On Monday, January 13, 2003, at 12:38 PM, John Peacock wrote:
Rather than relying on that assumption, it would be prudent to do some testing first. The current Math::BigInt module is much faster than the one that shipped with 5.6.x, and provides multiple backends for even faster performance. Since you are already loading overload.pm (I presume), the overall module speed will not be significantly different whether you load Math::BigInt or not.I think that this will be really cool. I look forward to your implementation of the TAI engine for the DateTime interface, John.
Alternatively, I am starting to work my way through compiling libtai under Win32 natively (it compiles just fine under Cygwin as well as all *nixes that support unsigned long long). Using TAI64 as the base storage method could be very fast on any platform without having to rely on M::BI for extended precision.
I'm not surprised to find you participating, BTW. What Dave proposes to do here is in line with what you did with version.pm (and good work, that, BTW). ;-)
David
--
David Wheeler AIM: dwTheory
[EMAIL PROTECTED] ICQ: 15726394
http://david.wheeler.net/ Yahoo!: dew7e
Jabber: [EMAIL PROTECTED]
