Re: [cdt-dev] Post 2.0 milestones

[Douglas Schaefer <dschaefe@xxxxxxxxxx>]

I haven't, but I love the quote there: "it's important to remember that 
the appropriate data structure or algorithm is all that separates you from 
success and absolute failure." Right now, that hits pretty close to home.

Using third party collection classes could get us a head start. We do have 
to be concerned with licensing, though. I also get the feel that our 
requirements are pretty peculiar and that hand-crafting our own data 
structures will get us the most bang for the buck. Something we can look 
into, though.

Now if only I can alloca in Java...

Thanks!
Doug Schaefer, IBM's Eclipse CDT Architect
Ottawa (Palladium), Ontario, Canada



Chris Wiebe <cwiebe@xxxxxxx> 
Sent by: cdt-dev-admin@xxxxxxxxxxx
06/21/2004 04:03 PM
Please respond to
cdt-dev


To
cdt-dev@xxxxxxxxxxx
cc

Subject
Re: [cdt-dev] Post 2.0 milestones






Has anyone looked into drop-in replacements for the java collections? 
There are some good, high-performance, low-footprint alternatives out 
there.

Here is a useful list:
http://www.manageability.org/blog/stuff/open-source-java-data-structures-and-algorithms


-Chris


Douglas Schaefer wrote:

> Without know when 2.0.1 or 2.1 are, it's hard to answer you question :P
> 
> As many of you have noticed, the performance of the parser is not where 
we 
> want it. The good news is that we're finally getting to the point where 
> the parser is pretty complete and features like content assist and open 
> declaration are working great (when you extend the timeout, that is).
> 
> We have a couple of strategies for addressing this in the short term and 

> they are aimed at speeding up the data structures we are using during 
the 
> parse. First is a rewrite of the scanner to use char[]'s and eliminate 
the 
> use of strings in the tokens (see Scanner2). Second is a replacement of 
> the HashMaps and related structures used by the symbol table to greatly 
> reduce the number of objects and to speed up creation and lookup times. 
We 
> have longer term strategies in our bag as well and are always open to 
help 
> and suggestions.
> 
> To follow our progress on this issue, we have a bug open 
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=59468. We have a pretty 
> aggressive target but comparing to the parse times of gcc, they should 
be 
> achievable. However, these changes will be destabilizing so the sooner 
we 
> can get them in, the sooner we can recover. But one thing is for 
certain, 
> we are close to greatness with these features (IMHO :-).
> 
> Doug Schaefer, IBM's Eclipse CDT Architect
> Ottawa (Palladium), Ontario, Canada
> 
> 
> 
> David Daoust/Ottawa/IBM@IBMCA 
> Sent by: cdt-dev-admin@xxxxxxxxxxx
> 06/21/2004 02:14 PM
> Please respond to
> cdt-dev
> 
> 
> To
> cdt-dev@xxxxxxxxxxx
> cc
> 
> Subject
> Re: [cdt-dev] Post 2.0 milestones
> 
> 
> 
> 
> 
> 
> 
> This is just what I have been looking for :-) 
> 
> In general the development team at IBM is planning to push on improving 
> parser performance and memory use in 2.0.1 -- I have moved a couple of 
> defects to 2.0.1 based on this. 
> 
> We are currently still poking away at performance/memory for 2.0, but we 

> do not want to make any major (destabilizing) changes until after 2.0 
GA. 
> 
> 
> Do people think that this is reasonable (should we be targeting 2.1 
> instead)? 
> 
>         - Dave 
> 
> 
> 
> Kleanthis Hapitas/Ottawa/IBM@IBMCA 
> Sent by: cdt-dev-admin@xxxxxxxxxxx 
> 06/21/2004 01:59 PM 
> 
> Please respond to
> cdt-dev
> 
> 
> To
> cdt-dev@xxxxxxxxxxx 
> cc
> 
> Subject
> [cdt-dev] Post 2.0 milestones
> 
> 
> 
> 
> 
> 
> 
> 
> 
> Hi all, 
> 
> I have requested that additional milestones get added to the CDT in 
> bugzilla, and they have just been created. There is now a 2.0.1 and a 
2.1 
> target milestone to choose from. I know we haven't had any concrete 
> discussions yet as to when the 2.0.1 or 2.1 delivery dates will be, but 
> with these milestones we can start to distinguish between bugs to be 
fixed 
> sometime in the near future and those that will be fixed whenever... 
> 
> My intention is to provide a target milestone (2.0.1) for defects that 
we 
> want to fix on the 2.0 stream after it goes GA at the end of the month, 
> and another (2.1) for the next release that comes out of the Head stream 

> after 2.0 reaches GA. 
> 
> Questions, comments, concerns, let me know... 
> 
> Kleo 
> 
> _______________________________________________
> cdt-dev mailing list
> cdt-dev@xxxxxxxxxxx
> http://dev.eclipse.org/mailman/listinfo/cdt-dev
> 
> 
_______________________________________________
cdt-dev mailing list
cdt-dev@xxxxxxxxxxx
http://dev.eclipse.org/mailman/listinfo/cdt-dev