Comparison of Toric results from transp with the latest stand alone Toric showed a significant reduction in the power deposited. Further investigations show a similar discrepancy, using the test datasets distributed with the code.
Toric 4.2 forms part of the current transp distribution. It may be run as part of the main transp executable, or as a stand alone version ( $XE/toric_main ). The version used here was from the transp build of 28-Sep-2007
Toric 5 was checked out from the svn repository on 16-Oct-2007
Both codes were built using Lahey/Fujitsu Fortran 95 Express Release L6.20d, gcc version 4.1.1 20070105, running on Fedora Core release 6 (Zod) workstations at JET, and linked to the same set of external ( BLAS, LAPACK etc ) libraries.
Two test runs supplied with Toric5 have been used - tests/Test1 & tests/Test1a -
"Test1 - Asdex Case
Mode Conversion on high field side.Low res, can be run serially in a few minutes.
nelm = 450, ntt = 64, For quick comparison to parallel runs. "Some complication results from a change in the plasma profile ordering between versions 4.2 & 5. tests/Test1/profnt_19314a is stored in the Toric 4.2 ordering ( Ni_2, Ti_i ), while ./Test1a/profnt_19314 is stored in the Toric 5 order - ( Ni_1, Ti_1, Ni_2 ). For the 'crossed version' runs ( test1/T5.0, test1a/T4.2 ) the appropriate profnt file was substituted for the one originally supplied in the tests subdirectory.
Print statements have been added to t4_mod_profnt, to verify that the correct distributions have been read - see logfiles ( below ).
In addition, the antenna length in Test1/torica.inp is 50cm, while the Test1a value is 90cm. The other namelist parameters used in these tests are unchanged.
Toric 4.2 reproduced the Test1 output fairly closely, while Toric5 reproduced the Test1a output exactly. The division of RF power between processes is consistent for all runs for the same test, but the Total Absorbed Power ( & all other power depositions ) for Toric5 are about 12% of the Toric 4.2 values.
Total Absorbed Power Sample OP Toric 4.2 Toric 5.0 T5/T4.2 Test1 1.2188E+00 1.1839E+00 1.3623E-01 0.1150 Test1a 3.0517E-01 2.3227E+00 3.0517E-01 0.1314 Total absorbed power ( MW/kA**2 )
Given the agreement between Toric 4.2 -Test1 sample output, & Toric5.0-Test1a, it seems likely that the codes which produced the sample outputs also show the 90% power loss between versions 4.2 & 5. Some change in the overall normalization between the versions would explain why the absolute power values are changed, while the relative values are not.Alternatively, it may be that some input parameter or distribution should be changed for Toric 5 ( which outputs SI quantities ), but has not been .
Test1 Sample, Toric 4.2, Toric 5 ; Field on axis ( plots ) Toric 4.2 , Toric 5
Test1a Sample, Toric 4.2, Toric 5
T4.0 / T5.3 comparison >>