Bug: Cholesky test on HIP sometimes failing due to pathological random matrices that create small roundoff errors

[shakedregev]

Description how to reproduce the bug

ctest -j

Check Testing/Temporary/LastTest.log in your build directory.

Running tests on HIP device:
--- ^[[1;32mPASS^[[0m: Test minimalCorrectness
--- ^[[1;32mPASS^[[0m: Test randomized n = 3
--- ^[[1;32mPASS^[[0m: Test randomized n = 10
--- ^[[1;32mPASS^[[0m: Test randomized n = 100
--- ^[[1;32mPASS^[[0m: Test randomized n = 1000
Test failed at index 0: expected 0.211033326206, got 0.211033326170
, difference 0.000000000036
Test failed at index 1: expected 0.287576402206, got 0.287576402216
, difference 0.000000000010              
Test failed at index 2: expected 0.382495751783, got 0.382495751786
, difference 0.000000000004

The code uses an old C style random number generator which is prone to errors and does not fix the seed, making it irreproducible.

Re::Solve version

develop

System and environment details

HIP, Frontier

Additional information

No response