Compiled
- To: mathgroup at smc.vnet.net
- Subject: [mg22168] Compiled
- From: bernd at bio.vu.nl (Bernd Brandt)
- Date: Thu, 17 Feb 2000 01:23:57 -0500 (EST)
- Sender: owner-wri-mathgroup at wolfram.com
Hi,
Does someone know when to really use Compiled -> False in Numerical Functions?
When using Numerical functions such as NIntegrate, FindMinimum and
NDSolve, with WorkingPrecision higher than MachinePrecision, I thought
it was needed to use Compiled->False.
However, in 3.9.3 Numerical Integration of the Math book:
In[10]:= NIntegrate[1/x, {x, 1, 2}, WorkingPrecision->40]
Here the Compiled -> False is omitted.
In 3.9.7:
NDSolve[{y'[x] == I/4 y[x], y[0] == 1}, y, {x, 1},
AccuracyGoal -> 20, PrecisionGoal -> 20,
WorkingPrecision -> 25]
Here the Compiled -> False is also omitted.
The help on Compiled states:
* You should set Compiled -> False if you need to use higher precision numbers.
The help on WorkingPrecision states:
NIntegrate[1/x, {x, 1, 2}, WorkingPrecision -> 40, Compiled -> False]
This seems to be in contradiction with the above examples.
Regards,
Bernd