Bill Allombert on Fri, 6 Dec 2002 23:18:25 +0100


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: CVS branch gmp-kernel-2-2-5


Hello developers,

I have commited my changes to the CVS and merged the change to the 
main branch to the gmp-kernel branch.

The recent real Karatsuba multiplication changes have not been merged, for
several reasons.

I want to review the TODO list:

> --- Only gmp INT are implemented. REAL are standard PARI reals.
> --- As a consequence, INT<-->REAL conversion are slower.
This is hard to do, and will not probably be done until Karim decide
if he want to change the internal real format.

> --- The /*HACK*/ in the code need to be investigated and guard values added,
> else stack overrun will happen.
To do. Surprisingly no such bugs were found so far...

> --- No Montgomery multiplication support. The GMP code for that is private to GMP.
I have a long-term plan to fix this, but don't hold your breath.

> --- Bit manipulation routine (bitand,bittest,etc...) need to be tested.
To do.

> --- diviiexact is just a call to divii.
I have been told that current gmp implementation of diviiexact has some
performance issue with very large operands, so it is not high priority.

> --- Configure --kernel=gmp must look harder to find libgmp.
That seems the most important if we are going to distribute it.

> --- Lots of things need tuning.
To do.

Cheers,
Bill