NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part2/14266

Internet Message Format | 1996-08-05 | 1.6 KB

Path: aadt.sdt.com!usenet
From: Larry Baker <[e-mail]>
Newsgroups: comp.lang.c++,comp.os.ms-windows.programmer.win32
Subject: Re: VC++ 4.0 memory allocation slower than in 2.x!!!
Date: Fri, 29 Mar 1996 16:56:35 -0800
Organization: SABRE Decision Technologies
Message-ID: <[e-mail]>
References: <[e-mail]> <4jgde4$[e-mail]>
NNTP-Posting-Host: parmail.sdt.com
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 2.0 (Win16; I)

Danny Brands wrote:
> Check the the latest issues of Byte. I think, in the march issue
> they have an article on this. They were running preformance analysis
> programs which did a lot of malloc stuff with unreproducable
> results. I appeared that it had to do with the allignment
> of the allocated block in memory if I understood it well.

What was happening was that malloc was returning
byte-aligned allocations.  This is a problem for things like
floats, doubles, ints and longs, as they have a significant
hardware-level performance penalty (on the Pentium, among others)
if they aren't aligned on a 32-bit word boundary.

After they used their own (modified) allocation routines (which
really just encapsulated calls to malloc/free, but gauranteed
word-aligned allocations) everything went back to normal.
I believe they posted their allocation routines to BIX or
elsewhere.

Note that this has nothing to do with the actual performance of
malloc and free themselves; this has to do with the performance of
the routines that use the memory allocated *by* malloc.

Cheers,

Larry Baker
[e-mail]