summaryrefslogtreecommitdiff
AgeCommit message (Collapse)AuthorFilesLines
2017-05-28xLAQRx, xTGSEN, IPARMQ: 'integer scalar', 'integer' -> 'INTEGER' in ↵Henning Thielemann19-145/+145
parameter comments
2017-05-28xGGESX, SELCTG parameter description: 'procedure)' -> 'a'Henning Thielemann4-4/+4
2017-05-28SSBGVX: add '*' for multiplication in array dimensions in parameter commentsHenning Thielemann1-2/+2
2017-05-28xLAQR2, xLAQR3, xGEJSV: add parentheses to dimensions of one-dimensional ↵Henning Thielemann8-10/+10
arrays in parameter comments
2017-05-28xBBCSD: dimenison -> dimensionHenning Thielemann4-4/+4
2017-05-24Merge pull request #157 from echeresh/e_lapackelangou40-64/+40
Fixing NaN-checking in LAPACKE interfaces for sub- and super-diagonal parameters
2017-05-24Fix lapacke *stegr, *stein, *bdsdc, *stev, *stevd: e has only (n-1) elementsecheresh16-16/+16
2017-05-24Fix lapacke_*con_3 and lapacke_*tri_3: e stores (n-1) input elements and ↵echeresh12-12/+24
offset depends on uplo
2017-05-24Fix lapacke_*trf_rk and lapacke_*sv_rk: remove nan-checking for e, it is ↵echeresh12-36/+0
output array
2017-05-22Merge pull request #156 from nareshkamboju/masterlangou1-11/+11
lapack_testing.py: print results in one line
2017-05-22lapack_testing.py: print results in one lineNaresh Kamboju1-11/+11
The purpose of the patches is to print lapack results in a single line test-case-name sub-tests-running pass/fail status and count Modified sub test case name to single word Nonsymmetric Eigenvalue Problem to Nonsymmetric-Eigenvalue-Problem Signed-off-by: Naresh Kamboju <naresh.kamboju@linaro.org>
2017-05-22propagate a72750e643027648736f5adb1f89514b1d6e8ae3 to SINGLE PRECISION filesJulien Langou2-4/+4
2017-05-22propagate the INTENT(IN) in the headerJulien Langou1-1/+1
2017-05-22Minor correction. I think this was the intent of the previous commit.Julien Langou1-1/+1
2017-05-22Merge pull request #155 from jeffhammond/fix_isnan_aliasinglangou2-2/+2
use INTENT(IN) to avoid violating Fortran aliasing rules
2017-05-22Merge pull request #140 from SoapGentoo/cmake-fixeslangou7-16/+12
Replace last occurrences of `include` install paths
2017-05-21Merge pull request #154 from echeresh/xlange_lapackelangou4-74/+92
Avoid conversion between layouts in lapacke_?lange_work
2017-05-21Avoid conversion between layouts in lapacke_?lange_workecheresh4-74/+92
2017-05-19use INTENT(IN) to avoid violating Fortran aliasing rulesJeff Hammond2-2/+2
see https://github.com/Reference-LAPACK/lapack/issues/152 for details. fixes issue #152
2017-05-18Fix #147: xlapy2 not propagating nansJulien Langou2-16/+43
xLAPY2 now returns a NaN whenever input variables X or Y are NaNs. The previous xLAPY2 was relying on FORTRAN INTRINSIC MAX and MIN to behave in a certain way with NaNs (i.e. return a NaN whenever X or Y are NaN) and this behavior is not observed on some (most?) compilers. We handle the NaN behavior of xLAPY2 by checking for NaNs at the start of the function. Thanks to Andreas Noack for providing report and sample code to demonstrate the problem.
2017-05-13Merge pull request #149 from JeffreySax/masterlangou11-49/+49
Bug fix in SLAMSWLQ; fix type mismatches.
2017-05-09Bug fix in SLAMSWLQ; fix type mismatches.JeffreySax11-49/+49
- Fix: SLAMSWLQ calls DGEMLQT instead of SGEMLQT. - Use dummy variable for ?LANGE work array argument in [CS]GETSLS. - Use 1-element array for workspace query in ?GETSLS. - Fix PARAMETER declaration type mismatches in ILASLC, IPARAM2STAGE, SGELQT3, ZHECON_3, ZHECON_ROOK, ZHETRI2X.
2017-05-04Merge pull request #144 from tmijieux/lapacke/fix_tpmqrt_work_allocationlangou4-4/+16
Fix issue #143 (Invalid ?tpmqrt workspace allocation)
2017-05-04Fix issue #143 (Invalid ?tpmqrt workspace allocation)Thomas Mijieux4-4/+16
2017-05-02Merge pull request #142 from iyamazaki/masterlangou6-48/+86
fix for issue #141 (how to detect singularity in T).
2017-05-01fix for issue #141 (how to detect singularity in T).iyamazaki6-48/+86
2017-04-15Replace last occurrences of `include` install pathsDavid Seifert7-16/+12
* Distributions might need to change the header dir * Also change the template paths in the `.pc` files to the idiomatic CMake `GNUInstallDirs` full paths, which are always correct, regardless of whether the user specified relative or absolute paths. This makes the build system somewhat easier and more idiomatic.
2017-04-13This is related to issue #42.Julien Langou1-24/+23
This is a fix for double precision. Need to fix S, C and Z now.
2017-04-13Fix #52Julien Langou2-4/+16
Added in comment of xLARRV Note: VU is currently not used by this implmentation of DLARRV, VU is passed to DLARRV because it could be used compute gaps on the right end of the extremal eigenvalues. However, with not much initial accuracy in LAMBDA and VU, the formula can lead to an overestimation of the right gap and thus to inadequately early RQI 'convergence'. This is currently prevented this by forcing a small right gap. And so it turns out that VU is currently not used by this implementation of DLARRV.
2017-04-01Merge pull request #138 from kbergstr/fixeslangou4-20/+20
Fixed formatting (indentation, spaces, style) in some routines
2017-04-01Merge pull request #137 from echeresh/xlarrx_fixlangou24-4476/+4618
Add quick return for *LARR* routines
2017-03-31Add quick return for *LARR* routineseugene.chereshnev24-4476/+4618
2017-03-30Line endings changed from win style to lnx styleKirana Bergstrom2-4472/+4472
2017-03-29Changed indendation for correct alignmentKirana Bergstrom2-16/+16
2017-03-28took out extra spaces, added workspace calculation conditionKirana Bergstrom2-3/+3
2017-03-28took out extra spaceskirana.bergstrom1-1/+1
2017-03-27Merge pull request #136 from echeresh/370_fixlangou55-530/+772
A number of fixes for routines and testing sources integrated in LAPACK 3.7.0
2017-03-27Fix external names and remove unused routines from external statementseugene.chereshnev17-67/+36
2017-03-27Remove redundant IF in TESTING/LIN/*chkrfp.feugene.chereshnev4-20/+0
2017-03-27Fix *chk{sb,hb}2stg.f: corrected name of subroutine in external sectioneugene.chereshnev3-3/+3
2017-03-27Align TESTING/{EIG,LIN}/chkxer.feugene.chereshnev1-20/+2
2017-03-27Align TESTING/{EIG,LIN}/*larhs.feugene.chereshnev6-10/+10
2017-03-24Align TESTING/{LIN,MATGEN}/*lahilb.feugene.chereshnev8-52/+52
2017-03-24Align TESTING/{EIG,LIN}/*get02.feugene.chereshnev4-6/+6
2017-03-24Fix *2st_kernels.f: unify doc commentseugene.chereshnev4-112/+283
2017-03-24Fix *lasyf_aa.f: replace N -> M in commentseugene.chereshnev4-108/+108
2017-03-24Fix *drvls.f: duplicate loop to make work query for each of test caseseugene.chereshnev4-132/+272
Workspace size is not monotonic function of input sizes so there should be loop over all test cases with workspace queries and finding maximum between them.
2017-03-23Merge branch 'master' of https://github.com/Reference-LAPACK/lapackJulien Langou82-834/+894
2017-03-23This is related to #135Julien Langou2-4/+4
Since the order of multiplication in Fortran is left to right, the expression MAXITR*N*N*UNFL first computes MAXITR*N*N as an INTEGER. This causes INTEGER overflow for N>=18919. To avoid the problem, rewrite as: MAXITR*(N*(N*UNFL))
2017-03-23Fix #135Julien Langou2-12/+45
Thanks Cezary Dendek!