Charles Oliver Nutter-2 wrote:
>
> Christian Seiler wrote:
>> I noticed that there is already an issue covering this topic:
>> http://jira.codehaus.org/browse/JRUBY-2184
>>
>> JRuby is on-par with simple models (a few ints) with my setup so I tried
>> with ten datetime columns. JRuby is about 20% slower, please look at the
>> comment which I've added. Probably not top prio for 1.1.3
>
> There have been some Time related perf improvements for 1.1.3, but I
> wouldn't expect they're helping you much. The perf problems with
> datetime probably lies in the conversion from JDBC types to Ruby types.
> At any rate...it's a priority, and hopefully we'll get to it soon. Any
> help narrowing down why it's slow would speed the process.
>
> - Charlie
>
I also checked Time.now, it's somewhat slower than on MRI. But 10.times
{puts Benchmark.measure {10000.times {Testm.new }}.total} doesn't even hit
the database and leaves all properties nil, so there shouldn't be any
conversion or creation of time objects.
I did a little CPU-profiling with the new VisualVM tool. I've attached the
results to Jira (did the Testm.new benchmark again). I don't see something
like a smoking gun, but for me it's hard to decrypt that stuff anyway ;-)
--
View this message in context:
http://www.nabble.com/JRuby-1.1.3-by-end-of-week...Nominate-problems-here...-tp18448971p18462573.html
Sent from the JRuby - Dev mailing list archive at Nabble.com.
---------------------------------------------------------------------
To unsubscribe from this list, please visit:
http://xircles.codehaus.org/manage_email
