Subject: Re: [PATCH v3 0/2] Fixup for discontiguous/sparse numa nodes
From: Satheesh Rajendran <sathnaga@linux.vnet.ibm.com>
To: acme@kernel.org, mingo@kernel.org, linux-kernel@vger.kernel.org,
        linux-perf-users@vger.kernel.org
Cc: srikar@linux.vnet.ibm.com, bala24@linux.vnet.ibm.com
In-Reply-To: <cover.1503310062.git.sathnaga@linux.vnet.ibm.com>
References: <cover.1503310062.git.sathnaga@linux.vnet.ibm.com>
Content-Type: text/plain; charset="UTF-8"
Date: Tue, 19 Sep 2017 15:04:38 +0530
Mime-Version: 1.0
Content-Transfer-Encoding: 8bit
Message-Id: <1505813678.16495.5.camel@linux.vnet.ibm.com>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 3080
Lines: 100

Hi Arnaldo,

Please let me know if any further comments.
Thanks in advance :-)

Regards,
-Satheesh.On Mon, 2017-08-21 at 15:45 +0530, sathnaga@linux.vnet.ibm.com wrote:
> From: Satheesh Rajendran <sathnaga@linux.vnet.ibm.com>
> 
> Certain systems would have sparse/discontinguous
> numa nodes.
> perf bench numa doesnt work well on such nodes.
> 1. It shows wrong values.
> 2. It can hang.
> 3. It can show redundant information for non-existant nodes.
> 
>  #numactl -H
> available: 2 nodes (0,8)
> node 0 cpus: 0 1 2 3 4 5 6 7
> node 0 size: 61352 MB
> node 0 free: 57168 MB
> node 8 cpus: 8 9 10 11 12 13 14 15
> node 8 size: 65416 MB
> node 8 free: 36593 MB
> node distances:
> node   0   8
>   0:  10  40
>   8:  40  10
> 
> Scenario 1:
> 
> Before Fix:
>  # perf bench numa mem --no-data_rand_walk -p 2 -t 20 -G 0 -P 3072 -T
> 0 -l 50 -c -s 1000
> ...
> ...
>  # 40 tasks will execute (on 9 nodes, 16 CPUs): ----> Wrong number of
> nodes
> ...
>  #    2.0%  [0.2
> mins]  1/1   0/0   0/0   0/0   0/0   0/0   0/0   0/0   4/1  [ 4/2 ]
> l:  0-0   (  0) ----> Shows info on non-existant nodes.
> 
> After Fix:
>  # ./perf bench numa mem --no-data_rand_walk -p 2 -t 20 -G 0 -P 3072
> -T 0 -l 50 -c -s 1000
> ...
> ...
>  # 40 tasks will execute (on 2 nodes, 16 CPUs):
> ... 
>  #    2.0%  [0.2 mins]  9/1   0/0  [ 9/1 ] l:  0-0   (  0)
>  #    4.0%  [0.4 mins] 21/2  19/1  [ 2/3 ] l:  0-1   (  1) {1-2}
> 
> Scenario 2:
> 
> Before Fix:
>  # perf bench numa all
>  # Running numa/mem benchmark...
> ....
> ...
>  # Running RAM-bw-remote, "perf bench numa mem -p 1 -t 1 -P 1024 -C 0
> -M 1 -s 20 -zZq --thp  1 --no-data_rand_walk"
> perf: bench/numa.c:306: bind_to_memnode: Assertion `!(ret)' failed.
> ------------> Got hung
> 
> After Fix:
>  # ./perf bench numa all
>  # Running numa/mem benchmark...
> ....
> ...
>  # Running RAM-bw-remote, "perf bench numa mem -p 1 -t 1 -P 1024 -C 0
> -M 1 -s 20 -zZq --thp  1 --no-data_rand_walk"
> 
>  # NOTE: ignoring bind NODEs starting at NODE#1
>  # NOTE: 0 tasks mem-bound, 1 tasks unbound
>          20.017 secs slowest (max) thread-runtime
>          20.000 secs fastest (min) thread-runtime
>          20.006 secs average thread-runtime
>           0.043 % difference between max/avg runtime
>         413.794 GB data processed, per thread
>         413.794 GB data processed, total
>           0.048 nsecs/byte/thread runtime
>          20.672 GB/sec/thread speed
>          20.672 GB/sec total speed
> 
> Changes in v2:
> Fixed review comments for function names and alloc failure handle
> 
> Changes in v3:
> Coding Style fixes.
> 
> 
> Satheesh Rajendran (2):
>   perf/bench/numa: Add functions to detect sparse numa nodes
>   perf/bench/numa: Handle discontiguous/sparse numa nodes
> 
>  tools/perf/bench/numa.c | 61
> +++++++++++++++++++++++++++++++++++++++++++------
>  1 file changed, 54 insertions(+), 7 deletions(-)
>