venado system specs

[rfhaque]

Description

This PR adds the system config specs for the LANL Venado system.

1. saxpy errors out since CUDA 12.5 does not support clang 18 as the host compiler. Adding "-allow-unsupported-compiler" does not work (or the flag isn't begin added correctly by spack)
2. caliper fails when apply a patch for the aarch64 architecture in spack

Dependencies: FIXME:Add a list of any dependencies.

Fixes issue(s): FIXME:Add list of relevant issues.

Type of Change

• {x} Adding a system, benchmark, or experiment
• { } Modifying an existing system, benchmark, or experiment
• { } Documentation update
• { } Build/CI update
• { } Benchpark core functionality

Checklist:

If adding/modifying a system:

• { } Create a new directory for the system and a new system.py file
• { } Add a new dry run unit test in .github/workflows
• { } System appears in System Specifications table in docs catalogue section

If adding/modifying a benchpark:

• { } Add a new application.py and (maybe) package.py under a new directory
  for this benchmark
• { } Configure an experiment
• { } Benchmark appears in Benchmarks and Experiments table in docs catalogue
  section

If adding/modifying a experiment:

• { } Extend experiment.py under existing directory for specific benchmark
• { } Define a single node and multi-node experiments

If adding/modifying core functionality:

• { } Update docs
• { } Update .github/workflows and .gitlab/ci unit tests (if needed)

[pearce8]

@rfhaque where is the PR checklist template? Those are helpful for reviewing. In particular, for adding a new system, we also need:
{ } Add a human-readable system description to systems/all_system_definitions
{ } Verify info added in all_system_definitions generates a new system column in Catalogue/System Specifications (https://software.llnl.gov/benchpark/system-list.html)
{ } Add a dry run of an existing benchmark using the new system definition

(@slabasan, if those are not listed in the PR template right now, we probably want to add them)

[pearce8]

@daboehme Can you please see the patch for Caliper on Arm? We are working with a Grace Hopper machine. It would be great if you can fix this in Caliper and put up a PR to benchpark to remove the patch (if this PR is already merged).

[daboehme]

This should be fixed in Caliper v2.12.1, which is the current default in upstream spack

[pearce8]

@daboehme thanks!

@rfhaque are you using the tip of develop of Caliper which should be v2.12.1?

[pearce8]

Also this patch @daboehme

[daboehme]

This one was only required in v2.9.0, which is quite outdated at this point

[pearce8]

@rfhaque can you please remove the patches and use the newest Caliper instead? Is there a reason you are not doing so?

[rfhaque]

@pearce8 This is just a temporary fix, till we pin the latest spack version.

[pearce8]

@rfhaque What is different about this spack package from upstream? We need to work with @dbo to review/upstream it.

[pearce8]

@rfhaque Do they call them "clusters" or "partitions"? Do you also need to set "partition" in the slurm script to correspond to which partition of the machine you are wanting to use?

@alexrlongne might know.

[rfhaque]

@pearce8 While technically these are hardware partitions, venado slurm defines its own "partitions" (standard, debug, gpu etc) corresponding to some set of resources/qos

[pearce8]

@rfhaque what does this extra_batch_opts do?