Michael Vincent van Rantwijk wrote:

> I mean what is usually the bottleneck for compilers?  HD, RAM, OS?

It's a combination of factors.

The build system itself involves launching subprocesses, checking 
datestamps and such. This is faster on Linux than it is on Windows, 
although MSYS is slightly better than Cygwin in this respect. (Interix 
is faster than MSYS, but unsupported.) In particular you'll see more of 
a relative benefit of a fast HD on Windows.

Compilation and linking is generally a mixture of CPU and HD. This is 
demonstrated by the fact that a parallel make is slightly faster than a 
standard make as while one file is being read from the disk the 
processor can usefully continue to compile a separate file.

However linking debug is the killer, particularly gklayout (non-static) 
or libxul (xulrunner static) or the final executable (static). For 
gklayout you must have 512MB of RAM - it can (eventually) be done in 
384MB but it wears your disk out (literally - I found out the hard way.) 
For the others, or if you want to use your computer while you're 
linking, you probably need twice as much.
_______________________________________________
Project_owners mailing list
[email protected]
http://mozdev.org/mailman/listinfo/project_owners

Reply via email to