.
Mon, Sep 25, 2:11 PM
Hahnfeld added a dependent revision for D38257: [OpenMP] Fix memory leak when translating arguments: D38258: [OpenMP] Fix passing of -m arguments to device toolchain.
Mon, Sep 25, 2:11 PM
Hahnfeld created D38259: [OpenMP] Fix translation of target args.
Mon, Sep 25, 2:10 PM
Hahnfeld created D38258: [OpenMP] Fix passing of -m arguments to device toolchain.
Mon, Sep 25, 2:10 PM
Hahnfeld created D38257: [OpenMP] Fix memory leak when translating arguments.
Mon, Sep 25, 2:09 PM
Hahnfeld added a comment to D38185: Implementation of OMPT as specified in OpenMP 5.0 Preview 1.

Some remarks from a high-level look at this revision. Although it might not be possible to split the OMPT changes, any preparation change should go into separate commits.

  1. Separate out the changes that are not related to OMPT at all (.clang-format, kmp_task_reduction_nest.cpp)
  2. Deleting the old header files and restricting OMPT to OpenMP 5.0 should be doable in a separate change
  3. Are there any changes that are active when OMPT is disabled? We might want to discuss them on their own, all other changes are neatly hidden behind #defines
Mon, Sep 25, 7:05 AM

Sep 22 2017

Hahnfeld added a comment to D38040: [OpenMP] Add an additional test for D34888.

The test is verifying whether the parameter is passed to the kernel correctly. I believe it was not passed as a reference before the patch.

Sep 22 2017, 6:50 AM

Sep 21 2017

Hahnfeld added a comment to D38040: [OpenMP] Add an additional test for D34888.

Hi Doru,

Sep 21 2017, 1:46 PM
Hahnfeld added a comment to D37810: [test] Enable LeakSanitizer on 64-bit Darwin ASan clang builds.

@gtbercea and I will take a look at the leak in D34784 next week and hopefully get it sorted out...

Sep 21 2017, 1:40 PM

Aug 29 2017

Hahnfeld added a comment to D34701: [openmp-target-tests] OpenMP 4.5 Target data test cases.
  1. Is there as specific reason for using both puts and printf? There are also calls to printf without formatting arguments...
  2. Do these tests actually pass with a current version of clang?
Aug 29 2017, 5:48 AM

Aug 14 2017

Hahnfeld added inline comments to D34784: [OpenMP] Add flag for specifying the target device architecture for OpenMP device offloading.
Aug 14 2017, 12:56 AM · Restricted Project
Hahnfeld added a comment to D29660: [OpenMP] Add flag for overwriting default PTX version for OpenMP targets.

Disabled openmp-offload.c on Linux again: https://reviews.llvm.org/rL310772

Aug 14 2017, 12:53 AM
Hahnfeld committed rL310817: Fix memory leak in ToolChain::TranslateOpenMPTargetArgs.
Fix memory leak in ToolChain::TranslateOpenMPTargetArgs
Aug 14 2017, 12:45 AM

Aug 11 2017

Hahnfeld removed a reviewer for D34158: For Linux/gnu compatibility, preinclude <stdc-predef.h> if the file is available : Hahnfeld.
Aug 11 2017, 7:11 AM

Aug 10 2017

Hahnfeld resigned from D34158: For Linux/gnu compatibility, preinclude <stdc-predef.h> if the file is available .
Aug 10 2017, 11:05 PM

Aug 2 2017

Hahnfeld added a comment to D34701: [openmp-target-tests] OpenMP 4.5 Target data test cases.

Hi Jonas

Did you have further comments on this commit?

Thanks
Sunita

Aug 2 2017, 12:13 AM

Aug 1 2017

Hahnfeld accepted D33227: [OpenMP] libomptarget: move debugging dumps under control of environment variable OMP_TARGET_DEBUG.

Given the lack of additional comments, LGTM. Please adapt the commit description to reflect the changes during review.

Aug 1 2017, 11:38 PM · Restricted Project

Jul 26 2017

Hahnfeld added a comment to D35719: [CMake] Disable building libomptarget and add CMake switch.

I've committed the patch without an explicit answer from the IBM guys to unblock the release

Jul 26 2017, 6:56 AM
Hahnfeld committed rL309115: [CMake] Disable building libomptarget and add CMake switch.
[CMake] Disable building libomptarget and add CMake switch
Jul 26 2017, 6:55 AM
Hahnfeld closed D35719: [CMake] Disable building libomptarget and add CMake switch by committing rL309115: [CMake] Disable building libomptarget and add CMake switch.
Jul 26 2017, 6:55 AM

Jul 23 2017

Hahnfeld added inline comments to D35719: [CMake] Disable building libomptarget and add CMake switch.
Jul 23 2017, 11:19 PM
Hahnfeld updated the diff for D35719: [CMake] Disable building libomptarget and add CMake switch.

Disable building by default.

Jul 23 2017, 11:15 PM

Jul 21 2017

Hahnfeld added inline comments to D35719: [CMake] Disable building libomptarget and add CMake switch.
Jul 21 2017, 1:07 AM

Jul 20 2017

Hahnfeld created D35719: [CMake] Disable building libomptarget and add CMake switch.
Jul 20 2017, 11:44 PM

Jul 13 2017

Hahnfeld committed rL307899: [GOMP] Fix (un)tied tasks with the GCC.
[GOMP] Fix (un)tied tasks with the GCC
Jul 13 2017, 3:38 AM

Jul 12 2017

Hahnfeld added a comment to D34876: [asan] Remove check for stack size.
In D34876#806869, @kcc wrote:

This check is a sanity check, please keep it, but feel free to change the constant to e.g. 0x100000000 (16x increase)

Jul 12 2017, 11:04 PM

Jul 5 2017

Hahnfeld added inline comments to D29658: [OpenMP] Customize CUDA-based tool chain selection.
Jul 5 2017, 11:25 PM
Hahnfeld added a comment to D29905: [OpenMP] Pass argument to device kernel by reference when map is used. .

Does this also include the fixes in the following revision?

https://reviews.llvm.org/D29905

Jul 5 2017, 11:15 PM

Jul 4 2017

Hahnfeld committed rL307127: [asan] Remove check for stack size.
[asan] Remove check for stack size
Jul 4 2017, 11:55 PM
Hahnfeld closed D34876: [asan] Remove check for stack size by committing rL307127: [asan] Remove check for stack size.
Jul 4 2017, 11:55 PM

Jul 3 2017

Hahnfeld added a comment to D34158: For Linux/gnu compatibility, preinclude <stdc-predef.h> if the file is available .

The other test that fails is my own new test! It fails because I don't know how to set it up so the test thinks it has a gcc toolchain with version > 4.8. I tried using gcc-toolchain set to various other Linux toolchains that i see in the test/Driver/Inputs - none of them cause the gcc version to be in the range. I also tried using -ccc-installation=Inputs/ which I see being used for gcc version parsing. How can I set up the test so that the GCCInstallation has a Version >= 4.8? I test the new functionality from the console on Linux and can confirm it's working.

Jul 3 2017, 7:33 AM

Jul 1 2017

Hahnfeld added a comment to D34876: [asan] Remove check for stack size.

What's the stack size on CentOS? What was the point of the original check?

Jul 1 2017, 3:52 AM
Hahnfeld updated the summary of D34876: [asan] Remove check for stack size.
Jul 1 2017, 3:52 AM