protze.joachim (Joachim Protze)
User

Projects

User does not belong to any projects.

User Details

User Since
Sep 24 2015, 4:45 AM (146 w, 5 d)

Recent Activity

Today

protze.joachim accepted D49418: Also support several images for elf.

The change is symetric with what is implemented for cuda.
LGTM

Tue, Jul 17, 8:38 AM
protze.joachim updated the summary of D49418: Also support several images for elf.
Tue, Jul 17, 4:06 AM

Thu, Jul 5

protze.joachim accepted D48984: [OpenMP] Define __STDC_FORMAT_MACROS in callback.h.

Lgtm

Thu, Jul 5, 11:57 AM
protze.joachim accepted D48932: [OMPT] Add synchronization to threads_nested.c testcase.

LGTM

Thu, Jul 5, 2:13 AM
protze.joachim accepted D48888: Dropped non-supoorted "--no-as-needed" flag from OMPT tests for macOS.

LGTM

Thu, Jul 5, 2:11 AM

Tue, Jul 3

protze.joachim added inline comments to D47824: [OMPT] Add testcases for thread_begin and thread_end callbacks.
Tue, Jul 3, 12:54 PM
protze.joachim requested changes to D48888: Dropped non-supoorted "--no-as-needed" flag from OMPT tests for macOS.

As discussed on the mailing list, the flag should only be dropped on Mac OS. My pragmatic solution would be:

Tue, Jul 3, 12:47 PM

Wed, Jun 20

protze.joachim added a comment to D47717: [OMPT] Make sure that OMPT is enabled in runtime entry points that access internals of the runtime.

I think we should change the name of the bit ompt_enabled.enabled to ompt_enabled.activated. That would make joachim’s distinction clear. I was commenting on the setting of OMP_TOOL=enabled. There is a difference between enabled and activated.

Wed, Jun 20, 8:40 AM

Mon, Jun 18

protze.joachim added a comment to D47717: [OMPT] Make sure that OMPT is enabled in runtime entry points that access internals of the runtime.

The OMPT tool can decide at multiple points to be inactive, here we look at:

Mon, Jun 18, 7:54 AM
protze.joachim added a comment to D48282: [OMPT] Use alloca() to force availability of frame pointer .

The problem, which gets visible in this test case is the use of builtin_frame_address(1), which is documented to be not safe.
Is there a better way to get the canonical frame address of the calling function? Also the address returned by
builtin_frame_address seems to be different from the canonical frame address. How can we get the requested address?

Mon, Jun 18, 7:28 AM

Jun 7 2018

protze.joachim added inline comments to D47903: [OpenMP] Use C++11 Atomics - barrier, tasking, and lock code.
Jun 7 2018, 2:19 PM · Restricted Project
protze.joachim added a comment to D47709: [OMPT] Fix OMPT callbacks for the taskloop construct and add testcase.

taskloop is actually a tasking construct and explicitly no worksharing construct. So, please move the test in the task directory.

Jun 7 2018, 8:54 AM

Jun 4 2018

protze.joachim added inline comments to D47709: [OMPT] Fix OMPT callbacks for the taskloop construct and add testcase.
Jun 4 2018, 1:40 PM

May 27 2018

protze.joachim accepted D47119: [OMPT] Fix test parallel/not_enough_threads.c.

LGTM

May 27 2018, 3:11 AM

May 10 2018

protze.joachim added a comment to D44881: [sanitizer_common] Ignore unloading of suppressed library.

@dvyukov thanks for the detailed reasoning.

May 10 2018, 11:27 AM

May 7 2018

protze.joachim created D46533: [OMPT] Provide the right thread_num for ancestor levels.
May 7 2018, 9:42 AM
protze.joachim accepted D43568: [OMPT] Rename ompt_frame_t to omp_frame_t.

Result of discussion in Tools WG was that the spec will keep omp_frame_t. I will submit this patch together with D43568

May 7 2018, 9:31 AM
protze.joachim created D46530: [OMPT] Rename ompt_wait_id to omp_wait_id.
May 7 2018, 9:28 AM
protze.joachim accepted D46260: [OMPT] Fix thread_num for implicit_task_end callbacks in nested parallel regions.

LGTM

May 7 2018, 5:46 AM

Apr 23 2018

protze.joachim added a comment to D45890: [OMPT] Add implementation and tests of Archer tool.

Reasons to keep the name archer (or archer-rt):

  • We already have that name, we don't need to come up with a new name :)
  • You can easily find related publications under that name, which explain the reasoning behind the library
  • The programmer should never see the name, once the tool is integrated into the runtime workflow
Apr 23 2018, 2:12 AM

Apr 20 2018

protze.joachim created D45890: [OMPT] Add implementation and tests of Archer tool.
Apr 20 2018, 10:01 AM

Apr 17 2018

protze.joachim updated the diff for D44881: [sanitizer_common] Ignore unloading of suppressed library.

I added a testcase. I added it just for Linux, because I have no machine ready to test dl-loading on other OS.

Apr 17 2018, 8:52 AM

Apr 11 2018

protze.joachim added a reviewer for D44881: [sanitizer_common] Ignore unloading of suppressed library: vitalybuka.
Apr 11 2018, 1:55 AM

Mar 25 2018

protze.joachim updated the diff for D15154: [sanitizer_common] Enable external tool to register PrintfAndReportCallback.

I updated this differential to only export the function.

Mar 25 2018, 2:27 PM
protze.joachim created D44881: [sanitizer_common] Ignore unloading of suppressed library.
Mar 25 2018, 2:15 PM

Feb 28 2018

protze.joachim accepted D43882: [OMPT] Fix interoperability test with GCC.

LGTM.

Feb 28 2018, 1:24 PM
protze.joachim added reviewers for D43897: [OMPT] Fix task-type test with GCC: hbae, Hahnfeld.
Feb 28 2018, 1:14 PM
protze.joachim created D43897: [OMPT] Fix task-type test with GCC.
Feb 28 2018, 1:13 PM
protze.joachim added inline comments to D42523: [OMPT] Add api_calls_misc.c testcase and rename api_calls.c testcase.
Feb 28 2018, 9:25 AM

Feb 23 2018

protze.joachim added a comment to D41942: [OMPT] Add interoperability testcase.

The idea was that the WAIT in line 11 should ensure that both initial threads arrived.
But actually the runtime is not initialized before line 11. To fix the race, we need to call into the runtime in a way, that makes both threads initial threads before the SIGNAL in line 9.

Feb 23 2018, 12:19 PM
protze.joachim accepted D43308: [OMPT] Fix parallel_data in implicit barrier-end.

LGTM

Feb 23 2018, 8:46 AM
protze.joachim accepted D43191: [OMPT] Fix test tasks/serialized.c with optimization.

LGTM

Feb 23 2018, 8:03 AM

Feb 22 2018

protze.joachim added a comment to D43568: [OMPT] Rename ompt_frame_t to omp_frame_t.

As far as I know, the current implementation is very close to the specification in TR6.
Here we have a tiny diff between the spec and the implementation. I agree that we should not apply this patch if we will roll back the change in the spec.

Feb 22 2018, 12:58 AM

Feb 19 2018

protze.joachim updated the diff for D43115: [OMPT] Fix inconsistent testcases.

Use the captured pattern as pointed out by Olga

Feb 19 2018, 3:24 AM

Feb 17 2018

protze.joachim added inline comments to D43308: [OMPT] Fix parallel_data in implicit barrier-end.
Feb 17 2018, 8:41 AM
protze.joachim requested changes to D43308: [OMPT] Fix parallel_data in implicit barrier-end.
Feb 17 2018, 6:46 AM
protze.joachim updated the diff for D43115: [OMPT] Fix inconsistent testcases.

The initial issue in this patch is resolved by D43195. So removed the additionally printed address for AARCH64, but still allow the testcases to match any printed address.

Feb 17 2018, 5:57 AM
protze.joachim added a comment to D41942: [OMPT] Add interoperability testcase.

I applied clang-format on commit.

Feb 17 2018, 1:45 AM

Feb 14 2018

protze.joachim added inline comments to D43191: [OMPT] Fix test tasks/serialized.c with optimization.
Feb 14 2018, 10:10 AM
protze.joachim added a comment to D43191: [OMPT] Fix test tasks/serialized.c with optimization.

This behavior is right (p.417,l.5):

Feb 14 2018, 10:10 AM

Feb 9 2018

protze.joachim created D43115: [OMPT] Fix inconsistent testcases.
Feb 9 2018, 3:26 AM

Feb 6 2018

protze.joachim added reviewers for D42472: [OMPT] Fix ompt_get_task_info() and add tests for it: hbae, omalyshe.
Feb 6 2018, 6:20 AM
protze.joachim added reviewers for D41904: [OMPT] Add tool_not_available testcase: hbae, omalyshe.
Feb 6 2018, 6:13 AM
protze.joachim added reviewers for D41942: [OMPT] Add interoperability testcase: hbae, omalyshe.
Feb 6 2018, 6:13 AM
protze.joachim added reviewers for D42166: [OMPT] Add tool_available_search testcase: hbae, omalyshe.
Feb 6 2018, 6:13 AM
protze.joachim added reviewers for D42162: [OMPT] Update api_calls testcase: omalyshe, hbae.
Feb 6 2018, 6:13 AM
protze.joachim added reviewers for D42169: [OMPT] Formatting: omalyshe, hbae.
Feb 6 2018, 4:54 AM
protze.joachim added reviewers for D42523: [OMPT] Add api_calls_misc.c testcase and rename api_calls.c testcase: hbae, omalyshe.
Feb 6 2018, 4:54 AM

Jan 30 2018

protze.joachim added a reviewer for D42709: [OMPT] Fix tool initialization returning 0: omalyshe.
Jan 30 2018, 2:33 PM
protze.joachim created D42709: [OMPT] Fix tool initialization returning 0.
Jan 30 2018, 2:18 PM

Jan 25 2018

protze.joachim abandoned D42432: [OMPT] Limit omp_control_tool testcase to 5.0.

ok

Jan 25 2018, 10:17 AM
protze.joachim updated the diff for D42432: [OMPT] Limit omp_control_tool testcase to 5.0.

NEEDS must be REQUIRES

Jan 25 2018, 6:49 AM
protze.joachim updated the summary of D42427: Fix broken OpenMP runtime test cases for Windows.
Jan 25 2018, 6:49 AM
protze.joachim updated the diff for D42427: Fix broken OpenMP runtime test cases for Windows.

NEEDS replaced by REQUIRES

Jan 25 2018, 6:46 AM
protze.joachim added a comment to D42427: Fix broken OpenMP runtime test cases for Windows.

It was defined in https://reviews.llvm.org/D42432

Jan 25 2018, 6:41 AM
protze.joachim added a comment to D42427: Fix broken OpenMP runtime test cases for Windows.

@omalyshe do you have any advice, what is needed to make the omp_control_tool symbol available on Windows?

My naive approach was to follow the things done for omp_get_thread_num.

Adding omp_control_tool to dllexports is a right way.

LGTM.

Jan 25 2018, 4:44 AM
protze.joachim added a comment to D42427: Fix broken OpenMP runtime test cases for Windows.

@omalyshe do you have any advice, what is needed to make the omp_control_tool symbol available on Windows?

Jan 25 2018, 4:23 AM
protze.joachim added inline comments to D42472: [OMPT] Fix ompt_get_task_info() and add tests for it.
Jan 25 2018, 2:10 AM

Jan 23 2018

protze.joachim added inline comments to D42427: Fix broken OpenMP runtime test cases for Windows.
Jan 23 2018, 10:33 AM
protze.joachim updated the diff for D42427: Fix broken OpenMP runtime test cases for Windows.

Assuming we have pthread on all platforms but Windows

Jan 23 2018, 10:02 AM
protze.joachim created D42432: [OMPT] Limit omp_control_tool testcase to 5.0.
Jan 23 2018, 9:54 AM
protze.joachim added reviewers for D42432: [OMPT] Limit omp_control_tool testcase to 5.0: Hahnfeld, jlpeyton.
Jan 23 2018, 9:54 AM
protze.joachim added a comment to D38185: Implementation of OMPT as specified in OpenMP 5.0 Preview 1.
In D38185#985047, @hans wrote:

Also, r317085 doesn't seem to have gone to any mailing list, which is unfortunate. Maybe because the committer was not subscribed and it got stuck in moderation.

Jan 23 2018, 8:53 AM
protze.joachim added reviewers for D42427: Fix broken OpenMP runtime test cases for Windows: jlpeyton, Hahnfeld, omalyshe.
Jan 23 2018, 8:52 AM
protze.joachim created D42427: Fix broken OpenMP runtime test cases for Windows.
Jan 23 2018, 8:49 AM

Jan 17 2018

protze.joachim added a comment to D42162: [OMPT] Update api_calls testcase.

LGTM now

Jan 17 2018, 3:33 AM
protze.joachim accepted D41896: [OMPT] Use fuzzy return addresses in lock testcases .

Still surprised that this is necessary for void function calls, but with that change the tests also succeed for intel compiler.

Jan 17 2018, 1:15 AM
protze.joachim accepted D41166: [OMPT] Rename ompt_mutex_impl_t to kmp_mutex_impl.

We discussed to relax the wording in the OpenMP spec, so that any implementation that provides ompt_mutex_impl_unknown with value 0 is complying. This will include #define, enum or "static const" values.

Jan 17 2018, 1:02 AM

Jan 11 2018

protze.joachim added a comment to D41817: [OMPT] Enable OMPT on 32-bit ARM machines.

Ah, I think I understand your plan now.

Jan 11 2018, 4:30 AM · Restricted Project
protze.joachim added inline comments to D41942: [OMPT] Add interoperability testcase.
Jan 11 2018, 4:18 AM
protze.joachim accepted D41167: [OMPT] Return appropiate values for ompt runtime entry points when the current thread is not an OpenMP thread.

LGTM

Jan 11 2018, 2:51 AM
protze.joachim accepted D41182: [OMPT] Add Workaround for Intel Compiler Bug.

LGTM

Jan 11 2018, 1:37 AM
protze.joachim added a comment to D41817: [OMPT] Enable OMPT on 32-bit ARM machines.
Library built with: Clang  | GCC
Testing with:              |
Clang               works  | fails
GCC                 fails  | fails
Jan 11 2018, 1:08 AM · Restricted Project

Jan 10 2018

protze.joachim accepted D41831: Minor code cleanup.

Ok, I think, I now understood the change.

Jan 10 2018, 4:37 AM

Jan 9 2018

protze.joachim added a comment to D41182: [OMPT] Add Workaround for Intel Compiler Bug.

@omalyshe AFAIK this issue is Case#: 03138964

Jan 9 2018, 9:32 AM
protze.joachim accepted D41853: [OMPT] Fix cast and printf of wait_id in lock test.

I would have expected, that

printf("%" PRIu64 ": &lock: %lu\n", ompt_get_thread_data()->value, (unsigned long) &lock);

would also fix the issue.

Jan 9 2018, 9:30 AM
protze.joachim added a comment to D41817: [OMPT] Enable OMPT on 32-bit ARM machines.

As I understand the issue, this is not a fault in the test code, but the value provided by parallel-begin is wrong? In that case, I suggest to pass NULL for this architecture/compiler.

Jan 9 2018, 9:02 AM · Restricted Project
protze.joachim added inline comments to D41831: Minor code cleanup.
Jan 9 2018, 6:56 AM
protze.joachim added inline comments to D41831: Minor code cleanup.
Jan 9 2018, 5:38 AM
protze.joachim added inline comments to D41831: Minor code cleanup.
Jan 9 2018, 5:28 AM

Jan 8 2018

protze.joachim added a comment to D41817: [OMPT] Enable OMPT on 32-bit ARM machines.

I didn't even know these tests are allowed to be compiled with anything else than clang 6.0+...

Jan 8 2018, 8:30 AM · Restricted Project

Jan 2 2018

protze.joachim updated the diff for D41508: [OMPT] Build runtime with OMPT support by default.

Updated the README

Jan 2 2018, 3:46 AM
protze.joachim updated the diff for D41508: [OMPT] Build runtime with OMPT support by default.

Updated the README

Jan 2 2018, 3:25 AM

Dec 24 2017

protze.joachim updated the diff for D41508: [OMPT] Build runtime with OMPT support by default.

Removed 32 bit ARM from the list of supported architectures until it is tested.

Dec 24 2017, 7:00 AM

Dec 23 2017

protze.joachim accepted D41165: [OMPT] Set frame address when creating a task with dependences.
Dec 23 2017, 6:50 AM
protze.joachim added a comment to D41165: [OMPT] Set frame address when creating a task with dependences.

Does the latest change assume current_task == new_taskdata->td_parent (which would make sense to me)? If not this change has more implications than the frame address.

Dec 23 2017, 6:50 AM
protze.joachim updated the diff for D41508: [OMPT] Build runtime with OMPT support by default.

I added the requirement for hardware architecture support to LIBOMP_HAVE_OMPT_SUPPORT.
This way one would also get an error when manually turing LIBOMP_OMPT_SUPPORT=on on an unsupported architecture.

Dec 23 2017, 1:58 AM

Dec 22 2017

protze.joachim added a comment to D41182: [OMPT] Add Workaround for Intel Compiler Bug.
icc bug

What bug it is and which version of ICC is affected?

Dec 22 2017, 8:49 AM

Dec 21 2017

protze.joachim created D41508: [OMPT] Build runtime with OMPT support by default.
Dec 21 2017, 11:33 AM
protze.joachim updated the summary of D41504: Remove unused positional argument for printf.
Dec 21 2017, 10:53 AM
protze.joachim created D41504: Remove unused positional argument for printf.
Dec 21 2017, 10:52 AM
protze.joachim created D41499: [OMPT] Fix failing test cases for gcc on Ubuntu.
Dec 21 2017, 10:01 AM
protze.joachim added inline comments to D41171: [OMPT] Handle null pointer in set_callback to improve performance.
Dec 21 2017, 6:24 AM
protze.joachim added a comment to D41165: [OMPT] Set frame address when creating a task with dependences.

This fixes the test with GCC which already had the enter address, right?

Dec 21 2017, 5:04 AM
protze.joachim requested changes to D41165: [OMPT] Set frame address when creating a task with dependences.

I think, the address needs to be reset on return. Please include commit 45a3a39 to this patch.

Dec 21 2017, 5:02 AM

Dec 13 2017

protze.joachim added a comment to D41171: [OMPT] Handle null pointer in set_callback to improve performance.

This change is more a performace optimization. (reset the bit in the bitmap to 0)
The callback was set to NULL before and all callback invocation checks for the NULL-pointer.

Dec 13 2017, 7:54 AM

Dec 8 2017

protze.joachim added inline comments to D40949: [OMPT] Add missing ompt_get_num_procs function.
Dec 8 2017, 4:03 AM

Dec 7 2017

protze.joachim updated the diff for D40949: [OMPT] Add missing ompt_get_num_procs function.

Thanks

Dec 7 2017, 7:03 AM
protze.joachim created D40949: [OMPT] Add missing ompt_get_num_procs function.
Dec 7 2017, 4:10 AM

Dec 5 2017

protze.joachim added a comment to D40386: [OpenMP] Add entry for Intel Compiler 18.

For released versions (at least for the current way of defining the macros) the following should be possible (not tested):

#define KMP_COMPILER "Intel C++ Compiler " stringer(__INTEL_COMPILER / 100) "." stringer(__INTEL_COMPILER % 100) "." stringer(__INTEL_COMPILER_UPDATE)

Maybe you need to assign the temporary values to other defines for the macro magic to work, I'm not too familiar with how it works...

Dec 5 2017, 5:33 AM