From: Tanaka Akira <akr@...>
Date: 2013-08-17T09:41:05+09:00
Subject: [ruby-core:56671] Re: [ruby-trunk - Feature #8796][Open] Use GMP to accelerate Bignum operations

2013/8/17 Eric Wong <normalperson@yhbt.net>:

> Is there more performance improvement without the conversion?
>
> How about push the conversion cost to legacy C API users to make
> Bignum faster for pure-Ruby use in a future patch?

It is same as ko1's idea.
I don't against it.
Feel free to implement and propose it.

However it has several difficulties.

* It is a big task.
  It need to implement all methods, not just slow methods.

* ABI incompatibility.
  ko1 tackles this in Feature #6083.

* LGPL
  It is no problem for me but I guess some people don't accept it.
  So we need to maintain non-GMP implementation anyway.
  Maintaining two implementations is troublesome.

* We cannot access internal of mpz_t.
  We may be limited to add new feature with optimal performance.
  (mpz_getlimbn and mpz_size may be enough?)

* It cannot embed small bignums.
  So it needs more memory allocation.
  (mpz_array_init may solve this problem?)
-- 
Tanaka Akira