Re: gcc compile-time performance

[Richard Earnshaw <rearnsha at arm dot com>]

> > <<Yes, that is alternative.  Will it be considerably faster than the current
> > FP emulation library doing the same?
> > >>
> > 
> > Oh yes, much faster, by a significant factor, which is all you need in this
> > case.
> Hmm, I still guess there are easier ways to cut of 1-1.5% (even with host FP
> the function was taking about 1% of compilation time due to CFG traversal and I
> doubt I can get faster.  However I can try to do that after finals if this is
> considered usefull.
> 
> (BTW I am still having problems to see it in the -Q dumps.  Curisously enoght
> if I enable -fprofile-arcs, I get the branch probabilities printed, but not
> without it).
> On what machine do you do the measurements?
> 
> Honza

In this case a 6 cpu P3 box.

Note, I see very bizarre numbers from the -Q dumps also -- I don't think 
the timers are returning small enough increments as the machines get 
faster -- note that often we are getting zero complete clock ticks within 
a single optimization.



cat /proc/cpuinfo 
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 10
model name      : Pentium III (Cascades)
stepping        : 1
cpu MHz         : 700.061
cache size      : 2048 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca 
cmov pat pse36 mmx fxsr sse
bogomips        : 1395.91

processor       : 1