profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/SimeonEhrig/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
Simeon Ehrig SimeonEhrig Dresden Computer Scientist at HZDR

SimeonEhrig/CUDA-Runtime-Interpreter 8

It's a prototype for an interpreter, which can interpret the host code of a CUDA Program, written with the runtime API.

ComputationalRadiationPhysics/xeus-cling-cuda-container 6

The repository contains container recipes to build the entire stack of Xeus-Cling and Cling including cuda extension with just a few commands.

ComputationalRadiationPhysics/Xeus-Cling-CUDA-Example 5

Game-of-Life example notebook for the xeus-cling-cuda kernel.

ComputationalRadiationPhysics/crp-container 3

Container receipes for different projects of the HZDR Computational Radiation Physics group.

SimeonEhrig/IncrementalPTXJIT 3

Incremental just-in-time compiler for nvidia ptx based on cling.

SimeonEhrig/cling 2

The cling C++ interpreter

SimeonEhrig/emacs-config 2

My personal emacs configuration

SimeonEhrig/cling-helper 1

A collection of C++ scripts for the Cling interpreter. The scripts provide easy access to internal information of the Cling. The main purpose is the development and debugging of the Cling.

SimeonEhrig/MyCMake 1

Contains different examples generated by learning CMake.

push eventSimeonEhrig/alpaka

Simeon Ehrig

commit sha c9680bdce314c111acee31e6c570ae4cf7a4b548

Fix alpaka_add_library for the HIP accelerator

view details

René Widera

commit sha 861d45a17dd5c22ef1e77d1a028e9a4757a39177

Run CUDA 9.2 test on HZDR CI Copy github actions CUDA 9.2 jobs over to HZDR gitlab CI. HZDR CI: - CUDA: use nvidia cuda container (do not install CUDA) - use Ubuntu 18.04 for CUDA 9.2 (Ubuntu 20.04 container does not exists) - do not test `ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLE` together with CUDA (reduce compile time) - OpenMP + CUDA is still tested with `ALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE`

view details

René Widera

commit sha 261bdf70f359b3d97dfdfb3cc2bd39ec0472c8d1

remove CUDA 9.2 Linux tests from Github actions

view details

push time in 17 hours

startedsignalapp/Signal-Desktop

started time in 5 days

startedplasma-umass/coz

started time in 6 days

PullRequestReviewEvent

pull request commentalpaka-group/alpaka

Run CUDA 9.2 tests on HZDR CI

It looks like the alpaka clang-cuda versions are built against GCC5 headers.

You can specify alternative headers clang should use:

clang++ --gcc-toolchain=/usr/local/gcc/6.4.0

Does you know, if g++ also use the libstdc++ version of the default host compiler, means for example if you compile with g++-8 in Ubuntu 16.04, it selects the libstdc++ of the g++-5? If yes, we should implement an mechanism to manual specific it.

psychocoderHPC

comment created time in 7 days

Pull request review commentalpaka-group/alpaka

Run CUDA 9.2 tests on HZDR CI

 .base:-  image: registry.gitlab.com/hzdr/crp/alpaka-group-container/alpaka-ci-${ALPAKA_CONTAINER_SUFFIX}:${ALPAKA_GITLAB_CI_CONTAINER_VERSION}   variables:     ALPAKA_DEBUG: 0-    BOOST_ROOT: /opt/boost/${ALPAKA_BOOST_VERSION}-    BOOST_LIBRARYDIR: /opt/boost/${ALPAKA_BOOST_VERSION}/lib-    ALPAKA_CI_BOOST_LIB_DIR: /opt/boost/${ALPAKA_BOOST_VERSION}+    ALPAKA_CI_BOOST_BRANCH: "boost-${ALPAKA_BOOST_VERSION}"+    BOOST_LIBRARYDIR: "/opt/boost/${ALPAKA_BOOST_VERSION}/lib"+    ALPAKA_CI_OS_NAME: "Linux"   script:-    # create fake sudo-    - cp ./script/gitlabci/fake_sudo.sh /bin/sudo-    - chmod +x /bin/sudo-    - mkdir -p ${ALPAKA_CI_CMAKE_DIR}-    - source ./script/install_cmake.sh+    - source ./script/before_install.sh+    - source ./script/install.sh     - source ./script/run.sh+  interruptible: true -.base_gcc:+.base_cuda_gcc:+  image: nvidia/cuda:${ALPAKA_CI_CUDA_VERSION}-devel-ubuntu${ALPAKA_CI_UBUNTU_VER}

What is wrong with the alpaka container? Can you please open an issue, that it can be fixed: https://gitlab.com/hzdr/crp/alpaka-group-container

psychocoderHPC

comment created time in 7 days

Pull request review commentalpaka-group/alpaka

Run CUDA 9.2 tests on HZDR CI

 echo ALPAKA_CI_BOOST_BRANCH_MINOR: "${ALPAKA_CI_BOOST_BRANCH_MINOR}" #------------------------------------------------------------------------------- # CUDA export ALPAKA_CI_INSTALL_CUDA="OFF"-if [ "${ALPAKA_ACC_GPU_CUDA_ENABLE}" == "ON" ]+if [[ "${ALPAKA_ACC_GPU_CUDA_ENABLE}" == "ON" && -z "${GITLAB_CI+x}" ]]

Why do you use the nvidia/cuda:{version}-devel-ubuntu{version} container and not the nvidia/cuda:{version}-base-ubuntu{version} container and install cuda afterwards, like you suggested for the original alpaka container?

psychocoderHPC

comment created time in 7 days

PullRequestReviewEvent
PullRequestReviewEvent

push eventSimeonEhrig/alpaka

René Widera

commit sha c7728fd42b4f26bd54775cb138ee69cfdcc5dc47

HIP: fix peer to peer memcopy Fix HIP p2p memcopy between devices within a single host system. This PR additionally refactored the CUDA p2p implementation. For a long time CUDA is not requiring explicit p2p copy calls, `cudaMemcpy*` is detecting if a copy operation must be performed between different devices and will to all required configurations.

view details

René Widera

commit sha 8e00ffeddc8d49f48668b7c67c6320305944d074

fix `ALPAKA_STATIC_ACC_MEM_CONSTANT` and `ALPAKA_STATIC_ACC_MEM_GLOBAL` memory - Usage of `ALPAKA_STATIC_ACC_MEM_CONSTANT` and `ALPAKA_STATIC_ACC_MEM_GLOBAL` was broken for clang CUDA and HIP - HIP: `hipGetSymbolAddress` interface was wrongly used Remove the possibility to initialize both types of memory during the definition. Co-authored-by: Simeon Ehrig <s.ehrig@hzdr.de>

view details

René Widera

commit sha 184644118b11821548a76f2ba019b58de1772638

update documentation

view details

René Widera

commit sha d1ccc7a8e11cb410b5aa072d7c70b161a949f69f

fix CUDA 11.3+ compile Fix error shown with CUDA 11.3+ ``` /builds/hzdr/crp/alpaka/example/reduce/src/reduce.cpp:93:38: error: '__T14' was not declared in this scope auto const taskKernelReduceMain(alpaka::createTaskKernel<Acc>( ~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ``` Some kind of error was already fixed with #1293.

view details

Jan Stephan

commit sha 195ee4dc3c516592cead9d429df2eba86bb3a103

Clean CI

view details

René Widera

commit sha ca03584c2526bb8b4f4594dea1844c4f984fee5b

fix `alpaka_compiler_option` Change cmake function `alpaka_compiler_option` to a macro else the variable will not be visible in the parameter scope.

view details

René Widera

commit sha 7a6da751aa953490cdf4d56aada89a80a63bbcc7

create view factory methods - add method `createView` and `createSubView` - move method `calculatePitchesFromExtents` from `ViewPlainPtr.hpp` into `view/Traits.hpp`

view details

René Widera

commit sha b322b917c8be1ab55f113ca60cedb01e2f33f9f3

use factories to create views INstead of explicit defining view types use the factories to simplify the view creation.

view details

René Widera

commit sha b5ed55a2bd660bec0339b68cb620833fa21f9725

fix cheatsheet link in the docs

view details

René Widera

commit sha c09d0fad0f7bf2cbda5a525ad197936ac44c3dbd

update cheatsheet with view factories

view details

René Widera

commit sha 84aab66e819078839019de07e26bb77b08100687

address review comments - update documentation - change argument order of `createView()`

view details

René Widera

commit sha 208eb3263845063db2c2217d8bcc1b50718274dd

fix warning in thread fence test Fix ``` alpaka/test/unit/mem/fence/src/FenceTest.cpp(49): warning #191-D: type qualifier is meaningless on cast type detected during: ... ```

view details

René Widera

commit sha 5b2426aa730ae708b6b1eac89bf4fe98f66dc471

fix Clang-CUDA math tests fix: #1020 The math tests sometimes failed because we using random input data for tests and `exp` results are not exact compared to `std::exp`.

view details

Bernhard Manfred Gruber

commit sha a1ec01202ac0888299505684e493340d2648e04e

implement tuple interface for Vec to allow structured bindings

view details

Jeffrey Kelling

commit sha 285066064dce08ff588cd59aa0ab0df998e835ac

Omp5: Implement createStaticDevMemView(...DevOmp5...) DevOmp5: Add matStatic()

view details

Jeffrey Kelling

commit sha 7497e83df9c2437290a3e5e0b96da1d9995d2ad1

DevOacc: move makeCurrent() to DevOaccImpl DevOaccImpl::ctor: Add makeCurrent() before parallel

view details

Jeffrey Kelling

commit sha b8c37843a1b89034f391b3e37b1eb97c8f431f4f

oacc/Copy: Fix device to device copy for OpenACC < 2.5

view details

Jeffrey Kelling

commit sha 7cd1f1f862bbc9992b8e3fb694de9791a5f03fac

Oacc: Implement createStaticDevMemView(...DevOacc...)

view details

Simeon Ehrig

commit sha 9ac94b8faf85a9bceb3c260ddb3a32599644d99f

Add initial GitLab CI - add .gitlab-ci.yml - create job templates - basic GCC job - basic CUDA job - add CI documentation Co-authored-by: René Widera <r.widera@hzdr.de>

view details

push time in 7 days

issue commentalpaka-group/alpaka

CMake Modernization

Okay, but I'm not sure if it will work, if link one target with two different backends. But like we already offline discussed, I think the behavior will be similar to the current cmake. It depends on the compiler, if e.g. OpenMP with CUDA is working together. But we get also a new ability:

add_executable(testExe main.cpp)

add_library(cudaKernels cudaKernels.cpp)
target_link_library(cudaKernels PRIVATE alpaka::cuda)

add_library(omp2Kernels omp2Kernels.cpp)
target_link_library(omp2Kernels PRIVATE alpaka::omp2)

target_link_library(testExe PRIVATE cudaKernels omp2Kernels)

This allows for example, to compile the code with the Clang compiler as host and device. At the moment, it does not work, because clang++ -fopenmp -xcuda is not working.

j-stephan

comment created time in 7 days

push eventSimeonEhrig/vikunja

Simeon Ehrig

commit sha 4dd3594802864b04f2f54f51db1d5fe093d0add1

Changed interface of transform and reduce functor - now, the alpaka acc object is required - transform: func(TData i) -> func(TAcc const &acc, TData i) - reduce: func(TData i, TData k) -> func(TAcc const &acc, TData i, TData k)

view details

Simeon Ehrig

commit sha 9c78145c4c19a3bee91d0f1d5660ac62bc829373

Merge pull request #27 from SimeonEhrig/addAccToAPI Add Alpaka `TAcc` to inferface of functors

view details

Simeon Ehrig

commit sha 83d54160607446e141f7f52cfdc6bc2d651672c6

Fix alpaka 0.7 version for CI

view details

Simeon Ehrig

commit sha 3b0b165eef39d629f5d47cda3ecbdc26e03c88d2

Merge pull request #28 from SimeonEhrig/fixTestAlpakaVersion Fix alpaka 0.7 version for CI

view details

push time in 7 days

issue commentalpaka-group/alpaka

CMake Modernization

2\. I originally had the idea to introduce back-end-specific targets. I believe we might be able to tackle this soon because `alpaka_add_{executable,library}` are much less complex than they once were. The only thing still done by these functions is to set the source file properties for a given target. We should investigate whether this can be solved by setting the `LINKER_LANGUAGE` property on the alpaka target.

For the user interface, do you mean something like:

add_executable(testExe main.cpp)
target_link_library(testExe PUBLIC alpaka::CudaACC)
j-stephan

comment created time in 8 days

PR opened alpaka-group/alpaka

Fix alpaka_add_library for the HIP accelerator

I got the error message, that HIP_ADD_LIBRARY is not defined, if I called the cmake function alpaka_add_library. After removing the HIP specific part in the cmake function alpaka_add_library everything works fine. The official HIP documentation also suggest to use add_library: https://rocmdocs.amd.com/en/latest/Installation_Guide/Using-CMake-with-AMD-ROCm.html#using-hip-in-cmake

I think HIP_ADD_LIBRARY is deprecated, because it was defined in the findHIP.cmake, which is not shipped with HIP installation anymore.

+1 -24

0 comment

1 changed file

pr created time in 8 days

create barnchSimeonEhrig/alpaka

branch : fixAddLibraryHIP

created branch time in 8 days

push eventalpaka-group/vikunja

Simeon Ehrig

commit sha 83d54160607446e141f7f52cfdc6bc2d651672c6

Fix alpaka 0.7 version for CI

view details

Simeon Ehrig

commit sha 3b0b165eef39d629f5d47cda3ecbdc26e03c88d2

Merge pull request #28 from SimeonEhrig/fixTestAlpakaVersion Fix alpaka 0.7 version for CI

view details

push time in 9 days

PR opened alpaka-group/vikunja

Fix alpaka 0.7 version for CI
+1 -1

0 comment

1 changed file

pr created time in 9 days

create barnchSimeonEhrig/vikunja

branch : fixTestAlpakaVersion

created branch time in 9 days

pull request commentalpaka-group/alpaka

HIP tests on HZDR CI

@psychocoderHPC Be aware, that our AMD has 2 graphics cards with 3 GPUs. Therefore you need to setup the environment variable HIP_VISIBLE_DEVICES like I did in cupla.

export HIP_VISIBLE_DEVICES =

  • 0 and 1: GPU 0 and 1 of the dual Fury Graphics card
  • 2: Vega 64
psychocoderHPC

comment created time in 9 days

pull request commentComputationalRadiationPhysics/picongpu

CI: use stages for downstream pipe

@psychocoderHPC I suggest, that you also implement a functionality to manual choose which jobs are executed or preferred. I think, this is really helpful, if a specific job crashes because of a bug. I already implemented a python function, which grep a list of job names from the commit message: https://gitlab.com/hzdr/crp/gitlab-ci-test/-/tree/jobControllCommitMessage/jobControllCommitMessage

You can integrate it in the job generator.

psychocoderHPC

comment created time in 12 days

startedagronholm/typeguard

started time in 14 days

pull request commentComputationalRadiationPhysics/picongpu

CI: compile on x86 runners

Tobias added an new tag to the CPU runners: cpuonly Please use it, to not block the GPU runners with CPU only jobs.

psychocoderHPC

comment created time in 14 days

push eventalpaka-group/vikunja

Simeon Ehrig

commit sha 4dd3594802864b04f2f54f51db1d5fe093d0add1

Changed interface of transform and reduce functor - now, the alpaka acc object is required - transform: func(TData i) -> func(TAcc const &acc, TData i) - reduce: func(TData i, TData k) -> func(TAcc const &acc, TData i, TData k)

view details

Simeon Ehrig

commit sha 9c78145c4c19a3bee91d0f1d5660ac62bc829373

Merge pull request #27 from SimeonEhrig/addAccToAPI Add Alpaka `TAcc` to inferface of functors

view details

push time in 15 days

PR merged alpaka-group/vikunja

Reviewers
Add Alpaka `TAcc` to inferface of functors
  • Changed interface of transform and reduce functor
  • transform: func(TData i) -> func(TAcc const &acc, TData i)
  • reduce: func(TData i, TData k) -> func(TAcc const &acc, TData i, TData k)
+100 -26

5 comments

8 changed files

SimeonEhrig

pr closed time in 15 days

pull request commentalpaka-group/vikunja

Add Alpaka `TAcc` to inferface of functors

Originally, I planed to make the Acc object optional but I don't know how. Maybe it is coming back in the future if I have more experience.

SimeonEhrig

comment created time in 15 days

push eventSimeonEhrig/cupla

Simeon Ehrig

commit sha 9a528229b96a1673a7f61f2a1be86539349584b6

Fix CMake, that alpaka can be included via add_subdirectory before cupla in a external project

view details

push time in 16 days

PR opened alpaka-group/cupla

Allow to include alpaka via add_subdirectory before cupla in an external project

With this PR, the following CMake is possible in a project, which uses cupla.

cmake_minimum_required(VERSION 3.18.0)
project(test)

add_subdirectory(alpaka)
add_subdirectory(cupla)

cupla_add_executable(${PROJECT_NAME} main.cpp)
+37 -20

0 comment

2 changed files

pr created time in 16 days

push eventSimeonEhrig/cupla

Simeon Ehrig

commit sha 2e43ec59c502b89725023c8fe8c55022674c6d54

Fix CMake, that alpaka can be included via add_subdirectory before cupla in a external project

view details

push time in 16 days

create barnchSimeonEhrig/cupla

branch : alpakaAddSubdirectory

created branch time in 16 days