Re: [PATCH] pooling NUMA scheduler with initial load balancing

From: Erich Focht (efocht@ess.nec.de)
Date: Wed Oct 09 2002 - 11:26:20 EST


On Tuesday 08 October 2002 21:44, Martin J. Bligh wrote:
> > Thanks for the hints, I cleaned up the first patch, too. No
> > CONFIG_NUMA_SCHED any more, switched to MAX_NUMNODES, including
> > asm/numa.h from asm/topology.h, so no need for you to see it.
>
> Cool. Well it compiles now, but:
>
> CPU 3 IS NOW UP!
> Starting migration thread for cpu 3
> Bringing up 4
> CP4 4ivi e e UPr: 0000
> ing
> igrU:i 4
> eaIP: p060:[<c0114231>] Not tainted
> EFLAGS: 00010046
> EIP is at calc_pool_load+0x109/0x120

This is strange. It works for me really reliably. I added a check
for non-online CPUs in calc_pool_load and changed the pool_lock to
be a spinlock. The patches are attached again.

Results so far (all based on 2.5.39 + ia64 + discontig, measured on
16 CPU Azusa), averaged over 4 measurements. The statistics show that
we need more measurements to get something meaningfull.

                   2.5.39 numa_sched+ilb
numa_test 4 4
AverageUserTime 31.18+-0.67 31.25+- 2.51
    ElapsedTime 41.59+-1.95 38.55+- 4.79
  TotalUserTime 124.78+-2.71 125.08+-10.03
                                                                     
numa_test 8 4
AverageUserTime 30.84+-0.51 28.94+- 0.04
    ElapsedTime 45.02+-1.54 30.00+- 0.24
  TotalUserTime 246.82+-4.02 231.66+- 0.34
                                                                     
numa_test 16 4
AverageUserTime 33.61+- 0.72 32.15+- 0.33
    ElapsedTime 47.16+- 0.95 37.45+- 3.20
  TotalUserTime 538.05+-11.58 514.72+- 5.25
                                                                     
numa_test 32 4
AverageUserTime 37.94+- 0.74 33.36+- 0.02
    ElapsedTime 84.83+- 1.28 68.78+- 0.18
  TotalUserTime 1214.37+-23.64 1068.06+- 0.53
                                                                     
numa_test 64 4
AverageUserTime 39.58+- 1.34 33.64+- 0.09
    ElapsedTime 168.04+- 4.77 142.84+- 5.44
  TotalUserTime 2533.82+-85.96 2153.69+- 5.58

Regards,
Erich





-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Tue Oct 15 2002 - 22:00:31 EST