X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
X-Recipient: djgpp AT delorie DOT com
Message-ID: <47C257CA.1090708@iki.fi>
Date: Mon, 25 Feb 2008 07:53:14 +0200
From: Andris Pavenis <andris DOT pavenis AT iki DOT fi>
User-Agent: Thunderbird 2.0.0.9 (X11/20071115)
MIME-Version: 1.0
To: djgpp AT delorie DOT com
Subject: Re: 64-bit int?  segment-offset pointers? and biosdisk.c
References: <4a-dnX9djZFzTSLanZ2dnUVZ_rmjnZ2d AT comcast DOT com> <200802250229 DOT m1P2TK93027736 AT envy DOT delorie DOT com> <47C247B5 DOT 598127DE AT yahoo DOT com>
In-Reply-To: <47C247B5.598127DE@yahoo.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Reply-To: djgpp AT delorie DOT com

CBFalconer wrote:
> DJ Delorie wrote:
>>> does gcc handle a 64-bit int in 2.03?  if so, what type is it called?
>> "long long"
>>
>>> I was hoping to see if I could rewrite the existing biosdisk library
>>> with the new int13h extensions (as of 1995).
>> Go ahead.  Whether it gets into the djgpp master sources depends, of
>> course, on the quality of your code (readability, maintainability,
>> backwards compatibility, etc).
>>
>> Even so, you can always publish your new code and let people download
>> it as an extension.
> 
> Well, why isn't my nmalloc incorporated then?  I spent some time
> making the additions requested (i.e. manuals, malldbg module, etc.)
> back then.  I know of no faults.  It should just drop in.
> 

Some time ago I did some testing with nmalloc. It helped when trying to build
development versions of gcc-4.3 (i didn't have get it completely built as
native compiler, but with Linux-to-DJGPP cross-compiler things could be better:
I have RPM, but I have not yet tested it). There were however stability
problems with nmalloc. My malloc stress tests (allocating huge amount of small
different size memory blocks and after that freeing them in random order)
crashed randomly when linked with nmalloc (while allocating blocks). There
were no crashes with v2.04 malloc.

As another stress test rebuilt GMP-4.2.1 with GCC-4.2.3, built program for
calculating huge amount of digits of pi. I have not yet tested how it works
when linked with nmalloc, but there were no problems to get 1.5e8 first digits.
It required about 30 minutes and about 1.5Gb of RAM. Should recheck when linked
with nmalloc.

Andris