On Wed, May 29, 2013 at 04:22:03PM +0200, Paolo Bonzini wrote: > Il 29/05/2013 10:09, Holger Hans Peter Freyther ha scritto: > >> > > >> > The testcase is simply "'foo' asFile readStream contents". > > Yes, I did that but compared to 3.2.5/master vs. patch I don't > > see much change in regard to the amount of fcntl. Which means > > there was something else (which I forgot now) that we considered > > inefficient. > > It's the amount of lseek and read, not fcntl.
I remember that we saw excess.. and I didn't see a difference as I was trying to read from /proc/cpuinfo and there is no difference. Reading from a real file gives me a difference. read(4, "!\n! OsmoBSC (0.9.14.182-edce) co"..., 1024) = 1024 read(4, "ty 30000\n location_area_code 82"..., 1024) = 1024 read(4, " phys_chan_config TCH/F\n hop"..., 1024) = 940 vs. read(3, "!\n! OsmoBSC (0.9.14.182-edce) co"..., 2988) = 2988 _______________________________________________ help-smalltalk mailing list [email protected] https://lists.gnu.org/mailman/listinfo/help-smalltalk
