module functions on the GPU

[Iximiel]

Description

Hi @gtribello, I am starting also to move some of the functionalities of function to the GPU.

I started with BESSEL but I think there are no test to check if I am doing things correctly, and I am sad since I like that refactor with no need of redeclaring A or B for each instance.

Being more serious: at time of writing this I did the "around trio" : "LESS_THAN BETWEEN MORE_THAN", I think I succeeded in porting them as standalone actions, but I am not sure how it will be possible to tag them with USEGPU in the shortcuts like DISTANCES ATOMS1=1,2 ATOMS2=5,4 BETWEEN1={GAUSSIAN LOWER=0.5 UPPER=1.0 LABEL=b1} BETWEEN2={GAUSSIAN LOWER=1.0 UPPER=2.0 LABEL=b2}. I think that the solution for that will need a separate PR/discussion

Target release

I would like my code to appear in release v2.11

Type of contribution

• changes to code or doc authored by PLUMED developers, or additions of code in the core or within the default modules
• changes to a module not authored by you
• new module contribution or edit of a module authored by you

Copyright

• I agree to transfer the copyright of the code I have written to the PLUMED developers or to the author of the code I am modifying.

• the module I added or modified contains a COPYRIGHT file with the correct license information. Code should be released under an open source license. I also used the command cd src && ./header.sh mymodulename in order to make sure the headers of the module are correct.

Tests

• I added a new regtest or modified an existing regtest to validate my changes.
• I verified that all regtests are passed successfully on GitHub Actions.

[gtribello]

Hi @Iximiel

I think you have worked this out already, but BESSEL is not something we are likely to need on the GPU. The reason it is implemented is described in the manual:

https://www.plumed.org/doc-master/user-doc/html/BESSEL/

It is tested in the PAMM module.

In terms of getting things like DISTANCES LESS_THAN to work, I wouldn't worry about it. This syntax is deprecated. There are some shortcuts where there might be value in adding the option in symfunc, but we can do them on a case-by-case basis.

[Iximiel]

Ok, then I remove the #ifdef openacc things from BESSEL

It is tested in the PAMM module.

If it pass the test there, I would keep it like this, so that it is not initializing/copying extra 55 numbers each time it is called

---

@carlocamilloni we should try to see if the chbevl function is faster with the constexpr arrays instead of the vectors for A and B in also SAXS, but that is a topic for another PR

[gtribello]

HI @Iximiel

To clarify, I think having the OpenACC in Bessel is not going to do any harm. My point was simply there is probably not a lot of people using this action for anything computationally intensive where the GPU is going to make a lot of difference. However, we can change the function and add ACC in there as long as it all just works easily. However, if it is a lot of effort (which I don't think it would be) then it can be abandoned.

[Iximiel]

HI @Iximiel

To clarify, I think having the OpenACC in Bessel is not going to do any harm. My point was simply there is probably not a lot of people using this action for anything computationally intensive where the GPU is going to make a lot of difference. However, we can change the function and add ACC in there as long as it all just works easily. However, if it is a lot of effort (which I don't think it would be) then it can be abandoned.

ok, so the code can remain like this, in future you can add the #ifdef __PLUMED_HAS_OPENACC #define __PLUMED_USE_OPENACC 1 #endif //__PLUMED_HAS_OPENACC to reactivate the GPU, as now we are saving some compilation time (nvc is SLOW) and avoiding extra errors.

At this point I think this is ok to be merged, if we merge #1292 before this I can rebase and update gpu.md with the new actions directly here

---

I found the chain of calls to test BESSEl: To test BESSEL you have to call KERNEL, but to test KERNEL you have to check PAMM. It might be better to have a something more focused on each step