Message-ID: <22a2dafb-de05-199b-54ed-0c3b24349826@linux.intel.com>
Date:   Fri, 16 Jun 2017 17:22:29 +0300
From:   Alexey Budankov <alexey.budankov@...ux.intel.com>
To:     Mark Rutland <mark.rutland@....com>
Cc:     Peter Zijlstra <peterz@...radead.org>,
        Ingo Molnar <mingo@...hat.com>,
        Arnaldo Carvalho de Melo <acme@...nel.org>,
        Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
        Andi Kleen <ak@...ux.intel.com>,
        Kan Liang <kan.liang@...el.com>,
        Dmitri Prokhorov <Dmitry.Prohorov@...el.com>,
        Valery Cherepennikov <valery.cherepennikov@...el.com>,
        David Carrillo-Cisneros <davidcc@...gle.com>,
        Stephane Eranian <eranian@...gle.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 1/n] perf/core: addressing 4x slowdown during
 per-process profiling of STREAM benchmark on Intel Xeon Phi

On 16.06.2017 17:08, Alexey Budankov wrote:
> On 16.06.2017 12:09, Mark Rutland wrote:
>> On Fri, Jun 16, 2017 at 01:10:10AM +0300, Alexey Budankov wrote:
>>> On 15.06.2017 22:56, Mark Rutland wrote:
>>>> On Thu, Jun 15, 2017 at 08:41:42PM +0300, Alexey Budankov wrote:
>>>>> This series of patches continues v2 and addresses captured comments.
>>
>>>>> Specifically this patch replaces pinned_groups and flexible_groups
>>>>> lists of perf_event_context by red-black cpu indexed trees avoiding
>>>>> data structures duplication and introducing possibility to iterate
>>>>> event groups for a specific CPU only.
>>>>
>>>> If you use --per-thread, I take it the overhead is significantly
>>>> lowered?
>>>
>>> Please ask more.
>>
>> IIUC, you're seeing the slowdown when using perf record, correct?
> 
> Correct. Specifically in per-process mode - without -a option.
> 
>>
>> There's a --per-thread option to ask perf record to not duplicate the
>> event per-cpu.
>>
>> If you use that, what amount of slowdown do you see?

After applying all three patches:

- system-wide collection:

[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 303.795 MB perf.data (~13272985 samples) ]
2162.08user 176.24system 0:12.97elapsed 18021%CPU (0avgtext+0avgdata 
1187208maxresident)k
0inputs+622624outputs (0major+1360285minor)pagefaults 0swaps

- per-process collection:

[ perf record: Woken up 5 times to write data ]
[ perf record: Captured and wrote 1.079 MB perf.data (~47134 samples) ]
2102.39user 153.88system 0:12.78elapsed 17645%CPU (0avgtext+0avgdata 
1187156maxresident)k
0inputs+2272outputs (0major+1181660minor)pagefaults 0swaps

Elapsed times look similar. Data file sizes differ significantly.

Test script:

#!/bin/bash

echo 0 > /proc/sys/kernel/watchdog
echo 0 > /proc/sys/kernel/perf_event_paranoid
/usr/bin/time 
"/root/abudanko/vtune_amplifier_2018_zip/bin64/amplxe-perf" record 
--per-thread [-a] -N -B -T -R -d -e 
cpu/period=0x155cc0,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x3c,in_tx=0x0,ldlat=0x0,umask=0x0,in_tx_cp=0x0,offcore_rsp=0x0/Duk,cpu/period=0x155cc0,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x0,in_tx=0x0,ldlat=0x0,umask=0x3,in_tx_cp=0x0,offcore_rsp=0x0/Duk,cpu/period=0x155cc0,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xc0,in_tx=0x0,ldlat=0x0,umask=0x0,in_tx_cp=0x0,offcore_rsp=0x0/Duk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x3,in_tx=0x0,ldlat=0x0,umask=0x8,in_tx_cp=0x0,offcore_rsp=0x0/ukpp,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x3,in_tx=0x0,ldlat=0x0,umask=0x1,in_tx_cp=0x0,offcore_rsp=0x0/ukpp,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x4,in_tx=0x0,ldlat=0x0,umask=0x2,in_tx_cp=0x0,offcore_rsp=0x0/ukpp,cpu/period=0x186a7,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x4,in_tx=0x0,ldlat=0x0,umask=0x4,in_tx_cp=0x0,offcore_rsp=0x0/ukpp,cpu/period=0x1e8483,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x3c,in_tx=0x0,ldlat=0x0,umask=0x0,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x1e8483,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xc2,in_tx=0x0,ldlat=0x0,umask=0x10,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xca,in_tx=0x0,ldlat=0x0,umask=0x4,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xca,in_tx=0x0,ldlat=0x0,umask=0x90,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x1e8483,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xc2,in_tx=0x0,ldlat=0x0,umask=0x1,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xc3,in_tx=0x0,ldlat=0x0,umask=0x4,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x4,in_tx=0x0,ldlat=0x0,umask=0x20,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x5,in_tx=0x0,ldlat=0x0,umask=0x3,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x1e8483,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xcd,in_tx=0x0,ldlat=0x0,umask=0x1,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x3,in_tx=0x0,ldlat=0x0,umask=0x4,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x86,in_tx=0x0,ldlat=0x0,umask=0x4,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x4,in_tx=0x0,ldlat=0x0,umask=0x10,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x4,in_tx=0x0,ldlat=0x0,umask=0x40,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x4,in_tx=0x0,ldlat=0x0,umask=0x80,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xc2,in_tx=0x0,ldlat=0x0,umask=0x40,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xc2,in_tx=0x0,ldlat=0x0,umask=0x20,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x5,in_tx=0x0,ldlat=0x0,umask=0x2,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xe6,in_tx=0x0,ldlat=0x0,umask=0x1,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xe7,in_tx=0x0,ldlat=0x0,umask=0x1,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xc3,in_tx=0x0,ldlat=0x0,umask=0x1,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0xc3,in_tx=0x0,ldlat=0x0,umask=0x2,in_tx_cp=0x0,offcore_rsp=0x0/uk,cpu/period=0x30d43,pc=0x0,any=0x0,inv=0x0,edge=0x0,cmask=0x0,event=0x4,in_tx=0x0,ldlat=0x0,umask=0x1,in_tx_cp=0x0,offcore_rsp=0x0/uk 
-- ./stream

>>
>> It might be preferable to not open task-bound per-cpu events on systems
>> with large cpu counts, and it would be good to know what the trade-off
>> looks like for this case.
>>
>>>>> +static void
>>>>> +perf_cpu_tree_insert(struct rb_root *tree, struct perf_event *event)
>>>>> +{
>>>>> +    struct rb_node **node;
>>>>> +    struct rb_node *parent;
>>>>> +
>>>>> +    WARN_ON_ONCE(!tree || !event);
>>>>> +
>>>>> +    node = &tree->rb_node;
>>>>> +    parent = *node;
>>>>
>>>> The first iteration of the loop handles this, so it can go.
>>>
>>> If tree is empty parent will be uninitialized what is harmful.
>>
>> Sorry; my bad.
>>
>> Thanks,
>> Mark.
>>
> 
> 

Powered by blists - more mailing lists