LFortran now supports all intrinsic functions

If the community is interested I can move the implementations of sum and dot_product in fast_math to stdlib. From the tests in the library, this reduces the choices to two: fastest (chunked), accurate (chunked kahan). Even the kahan one can be faster than the intrinsic (according to the test I’ve done with gfortran 13; with intel is slower)

2 Likes