Re: [PATCH] sched: fair: don't depend on wake_wide if waker and wakee are already in same LLC

[Mel Gorman]

On Wed, May 26, 2021 at 09:38:19PM +0000, Song Bao Hua (Barry Song) wrote:
> > And no supportive numbers...
> 
> Sorry for the confusion.
> 
> I actually put some supportive numbers at the below thread which
> derived this patch:
> https://lore.kernel.org/lkml/bbc339cef87e4009b6d56ee37e202daf@xxxxxxxxxxxxx/
> 
> when I tried to give Dietmar some pgbench data in that thread,
> I found in kunpeng920, while software ran in one die/numa with
> 24cores sharing LLC, disabling wake_wide() brought the best
> pgbench result.
> 
>                 llc_as_factor          don't_use_wake_wide
> Hmean     1     10869.27 (   0.00%)    10723.08 *  -1.34%*
> Hmean     8     19580.59 (   0.00%)    19469.34 *  -0.57%*
> Hmean     12    29643.56 (   0.00%)    29520.16 *  -0.42%*
> Hmean     24    43194.47 (   0.00%)    43774.78 *   1.34%*
> Hmean     32    40163.23 (   0.00%)    40742.93 *   1.44%*
> Hmean     48    42249.29 (   0.00%)    48329.00 *  14.39%*
> 
> The test was done by https://github.com/gormanm/mmtests
> and
> ./run-mmtests.sh --config ./configs/config-db-pgbench-timed-ro-medium test_tag
> 

Out of curiousity, I briefly tested this on a Zen2 machine which also
has multiple LLCs per node. Only tbench4 was executed and I cancelled
the other tests because of results like this

tbench4
                          5.13.0-rc2             5.13.0-rc2
                             vanilla sched-nowakewidellc-v1r1
Hmean     1        349.34 (   0.00%)      334.18 *  -4.34%*
Hmean     2        668.49 (   0.00%)      659.12 *  -1.40%*
Hmean     4       1307.90 (   0.00%)     1274.35 *  -2.57%*
Hmean     8       2482.08 (   0.00%)     2377.84 *  -4.20%*
Hmean     16      4460.06 (   0.00%)     4656.28 *   4.40%*
Hmean     32      9463.76 (   0.00%)     8909.61 *  -5.86%*
Hmean     64     15865.30 (   0.00%)    19682.77 *  24.06%*
Hmean     128    24350.06 (   0.00%)    21593.20 * -11.32%*
Hmean     256    39593.90 (   0.00%)    31389.33 * -20.72%*
Hmean     512    37851.54 (   0.00%)    30260.23 * -20.06%*

-- 
Mel Gorman
SUSE Labs