NBLK: Difference between revisions
No edit summary |
No edit summary |
||
(One intermediate revision by the same user not shown) | |||
Line 23: | Line 23: | ||
{{sc|NBLK|Examples|Examples that use this tag}} | {{sc|NBLK|Examples|Examples that use this tag}} | ||
---- | ---- | ||
[[Category:INCAR]][[Category:Performance]] | [[Category:INCAR tag]][[Category:Performance]] |
Latest revision as of 14:47, 8 April 2022
NBLK = [integer]
Default: NBLK | = -1 | VASP.4.6 |
= 256 | in VASP.5.2, if dfast |
Description: NBLK determines the blocking factor in many BLAS level 3 routines.
In some cases, VASP has to perform a unitary transformation of the current orbitals. This is done using a work array CBLOCK and the following FORTRAN code:
DO 100 IBLOCK=0,NPL-1,NBLK ILEN=MIN(NBLK,NPL-IBLOCK) DO 200 N1=1,N DO 200 M=1,ILEN CBLOCK(M,N1)=C(M+IBLOCK,N1) C(M+IBLOCK,N1)=0 200 CONTINUE C C(IBLOCK+I,N)=SUM_(J,K) CH(I,K) CBLOCK(K,N) CALL ZGEMM ('N', 'N', ILEN, N, N, (1.,0.), CBLOCK, NBLK, CH, N, & (1.,0.), C(IBLOCK+1,1), NDIM) 100 CONTINUE
ZGEMM is the matrix matrix multiplication command of the BLAS package. The task performed by this call is indicated by the comment line written above the ZGEMM call. Generally NBLK=16 or 32 is large enough for super-scalar machines. A large value might be necessary on vector machines for optimal performance (NBLK=128).