History of LAPACK releases Version 1.0 : February 29, 1992 Revised, Version 1.0a: June 30, 1992 Revised, Version 1.0b: October 31, 1992 Revised, Version 1.1: March 31, 1993 Version 2.0: September 30, 1994 Version 3.0: June 30, 1999 Update, Version 3.0: October 31, 1999 Update, Version 3.0: May 31, 2000
To better address the needs of our LAPACK users, we're in the process of creating this RELEASE_NOTES webpage. Users will be able to download updates/patches to LAPACK in a variety of ways (Unix patch, tar-gzip file), and will no longer need to wait for an update/release, or make the changes by-hand.
This file contains:
LAPACK has been tested on a variety of architectures. However, keep in mind that we do not have access to every possible type of machine, and the regrettable bug report will appear. We apologize for any inconvenience this may cause to our users, and strongly encourage users to contact us if you suspect an abnormality. Please contribute "machine-specific installation.hints.gz" for architectures that we do not have listed. These "hints" are of great benefit to many users.
LAPACK test suite failure when linking with ATLAS: The LAPACK test suite assumes that diagonal elements are not inverted, and then tests only the 'Left' case of xTRSM, since the 'Right' case of the reference implementation of xTRSM inverts the diagonal elements. In the implementation of xTRSM in ATLAS, inversion is used for both cases. Thus, the LAPACK test suite (./xlintsts < stest.in ) will fail when linking with ATLAS. The LAPACK test suite needs to be modified to allow for the inversion of diagonal elements, as ATLAS's implementation of xTRSM is valid.
Errata in LAPACK Users' Guide, Third Edition
Machine-specific Installation Hints
The following bug fixes have NOT yet been incorporated in a release/update of LAPACK.
UPDATES (Version 3.0) October 31, 1999 and May 31, 2000
The modifications listed below are included in the file: http://www.netlib.org/lapack/update.tgz as well as the http://www.netlib.org/lapack/lapack.tgz.Note: Unless otherwise stated, changes to single precision routines (names beginning with S or C) apply also to the corresponding double precision routines (names beginning with D and Z).
To view/download individual updated files from this list, please consult
the individual files on netlib.
http://www.netlib.org/lapack/single
http://www.netlib.org/lapack/double
http://www.netlib.org/lapack/complex
http://www.netlib.org/lapack/complex16
------------- ------- --------------------------
DIRECTORY ROUTINE DESCRIPTION OF CHANGE
------------- ------- --------------------------
LAPACK/ Makefile Added ./ for executables
LAPACK/
BLAS/TESTING/ cblat3.f Missing CMPLX intrinsic in assignment statement
Lines 1995-1996
ALPHA = ( ONE, -ONE )
BETA = ( TWO, -TWO )
should be
ALPHA = CMPLX( ONE, -ONE )
BETA = CMPLX( TWO, -TWO )
SRC/
Makefile Move slarrb.o, slarre.o, and slarrf.o to SCLAUX
Move dlarrb.o, dlarre.o, and dlarrf.o to DZLAUX
cheevr.f Line 310 should instead be
RWORK(1) = LRWMIN
chetd2.f Error in zero'ing of imaginary parts of diagonal
elements
chbgvd.f Incorrect workspace calculation for LLWK2 and
LLRWK when N=1.
chpgvd.f Missing LQUERY return
cstegr.f Replace SSTEGR->CSTEGR in call to XERBLA
sbdsdc.f Workspace correction
Line 113, 2*N --> 4*N
Documentation correction IWORK dimension
sbdsqr.f/ Workspace correction
cbdsqr.f Line 99, 2*N --> 4*N
sgeev.f Corrected error with LQUERY and WORK(1)
sgelsd.f/ Leading comment correction for A and LWORK and
cgelsd.f workspace correction
sgelss.f/ Workspace correction
Line 178, 5*N-4 --> 5*N
Line 194, 5*M-4 --> 5*M
cgelss.f Workspace correction
Line 95, 5*MIN(M,N)-1 --> 5*MIN(M,N)
Line 174, 5*N-1 --> 5*N
Line 204, 5*M-1 --> 5*M
Line 224, 5*M-1 --> 5*M
sgesdd.f/ Possible MAXWRK uninitialized and
cgesdd.f and LWORK=-1 query insertion
sgesvd.f/ Workspace correction in comments
Line 118, 5*MIN(M,N)-4 --> 5*MIN(M,N)
Line 217, MAX(3*N,5*N-4) --> 5*N
Line 390, MAX(3*M,5*M-4) --> 5*M
cgesvd.f Workspace correction in comments
Line 123, MAX(3*MIN(M,N),5*MIN(M,N)-4) --> 5*MIN(M,N)
Line 224, MAX(3*N,5*N-4) --> 5*N
Line 386, MAX(3*M,5*M-4) --> 5*M
sgtsv.f Leading comment correction for B
slaed2.f Error in routine name passed to XERBLA
slaed4.f Corrected error in rootfinding
slals0.f/ Out-of-range array reference
clals0.f
slalsd.f/ Leading comment correction for LWORK and a
clalsd.f call to SLASRT was added
slarzb.f/ Out-of-range array reference
clarzb.f
slasd2.f Fix out-of-range memory reference when N=0
slasd3.f Fix out-of-range memory reference when N=0
slasd4.f Cray compiler workaround on lines 596-597 and 607-608
slasda.f Leading comment correction for size of WORK
slasdq.f Workspace correction
Line 115, max(1,4*N-4) --> 4*N
and incorrect routine name passed to XERBLA
slasq1.f Workspace correction
Line 53, 2*N --> 4*N
and modifications for IEEE
slasq2.f Fix out-of-range error for N=1 and modifications
for IEEE
slasq3.f Modifications for IEEE and index correction
I0-->N0 on line 193
slasq4.f Modifications for IEEE
slasq5.f Modifications for IEEE
slasq6.f Modifications for IEEE
sormrz.f Missing LQUERY return
spteqr.f/ Workspace correction
cpteqr.f Line 78, 2*N --> 4*N
sstevr.f Documentation correction and call to SSTEGR
ssyevr.f/ Documentation correction and call to SSTEGR
cheevr.f
TESTING/
Makefile Added ./ for executables
sgd.in/ Fixed NSIZE line for xGX
cgd.in Replace line 18 with
5 Largest matrix dimension (0 <= NSIZE <= 5)
TESTING/MATGEN/
slatm6.f Modification for SGESVD workspace change
TESTING/LIN/
schkgb.f/ Missing integer item in 9996 FORMAT statement
cchkgb.f
schkge.f/ Correction to FORMAT statements
cchkge.f
schkpb.f/ Removal of XTYPE='C' line
cchkpb.f
schkpo.f/ Removal of XTYPE='C' line
cchkpo.f
schkpp.f/ Removal of XTYPE='C' line
cchkpp.f
schksp.f/ Removal of XTYPE='C' line
cchksp.f
schksy.f/ Removal of XTYPE='C' line
cchksy.f
schktb.f/ Correction to FORMAT statements
cchktb.f
schktp.f/ Correction to FORMAT statements
cchktp.f
schktr.f/ Correction to FORMAT 9999 and 9998 statements
cchktr.f
sdrvls.f/ Correction to argument passed to INTRINSIC and
cdrvls.f remove a call to SLASRT in TEST 15;
Correction to leading comments (IWORK)
TESTING/EIG/
schkhs.f Fixed out-of-range error for N=0
sdrgsx.f/ BDSPAC correction for xGESVD workspace and
cdrgsx.f initialize IFUNC = 0
sdrvbd.f/ IWTMP correction for xGESVD workspace
cdrvbd.f
sdrvst.f/ Out-of-range array reference.
cdrvst.f
cerrbd.f Correction to summary line
serred.f/ Correction to summary line
cerred.f
TIMING/
Makefile Erroneous sgeptim.in instead of SGEPTIM.in on
line 253
TIMING/LIN/
stimls.f/ Correction to size of LWORK for xGELSD
ctimls.f
stimq3.f/ Correction Q3->QP before ATIMIN call
ctimq3.f
TIMING/LIN/LINSRC/
sgelsd.f/ Same correction as in SRC
cgelsd.f
sgelss.f/ Same correction as in SRC
cgelss.f
slals0.f/ Same correction as in SRC
clals0.f
slalsd.f/ Same correction as in SRC
clalsd.f
TIMING/EIG/
stim22.f/ Disable timing of SSTEGR if non-IEEE-754 compliant
ctim22.f
TIMING/EIG/EIGSRC/
sbdsdc.f Same correction as in SRC
sbdsqr.f/ Same correction as in SRC
cbdsqr.f
sgesdd.f/ Same correction as in SRC
cgesdd.f
slaed2.f Same correction as in SRC
slasd2.f Same correction as in SRC
slasd3.f Same correction as in SRC
slasd4.f Same correction as in SRC
slasdq.f Same correction as in SRC
slasq1.f Same correction as in SRC
slasq2.f Same correction as in SRC
slasq3.f Same correction as in SRC
slasq4.f Same correction as in SRC
slasq5.f Same correction as in SRC
slasq6.f Same correction as in SRC
spteqr.f/ Same correction as in SRC
cpteqr.f