- edited description
GTSAM 4.0 Release Milestones
The below issues should be fixed for 4.0 release (and possibly an intermediate hotfix):
Additionally, the following branches should be merged into develop when they are finished
- feature/ordering
- feature/BAD
Finally, this release will see the move to public gtsam!
Any other issues/features should be addressed?
Reference: @fdellaert @jdong37 @cbeall3 @lucacarlone @thduynguyen @ydjian
Comments (23)
-
reporter -
reporter - edited description
-
reporter - edited description
-
Sorta thinking we should resolve this and then release a maintenance version 3.whatever...
-
BTW about to add another issue.
-
reporter We can leave this open for a more long term plan this semester and do a targeted maintenance version to address bug fixes.
-
OK, which ones are issued to be fixed before release 3.X?
-
reporter I would classify everything that crashes/ fails to compile, or fails unit tests. It seems all the issues other than issue
#130qualify as that, I'll move to a separate issue. -
reporter - edited description
-
reporter - edited description
-
I have to stop coding for a while. But here’s my current thinking (best read in an Evernote that has some links
We are using ptr_map for Values. Could still be OK, but internally we simply create a Chart object. DerivedValue and Value will no longer be visible outside. With a chart and traits, might not need to. @mikebosse if you have time to try this please go ahead, otherwise it'll have to wait until I'm a coding funk again.
Erasing types is cool. Conclusions in part IV by (very cool) Andrzej. Some timings are here. Might or might not need any(!) of this, as ptr_map does trick.
I still need to improve performance of expressions by trying unordered. I created an issue so @stan84 might have a go at this.
Values compatibility, trait implementations, etc can all benefit from SFINAE or tag forwarding.
I want to use Fusion more in Expressions, even to the point of having no custom classes for functions: everything is deduced from function passed. Possible? After @stan84 is successful with
unordered
, it might be something he might be interested in tackling. -
reporter - edited description
-
@stan84: it would be great to include smart factors in GTSAM 4.0 after your fixes! let's discuss this during tomorrow's meeting.
-
@lucacarlone Sure!
-
Hi guys, given the slowness with incorporating our special matrix/vector initialization patch into Eigen, I suggest that I remove that feature for the 4.0 release. I'm glad to do that, i.e. refactor GTSAM (mostly the unit tests).
An update on the patches - some internal expression template changes in the latest dev version of Eigen render our patches not working, and so far I haven't been able to figure it out, nor have I heard back on the message I sent the Eigen developers. Though, even after it's fixed and they incorporate the patch, it will take quite a while for the next Eigen version to filter down into system installs for Ubuntu. So, it was probably a mistake to make these changes in the first place (sorry!).
-
Richard, does that mean the solution I suggested, having a header-file patch in GTSAM-Eigen-includes.h, has no chance of working?
-
Hi Frank, I replied about that earlier but it sounds like I misunderstood what you meant when you said "header file patches" previously, since it sounds like you're not talking about patching the system-installed Eigen. Can you explain what you mean by a header-file patch? Are you thinking of including some special code in GTSAM-Eigen-includes.h that somehow modifies the << operator?
-
Yes, exactly :-)
-
reporter Definitely need eyes on my METIS pull request so we can get it integrated with the BAD changes and see if any build/execution issues exist.
-
- edited description
-
- changed status to wontfix
Superseded by Roadmap in wiki
-
Hey guys, just curious what is the planned release date for 4.0?
-
Nothing scheduled yet. When we have volunteers helping us out with the crucial things :-)
Best Frank
- Log in to comment