ClangBSD build failures

Dominic Fandrey kamikaze at
Tue Apr 27 09:16:49 UTC 2010

On 27/04/2010 10:05, Roman Divacky wrote:
> On Tue, Apr 27, 2010 at 08:21:41AM +0200, Dominic Fandrey wrote:

>> An interesting result is that buildkernel with clang takes longer:
>> CC=clang
>> time -l make buildkernel
>>       921.31 real       802.25 user       114.93 sys
>> time -l make buildkernel -j3
>>       645.17 real       838.46 user       143.03 sys
>> CC=cc
>> time -l make buildkernel
>>       877.14 real       757.42 user       115.11 sys
>> time -l make buildkernel -j3
>>       628.32 real       798.03 user       149.52 sys
> fwiw.. these are my times:
> clang:
> 403.342u 42.516s 6:53.30 107.8% 21957+2248k 33+56671io 364pf+0w
> gcc:
> 451.952u 42.860s 7:23.16 111.6% 6564+2012k 78+43200io 3pf+0w
> note that clang build had more page faults thus would be a little faster
> without them

Nice compile times, and thank you for destroying my illusions that
my Core2Duo notebook performs quite decently. :(

The difference is alarmingly huge. I wonder whether the memory disk
actually hurts performance. I will have to test this.

I normally use ccache, and am used to a lot faster buildkernels and
buildworlds, but I turned this off for the performance tests. So
this didn't alarm me until I saw your measurements.


A: Because it fouls the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail? 

More information about the freebsd-current mailing list