Re: Quick integral.

*To*: mathgroup at smc.vnet.net*Subject*: [mg73441] Re: [mg73393] Quick integral.*From*: jmt <jmt at dxdydz.net>*Date*: Fri, 16 Feb 2007 01:06:57 -0500 (EST)*References*: <200702151001.FAA05398@smc.vnet.net>

On Thursday 15 February 2007 11:01, Jouvenot, Fabrice wrote: > The purpose of our work is to have something really quick. > We made some simplifications, approximations, etc. > > We have now difficulties to accelerate it. > We think we can run it faster if we ask Mathematica to calculate our > integrals with only 2 significant numbers. But we tuned different > parameters of Intergral to have a rude approximation of the integral and > so have the calculus quicker than before. > But (we are beginers with mathematica (but not in programing)) it do not > seems to be less precise or faster. > As far as I understand, you can extend the precision of your computation, but can hardly restraint it, because a restrained precision will finally always use the default plateform floating point facilities (see $MachinePrecision), these being faster than any other implementation you can make. May I suggest you change the strategy for computing your numerical integral ? -- -------------------------------------- Informatique scientifique et technique http://www.dxdydz.net Jean-Marie Thomas mailto:jmt at dxdydz.net +33 (0)3 88 32 93 64

**References**:**Quick integral.***From:*"Jouvenot, Fabrice" <F.Jouvenot@liverpool.ac.uk>