| RE: [cdt-dev] Improve Indexer on large project |
From: cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Schorn, Markus
Sent: Friday, February 19, 2010 9:59 AM
To: CDT General developers list.
Subject: RE: [cdt-dev] Improve Indexer on large project... so after all you were running low on virtual memory.Markus.
Dear Markus and others,
From: cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Toan Pham
Sent: Thursday, February 18, 2010 7:05 PM
To: CDT General developers list.
Subject: Re: [cdt-dev] Improve Indexer on large project
Importance: Low
* workspace on local disk?
Yes* running low on virtual memory?
No* attempting content assist while indexer or a search is running? If so is it faster otherwise?
No* do you use multiple projects with project references amoung them? If so, is it faster with a single project?
No* do you use a prebuilt index?
No.
My project size is 3.9 gig after running make. (ps. i do not want to run make clean b/c it takes too long to build my project from scratch again). When i run full index, while the indexer is running it reported that there are 8,459 source files. I do not see anywhere in which the indexer reports the number of identifiers, declarations, etc.....
Also, i found out why it was taking too long for content assistance to run. The default heap size for eclipse is only 50Megs. I changed the default heap size to 700M in the /usr/lib/eclipse/eclipse.ini file and that did not effect anything, and that was when i started this thread thinking that the heap size has been pre-allocated. Until recently, I then set the default heap size using params "-vmargs -Xmx800M" and eclipse is 30 times faster. Thank you everyone for helping,
Thanks,
Toan
On Thu, Feb 18, 2010 at 7:25 AM, Schorn, Markus <Markus.Schorn@xxxxxxxxxxxxx> wrote:
Hi Toan,you can check (a few have already be mentioned) the following:* workspace on local disk?* running low on virtual memory?* attempting content assist while indexer or a search is running? If so is it faster otherwise?* do you use multiple projects with project references amoung them? If so, is it faster with a single project?* do you use a prebuilt index?If your project is really large (did not sound like it) you can try to increase the database cache size in the indexerpreferences.Markus.
From: cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Elmenthaler, Jens
Sent: Thursday, February 18, 2010 8:00 AMSubject: RE: [cdt-dev] Improve Indexer on large project
To: CDT General developers list.
Importance: LowWe might have a much larger code base, building a clean index takes 60min, but we don’t have any issues with content assist. There must be something else.
Jens.
From: cdt-dev-bounces@xxxxxxxxxxx [mailto:cdt-dev-bounces@xxxxxxxxxxx] On Behalf Of Toan Pham
Sent: Mittwoch, 17. Februar 2010 22:54
To: CDT General developers list.
Subject: Re: [cdt-dev] Improve Indexer on large project
What i am more concern with is lookup time for auto completion, although i've experienced slow indexing time for my project, > 10 mins. Ctag lookup would be simpler and faster.
The auto completion feature is almost unusable for large projects.On Wed, Feb 17, 2010 at 3:59 PM, Doug Schaefer <cdtdoug@xxxxxxxxx> wrote:
The time to index is independent of the lookup time for auto completion. Indexing is done as you save files or on import of existing files. Changing indexing strategies won't help you if the look up is too slow. Unless our database look up is too slow, but it's built for speed and something like ctags is likely to be really slow (and highly inaccurate since it doesn't do a real parse).
Is the content assist always slow or is it in only certain files? If it's always slow (i.e. in every file in your project), then it points to a problem in the database. If it's only in certain files, then it's a problem in CDT's parser used to determine the context of the assist.
Doug.
On Wed, Feb 17, 2010 at 3:30 PM, Toan Pham <tpham3783@xxxxxxxxx> wrote:
Hi,
I am working with a large project and find that the eclipse's indexer takes a long time (10 secs or more) to look for auto completion tags.
I do not know what happen to it; probably the algorithm for the indexer is inefficient. I am not sure if anyone has thought of the following methods
to improve performance on intelli-sense feature:
1. Use Ctag database - Having the ability to load a ctag database (similar to a feature in Ajunta) and index based on the ctag database only. I can imagine this is a nice must-have feature because it speeds up eclipse and also making eclipse more valuable.
2. Path Filter indexer - Having the ability to filter out directory paths that the indexer should not be indexing would also speed things up.
I like eclipse and hope that developers are improving the indexer in any way. These are my thoughts and hope that they'll consider them.
Thanks,
Toan_______________________________________________
cdt-dev mailing list
cdt-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cdt-dev
_______________________________________________
cdt-dev mailing list
cdt-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cdt-dev
_______________________________________________
cdt-dev mailing list
cdt-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/cdt-dev