These are chat archives for symengine/symengine

24th
Apr 2016
Francesco Biscani
@bluescarni
Apr 24 2016 14:30
Piranha is probably going to be rather slow for univariate and short polynomials
it has lots of overhead for those specific use cases
Shivam Vats
@shivamvats
Apr 24 2016 17:26
@bluescarni Which algorithms do you use for trigonometric series expansion in Piranha? Our new implementation is around 2x slower than Piranha's. We are wondering if the bottleneck is in our polynomial operations or our expansion algorithms.
Francesco Biscani
@bluescarni
Apr 24 2016 18:24
@shivamvats piranha does not offer that type of capability directly (at least not yet), it just gives you some low-level primitives for working with multivariate polynomials
Francesco Biscani
@bluescarni
Apr 24 2016 18:32
if I recall correctly @isuruf and @rwst were involved in the symengine series expansion work?
Isuru Fernando
@isuruf
Apr 24 2016 18:35
@shivamvats, @irislq, your conclusions in the wiki are wrong. Improving the series expansion code will improve both Piranha implementation and your implementation, although there won't be any change in Flint implementation as it uses Flint operations.
Slowdown is due to polynomial operations
One bottleneck I can see is that, the polynomials are immutable in your implementation whereas Piranha is not
Iris Lui
@irislq
Apr 24 2016 22:31
@isuruf Why would our polynomials being immutable cause a bottleneck?