Re: Scalable memory manager for Delphi.

From: Nelson Esmai (esmai_at_esmai.com)
Date: 02/16/04 

Date: Mon, 16 Feb 2004 15:43:47 -0600

The current manager is labeled "beta" mainly because we're waiting to see IN
WHAT WAY it can be sold. It will be competitively priced with other
managers on the market and MIGHT be free for non-commercial use.
Performance should be equal or better than any other manager on the market.
I can do some comparisons to the MultiMM unit on optimalcode.com (now
expired from what I understand).

We found performance to be generally equal to MultiMM yet offers several
advantages.
1. MultiMM is mostly stable, but not entirely. Our main client relied on it
for high-stress web services until this new manager replaced it. MultiMM
wrongly assumes that the thread that allocated a block of memory will always
be the thread to free it and therefore may place incorrect locks and cause
memory corruption. The client survived only by having two copies of the app
running simultaneously and comparing results.
2. Earlier versions of MultiMM could sometimes leak large chunks of memory
when threads were created and destroyed. The memory would be unrecoverable.
3. MultiMM cannot free memory. It allocates memory to peak. BigBrain frees
memory and has roughly equivalent performance.
4. Statistics. BigBrain can give you statistics useful for detecting memory
leaks, about memory overhead, allocated, free, and wasted space... for every
thread.

Our benchmarks are still a bit preliminary but included sample apps can let
you roll your own benchmark and do your own evaluation. Single thread
performance showed 86% improvement on a dual PIII and around 400%
improvement on a Dual Athlon MP. The standard memory manager scalability
goes down with added threads, while BigBrain's scalability goes up or stays
linear. (Stats so far are on www.digitaltundra.com). We've done extensive
Xeon testing with it, however, the latest demo app hasn't been run on Xeons
for a fair comparison.

"Kevin Powick" <nospam@nomail.com> wrote in message
news:MPG.1a9af88f7cc1981c989725@newsgroups.borland.com...
> In article <40311e98$1@newsgroups.borland.com>, esmai@esmai.com says...
>
> > After many year of pain and suffering. A memory manager that is both
fast,
> > scalable, and reliable: http://www.digitaltundra.com
>
> What about pricing and comparisons to other mem managers that are
> already out there?
>
> Is this only beneficial to multi-threaded apps, or will any/all apps
> benefit?
>
> --
> Kevin Powick

Relevant Pages

  • Re: WebBrowser control - High memory use
    ... I believe i was the one saying "my customers are happy now" as this was were ... Now minimize the app, wait until the app memory drops, now maximize the ... Task manager is entirely appropriate for checking *real* ...
    (microsoft.public.dotnet.languages.vb)
  • Re: WebBrowser control - High memory use
    ... Task manager is entirely appropriate for checking *real* ... is that it doesn't show much memory is shared). ... corrected the app will crash. ...
    (microsoft.public.dotnet.languages.vb)
  • Re: WebBrowser control - High memory use
    ... Task manager is entirely appropriate for checking *real* ... is that it doesn't show much memory is shared). ... corrected the app will crash. ...
    (microsoft.public.dotnet.languages.vb)
  • Re: WebBrowser control - High memory use
    ... Now minimize the app, wait until the app memory drops, now maximize the ... Task manager is entirely appropriate for checking *real* ...
    (microsoft.public.dotnet.languages.vb)
  • Re: WebBrowser control - High memory use
    ... Now minimize the app, wait until the app memory drops, now maximize the app and see the memory usage ... Task manager is entirely appropriate for checking *real* ...
    (microsoft.public.dotnet.languages.vb)
Loading