-
Notifications
You must be signed in to change notification settings - Fork 42
SNL sparta #1033
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
SNL sparta #1033
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
- Sparta dry run should be passing
- Add Sparta+openmp and Sparta+rocm to CI on Dane and Tuo or Tioga
- Since this will likely go in after the Spack 1.0 support, please update the package to meet Spack 1.0 requirements.
scheibelp
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
(this looks compatible with #953)
| BENCHMARK: [amg2023, kripke, laghos, raja-perf] | ||
| VARIANT: [+rocm] | ||
| GPUMODE: SPX | ||
| # sparta needs specific variant |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@michaelmckinsey1 What do you mean by "Sparta needs a different version of rocm"?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
That's not what this is saying. This benchmark can not be defined the same as the other rocm benchmarks because it requires a specific variant when running with rocm, i.e. fft_kokkos=hipfft. That is why I was thinking this is bad practice to encode this variant in the experiment. I was thinking this is something you would set in the package.py, like depends_on("hipfft", when="rocm"), but after discussing with @scheibelp and @rfhaque yesterday I'm not sure it's so simple.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@scheibelp we need help with this
experiments/sparta-snl/experiment.py
Outdated
| if self.spec.satisfies("+openmp"): | ||
| kokkos_mode += "t {n_threads_per_proc}" | ||
| if self.spec.satisfies("+rocm") or self.spec.satisfies("+cuda"): | ||
| kokkos_mode += " g {n_gpus}" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
g should probably be 1 -- or n_gpus/m_gpus_per_node
.github/utils/dryruns.py
Outdated
| "sparta-snl+openmp aws-pcluster instance_type=c6g.xlarge", | ||
| "sparta-snl+openmp aws-pcluster instance_type=c4.xlarge", | ||
| "sparta-snl+openmp generic-x86", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
FYI This is to fix dryrun error where this particular experiment is generated asking for too many cores for these particular systems.
|
@michaelmckinsey1 do you know why lint is failing? |
|
It says |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## develop #1033 +/- ##
===========================================
+ Coverage 65.31% 65.44% +0.12%
===========================================
Files 44 44
Lines 3241 3241
Branches 256 256
===========================================
+ Hits 2117 2121 +4
+ Misses 1117 1113 -4
Partials 7 7 🚀 New features to boost your workflow:
|
This PR adds the Sparta application (docs: Adding a Benchmark)
repo/sparta-snl/package.pyto specify buildrepo/sparta-snl/application.pyto specify runtime parametersexperiments/sparta-snl/experiment.pyto specify the experiment to run