Re: [boost] [Review Request] Multiprecision Arithmetic Library

Subject: Re: [boost] [Review Request] Multiprecision Arithmetic Library
From: Christopher Kormanyos (e_float_at_[hidden])
Date: 2012-04-07 19:51:13

<snip>


> Ah, good point, oversight on my part.  I've just committed an update
> that uses your 1100 digit values to initialize when possible.
> John. 

Hi John,

This caused a weird repercussion. When we define constants with
hundreds or thousands of digits of precision in Boost.Multiprecision,
this is in conflict with boost::math::constants, which has only
tens or maybe a hundred digits.

I instantiated an mp_number<cpp_dec_float> from the value of
boost::math::constants::e() today and got only maybe 100 decimal
digits of precision.

This is a very subtle, hard to catch detail. It's easy to see in the source,
but confusing when interacting with Boost.Multiprecision and Boost.Math.
I don't know where to patch this one. Unfortunately, we might need to
add more digits to Boost.Math. This seems to be string-based and can
potentially be synchronized with our new 1100 digit strings.  Then we might
want to consider a common string pool for Boost.Math and Boost.Multiprecision
so as to avoid multiple upkeep of strings.

Eventually, I will *hammer* Boost.Math with 1,000 digits. We will
find issues.

Best regards, Chris.