[erlang-questions] Arch Linux patches?

Antonio SJ Musumeci <>
Wed Nov 13 19:11:24 CET 2013


Any developments? Curious to find the reason for this.

On Fri, Nov 1, 2013 at 7:37 AM, Aaron France <> wrote:
> Are they both using the exact same commit?
>
>
> On 01/11/13 11:18, Loïc Hoguin wrote:
>>
>> I'm sure there's a reasonable explanation.
>>
>> Trying with Cowboy's hello_world package, the dumbest benchmark you can
>> do. Using siege for this, using the default conf file except verbose set to
>> false.
>>
>> To compile the example:
>>
>> % make
>> % ./_rel/bin/hello_world_example console
>>
>> You will have to recompile it when switching Erlang installs of course, as
>> it's a release and the VM files are copied.
>>
>> Pacman install:
>>
>> % erl
>> Erlang R16B02 (erts-5.10.3) [source] [64-bit] [smp:4:4] [async-threads:10]
>> [hipe] [kernel-poll:false]
>> ...
>> % siege -b -c 100 -t 5s http://127.0.0.1:8080
>> ** SIEGE 3.0.5
>> ** Preparing 100 concurrent users for battle.
>> The server is now under siege...
>> Lifting the server siege...      done.
>>
>> Transactions:              186841 hits
>> Availability:              100.00 %
>> Elapsed time:               14.41 secs
>> Data transferred:            2.14 MB
>> Response time:                0.01 secs
>> Transaction rate:        12966.07 trans/sec
>> Throughput:                0.15 MB/sec
>> Concurrency:               99.52
>> Successful transactions:      186841
>> Failed transactions:               0
>> Longest transaction:            0.06
>> Shortest transaction:            0.00
>>
>> Kerl install (no option, just build, install, activate, and of course
>> example rebuilt from scratch):
>>
>> % erl
>> Erlang R16B02 (erts-5.10.3) [source] [64-bit] [smp:4:4] [async-threads:10]
>> [hipe] [kernel-poll:false]
>> ...
>> % siege -b -c 100 -t 5s http://127.0.0.1:8080
>> ** SIEGE 3.0.5
>> ** Preparing 100 concurrent users for battle.
>> The server is now under siege...
>> Lifting the server siege...      done.
>>
>> Transactions:              121051 hits
>> Availability:              100.00 %
>> Elapsed time:               14.37 secs
>> Data transferred:            1.39 MB
>> Response time:                0.01 secs
>> Transaction rate:         8423.87 trans/sec
>> Throughput:                0.10 MB/sec
>> Concurrency:               99.65
>> Successful transactions:      121051
>> Failed transactions:               0
>> Longest transaction:            0.04
>> Shortest transaction:            0.00
>>
>> Your guess is as good as mine.
>>
>> On 11/01/2013 10:28 AM, Aaron France wrote:
>>>
>>> Hi,
>>>
>>> I'd hate to sound like a party-pooper but I'm very much inclined to say
>>> that this endeavour is not a very fruitful one.
>>>
>>> Arch essentially just packages upstream. It's quite likely any
>>> performance gains you see are simply factors mainly a newer kernel and
>>> more up-to-date packages.
>>>
>>> That said, I'll happily entertain the idea that Arch is somehow a
>>> performance distro.
>>>
>>> Aaron
>>>
>>>
>>> On 01/11/13 06:07, Dmitry Kolesnikov wrote:
>>>>
>>>> What are test cases you run to validate performance? And What was
>>>> environment?
>>>>
>>>>
>>>> Best Regards,
>>>> Dmitry >-|-|-*>
>>>>
>>>>
>>>>> On 31.10.2013, at 23.22, Loïc Hoguin <> wrote:
>>>>>
>>>>> I know the Makefile doesn't seem to do anything but last I checked
>>>>> (been a while) I had better performance with the precompiled version
>>>>> compared to compiling with kerl (I'm not sure if I tried with manual
>>>>> compilation).
>>>>>
>>>>> I'll check again tomorrow.
>>>>>
>>>>>> On 10/31/2013 10:04 PM, Aaron France wrote:
>>>>>> Demonstrably false.
>>>>>>
>>>>>> ArchLinux uses https://gist.github.com/AeroNotix/7257133 this to build
>>>>>> it's package, no patches, no wicked switches, just a plain makefile.
>>>>>>
>>>>>> Aaron
>>>>>>
>>>>>>
>>>>>> On Thu, Oct 31, 2013 at 9:51 PM, Olivier Boudeville
>>>>>> < <mailto:>>
>>>>>> wrote:
>>>>>>
>>>>>>      Hi,
>>>>>>
>>>>>>      I may be wrong but I think that, some time ago, someone (maybe
>>>>>>      Loïc?) mentioned incidentally in this mailing list that the Arch
>>>>>>      Linux version of Erlang (obtained through pacman) was reported to
>>>>>>      be, at least in some cases, significantly more efficient than the
>>>>>>      stock, official version (the trouble is that I can't find that
>>>>>>      message from the list archives).
>>>>>>
>>>>>>      I was wondering if it had been confirmed, and, if yes, if there
>>>>>> were
>>>>>>      some patches sent upstream by the Arch maintainers that could be
>>>>>>      fruitfully applied to the official sources?
>>>>>>
>>>>>>      Thanks for any information!
>>>>>>
>>>>>>      Cheers,
>>>>>>
>>>>>>      Olivier.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>      _________________________________________________
>>>>>>      erlang-questions mailing list
>>>>>>       <mailto:>
>>>>>>      http://erlang.org/mailman/__listinfo/erlang-questions
>>>>>> <http://erlang.org/mailman/listinfo/erlang-questions>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> erlang-questions mailing list
>>>>>> 
>>>>>> http://erlang.org/mailman/listinfo/erlang-questions
>>>>>
>>>>>
>>>>> --
>>>>> Loïc Hoguin
>>>>> Erlang Cowboy
>>>>> Nine Nines
>>>>> http://ninenines.eu
>>>>> _______________________________________________
>>>>> erlang-questions mailing list
>>>>> 
>>>>> http://erlang.org/mailman/listinfo/erlang-questions
>>
>>
>>
>
> _______________________________________________
> erlang-questions mailing list
> 
> http://erlang.org/mailman/listinfo/erlang-questions



More information about the erlang-questions mailing list