On Wednesday, July 07, 2010 04:56:54 pm Brice Goglin wrote: > Le 07/07/2010 13:48, Jirka Hladky a écrit : > > I know that such mapping is artificial. However, it enables me to see > > that hyper threading is enabled. I use it in my reports only. > > Just compare the number of cores and the number of PUs. It should be > enough to know if each core has multiple threads. On some architectures > (power7), this may require recent kernels before everything shows up > properly. But on x86 and ia64, I think things should be fine. Yes, I'm already doing this:-)
> > > I wonder if some similar concept exists in hwloc. To be honest, I don't > > know what the future of "thread id" in /proc/cpuinfo file is. It seems > > to be only in /proc/cpuinfo for ia64. > > We currently have no room to store this thread id since it would likely > to go in the physical index of PU objects but we already store the > logical processor index there. But we could add some PU-specific > attributes to store it if somebody really needs this thread id. > > /proc/cpuinfo is highly non portable anyway, so it's not clear we really > want to look closely at what it contains. And parsing /sys is much > easier than /proc/cpuinfo. Do you see this thread_id in > /sys/devices/system/cpu/cpu%d/topology/ ? I agree that /proc/cpuinfo is non portable. That's why we are moving to hwloc:-) I have checked /sys/devices/system/cpu/cpu%d/topology/ and there is no thread_id. So I my conclusion is that that it does not make sense to look into it further. Thanks! Jirka
