This is a report showing runs performed recently for the CCM-3 (Parallel Tight-Binding Molecular Dyanamic) CHSSI program:
Developer Date Time CPU Program Comments
Michael Mehl 2 June 1998 1100 Eastern Code 6690 ½ Origin (Patterson) static Ran static manual example 1 using serial, 2, 4, 6, and 8 processors. Results agreed with Absoft serial and IBM RS6000 serial to output precision in the SKENG file.
Michael Mehl 2 June 1998 1100 Eastern Code 6690 ½ Origin (Patterson) static Ran static manual example 7 using serial, 2, 4, 6, and 8 processors. Results agreed with Absoft serial and IBM RS6000 serial to output precision in the SKENG and QLMT files.
Michael Mehl 2 June 1998 1130 Eastern ARL SGI Origin static Test of static manual example1 Ran on serial, 2, 4, and 8 processors. All equivalent to previous runs to SKENG precision. Directories /home/navy/mehl/tbpar/static/example1/eckert.* where * = serial, par2, par4, par8
Michael Mehl 2 June 1998 1151 Eastern ARL SGI Origin static Test of static manual example3 Ran on serial, 2, 4, and 8 processors. All equivalent to previous runs to SKENG precision. Directories /home/navy/mehl/tbpar/static/example3/eckert.* where * = serial, par2, par4, par8
Michael Mehl 2 June 1998 1156 Eastern ARL SGI Origin static Test of static manual example7 Ran on serial, 2, 4, and 8 processors. All equivalent to previous runs to QLMT precision. Directories /home/navy/mehl/tbpar/static/example7/eckert.* where * = serial, par2, par4, par8
Michael Mehl 9 June 1998 1345 Eastern ARL SGI Origin static Test example I on SGI again. Works fine.
Michael Mehl 18 June 1998 1330 Eastern Linux Serial Computer static Preliminary setup of a script to minimize two components of a lattice. Specifically, this was for C44 in hcp Titanium. The results will eventually be part of the static manual (Example 10).
Michael Mehl 19 June 1998 1300 Eastern Linux Serial Computer static Final test of a script to minimize two components of a lattice. Specifically, this was for C44 in hcp Titanium. The results will eventually be part of the static manual (Example 10). Many calculations (about 60) are included in this step.
Michael Mehl 19 June 1998 1755 Eastern Linux Serial Computer fit New fit for Cu, based on PW91 LDA/LAPW data for fcc, bcc, and sc. Will eventually be used as part of a new, improved CuAu fit. This was a preliminary setup run.
Michael Mehl 19 June 1998 2300 Eastern Linux Serial Computer fit Continuation of the new fit for Cu, based on PW91 LDA/LAPW data for fcc, bcc, and sc. First try at fit -- reasonably good, except at large volume end for fcc and bcc.
Michael Mehl 22 June 1998 1242 Eastern Code 6690 ½ Origin (Patterson) fit Continuation of Cu fit, this time on Patterson (SGI Origin). Used symmetry at Gamma point only. Fit seems to work, but still has a rather large error at large volumes.
Michael Mehl 22 June 1998 1355 Eastern Code 6690 ½ Origin (Patterson) fit Continuation of Cu fit, using higher weights for fcc, bcc, and sc energies as compared to band structures and the "atomic" energy.. Still has a rather large error at large volumes.
Michael Mehl 22 June 1998 1429 Eastern Code 6690 ½ Origin (Patterson) fit Continuation of Cu fit. Use weight of 5000 for fcc and bcc energies, 500 for atomic an sc energies. Still has a rather large error at large volumes.
Michael Mehl 22 June 1998 1510 Eastern Code 6690 ½ Origin (Patterson) fit Tried to get a better fit by lowering lambda, freezing it and letting the onsite terms adjust. Not too successful, but will use for the next run.
Michael Mehl 22 June 1998 1545 Eastern Code 6690 ½ Origin (Patterson) fit Fixed lambda from the previous run, allowed everything to change. Not much movement.
Michael Mehl 22 June 1998 1655 Eastern Linux Serial Computer fit Try alternate run: start from scratch fitting routine. First fit only constant on-site terms. (Of course it is a bad fit right now.)
Michael Mehl 22 June 1998 1655 Eastern Linux Serial Computer fit Continue alternate alternate run: Using 500/5000 weights. Fit all onsite terms (eg=t2g). Got F=5.98 (Still an obviously bad fit.) Best 500/5000 fit is currently F=.139, before we started fooling with lambda.
Michael Mehl 22 June 1998 1913 Eastern Code 6690 ½ Origin (Patterson) fit After fixed lambda run, freed lambda and everything else. Get lambda = 1.345, F = 4.084, and rms energy error = 0.0066 For comparison, in best run lambda 1.44, F=.139 rms=.0013.