You can probably parallelize the findOverlaps function, but you'd have
to write the code yourself, and that code would be mostly bookkeeping
code to get the indices right. Maybe there's a case for adding a
parallelized findOverlaps function to BiocParallel?
You can't parallelize the disjoin op
On Thu, Nov 15, 2012 at 11:00 AM, Martin Morgan wrote:
> On 11/15/2012 10:53 AM, Henrik Bengtsson wrote:
>
>> Is there any write up/discussion/plans on the various types of
>> parallel computations out there:
>>
>> (1) one machine / multi-core/multi-threaded
>> (2) multiple machines / multiple pr
should approaches to fault-tolerance/recovery/debugging be a topic here?
On Thu, Nov 15, 2012 at 1:53 PM, Henrik Bengtsson wrote:
> Is there any write up/discussion/plans on the various types of
> parallel computations out there:
>
> (1) one machine / multi-core/multi-threaded
> (2) multiple mach
Personally, having used memcached in the past for distributed shared memory
caching, I am most interested in 3) and doRedis. Many cluster/batch
processing systems are a colossal PITA, and a worker queue would go a long
way towards fixing that. Less checkpointing, more results... I hope.
As an as
On 11/15/2012 10:53 AM, Henrik Bengtsson wrote:
Is there any write up/discussion/plans on the various types of
parallel computations out there:
(1) one machine / multi-core/multi-threaded
(2) multiple machines / multiple processes
(3) batch / queue processing (on large compute clusters with many
On 11/15/2012 6:21 AM, Kasper Daniel Hansen wrote:
I'll second Ryan's patch (at least in principle). When I parallelize
across multiple cores, I have always found mc.preschedule to be an
important option to expose (that, and the number of cores, is all I
use routinely).
Yes, Ryan provided a pu
Is there any write up/discussion/plans on the various types of
parallel computations out there:
(1) one machine / multi-core/multi-threaded
(2) multiple machines / multiple processes
(3) batch / queue processing (on large compute clusters with many users).
(4) ...
Are we/you mainly focusing on (1
I'll second Ryan's patch (at least in principle). When I parallelize
across multiple cores, I have always found mc.preschedule to be an
important option to expose (that, and the number of cores, is all I
use routinely).
Kasper
On Wed, Nov 14, 2012 at 7:14 PM, Ryan C. Thompson wrote:
> I just su
