framework: For cuda native events - extend the string length in PAPI_event_code_to_name to be PAPI_2MAX_STR_LEN

[Treece-Burgess]

Pull Request Description

The Perfworks Metrics API can have metrics that are greater than PAPI_MAX_STR_LEN. For example:

cuda:::TPC.TriageCompute.tpc__l1tex_sram_bytes_mem_untagged_data_shared_allocated_compute_realtime.peak_sustained_active.per_second - 131 characters
cuda:::TPC.TriageCompute.tpc__l1tex_sram_bytes_mem_untagged_data_shared_allocated_compute_realtime.peak_sustained_elapsed.per_second - 132 characters

This leads to PAPI_event_code_to_name returning PAPI_ENOMEM. This behavior can be seen on Hopper1 at Oregon (1 * GH200) with the test test_multipass_event_fail.cu.

This PR resolves this behavior by using PAPI_2MAX_STR_LEN if the eventcode provided by the user is found to be associated with the cuda component.

Testing

Testing was done on Hopper1 at Oregon (CPU: ARM Neoverse-V2 , GPU: 1 * GH200) with Cuda Toolkit 12.8.1.

• PAPI build: ✅
• PAPI utilities*: ✅
• net component tests: ✅
• coretemp component tests: ✅
• cuda component tests: ✅

* - papi_command_line, papi_component_avail, papi_native_avail

I have also verified that the ctests and ftests that call PAPI_event_code_to_name do not use cuda component native events, rather they use the perf_eventcomponent native events.

Note: This is a temporary fix for this issue and a longer term fix will be applied at a later date.

Author Checklist

• Description
  Why this PR exists. Reference all relevant information, including background, issues, test failures, etc
• Commits
  Commits are self contained and only do one thing
  Commits have a header of the form: module: short description
  Commits have a body (whenever relevant) containing a detailed description of the addressed problem and its solution
• Tests
  The PR needs to pass all the tests