Talk:Adaptive quadrature

Code is wrong? Tolerances need to halved?
Should the tolerance be halved when re-calling integrate?

i.e.   Q = integrate(f,a,m,tau/2) + integrate(f,m,b,tau/2) rather than Q = integrate(f,a,m,tau) + integrate(f,m,b,tau)?

This would make sure all tolerances sum to tau. —Preceding unsigned comment added by 130.88.47.50 (talk) 13:44, 27 October 2009 (UTC)
 * I noticed this too, and came on this talk page to see if anyone had given a reason already why not. I'll add it in. Timeroot (talk) 01:18, 12 October 2010 (UTC)