search

paboyle / Grid

Data parallel C++ mathematical object library
GNU General Public License v2.0
149 stars 106 forks source link

Milestones? #45

Closed aportelli closed 7 years ago

aportelli commented 7 years ago

I have started adding the '0.6.0' milestone to some issues. I think this system is useful to plan releases. Tell me if you think it is inconvenient.

paboyle commented 7 years ago

Where do we centrally collate/query all the work we think is needed for 0.6.0 ?

I agree that organisation is good -- just don't see where I browse the work list and process.

aportelli commented 7 years ago

There are many ways to do that, you can just click on the "milestone" button in the main issue menu. In the issue filters you can select only issues corresponding to one milestone. Finally, if you are browsing an issue which is flagged with a milestone then you can see the milestone progress on the right and click on it to see the corresponding issues. I hope that helps.

paboyle commented 7 years ago

Leaving what I would view as desirable for v0.6.0. Please add other comments in this thread.

-- FFT

-- Gamma rework to complete basis, work with Antonin on meson trace code to make it look pretty.

-- Doxygen avalable online somehow

-- Clean up mixed precision

-- Thread safe & threaded version of code in Lattice_transfer

-- Check for no call of peek/poke from threaded regions

-- Optimised precision changing

-- Assembly "Ddag" in KNL

-- Generic fermion reps

-- Applications directory and common HMC, RHMC

-- Move Lanczos to use Eigen matrix types. Replace "DenseMatrix" with Eigen types.

-- PAB to audit core code (lib, but not down lib/qcd tree).

aportelli commented 7 years ago

Thanks for the list it is nice to have a roadmap. We are now ~100 commits since 0.5.1 and important changes have been made. I thought that maybe it would be good to release 0.6 rather soon (i.e. in the next 2 weeks)? Many of the features you listed can represent a significant amount of time and could constitue good goals for the next version. I would suggest:

0.6:

0.7:

What do you think?

aportelli commented 7 years ago

Just to keep track of things, I have started adding ideas of things to do for the first "final" version (1.0).

paboyle commented 7 years ago

I'm worried that we will run into 1.0 too soon....

I'd vote for 0.5.3 quickly and the extended features in a 0.6.0 ?

Doesn't change the practice, but numbering would leave me more comfortable.

aportelli commented 7 years ago

That sounds good. My point was just on trying to have regular checkpointing of versions we think are usable, keeping master alive. I don't have a strong opinion on the numbering scheme.

aportelli commented 7 years ago

Hi I just want to keep this issue alive, we are slowly going into not releasing versions, master is totally outdated now. Shall we push onto releasing 0.6.0?