These are chat archives for symengine/symengine

6th
Nov 2015
Isuru Fernando
@isuruf
Nov 06 2015 02:12
@bluescarni, is there a function in mp_integer to do 1 << a?
Ralf Stephan
@rwst
Nov 06 2015 07:27
@certik here the revised numbers plus k_monomial:
                                N=100   N=1000      

flint fmpq_poly_mullow()         0.29       91
flint fmpq_poly_mul()            0.27       88
piranha psin*pcos trunc.         0.7       690
piranha psin*pcos no trunc.      1.0       975
piranha-kronecker trunc.         0.64      680
piranha-kronecker no trunc.      1.15      925
pari sin(x)*cos(x)              11        1400
Ralf Stephan
@rwst
Nov 06 2015 07:38
The polynomials are probably already too dense for piranha to show its muscles.
Ralf Stephan
@rwst
Nov 06 2015 08:41
The numbers are very sensitive to the coefficient size, e.g., with completely dense polynomials and if all numerators/denominators fit in a short then flint uses 0.027ms/8.4ms while piranha-short-trunc uses 1.8ms/200ms.
Ralf Stephan
@rwst
Nov 06 2015 08:53
Since with function expansions dense series are usually created I just decided to use flint for my implementation in symengine. Also I don't see the general need for Puiseux series---special cases where these are needed would then have to use piranha.
Ralf Stephan
@rwst
Nov 06 2015 13:38
Of course I totally forgot that this all only is applicable to the univarate case...