www.digitalmars.com         C & C++   DMDScript  

digitalmars.D - nedmalloc

reply "Craig Black" <cblack ara.com> writes:
I don't have a lot of free time to work on stuff like this, so it took me
forever.  I finally got nedmalloc compiled with the Digital Mars C compiler.
This is a good thing because now I can use nedmalloc with DMD.  For those
who want to know how I did it, here's how.

First of course, you need to download DMC and nedmalloc.

Then you need to download the latest Microsoft Platform SDK.  When you go to
download it there is a lot of stuff you don't need.  If you download all of
it, it will take a while.

Copy the include and lib directory from the SDK to the DMC directory
(Overwrite the old files).

You have to get the coffimplib utility from Digital Mars.  It is available
in the extented utility package.  It costs $15.  Walter will email it to you
when you order it.

Run the coffimplib utility on kernel32.lib and user32.lib.
coffimplib kernel32.lib -f
coffimplib user32.lib -f

If you are wanting to use DMC for anything else, you will most likely have
to convert more lib files, but only the above two are required to compile
nedmalloc.

Compile nedmalloc with DMC warnings off (-w).

Note that I haven't done any testing yet.  I was just happy that I finally
got it compiled.  Perhaps I can find some time for testing the next couple
weeks.

-Craig
Jan 22 2007
parent reply Zz <zz a.zz> writes:
Craig Black wrote:
  > Run the coffimplib utility on kernel32.lib and user32.lib.

 coffimplib kernel32.lib -f
 coffimplib user32.lib -f
 


I got it to work under dmc by adding some missing stuff into the win 
headers and compiled the test program that comes with nedmalloc just to 
confirm that everthing works, there was a big difference.

Look at a previous post about 2 weeks ago on how I did it.

Zz
Jan 23 2007
parent reply "Craig Black" <cblack ara.com> writes:
Thanks.  I missed your post.  That's definitely the easier way to fix it.  I 
wonder if we could get nedmalloc into a standard library somehow?  I think 
the license would allow it.

-Craig

"Zz" <zz a.zz> wrote in message news:ep5r87$1f4c$1 digitaldaemon.com...

 Craig Black wrote:
  > Run the coffimplib utility on kernel32.lib and user32.lib.

 coffimplib kernel32.lib -f
 coffimplib user32.lib -f


 I got it to work under dmc by adding some missing stuff into the win 
 headers and compiled the test program that comes with nedmalloc just to 
 confirm that everthing works, there was a big difference.

 Look at a previous post about 2 weeks ago on how I did it.

 Zz
Jan 24 2007
parent reply janderson <askme me.com> writes:
Craig Black wrote:

 Thanks.  I missed your post.  That's definitely the easier way to fix it.  I 
 wonder if we could get nedmalloc into a standard library somehow?  I think 
 the license would allow it.
 
 -Craig
 


That would be awesome.  Does the GC get any advantage from Ned-malloc?

-Joel
Jan 24 2007
parent "Craig Black" <cblack ara.com> writes:
"janderson" <askme me.com> wrote in message 
news:ep83d9$2h25$2 digitaldaemon.com...

 Craig Black wrote:

 Thanks.  I missed your post.  That's definitely the easier way to fix it. 
 I wonder if we could get nedmalloc into a standard library somehow?  I 
 think the license would allow it.

 -Craig


 That would be awesome.  Does the GC get any advantage from Ned-malloc?

 -Joel


No GC would not be faster. Only manual allocation (new/delete).  However, 
since Walter is working on the GC, I think it will eventually have 
competitive performance.  That being the case, I doubt the GC will ever come 
close to nedmalloc's performance.  Thus, most common things could be done 
with GC, but stuff that is very performance critical could be done with 
nedmalloc.

-Craig
Jan 25 2007