-
Notifications
You must be signed in to change notification settings - Fork 39
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
regression-test: crestart: Test failed. #2
Comments
Hej! |
git clone this repo Yesterday: crestart: Test failed.
(Working directory: ./Regression, Output file: crestart.megaTest.out )
For selector {'#_1': 11001, '#_2': 128}:
Headers ['#_3', '#_4', '#_5', '#_6']
Expected value: [0.0, 0.0, 0.0, 0.0]
Extracted value: [1.96354816e-05, -1.82671251e-05, -3.25996241e-06, 2.70167872e-17]
-O0 -ip -xHost could pass, but maybe not need by all subroutine ? |
Aha, this seems to be a precision error. Sometimes the compiler does some optimizations when you set I would not worry about it unless you have other tests failing. I would raise the point with the developer of that routine to eliminate some of the possible sources of confusion. Hope it helps! |
Thanks! BTW: could you explain about why use -mkl=parallel by default ? no mkl routines called from opemp parallel region? (read the code need time, better somebody could tell me :) ) and could I find any speed up report somewhere of uppasd ? Thanks again! |
I think @lbergqv or @anbergman are better suited to answer the However, they are a bit out of date, as we have made several performance improvements since then. |
It's true that no mkl routines are called from openmp regions however -mkl-parallel gives the threaded version that may or may not give a small performance gain (none of the mkl routines are critical by any means anyway). However, some computers (the KNL nodes on Kebnekaise, HPC2N to be precise) requires -mkl-parallel so it has been left as default. |
Thanks Lars ! |
And for the crestart issue, that comes from bad numerics but should not pose any problems for the core routines of the code. The crestart concerns calculated chiralities but the methodology is not perfectly stable so it is not very surprising nor alarming to see this kind of differences. |
On NSC Tetralith: buildenv-intel/2018b-eb
ifort (IFORT) 18.0.3 20180410
with default setting: make ifort
Best Regards!
WEI
The text was updated successfully, but these errors were encountered: