[Tests] Fix precision constraints for FP math operations
Opened this issue · 0 comments
edanor commented
Originally reported by: edanor (Bitbucket: edanor, GitHub: edanor)
Some floating point operations rely on approximation. The problem is to test such functions against standard version with automatically generated data inputs.
What should be done is:
- design a code pattern for calculating 'ULP distance' from the reference value,
- create a test fail/pass threshold value expressed in this ULP distance,
- report the ulp precision (for selected tests only) together with test result.