On 2012-08-22 at 05:04:59 [+0200], Joseph Prostko <joe.prostko@xxxxxxxxx> wrote: > First off, sorry for the delay in getting back to this thread, as I > see nobody has commented since I was last checking on it. I was in a > bad car wreck last Monday, and am finally getting back into the swing > of things. I hope everything's OK (besides the car, obviously). > On Tue, Aug 14, 2012 at 5:45 AM, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote: > > >> On 14/08/2012 07:26, Fredrik Holmqvist wrote: > >> > 2012/8/13 Joseph Prostko <joe.prostko@xxxxxxxxx>: > >> >> Yeah, I think for consistency, either we should use -O in both places, > >> >> or else -O3 -fno-strict-aliasing. [...] > >> > IIRC correctly I switched that from -O to -O3 long ago, in the hope > >> > that it would speed up build times. > > > If you build a complete image, a few seconds don't make much of a > > difference. Most of the time, however, developers only build one or two > > components repeatedly, so that the jam warm-up time becomes relevant. So > > some performance measuring would certainly be appreciated. > > Would just using something like the "time" command under identical > circumstances be what you had in mind, or did you have some more > sophisticated benchmarking metric in mind? "time" is fine. Running it more than once to average things would be a good idea, though. CU, Ingo