Re: [Mingw-msys] Slow start of make in MinGW/MSYS?

[Keith M. <kei...@us...>]

On Friday 17 August 2007 19:37, Ulf Nyman wrote:
> I have found that the make (GNU Make version 3.79.1, have try other
> version) is very slow in the start of building a source.

Yes.  Woe32 *is* slow.  Ask Bill to explain why he can't deliver 
performance which comes even close to that of GNU/Linux.

> It can take up to 20 sec to start the buildings!

I suspect that it's actually slow file system performance that's the 
culprit.  As a crude benchmark, several months ago I ran a timed 
sequence of 100 pdfroff cycles, (a shell script running a groff through 
GhostScript pipeline), to format and emit a 12 page PDF document;  
IIRC, this took the following times for the 100 cycles:

  ~1.5 mins, on 650MHz AMD-Duron, with 80GB ATA-66 HD, SUSE-10.0
  ~35 mins, on 1.7GHz Pentium-4, with similar HD, MSYS-Win2K
  ~1 hr, on same Pentium-4 box, but with Cygwin instead of MSYS.

> With thes result I think the problem is start of the make itself in
> MinGW/MSYS.

I think the problem is in the Woe32 file system itself; specifically in 
the time it takes initially, to read the Makefile, or script.

> Have some one else found simiar problems?

Yes, but I don't think there's much we can do about it; we just have to 
live with it.

> Some tip to solve this (spped up the make)

If you want the speed of GNU/Linux, then use GNU/Linux.  If you need to
develop for Woe32, then either live with the performance penalty, or do 
as I do, and use a cross-compiler hosted on GNU/Linux.

Regards,
Keith.