[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <46008ac6368e170f287d946103dc1dba@smtp.hushmail.com>
Date: Fri, 22 May 2015 18:01:39 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: autotune_run problem
On 2015-05-22 14:22, Agnieszka Bielec wrote:
> 2015-05-22 2:56 GMT+02:00 magnum <john.magnum@...hmail.com>:
>> On 2015-05-22 02:00, Agnieszka Bielec wrote:
>>> I discovered that for different set of arguments is running different
>>> algorithm to determine when tuning for gws is stopped
>>> If i make
>>> autotune_run(self, 1000, 0, 500);
>
>>> computed gws is optimal on my laptop and --dev=1 but not in --dev=5,
>>> it prints exceed for the optimal value and setting highest
>>> duration_time doesn't work
>>
>>
>> Did you try setting it to 1000 instead of 500? If that works better you
>> should implement a split kernel though. A full second duration is way to
>> long.
>
> thanks! I was only testing the last argument for autotune_run(self, 1,
> 1000, 100000); i don't know why
> with 1000 works so far for parallel
> but I discovered that pomelo for bigger costs (5, even not so big) can
> be faster 10 times than actually after setting the last argument to
> 100000 (x4 faster)
> and when I also modify 1.8 to 1.1 is x10 faster! (I must make speed
> tests again :<)
> I tested this on my laptop and for --cost=5:5,5:5
> yesterday I was thinking that there are 2 various algorithms for
> determining when autotune_run stops but it seems that two of them
> checks if we exceeded limits
> it is possible to disable checking based on time?
This is confusing. Using 100000 makes no sense! It will be parsed as ns
so equals 100us limit for a single kernel duration. That's too short!
static void autotune_run(struct fmt_main * self, unsigned int rounds,
size_t gws_limit, unsigned long long int max_run_time)
I believe 'rounds' is purely cosmetical (it's used to show the --verb=5
figures). 'gws_limit' should be 0 for no limit. And max_run_time should
either be 1000 or 10000000000.
magnum
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
