Comments (10)
I agree that there should be better documentation for building and running tests in oneMKL.
It also seems that there isn't support (as far as I can tell) for building with non standard ROCm installation. If the build process was better documented, the documentation effort could potentially also act as a survey on what works and what doesn't. This would be valuable IMO
from onemkl.
Hi @hjabird We are looking for community support for this project. Are you (or any others) willing to pitch in and help improve the documentation?
from onemkl.
Hi Craig, yes we will try to help with some of these issues when time permits. It would be useful to have some feedback before we start spending time on it.
To continue the discussion from #458 (comment) I was thinking it could make sense to document some configurations that are expected to work but not regularly tested. This could help users try some configurations and report issues. I think the combination of all the domains, platforms, compilers and OS lead to a number of configurations too high to be regularly tested. I'm curious to have @mmeterel's opinion on this.
from onemkl.
Hi Craig, yes we will try to help with some of these issues when time permits. It would be useful to have some feedback before we start spending time on it.
To continue the discussion from #458 (comment) I was thinking it could make sense to document some configurations that are expected to work but not regularly tested. This could help users try some configurations and report issues. I think the combination of all the domains, platforms, compilers and OS lead to a number of configurations too high to be regularly tested. I'm curious to have @mmeterel's opinion on this.
@Rbiessy I support adding documentation for untested configurations as long as how to use them are clearly documented and reproducible by internal and external developers. Main concern with not testing is, as the repo keeps evolving, how will we ensure these untested configurations will continue working? What do you think for such cases?
Another point about documenting which configs are tested and which are not. I support being transparent to users by providing this information (like our AMDs backend are not tested in CI yet), but we should make plans to close the gaps in CI in a timely manner. (For example adding AMD backends in CI is in progress)
from onemkl.
I don't have a good solution to ensure that untested configurations will continue working. I would say that if people are trying to use them we will eventually get an issue if something is broken. That is my understanding of a community-driven project.
Regarding the CI I have some doubts about how much details we should share. I believe that Alexey (@toxicscum) is working on closing the gap but I doubt everything in the README can be tested.
from onemkl.
I don't have a good solution to ensure that untested configurations will continue working. I would say that if people are trying to use them we will eventually get an issue if something is broken. That is my understanding of a community-driven project.
Regarding the CI I have some doubts about how much details we should share. I believe that Alexey (@toxicscum) is working on closing the gap but I doubt everything in the README can be tested.
Relying on user input when/if something is broken degrades the quality of the project/repo imho. One option would be, guaranteeing all the mentioned configurations work in every release. Then, at least we have a time stamp where the user can get a stable repo.
Regarding details about CI: When we mention about a configuration and say it is not tested regularly, we are already sharing this information. Ideally, in a repo like this, all CI should be public IMHO.
from onemkl.
So we've had internal discussions about the release of oneMKL. We are planning to create releases of oneMKL interface and ship them with the oneAPI base toolkit. These will be tested by the oneAPI core team at Codeplay using icpx
and the Codeplay Nvidia and AMD plugins. As I far as I know this is the only recommended way for users to get DPC++ running on all hardware. I am planning to start a separate discussion on the oneMKL interface release eventually.
I hope it makes more sense why we think we should document some configurations that are expected to work but may not be regularly tested yet.
from onemkl.
So we've had internal discussions about the release of oneMKL. We are planning to create releases of oneMKL interface and ship them with the oneAPI base toolkit. These will be tested by the oneAPI core team at Codeplay using
icpx
and the Codeplay Nvidia and AMD plugins. As I far as I know this is the only recommended way for users to get DPC++ running on all hardware. I am planning to start a separate discussion on the oneMKL interface release eventually.I hope it makes more sense why we think we should document some configurations that are expected to work but may not be regularly tested yet.
This is news to me (releasing oneMKL interfaces as part of base tool kit) Is this in replacement of the releases done in oneMKL interfaces repo or in addition to it? Tagging Irina and Maria for their information and inputs. @mkrainiuk @itopinsk
from onemkl.
I would prefer to keep the discussion on releases separate. I will try to start the discussion this week.
from onemkl.
Related Issues (20)
- const scaling paramters not supported for gemm_batch HOT 1
- oneMKL's CMake Does Not Detect SYCL Support when Compiled with `intel/llvm` HOT 1
- Could NOT find CBLAS (missing: CBLAS_file) HOT 2
- -DENABLE_CUFFT_BACKEND missing from documentation HOT 1
- Outdated conan installation instructions HOT 2
- Consider creating oneMKL releases HOT 5
- dlopen libOpenCL by SONAME to support split prefix distributions HOT 3
- Kernel verification failure during execution of axpby with complex float inputs HOT 6
- Request to implement lapack::gesv HOT 3
- Missing function rng::device::generate_single HOT 1
- Request to add compute_mode parameter for BLAS level 3 APIs (and their extensions)
- mklcpu backend is broken with adaptiveCPP
- [BLAS] Netlib backend fails to compile with AdaptiveCpp
- [BLAS] rocBLAS backend test can fail with HSA_STATUS_ERROR_INVALID_PACKET_FORMAT
- [DFT][MKLGPU] Use FWD_ and BWD_STRIDES for MKLGPU 2024.1
- [DFT] Use FWD_ and BWD_STRIDE API
- [BLAS] Wrong namespace usage in BLAS Documentation HOT 5
- pthreads not found HOT 3
- [DFT] compute_forward (forward_ip_rr) function template in forward.hpp overrides compute_forward (forward_op_cc) if in and out types are the same HOT 9
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from onemkl.