phynexis-deps/lapack
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
3,428 Commits 10 Branches 16 Tags
9aed94de6d78beddb218baaebce809193b8160fc
Commit Graph
3428 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
scr2016 43eb003a74 Moved ABSTOL and RELTOL verification in DLAQP3RK in ELSE clause of IF(I.EQ.1) 2023-11-14 04:42:06 -08:00
scr2016 9c0a21ebd3 changed comments and variable names and assigment for KB, IF, KF in DLAQP3RK 2023-11-14 04:42:06 -08:00
scr2016 7bc3615d79 removed redundant IF(.NOT.DONE.) statement at the end of the DGEQP3RK routine 2023-11-14 04:42:06 -08:00
scr2016 762496f6fc moved zero submatrix check in (if(ioffset ==0) else) clause into else part in DLAQP3RK 2023-11-14 04:42:06 -08:00
scr2016 9de1ee7775 changed indent in in one line in SRC/dlaqp2rk.f 2023-11-14 04:42:06 -08:00
scr2016 d152705578 added zero submatrix check to DLAQP3RK and changed the position of DONE parameter in the calling sequence 2023-11-14 04:42:06 -08:00
scr2016 b943dee7c7 in dlaqp3rk.f added comments TODO 2023-11-14 04:42:06 -08:00
scr2016 46f90f6f87 in dlaqp2rk.f moved ZERO submatrix condition check under if statement, when I!=1 2023-11-14 04:42:06 -08:00
scr2016 3fc469a5dd changes in dlaqp2rk.f IF( IOFFSET.EQ.0 .AND. K.EQ.1 ) THEN into IF( I.QE.1 ) THEN and F( MAXC2NRMK.EQ.ZERO ) THEN into IF( I.NE.1 .AND. MAXC2NRMK.EQ.ZERO ) THEN 2023-11-14 04:42:06 -08:00
scr2016 73bfa95e71 replaced MAXK with KMAX in dgeqp3rk.f 2023-11-14 04:42:06 -08:00
scr2016 8da81bc473 replaced MAXK with KMAX in glaqp3rk.f 2023-11-14 04:42:06 -08:00
scr2016 4c09d8efbf changed comments in SRC/dlaqp2rk.f 2023-11-14 04:42:06 -08:00
scr2016 e965983eb9 Added zero submatrix optimization and changed MAXK into KMAX in dlaqp2k.f 2023-11-14 04:42:06 -08:00
scr2016 f0b12a2193 fixed a bug in dgeqp3rk.f in quick return if possible for the whole original matrix when MAXC2NRM<= ABSTOL and ONE <=RELTOL 2023-11-14 04:42:05 -08:00
scr2016 8d5c716a64 fixe the comment in dgeqp3rk.f and changed design in dlaqp2rk.f to return from the loop instead of exiting the loop when ABSTOL or RELTOL criterion is satisfied 2023-11-14 04:42:05 -08:00
scr2016 f8edd87189 added quick return if possible at the beginning for 1) MAXK = 0; 2) ABSTOL and RELTOL criteria satisfaction for the whole original matrix A 2023-11-14 04:42:05 -08:00
scr2016 1287f67c9e added quick return if possible for MAXK = 0 in dgeqp3rk.f 2023-11-14 04:42:05 -08:00
scr2016 614452f28e changed KMAX into MAXK in dlaqp2rk.f, dgeqp3rk.f 2023-11-14 04:42:05 -08:00
scr2016 f5a907d796 changed comments for JPIV in dgeqp3rk.f dlaqp2rk.f dlaqp3rk.f 2023-11-14 04:42:05 -08:00
scr2016 e8517680f9 changed order of code blocks at the beginning in dgeqp3rk.f and added comments at the end of dlaqp2rk.f about setting MAXC2NRMK, RELMAXC2NRMK and TAU 2023-11-14 04:42:05 -08:00
scr2016 6e9d74860c Changed KP0 to KP1 in dgeqp3rk.f dlaqp2rk.f dlaqp3rk.f 2023-11-14 04:42:05 -08:00
scr2016 dd4c896894 added IMPLCIT NONE to dgeqp3rk.f dlaqp2rk.f dlaqp3rk.f and fixed comments in dlaqp3rk.f 2023-11-14 04:42:05 -08:00
scr2016 ba5afcdadb modified comments in TESTING/LIN/alareq.f TESTING/EIG/alareq.f TESTING/EIG/alarqg.f 2023-11-14 04:42:05 -08:00
scr2016 f238faeafb fixes to implicit declarations 2023-11-14 04:42:05 -08:00
scr2016 706fd1a7f5 added dgeqp3rk.f dlaqp2rk.f dlaqp3rk.f 2023-11-14 04:42:05 -08:00
scr2016 af1bb9e4b4 changed comment in dchkq3 2023-11-14 04:42:05 -08:00
langou ae9c818530 Merge pull request #925 from angsch/master 
Add Anderson's tests of NRM2
 2023-11-12 08:43:44 -07:00
Angelika Schwarz 26df4b3599 Improve testing of NRM2 
Add tests covering arrays with extreme numbers provided in
Anderson E. (2017)
Algorithm 978: Safe Scaling in the Level 1 BLAS
ACM Trans Math Softw 44:1--28
https://doi.org/10.1145/3061665

[D,S]NRM2 coverage improves
lines     59.1% -> 100.0%
branches  30.0% ->  90.0%.

[DZ,SC]NRM2 coverage improves
lines     67.0% -> 100.0%
branches  41.7% ->  91.7%
 2023-11-12 07:44:37 +01:00
Weslley S. Pereira f568a60945 Merge pull request #936 from Reference-LAPACK/DMD_improve_comments 
make the DMD comments in the header Doxygen compliant
 2023-11-11 14:47:37 -07:00
Julien Langou 466401c52b defgroup for DMD in groups-usr.dox 2023-11-11 14:44:58 -07:00
Julien Langou 3ac3bf9fa4 some more improvements to the doxygen documentation 2023-11-11 14:40:20 -07:00
Julien Langou 5d35af84fd add Tennessee, Berkeley, Colorado and NAG 2023-11-11 10:42:03 -07:00
Weslley S. Pereira 2382be4f23 Merge pull request #926 from weslleyspereira/fix-nan-propag-gecon 
GECON returns 1 if RCOND is NaN
 2023-11-11 10:35:32 -07:00
langou 6261d629e1 Merge pull request #930 from christoph-conrads/917-sorcsd2by1-computes-inaccurate-result 
Make vector orthogonalization more reliable
 2023-11-11 09:46:20 -07:00
Julien Langou a383745e50 make the DMD comments in the header Doxygen compliant 2023-11-10 20:36:27 -07:00
Christoph Conrads 64897c4474 xLARFGP: avoid overflows 
Avoid overflows when XNORM << ABS(ALPHA) << 1.

fixes #934
 2023-11-10 20:08:50 +01:00
Weslley S. Pereira d469b91278 Merge pull request #929 from martin-frbg/appv-flang11 
Appveyor CI: Expressly request flang 11.0.1 from conda
 2023-11-10 09:25:50 -07:00
Weslley S. Pereira 43a55bc960 Merge pull request #933 from Reference-LAPACK/issue931_timdavis__incorrect_blas_level_in_dtrsv 
fix #931
 2023-11-10 09:02:14 -07:00
Julien Langou 44e942df6d fix #931 
fix typo for the blas level of BLAS/SRC/DTRSV
 2023-11-10 08:49:41 -07:00
Christoph Conrads 5b0bc5e555 xORBDB5/xUNBDB5: ensure xORBDB6 input has unit norm 
Call sites may run into problems when this subroutine computes nonzero
vectors that are very small in norm.
 2023-11-10 16:29:41 +01:00
Christoph Conrads 7eb4057d6d xORBDB5/xUNBDB5: detect numerically zero vectors 
Do not call xORBDB6/xUNBDB6 with input vectors x that are numerically
zero because this can cause problems at the call sites (xLARFGP
computations might underflow) leading to, e.g., the computation of
nonunitary matrices in the xORCSD2BY1/xUNCSD2BY1 output.
 2023-11-09 17:14:58 +01:00
Christoph Conrads 80d7c69a52 xORBDB5/xUNBDB5: introduce real zero parameter 
* introduce a parameter representing a real zero for consistency
* stop comparing real values with complex zero
 2023-11-09 17:14:58 +01:00
Christoph Conrads ad3d9b8fc4 xORBDB6/xUNBDB6: use safe termination criterion 
With a modest amount of random testing on a work station, it is possible
to detect the loss of orthogonality caused by the old early termination
criterion.
 2023-11-09 17:14:58 +01:00
Christoph Conrads bc2b374725 xORBDB6/xUNBDB6: do not require unit-norm vector 
Do not assume that the vector x passed to these subroutines has
Euclidean norm one (the norm is almost surely smaller than one when
xORBDB5/xUNBDB5 calls this subroutine for the first time).

This commit fixes the occasional inaccurate results computed by
xORCSD2BY1/xUNCSD2BY1.
 2023-11-09 17:14:58 +01:00
Weslley S Pereira 68eafc69a8 Makes NaN an illegal value for ANORM on input of GECON. Let NaNs propagate in the computation of RCOND 2023-11-09 09:10:49 -07:00
Martin Kroeker eaa105b0e0 Expressly request flang 11.0.1 from conda 2023-11-08 23:35:08 +01:00
langou 8ecaaf9869 Merge pull request #928 from christoph-conrads/fix-critical-xUNBDB6-typo 
xORBDB6/xUNBDB6: fix a constant
 2023-11-06 13:47:02 -07:00
Christoph Conrads e0370cabce xORBDB6/xUNBDB6: fix a constant 
The commits listed below updated xORBDB6/xUNBDB6 with the aim of
improving numerical stability by avoiding the use of squared values.
With the exception of SORBDB6 these commits did not properly up a magic
constant called `ALPHA` in the code. SORBDB6 was initially updated
correctly before commit a015b21445
inserted an incorrect value again.

The incorrect magic constant may lead to early returns with only
partially orthogonalized output vectors.

Relevant commits:
* 54b3964b0f
* 6479e0f531
* 94419e8bfa
 2023-11-06 19:25:22 +01:00
Weslley S. Pereira ada20a01db Merge pull request #927 from pleshkovvli/fix-complex-bdsqr-maxit 
Fix complex BDSQR maxit
 2023-11-03 08:36:29 -06:00
Vladimir Pleshkov da7f21005e Fix complex BDSQR maxit 2023-11-03 10:29:44 +07:00