On Mon, 24 Nov 2003 15:39:35 +0100, "Miroslav Fidler" <cxl@...>
>> On Mon, 24 Nov 2003 13:35:07 +0100, you wrote:
>> >Why is LD/mingw linking static executable three times slower than
>LD/Linux linking the same executable on the same machine ?!
>> The R project (www.r-project.org) found that the default malloc was
>> very slow in the kinds of allocations that R uses, and replaced it
>> with Doug Lea's (available here:
>> Unfortunately, the startup code in MinGW makes an internal assumption
>> that the standard malloc is being used, and triggers an error if it's
>> not. The suggested workaround is to rename the new malloc, and in all
>> of your code define malloc to be the new name.
> Now I am a little but confused. So the problem is that malloc used in LD
>is slow and recompilation of LD using some better memory management would
>lead to much faster LD ? That simply LD spends way too much time in
> So that some #define xmalloc, #define free should help LD ?
I'm not saying that's the problem, but it's a possibility. If you're
concerned about the speed difference, it might be worth a try.