I am using the runtests.csh script to test the compiled code on our cluster. For some of the tests, I get a rather larger deviation from the reference. Before I dig deeper into the issue, would it be possible to confirm that references are correct? For example in the case of the EXAFS/SF6 example (which has a very large deviation), and I see that there was a commit ba2b82a that changed some of the input parameters.
Below is the summary of running the test:
SUMMARY OF TEST RESULTS:
========================
Test Started Finished Crash Avg.Deviation Max.Deviation
----------------------------------------------------------------------------------------------
COMPTON/Cu/ 11/15/21 10:45:16 10:45:49 No 0.002% 0.00%
DANES/BN/ 11/15/21 10:46:21 10:47:21 No 0.001% 0.02%
DANES/Cu/ 11/15/21 10:47:22 10:47:51 No 0.020% 0.04%
DANES/GeCl_4/ 11/15/21 10:47:53 10:48:11 No 0.069% 0.30%
DEBYE/EM/Cu/ 11/15/21 10:48:14 10:49:43 No 2.056% 21.50%
DEBYE/EM/Zn_Tetraimidazole/ 11/15/21 10:49:52 10:50:12 No 0.994% 5.07%
DEBYE/RM/Cu/ 11/15/21 10:50:19 10:51:45 No 2.180% 21.87%
DEBYE/RM/Zn_Tetraimidazole/ 11/15/21 10:51:55 10:52:11 No 0.590% 5.46%
ELNES/Cu/ 11/15/21 10:52:24 10:54:35 No 0.018% 0.04%
EXAFS/Cu/ 11/15/21 10:54:37 10:54:55 No 0.887% 8.52%
EXAFS/Cu_SCF/ 11/15/21 10:55:08 10:55:58 No 0.001% 0.01%
EXAFS/GeCl_4/ 11/15/21 10:56:09 10:56:27 No 0.528% 3.26%
EXAFS/SF6/ 11/15/21 10:56:39 10:56:56 No 11.441% 244.89%
EXAFS/YBCO/ 11/15/21 10:57:09 10:57:27 No 0.000% 0.00%
EXELFS/Cu/ 11/15/21 10:57:33 10:59:03 No 0.027% 0.21%
FPRIME/GeCl4/ 11/15/21 10:59:10 10:59:29 No 1.846% 17.26%
HUBBARD/CeO2/ 11/15/21 10:59:32 12:33:44 No 64.443% 1623.82%
HUBBARD/NiO/ 11/15/21 12:33:46 12:34:36 No 29.448% 233.86%
KSPACE/Cr2GeC/ 11/15/21 12:34:38 12:44:15 No 6.997% 29.28%
KSPACE/Graphite/ 11/15/21 12:44:18 12:48:14 No 1.179% 3.37%
MPSE/Cu/ 11/15/21 12:48:17 12:49:20 No 0.147% 1.34%
MPSE/Cu_OPCONS/ 11/15/21 12:49:23 12:52:25 No 0.002% 0.01%
NRIXS/GeCl_4/ 11/15/21 12:52:27 12:52:46 No 0.421% 1.33%
NRIXS/MgB2/ 11/15/21 12:52:49 12:53:33 No 1.855% 5.65%
XANES/BN/ 11/15/21 12:53:36 12:54:40 No 0.000% 0.00%
XANES/Cu/ 11/15/21 12:54:42 12:55:10 No 0.120% 1.14%
XANES/GeCl_4/ 11/15/21 12:55:13 12:55:31 No 0.184% 0.78%
XES/BN/ 11/15/21 12:55:34 12:56:40 No 0.000% 0.00%
XES/Cu/ 11/15/21 12:56:47 12:57:15 No 0.219% 0.81%
XES/GeCl_4/ 11/15/21 12:57:18 12:57:35 No 0.289% 0.68%
XMCD/Gd_L1/ 11/15/21 12:57:38 12:58:07 No 123.217% 759.12%
XMCD/MnF2_SPXAS/ 11/15/21 12:58:10 12:58:30 No 1121.986% 52139.89%
XNCD/LiIO3/ 11/15/21 12:58:33 12:59:13 No 0.000% 0.00%
I am using the
runtests.cshscript to test the compiled code on our cluster. For some of the tests, I get a rather larger deviation from the reference. Before I dig deeper into the issue, would it be possible to confirm that references are correct? For example in the case of the EXAFS/SF6 example (which has a very large deviation), and I see that there was a commit ba2b82a that changed some of the input parameters.Below is the summary of running the test: