Re: [Libunwind-devel] unw_backtrace(): Performance slowdown for ARM?

[Luke Diamand]

On 16 January 2018 at 18:19, Dave Watson <address@hidden> wrote:
> On 01/16/18 09:34 AM, Luke Diamand wrote:
>> On 16 January 2018 at 03:33, Luke Diamand <address@hidden> wrote:
>> > Hi!
>> >
>> > I could be wrong about this, but I'm finding a roughly 10x slowdown on
>> > ARM between 1.2.1 and current HEAD (e1ca8748).
>> >
>> > I have a small test program here:
>> >
>> > https://gist.github.com/luked99/7c67d94da5c33cac5a1b41a500013355
>
> I think there is a similar perf test program under tests/, 'make perf'
>

I tried Lperf-trace. On 1.2.1 I get this output from my RPI2:

$ time Lperf-trace
unw_getcontext : cold avg=  219.345 nsec, warm avg=  119.209 nsec
unw_init_local : cold avg= 5741.119 nsec, warm avg= 5378.723 nsec
no cache        : unw_step : 1st=32499.151 min=13745.056 avg=14395.800 nsec
global cache    : unw_step : 1st=13904.751 min=13754.053 avg=14391.024 nsec
per-thread cache: unw_step : 1st=13972.228 min=13754.053 avg=14396.393 nsec
52.73user 0.48system 0:53.22elapsed 99%CPU (0avgtext+0avgdata 68056maxresident)k
0inputs+0outputs (0major+16761minor)pagefaults 0swaps


With 1.3 (e1ca8748) it produced the first two lines:

unw_getcontext : cold avg=  238.419 nsec, warm avg=  119.209 nsec
unw_init_local : cold avg= 5288.124 nsec, warm avg= 4880.428 nsec
default         : unw_step : 1st=248018.301 min=185678.590 avg=191420.175 nsec

but then is either very slow, or is stuck. I killed it after about 5 minutes.