Local ISDF module (with k-sampling) for PBC mean-field calculation. #62
Conversation
|
The code in python seems only using the np.fft and the fft/ifft code from |
The calculation of V matrix (which is the counterpart of auxiliary Coulomb matrix in DF) heavily depends on fft/ifft (please see pbc_isdf_V.c) and is computationally demanding. It fftw3.h is not allowed to be included in pyscf-forge, I can rewrite all the involved code using np.fft/np.ifft or other python wrapper for fftw3 to see whether this will influence the performance. |
|
I think the normal way is to at least have numpy/scipy (the standard pyscf
dependencies) as a fallback.
…On Fri, Aug 16, 2024 at 8:40 AM ZhangNing ***@***.***> wrote:
The code in python seems only using the np.fft and the fft/ifft code from
pyscf.lib. Is the fftw wrapper fft.c still needed?
The calculation of V matrix (which is the counterpart of auxiliary Coulomb
matrix in DF) heavily depends on fft/ifft (please see pbc_isdf_V.c) and is
computationally demanding. It fftw3.h is not allowed to be included in
pyscf-forge, I can rewrite all the involved code using np.fft/np.ifft or
other python wrapper for fftw3 to see whether this will influence the
performance.
—
Reply to this email directly, view it on GitHub
<#62 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAN5NRFVH6AEN3EVNPGQUTTZRYMPNAVCNFSM6AAAAABMNEUTLGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEOJTG4ZTMNBQGM>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
|
Got it! Thanks for suggesting that.
gkc1000 ***@***.***> 于2024年8月16日周五 08:47写道：
… I think the normal way is to at least have numpy/scipy (the standard pyscf
dependencies) as a fallback.
On Fri, Aug 16, 2024 at 8:40 AM ZhangNing ***@***.***> wrote:
> The code in python seems only using the np.fft and the fft/ifft code
from
> pyscf.lib. Is the fftw wrapper fft.c still needed?
>
> The calculation of V matrix (which is the counterpart of auxiliary
Coulomb
> matrix in DF) heavily depends on fft/ifft (please see pbc_isdf_V.c) and
is
> computationally demanding. It fftw3.h is not allowed to be included in
> pyscf-forge, I can rewrite all the involved code using np.fft/np.ifft or
> other python wrapper for fftw3 to see whether this will influence the
> performance.
>
> —
> Reply to this email directly, view it on GitHub
> <#62 (comment)>,
> or unsubscribe
> <
https://github.com/notifications/unsubscribe-auth/AAN5NRFVH6AEN3EVNPGQUTTZRYMPNAVCNFSM6AAAAABMNEUTLGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEOJTG4ZTMNBQGM>
> .
> You are receiving this because you are subscribed to this thread.Message
> ID: ***@***.***>
>
—
Reply to this email directly, view it on GitHub
<#62 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AHJIAW2W6LQPBAPOQWYIU3DZRYNKTAVCNFSM6AAAAABMNEUTLGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEOJTG42DMOJVGI>
.
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
|
Please also clean up the style issues. You can run the flake8 locally |
|
|
||
| del mf | ||
| del pbc_isdf_info | ||
| exit(1) |
There was a problem hiding this comment.
Is there a reason why you exit(1) in all of the examples? This signifies an error has occurred which should not be the case if it reaches the end of the code.
There was a problem hiding this comment.
Thanks for pointing out this issue. It is added just because I do not want run all the for-loops when I am testing the code. It should not appear here and I will comment this out.
| @@ -0,0 +1,19 @@ | |||
| # Copyright 2014-2018 The PySCF Developers. All Rights Reserved. | |||
There was a problem hiding this comment.
Please update the copyright to 2014-2024 here and for the rest of the files.
There was a problem hiding this comment.
Thanks, I will change all of them.
|
Will you be providing any tests for this module? |
I have some test files within my local branch of ISDF module, I can add them back if needed. |
1.Local Interpolative separable density fitting module for PBC mean-field calculation.
2.Supporting robust fitting, k-point, MPI.
3.L-ISDF supports ao2mo but ao2mo_7d has not been implemented yet.