| #
7b5fd022
|
| 28-Oct-2025 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'jolivet/housekeeping' into 'main'
Minor housekeeping
See merge request petsc/petsc!8811
|
| #
ab701a13
|
| 27-Oct-2025 |
Pierre Jolivet <pierre@joliv.et> |
checkbadSource: check style of config/*
|
| #
7b2fcb5d
|
| 28-Jul-2025 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'jolivet/proper-capitalization' into 'main'
configure: fix capitalization of multiple packages and PETSc itself
See merge request petsc/petsc!8597
|
| #
f605775f
|
| 26-Jul-2025 |
Pierre Jolivet <pierre@joliv.et> |
configure: fix capitalization of multiple packages and PETSc itself
|
| #
0d821f60
|
| 30-Apr-2025 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'release'
|
| #
d7daa58d
|
| 29-Apr-2025 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'jolivet/fix-gcc15' into 'release'
Fix for GCC 15 switching from -std=gnu17 to -std=gnu23
See merge request petsc/petsc!8352
|
| #
145e2b68
|
| 28-Apr-2025 |
Pierre Jolivet <pierre@joliv.et> |
Fix for GCC 15 switching from -std=gnu17 to -std=gnu23
|
| #
f0cb6cbe
|
| 26-Sep-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'balay/migrate-ftp' into 'main'
docs, scripts: update from ftp.mcs.anl.gov to web.cels.anl.gov
See merge request petsc/petsc!6889
|
| #
30c3248d
|
| 25-Sep-2023 |
Satish Balay <balay@mcs.anl.gov> |
externalpackages: fix URLs from ftp.mcs.anl.gov to web.cels.anl.gov
|
| #
8a85f5a8
|
| 16-May-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge remote-tracking branch 'origin/release'
|
| #
6af05d5c
|
| 16-May-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'tisaac/fix-blis-for-f2cblaslapack' into 'release'
Configure blis to be compatible with f2cblaslapack in complex arithmetic
See merge request petsc/petsc!6460
|
| #
cee9ab54
|
| 12-May-2023 |
Toby Isaac <toby.isaac@gmail.com> |
Configure blis to be compatible with f2cblaslapack in complex arithmetic
Apparently compilers differ in how they handle complex number return types from fortran: some return on the stack (GNU) other
Configure blis to be compatible with f2cblaslapack in complex arithmetic
Apparently compilers differ in how they handle complex number return types from fortran: some return on the stack (GNU) others insert an additional argument (Intel). f2cblaslapack is fortran that has been translated into C using the latter style. But this means that if f2cblaslapack is going to call blis, blis has to make sure its blas compatibility layer matches. If we don't tell blis which format to use, it guesses using the fortran compiler that it knows about, and it might guess wrong. So we add a flag to blis configuration to control the complex return type, and we have f2cblaslapack set the default value of that flag when --download-f2cblaslapack is used.
Without this change, 'mat_tests-ex120_*' tests are broken if blis returns complex numbers on the stack. We haven't detected this yet because we do not have a blis/f2cblaslapack/complex arch in CI.
show more ...
|
| #
dab44c92
|
| 04-Jan-2023 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'jolivet/fix-f2cblaslapack-valgrind' into 'main'
f2cblaslapack: fix Valgrind error
See merge request petsc/petsc!5949
|
| #
05e09756
|
| 01-Jan-2023 |
Pierre Jolivet <pierre@joliv.et> |
f2cblaslapack: fix Valgrind error
Conditional jump or move depends on uninitialised value(s) at 0x1895D4: ilaenv_ (ilaenv.c:427) by 0x8AE1D3F: zsytrf_ (zsytrf.c:268) by 0x5CBCB8F: MatCholes
f2cblaslapack: fix Valgrind error
Conditional jump or move depends on uninitialised value(s) at 0x1895D4: ilaenv_ (ilaenv.c:427) by 0x8AE1D3F: zsytrf_ (zsytrf.c:268) by 0x5CBCB8F: MatCholeskyFactor_SeqDense (dense.c:918) by 0x5CBE11F: MatCholeskyFactorNumeric_SeqDense (dense.c:950) by 0x64CCB4B: MatCholeskyFactorNumeric (matrix.c:3390) by 0x7AC29B7: PCSetUp_Cholesky (cholesky.c:121) by 0x7F8FA2F: PCSetUp (precon.c:994) by 0x785EBD3: KSPSetUp (itfunc.c:405) by 0x168F46FB: STSetUp_Sinvert (sinvert.c:116) by 0x1696AB5F: STSetUp (stsolve.c:524) by 0x16DADF67: EPSSetUp (epssetup.c:381) by 0x16DBBB3F: EPSSolve (epssolve.c:133) Uninitialised value was created by a stack allocation at 0x188C50: ilaenv_ (ilaenv.c:194)
show more ...
|
| #
d3f926d8
|
| 23-Oct-2022 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'barry/2022-09-03/add-emcc' into 'main'
Get PETSc building and running with emcc
See merge request petsc/petsc!5588
|
| #
56eb276f
|
| 08-Oct-2022 |
Barry Smith <bsmith@mcs.anl.gov> |
Fixes for toclapack.sh
Fix f2c generation of xerbla to be void not int, since used as void in dgeev.c Remove a variety of lapack routines that cannot work because they depend on F90 routine Slight c
Fixes for toclapack.sh
Fix f2c generation of xerbla to be void not int, since used as void in dgeev.c Remove a variety of lapack routines that cannot work because they depend on F90 routine Slight cleanup of toclapack.sh script
Commit-type: bug-fix, portability-fix, external package /spend 10m
show more ...
|
| #
9f7a4602
|
| 16-Sep-2022 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'barry/2022-09-08/update-f2cblaslapack' into 'main'
update toclapack.sh
See merge request petsc/petsc!5611
|
| #
b50f998a
|
| 15-Sep-2022 |
Jose E. Roman <jroman@dsic.upv.es> |
Upgrade f2cblaslapack to LAPACK version 3.8.0
- Recent versions of LAPACK have replaced DIMAG by AIMAG and this causes errors in f2c. The fix is to revert this change in all precisions but singl
Upgrade f2cblaslapack to LAPACK version 3.8.0
- Recent versions of LAPACK have replaced DIMAG by AIMAG and this causes errors in f2c. The fix is to revert this change in all precisions but single. - Some of the new routines introduced in recent LAPACK use the colon notation for index ranges, which is not supported by f2c. As a workaround we omit them by adding them to the black list. They are: * _gejsv: preconditioned Jacobi SVD. Usage of index ranges is trivial and could be fixed easily. * _bdsvdx: SVD of bidiagonal matrix. Extensive use of index ranges. It is called by the driver routine _gesvdx, hence it is removed also.
The current toclapack.sh script cannot be used for lapack-3.9.0 and later due to additional issues
show more ...
|
| #
364b7219
|
| 09-Sep-2022 |
Barry Smith <bsmith@mcs.anl.gov> |
Changed f2clapack.py to generate functions that return void instead of the previous meaningless int
Add some support to toclapack.sh for later Lapack releases, but it still cannot be used with versi
Changed f2clapack.py to generate functions that return void instead of the previous meaningless int
Add some support to toclapack.sh for later Lapack releases, but it still cannot be used with versions after 3.4.2
They introduced a lot of new code that is apparently post-f77 but that can be possibly converted back to f77 with the changes to toclapack.sh
But they also changed a bunch of routines, like dnrm2.f90 that use much more of 90 constructs and cannot be converted back to f77. It might be possible to simply replace these with older versions of the f2c codes, but I am not going to do that now.
Commit-type: external-package /spend 7h
All the time wasted trying to update to later Lapack releases
show more ...
|
| #
5f2abdd3
|
| 25-Aug-2022 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'jolivet/feature-ksphpddm-fp16-ci' into 'main'
Partial support for __fp16 in KSPHPDDM
See merge request petsc/petsc!5543
|
| #
4c92ab2f
|
| 21-Aug-2022 |
Pierre Jolivet <pierre@joliv.et> |
configure: use qhlib rule in f2cblaslapack.py
|
| #
9e517322
|
| 21-Aug-2022 |
Pierre Jolivet <pierre@joliv.et> |
Sys: expose MPIU___FP16 to users
|
| #
f646d225
|
| 18-Aug-2022 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'jolivet/feature-expose-mpiu___float128' into 'main'
Sys: expose MPIU_SUM___FLOAT128 to users
See merge request petsc/petsc!5536
|
| #
d6d99c17
|
| 17-Aug-2022 |
Pierre Jolivet <pierre@joliv.et> |
configure: add --with-f2cblaslapack-[fp16,float128]-bindings
|
| #
76b1de3d
|
| 01-May-2022 |
Satish Balay <balay@mcs.anl.gov> |
Merge branch 'jolivet/housekeeping' into 'main'
General housekeeping
See merge request petsc/petsc!5192
|