I noticed, when I build fpc, then I get cpu usage between 70% and 100%. So it appears that both cores of my pentium are used.

But when a package is compiled using fpmake, then it drops to 50%, so apparently only one core is used?

Is that correct? Does the old Makefile driven compilation use both/all cores, while packages done via fpmake, are using one core only?

I noticed, when I build fpc, then I get cpu usage between 70% and 100%. So it appears that both cores of my pentium are used.

No, it can simply mean that the OS moves the process around, which can for example reduce cooling power.

AFAIK the makefile does not use parallel compilation.

Mattias

Yes it does, at least for compiling FPC itself (or rather parts thereof
- but which parts I don't know, probably the FCL).

eg:
make -j 5 all

Then when looking at the CPU monitor from Gnome, I see heavy CPU load on one core only. After some initial compilation on one core, it will switch to using all 4 cores in my system. This shaves about 30 seconds off the time for a clean compile of FPC on my system (if I remember correctly).

I think it was Michael van Canneyt who told me about this switch a while back, and that it indeed works with FPC too. How you apply this to your own work / projects, I have no idea.

Regards,
- Graeme -

Not completely, when you use 'make -Jx' with x>1 then the old makefiles can use multiple threads to actually compile. But it could be that some overhead is done in a second thread?

As fpmake is now used in trunk the biggest disadvantage is that for each unit that is compiled a instance of the compiler is invoked. If you add '-bu' to the fpmake commandline, it outperforms the 'older' makefiles.

There's also a multi-threaded version of fpmake, but that's not comitted yet.

Joost