Re: Ada implementation of Mersenne Twister (MT19937)

[Adrian Hoe <byhoe@greenlime.com>]

Adrian Hoe wrote:
> 
> Hello,
> 
> I could not find an Ada implementation of MT19937 pseudo random number
> generator by searching CLA. So, I decided to port Makoto Matsumoto and
> Takuji Nishimura's work to Ada.
> 
> Thought someone might be interested to use MT19937. You can download
> the source from my web site at http://greenlime.com/users/adrian.hoe


Something interesting I found in Ada's while-loop and for-loop. I don't
know if this have been discussed in CLA because I could not find
anything related by searching with Google.

My implementation of test3.adb uses while..loop (you can download
adamt19937 source from my home page) and Christoph modified the while
loop to a for..loop.

The time recorded on my Linux machine using my implementation
(while..loop) is 57seconds. But using Christoph for..loop, I recorded
107seconds on my same Linux machine (Christoph recorded 116seconds on
WIndows98).

It seems like for..loop is slower than while..loop. I did not realize
the difference between them until I tested Christoph's test3. I thought
it was WIndows98 problem.

Any discussion?
-- 
                              -- Adrian Hoe
                              -- http://greenlime.com/users/adrian.hoe