User Tools

Site Tools


readme.pcg2

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
Next revisionBoth sides next revision
readme.pcg2 [2016/10/21 22:17] – [Options] shogoreadme.pcg2 [2020/07/09 19:54] – [Summary] shogo
Line 5: Line 5:
 \\ \\
 BLUP90IOD2 is a BLUP program for solving a large number of mixed model equations, using preconditioned conjugate gradient (PCG). Another version for heterogeneous residual variances is also available. BLUP90IOD2 is a BLUP program for solving a large number of mixed model equations, using preconditioned conjugate gradient (PCG). Another version for heterogeneous residual variances is also available.
 +\\
 +BLUP90IOD2 has 2 additional versions: BLUP90IOD2OMP1 with OpenMP and BLUP90IOD2MPI with MPI. The OpenMP version is faster running with multiple cores. The MPI version has the same computing speed as the OpenMP version with less memory when we use SSD.
 \\ \\
 See PREGSF90 with genotypes (SNP) for options. See PREGSF90 with genotypes (SNP) for options.
Line 14: Line 16:
 OPTION conv_crit 1e-12 OPTION conv_crit 1e-12
 </file> </file>
-Set convergence criteria (deault 1e-12).+Set convergence criterion (default 1e-12). Usually, a stricter criterion (e.g., 1e-14 to 1e-16) is needed with a single-trait model than a multiple-start model.
 <file> <file>
 OPTION maxrounds 10000 OPTION maxrounds 10000
 </file> </file>
-Set maximum number of rounds (default 1000).+Set maximum number of rounds (default 5000).
 <file> <file>
 OPTION blksize 3 OPTION blksize 3
 </file> </file>
-Set block size for preconditioner (default 1).+Set block size for preconditioner (default 1) to accelerate convergence (usually 2 to 5 times faster). For a multiple-trait model, the number can be the number of traits.
 <file> <file>
 OPTION init_eq 10 OPTION init_eq 10
 </file> </file>
-Set the number of effects to be solved directly (default=0).+Set the number of effects to be solved directly (default=0) to accelerate convergence (usually, 10% to 50% faster).
 <file> <file>
 OPTION solv_method FSPAK OPTION solv_method FSPAK
Line 34: Line 36:
 OPTION tol 1d-12 OPTION tol 1d-12
 </file> </file>
-Tolerance to get a positive definite matrix (default=1d-12).+Tolerance to get a positive definite covariance component matrix (default=1d-12).
 <file> <file>
 OPTION residual OPTION residual
Line 42: Line 44:
 OPTION avgeps 50 OPTION avgeps 50
 </file> </file>
-Using the last 50 average eps for convergence.+Using the last 50 average eps for convergence, depending on the fluctuation of convergence. When using this option, the convergence criterion can be loosen (e.g., 1e-12 -> 1e-10, 1e-14 -> 1e-12), also depending on how convergence fluctuates.
 <file> <file>
 OPTION cont 1 OPTION cont 1
 </file> </file>
-To restart the program from the previous solutions.+To restart the program from the last solutions.
 <file> <file>
 OPTION missing -999 OPTION missing -999
Line 69: Line 71:
 \\ \\
 //For information on availability of BLUP90IOD, please contact Ignacy Misztal (ignacy@uga.edu) at the University of Georgia.// //For information on availability of BLUP90IOD, please contact Ignacy Misztal (ignacy@uga.edu) at the University of Georgia.//
- 
readme.pcg2.txt · Last modified: 2024/03/25 18:22 by 127.0.0.1

Donate Powered by PHP Valid HTML5 Valid CSS Driven by DokuWiki