Dear LAPACK team,
Currently, LAPACK contains many flavors for driver routines to solve
eigenproblems - for example:
DSYEV{D/R/X}
DSTEV{D/R/X}
DSPEV{D/X}
What I am missing here is a driver "DSPEVR". I can confirm that DSYEVR performs
very well in comparison to DSYEVX - in particular when a BLAS with SMP support
is employed. Now I am dealing with very large matrices which means that saving
memory becomes a major issue. For this reason I was considering to store my
real symmetric matrices switching from full size to packed form. So I was
wondering whether there is a DSPEVR() included in LAPACK ... the only hint I
found is contained in
TESTING/EIG/ddrvst.f? (line 405 and below):
00405 * To be added in 1999
00406 *
00407 * 79= | A - U S U' | / ( |A| n ulp ) DSPEVR('L','V','A', ... )
00408 * 80= | I - U U' | / ( n ulp ) DSPEVR('L','V','A', ... )
00409 * 81= |D(with Z) - D(w/o Z)| / (|D| ulp) DSPEVR('L','N','A', ... )
00410 * 82= | A - U S U' | / ( |A| n ulp ) DSPEVR('L','V','I', ... )
00411 * 83= | I - U U' | / ( n ulp ) DSPEVR('L','V','I', ... )
00412 * 84= |D(with Z) - D(w/o Z)| / (|D| ulp) DSPEVR('L','N','I', ... )
00413 * 85= | A - U S U' | / ( |A| n ulp ) DSPEVR('L','V','V', ... )
00414 * 86= | I - U U' | / ( n ulp ) DSPEVR('L','V','V', ... )
00415 * 87= |D(with Z) - D(w/o Z)| / (|D| ulp) DSPEVR('L','N','V', ... )
Can you tell me whether DSPEVR ever existed in earlier versions of LAPACK? If
not, are there any future plans to include a DSPEVR driver in LAPACK?
Thank you and best regards -
Christian Tuma
?
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.eecs.utk.edu/mailman/private/lapack/attachments/20110117/a4b24248/attachment.html
|