[SCIFIO] SCIFIOImgCells.java
Mark Hiner
hiner at wisc.edu
Mon Sep 21 11:38:48 CDT 2015
Hi Tobias,
>All the figuring out who is responsible and whether it is already managed
etc., that looks really expensive to do for each Cell if you have a lot of
small Cells as in BDV.
There will always be overhead for plugin-based matching. Ideally it
shouldn't be more than a couple of hash map lookups. As far as complexity
scaling with number of plugins of a given type, that's a general concern
which will likely necessitate more indexing support in SciJava eventually.
Also the ref manager should be a HandlerService.
The alternative I think is for cells to know that they need a phantom
reference that can cache them, and to always construct it when a cell is
constructed. Since you have to construct and queue a reference either way,
in theory I wouldn't expect compatibility determination to be the
performance bottleneck. I have not profiled the performance of the cache
service but if doing so did reveal compatibility determination to be a
problem, then I would also expect it to be resolved with better caching and
indexing.
Anyway, for technical stuff like this it's best we discuss on the mailing
list or GitHub.
If you notice problems while you're looking at the code, it's more
efficient for you to commit a fix directly since you have commit rights. If
the fix is not clear, then I would open a GitHub issue
<https://github.com/scifio/scifio/issues>.
Hope that helps.
Thanks,
Mark
On Fri, Sep 18, 2015 at 11:29 AM, Tobias Pietzsch <pietzsch at mpi-cbg.de>
wrote:
> Hi Mark,
>
> did you ever benchmark how much time is spent in
> DefaultRefManagerService.manage()
> All the figuring out who is responsible and whether it is already managed
> etc., that looks really expensive to do for each Cell if you have a lot of
> small Cells as in BDV.
>
> best regards,
> Tobias
>
> On 18 Sep 2015, at 17:24, Tobias Pietzsch <pietzsch at mpi-cbg.de> wrote:
>
> member variable DefaultRefManagerService.disposed
>
> https://github.com/scifio/scifio/blob/master/src/main/java/io/scif/refs/DefaultRefManagerService.java#L75
> is named and documented in the wrong way (should be renamed to
> “notDisposed” or “running” or similar), or used in the wrong way in
>
> https://github.com/scifio/scifio/blob/master/src/main/java/io/scif/refs/DefaultRefManagerService.java#L143
>
> https://github.com/scifio/scifio/blob/master/src/main/java/io/scif/refs/DefaultRefManagerService.java#L150
>
> On 18 Sep 2015, at 14:07, Tobias Pietzsch <pietzsch at mpi-cbg.de> wrote:
>
> Javadoc for class
>
> https://github.com/scifio/scifio/blob/master/src/main/java/io/scif/refs/CleaningRef.java#L41
> and method
>
> https://github.com/scifio/scifio/blob/master/src/main/java/io/scif/refs/CleaningRef.java#L49
> are contradictory.
>
> On 17 Sep 2015, at 18:18, Mark Hiner <hiner at wisc.edu> wrote:
>
> Hi Tobi,
>
> I'll be back Monday.
>
> There's are eclipse config files in scifio.git/config but I think maybe
> they should just be removed in favor of those in imagej/imagej. Either way
> a code cleanup commit wouldn't hurt.
>
> Feel free to commit any fixes directly if you'd like. A lot of those
> classes will probably change with a refactoring to rely more on imglib2
> classes. Otherwise, I will try to remember next week.
>
> On Thu, Sep 17, 2015 at 11:02 AM, Tobias Pietzsch <pietzsch at mpi-cbg.de>
> wrote:
>
>> Hi Mark,
>>
>> I’m currently digging through SCIFIOCellCache, MapDBCache, etc.
>> I guess I’ll be bugging you with questions over the next days...
>> ...respectively when you are back from your vacation. Please feel free to
>> postpone answering until after vacation. I’ll just send emails with
>> everything I notice so that I don’t forget. When will you be back?
>>
>> Could you please add me as an author of the
>> https://github.com/scifio/scifio/blob/master/src/main/java/io/scif/img/cell/SCIFIOImgCells.java
>> class (because this is basically the same as
>> https://github.com/bigdataviewer/bigdataviewer-core/blob/master/src/main/java/bdv/img/cache/VolatileImgCells.java
>> from BDV)?
>>
>> Something seems to have gone wrong with auto-formatting the comments here
>>
>> https://github.com/scifio/scifio/blob/master/src/main/java/io/scif/img/cell/SCIFIOCell.java#L60-L64
>> and here
>>
>> https://github.com/scifio/scifio/blob/master/src/main/java/io/scif/img/cell/SCIFIOCell.java#L69-L71
>>
>> Is there an eclipse style file for SCIFIO?
>>
>> best regards,
>> Tobias
>>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://scif.io/pipermail/scifio/attachments/20150921/7691dc6f/attachment.html>
More information about the SCIFIO
mailing list