We're seeing somewhat higher number of OpenBLAS test suite failures on ARM neoverse_v1 in:
For now, we've increased the number of accepted failures, but we'll need to investigate this issue in more detail. Are these test just out of numerical bounds? Or way off?
See also:
It did not seem to be a problem however in:
which build OpenBLAS 0.3.15 on ARM neoverse_v1 (see here)
We're seeing somewhat higher number of OpenBLAS test suite failures on ARM neoverse_v1 in:
For now, we've increased the number of accepted failures, but we'll need to investigate this issue in more detail. Are these test just out of numerical bounds? Or way off?
See also:
It did not seem to be a problem however in:
which build OpenBLAS 0.3.15 on ARM neoverse_v1 (see here)