Optimize building

[stefdoerr]

As per Alberto's request speed up

1. Solvation (divide and conquer approach to joining segments)
2. ~PDB writing (numba)~

So that it doesn't take 2 hours on 1million atoms systems (which apparently is a real case scenario :stuck_out_tongue: )

[cuzzo87]

Just some more information. I tested the same system with the same parameter with Amber and Charmm Amber --> 2h 30min Charmm --> 50 min

I attach the logs log_AMBER.log log_CHARMM.log

[giadefa]

I don't think that the problem is solvate. It's tleap.

Solvate finishes quickly

On 27 September 2017 at 16:22, Alberto notifications@github.com wrote:

Just some more information. I tested the same system with the same parameter with Amber and Charmm Amber --> 2h 30min Charmm --> 50 min

I attach the logs log_AMBER.log https://github.com/Acellera/htmd/files/1337512/log_AMBER.log log_CHARMM.log https://github.com/Acellera/htmd/files/1337513/log_CHARMM.log

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/Acellera/htmd/issues/456#issuecomment-332537702, or mute the thread https://github.com/notifications/unsubscribe-auth/AHaqOir16ryAw5jfkGSB42YFmvGiWHskks5smlowgaJpZM4Plxyi .

-- http://www.acellera.com

   <https://twitter.com/acellera>

https://www.youtube.com/user/acelleracom https://www.linkedin.com/company/2133167?trk=tyah&trkInfo=clickedVertical%3Acompany%2CclickedEntityId%3A2133167%2Cidx%3A2-1-2%2CtarId%3A1448018583204%2Ctas%3Aacellera https://www.acellera.com/md-simulation-blog-news/ http://is.gd/1eXkbS

[giadefa]

What I mean is the building that takes time not solvating.

On 27 September 2017 at 17:02, Gianni De Fabritiis < g.defabritiis@acellera.com> wrote:

I don't think that the problem is solvate. It's tleap.

Solvate finishes quickly

On 27 September 2017 at 16:22, Alberto notifications@github.com wrote:

Just some more information. I tested the same system with the same parameter with Amber and Charmm Amber --> 2h 30min Charmm --> 50 min

I attach the logs log_AMBER.log https://github.com/Acellera/htmd/files/1337512/log_AMBER.log log_CHARMM.log https://github.com/Acellera/htmd/files/1337513/log_CHARMM.log

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/Acellera/htmd/issues/456#issuecomment-332537702, or mute the thread https://github.com/notifications/unsubscribe-auth/AHaqOir16ryAw5jfkGSB42YFmvGiWHskks5smlowgaJpZM4Plxyi .

-- http://www.acellera.com

   <https://twitter.com/acellera>

https://www.youtube.com/user/acelleracom https://www.linkedin.com/company/2133167?trk=tyah&trkInfo=clickedVertical%3Acompany%2CclickedEntityId%3A2133167%2Cidx%3A2-1-2%2CtarId%3A1448018583204%2Ctas%3Aacellera https://www.acellera.com/md-simulation-blog-news/ http://is.gd/1eXkbS

-- http://www.acellera.com

   <https://twitter.com/acellera>

https://www.youtube.com/user/acelleracom https://www.linkedin.com/company/2133167?trk=tyah&trkInfo=clickedVertical%3Acompany%2CclickedEntityId%3A2133167%2Cidx%3A2-1-2%2CtarId%3A1448018583204%2Ctas%3Aacellera https://www.acellera.com/md-simulation-blog-news/ http://is.gd/1eXkbS

[stefdoerr]

"Building takes time" includes everything in the building process, including writing PDB files with a million atoms. We can't improve tleap but we can improve the writers

On Wed, Sep 27, 2017 at 5:05 PM, Gianni De Fabritiis < notifications@github.com> wrote:

What I mean is the building that takes time not solvating.

On 27 September 2017 at 17:02, Gianni De Fabritiis <

g.defabritiis@acellera.com> wrote:

I don't think that the problem is solvate. It's tleap.

Solvate finishes quickly

On 27 September 2017 at 16:22, Alberto notifications@github.com wrote:

Just some more information. I tested the same system with the same parameter with Amber and Charmm Amber --> 2h 30min Charmm --> 50 min

I attach the logs log_AMBER.log https://github.com/Acellera/htmd/files/1337512/log_AMBER.log log_CHARMM.log https://github.com/Acellera/htmd/files/1337513/log_CHARMM.log

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/Acellera/htmd/issues/456#issuecomment-332537702, or mute the thread https://github.com/notifications/unsubscribe-auth/ AHaqOir16ryAw5jfkGSB42YFmvGiWHskks5smlowgaJpZM4Plxyi .

-- http://www.acellera.com

https://twitter.com/acellera https://www.youtube.com/user/acelleracom https://www.linkedin.com/company/2133167?trk=tyah& trkInfo=clickedVertical%3Acompany%2CclickedEntityId% 3A2133167%2Cidx%3A2-1-2%2CtarId%3A1448018583204%2Ctas%3Aacellera https://www.acellera.com/md-simulation-blog-news/ http://is.gd/1eXkbS

-- http://www.acellera.com

https://twitter.com/acellera https://www.youtube.com/user/acelleracom https://www.linkedin.com/company/2133167?trk=tyah& trkInfo=clickedVertical%3Acompany%2CclickedEntityId% 3A2133167%2Cidx%3A2-1-2%2CtarId%3A1448018583204%2Ctas%3Aacellera https://www.acellera.com/md-simulation-blog-news/ http://is.gd/1eXkbS

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/Acellera/htmd/issues/456#issuecomment-332551592, or mute the thread https://github.com/notifications/unsubscribe-auth/AHkVgrlgrF2rvK8QVQ1iJy3XqEpadm5Wks5smmO1gaJpZM4Plxyi .

[j3mdamas]

I was the one asking to Alberto to do the comparison with CHARMM. It's clear here that there is something we cannot solve: tleap being slower than psfgen. We already had this idea from the OPM paper.

[stefdoerr]

I give up on point 2. Numba can't write files so the only option is making a C PDB writer which I'm not in the mood to do any time soon. 11% of the time is wasted on formatting the coordinates with the format83 method. Could be avoided but it would still be more or less similar performance.

Point 1 can still be done to cut it down by 6 minutes out of 2 hours which is debatable if necessary.

[stefdoerr]

If you send me a script and files to replicate the exact charmm build I'll profile it and see if anything can be improved. Otherwise I'll close the issue probably as being due to tleap

[stefdoerr]

@cuzzo87 that was you I was referring to in the last message

[stefdoerr]

I'll just close this till we hit it again. I don't have any real case scenarios for this problem