Add test for number of threads of OpenBLAS

[filikat]

Description

• Add openblas_get_num_threads, to query the number of threads used by OpenBLAS.
• After setting the number of threads for OpenBLAS to 1, check that indeed OpenBLAS is using 1 thread. This may catch errors in how the flag HIPO_USES_OPENBLAS is passed to HiPO.
• If BLAS_LIBRARIES is used and the provided library contains the string openblas, set the BLAS metadata so that HiPO knows that OpenBLAS is being used.

Checklist

• I have read the contributing guidelines
• This PR targets the latest branch
• Tests are passing
• Documentation was updated where relevant
• This PR is not primarily AI-generated (per the AI contributions policy in CONTRIBUTING.md)

[codecov]

Codecov Report

❌ Patch coverage is 0% with 20 lines in your changes missing coverage. Please review.
✅ Project coverage is 73.56%. Comparing base (c9e05bc) to head (e44f9fe).
⚠️ Report is 69 commits behind head on latest.

Files with missing lines	Patch %	Lines
highs/ipm/IpxWrapper.cpp	0.00%	14 Missing ⚠️
highs/ipm/hipo/ipm/FactorHighsSolver.cpp	0.00%	4 Missing ⚠️
extern/HighsExtrasExternalDeps.h	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           latest    #3293      +/-   ##
==========================================
+ Coverage   73.31%   73.56%   +0.24%     
==========================================
  Files         446      446              
  Lines      108397   109543    +1146     
  Branches    17360    17547     +187     
==========================================
+ Hits        79476    80588    +1112     
- Misses      28645    28678      +33     
- Partials      276      277       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mathgeekcoder]

Quick question: are you wanting to use openblas_get_num_threads for anything else?

An alternate implementation would be to modify the set_num_threads to also return the number of threads:

int openblas_set_num_threads(int num_threads) {
#if defined(HIPO_USES_OPENBLAS)
  openblas_set_num_threads(num_threads);
  return openblas_get_num_threads();
#else
  return num_threads;
#endif
}

This would make most of the other changes unnecessary.

[filikat]

openblas_get_num_threads is not used anywhere else at the moment, so I can remove it and follow your suggestion. I would still leave the check of whether the BLAS provider and HIPO_USES_OPENBLAS agree. If OpenBLAS is used, but HIPO_USES_OPENBLAS is not set, the performance deteriorates considerably because the number of threads is not set to 1. These changes are meant to guarantee that it doesn't happen.

[mathgeekcoder]

Looking better! I've added a couple more optional suggestions

[mathgeekcoder]

Perhaps we can include something like this in stringutil.h?

#if defined(_WIN32) || defined(_WIN64)
#include <string.h>  // _stricmp
#else
#include <strings.h>  // strcasecmp
#endif

inline int stringcasecmp(const char* lhs, const char* rhs) {
#if defined(_WIN32) || defined(_WIN64)
  return _stricmp(lhs, rhs);
#else
  return strcasecmp(lhs, rhs);
#endif
}

Then this becomes:

static bool usingAppleBlas() {
   return stringcasecmp(HighsExtras::blas::getInfo()->provider, "apple") == 0;
}

It should be faster and cleaner than creating a string, converting to lower and comparing.

[mathgeekcoder]

Alternatively, we could modify HighsExtrasFeatureInfo:

bool HighsExtrasFeatureInfo::isProvider(const char* check) {
   return stringcasecmp(provider, check) == 0;
}

such that:

HighsExtras::blas::getInfo()->isProvider("apple");

[filikat]

Yes, this looks neat, I will make the changes

[filikat]

We need to check for substrings, not exact match though (so that "Apple Accelerate" matches "apple"). I replaced strcasecmp with strcasestr. Unfortunately there seems to be no equivalent on Windows. One possible solution is to use _strnicmp iteratively. I am not sure of whether this is fast, but it shouldn't be a big deal. The providers are short and need to be checked only a couple of times.

[filikat]

I'm having some issues to include stringutil.h in HighsExtrasApiBinding.h in a way that works everywhere, in order to implement isProvider. I will remove isProvider for now.

[mathgeekcoder]

As above

static bool usingOpenBLAS() {
   return stringcasecmp(HighsExtras::blas::getInfo()->provider, "openblas") == 0;
}

[mathgeekcoder]

Sorry, can you explain why this is needed?

[filikat]

If I use OpenBLAS by passing libopenblas.so to BLAS_LIBRARIES, CMake doesn't know which vendor is being used and so it doesn't set the BLAS metadata (i.e., the provider is unknown, etc). Crucially, this means that HIPO_USES_OPENBLAS is not set, and OpenBLAS runs multi-threaded, with a substantial reduction in performance.
Maybe there are better ways of achieving this.