Project

General

Profile

Task #1424

Feature #1292: mdrun features to deprecate for 5.0

Feature #1500: Post-5.0 feature clean-up plan

remove mdrun -testverlet option

Added by Szilárd Páll almost 6 years ago. Updated almost 5 years ago.

Status:
Closed
Priority:
Normal
Assignee:
Category:
mdrun
Target version:
Difficulty:
uncategorized
Close

Description

The Verlet scheme will be default from now, so there is no need to keep this option.

Associated revisions

Revision 772f019a (diff)
Added by Mark Abraham almost 5 years ago

Remove mdrun -testverlet

This was only intended for quick performance testing of old .tpr files
during the transition period. The window where that was useful has
passed, and ongoing abuse of it has been observed. There is no need to
preserve this until the formal removal of the group scheme.

Fixes #1424

Change-Id: I589a8e316beeba6819cd01d9655bfc069bcbb174

History

#1 Updated by Szilárd Páll almost 6 years ago

Any objections?

#2 Updated by Berk Hess almost 6 years ago

This option was mainly to quickly test old tpr files people might already with the Verlet scheme. I don't know if this need has changed right now. We should certainly remove it for 5.1, but for 5.0 it might still be useful.

#3 Updated by Szilárd Páll almost 6 years ago

Berk Hess wrote:

This option was mainly to quickly test old tpr files people might already with the Verlet scheme. I don't know if this need has changed right now. We should certainly remove it for 5.1, but for 5.0 it might still be useful.

Wasn't a problem that -testverlet required making some assumptions that otherwise grompp would do (or warn the user about)? My concern is that -testverlet is being abused as the easy way to use GPUs and although we warn that this is not suitable for production, I've already seen this information spreading on the Internet with the warning stripped away, this page uses it in a "Sample batch script for Gromacs 4.6.1": http://biowulf.nih.gov/apps/gromacs-gpu.html

#4 Updated by Mark Abraham almost 6 years ago

People have had a year to start to get with the times. I'd be happy to remove it, but I suspect there will be a few people who have been ignoring 4.5 and 4.6 because clearly GROMACS progress is slower than AMBER, etc. Backward compatibility is a PITA.

Abuse can be dealt with in other ways :)

#5 Updated by Szilárd Páll almost 6 years ago

Mark Abraham wrote:

People have had a year to start to get with the times. I'd be happy to remove it, but I suspect there will be a few people who have been ignoring 4.5 and 4.6 because clearly GROMACS progress is slower than AMBER, etc. Backward compatibility is a PITA.

Well, if they've been ignoring it, they can use -testverlet with their old tpr-s with 4.6.x, can't they? I don't see much gain in providing backward compatibility with a convenience option like -testverlet, do you?

Abuse can be dealt with in other ways :)

How? One can start mailing to the biowulf.nih.gov maintainers that their howto is incorrect, but that's not very productive, I'd say.

#6 Updated by Mark Abraham almost 6 years ago

Szilárd Páll wrote:

Mark Abraham wrote:

People have had a year to start to get with the times. I'd be happy to remove it, but I suspect there will be a few people who have been ignoring 4.5 and 4.6 because clearly GROMACS progress is slower than AMBER, etc. Backward compatibility is a PITA.

Well, if they've been ignoring it, they can use -testverlet with their old tpr-s with 4.6.x, can't they? I don't see much gain in providing backward compatibility with a convenience option like -testverlet, do you?

The point of the feature was to let people try old tprs with new code. If people switching to 5.0 might have old tprs because they're switching from some prehistoric version, then the backward compatibility aspect is still relevant.

Abuse can be dealt with in other ways :)

How? One can start mailing to the biowulf.nih.gov maintainers that their howto is incorrect, but that's not very productive, I'd say.

I'd say that's maximally productive. Somebody there probably learns something, and people reading their future/updated work is less likely to unknowingly do the wrong thing with 4.6 or 5.0. The legislative solution (remove from 5.0) doesn't address the ongoing misuse of the old version.

#7 Updated by Szilárd Páll almost 6 years ago

Mark Abraham wrote:

Szilárd Páll wrote:

Mark Abraham wrote:

People have had a year to start to get with the times. I'd be happy to remove it, but I suspect there will be a few people who have been ignoring 4.5 and 4.6 because clearly GROMACS progress is slower than AMBER, etc. Backward compatibility is a PITA.

Well, if they've been ignoring it, they can use -testverlet with their old tpr-s with 4.6.x, can't they? I don't see much gain in providing backward compatibility with a convenience option like -testverlet, do you?

The point of the feature was to let people try old tprs with new code. If people switching to 5.0 might have old tprs because they're switching from some prehistoric version, then the backward compatibility aspect is still relevant.

Sure. I just don't like us catering for laziness.

Abuse can be dealt with in other ways :)

How? One can start mailing to the biowulf.nih.gov maintainers that their howto is incorrect, but that's not very productive, I'd say.

I'd say that's maximally productive. Somebody there probably learns something, and people reading their future/updated work is less likely to unknowingly do the wrong thing with 4.6 or 5.0. The legislative solution (remove from 5.0) doesn't address the ongoing misuse of the old version.

Mailed them. Now I dislike catering for laziness even more. :P

#8 Updated by Mark Abraham almost 6 years ago

Ja, I hate laziness too!

I know I have to actively repress my desire to get rid of all old stuff at the first provocation. It is amusing that we often disagree on what old stuff we should remove, and are often on opposite sides of the debate! :-)

#9 Updated by Szilárd Páll almost 6 years ago

biowulf.nih admins got their guide fixed. I guess this can be closed.

#10 Updated by Mark Abraham almost 6 years ago

  • Target version changed from 5.0 to 5.x

Great! Let's retarget for 5.x so we know to reconsider actually removing -testverlet for 5.1.

#11 Updated by Mark Abraham over 5 years ago

  • Parent task set to #1500

#12 Updated by Szilárd Páll almost 5 years ago

  • Target version changed from 5.x to 5.1

I'm adding this explicitly to 5.1 because sadly I keep seeing people use -testverlet even in production runs.

#13 Updated by Mark Abraham almost 5 years ago

Szilárd Páll wrote:

I'm adding this explicitly to 5.1 because sadly I keep seeing people use -testverlet even in production runs.

Agreed

#14 Updated by Gerrit Code Review Bot almost 5 years ago

Gerrit received a related patchset '1' for Issue #1424.
Uploader: Mark Abraham ()
Change-Id: I589a8e316beeba6819cd01d9655bfc069bcbb174
Gerrit URL: https://gerrit.gromacs.org/4353

#15 Updated by Mark Abraham almost 5 years ago

  • Status changed from New to Fix uploaded

#16 Updated by Mark Abraham almost 5 years ago

  • Status changed from Fix uploaded to Closed
  • Assignee changed from Szilárd Páll to Mark Abraham

Also available in: Atom PDF