| ec1065ed | 05-Oct-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
commit after merge
Hg-commit: e4ceadb2454090be7d46b3718e609756791804d0 |
| 8e80dd90 | 28-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
turn on PETSC_USE_BACKWARD_LOOP except on BlueGene
Hg-commit: d30d7ca3ce0043adca70a94eb85508a992cb05b5 |
| 4fd072db | 28-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
added MAT_SHARE_NONZERO_PATTERN for SeqBAIJ matrices in MatDuplicate()
added a couple missing PetscLogObjectMemory() in SeqBAIJ
Hg-commit: 3fbba1b9cd2771828452e6c3179c885256f18b67 |
| 783ef271 | 25-Sep-2009 |
Hong Zhang <hzhang@mcs.anl.gov> |
add icc(0) with matorderings under new data structure
Hg-commit: 5b99ce0799b371a259538425a489fed8c02ef64a |
| c760cd28 | 25-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
finished MAT_SHARED_NONZERO_PATTERN for MatDuplicate_SeqSBAIJ()
Hg-commit: 6b08b2ef797929b991322cb1faecd54bba1dc198 |
| 4da8f245 | 25-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
basic support of MAT_SHARE_NONZERO_PATTERN for SBAIJ matrix to share i,j, and jshort arrays
Hg-commit: bfb302adeda62ffaa7f0c715d88e9683bab6b787 |
| 9f892c73 | 24-Sep-2009 |
Hong Zhang <hzhang@mcs.anl.gov> |
cleanup MatSolve_SeqSBAIJ_1_NaturalOrdering_newdatastruct()
Hg-commit: 094a2f0b282a30d5056c3ae5be34f0bd7efbd127 |
| 5f44c854 | 24-Sep-2009 |
Hong Zhang <hzhang@mcs.anl.gov> |
support icc(0) with new datastructure
Hg-commit: 167d592268ba59b7046ee8a733239adc59cae028 |
| b5b17502 | 22-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
more work on MatRelax_SeqSBAIJ()
Hg-commit: 01929b8375460f52e65dbdba6d585594e066aa37 |
| 70dcbbb9 | 22-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
combined MatMult/Relax_SeqSBAIJ_1 with unsigned short version so no duplicate code we forget to fix
Hg-commit: da11a588fc5c37c4b5b7bc6428663222433174fc |
| 46c5371b | 11-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
fixed typos in update to MatRelax_SeqSBAIJ()
Hg-commit: 30497bbc5acc0b902dea4fa5526e23e51852cc0c |
| c0774073 | 11-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
MatRelax_SeqSBAIJ() backward solve with no forward solve now moves backwards through each row. This gives 10% faster Eisenstat on my Laptop
Hg-commit: 500e96e99903dac047c2360d5dd56a02ca4df47d |
| df826632 | 08-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
minor improvements/code modifications to MatRelax_SeqSBAIJ()
Hg-commit: 9de31ba8922331ba61bee731740c3ae28ff76b36 |
| 887ee2ca | 01-Sep-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
fixed Eisentat preconditioner for AIJ and SBAIJ to work for all omega, instead of just omega== 1
Hg-commit: d40d325f2fcbde49d908cb600d0aae362304a789 |
| a2b30743 | 20-Aug-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
added support for nonzero initial guess for Eisentat preconditioner; what UNIC needs and I forgot to put in before.
Hg-commit: c320a1da02e44812ec1e106a6b5c452116794b2e |
| a1c3900f | 19-Aug-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
finished adding -ksp_cg_single_reduction
fixed bug in MatDuplicate() for SBAIJ with jshort
Hg-commit: ab3bd021969125c78e64ec0e5509cc65844f43bc |
| 38702af4 | 11-Aug-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
for sbaij matrices with less than 2^16 columns use unsigned short to short j indices. Saves 5-10 percent on my laptop in MatRelax() saves nothing in MatMult() why?
Hg-commit: 620bb33c258dc0ffc4bc183
for sbaij matrices with less than 2^16 columns use unsigned short to short j indices. Saves 5-10 percent on my laptop in MatRelax() saves nothing in MatMult() why?
Hg-commit: 620bb33c258dc0ffc4bc183e43f256bab9918a06
show more ...
|
| 1c3996cd | 10-Aug-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
added PETSC_Prefetch() for Intel compiler with configure check for _mm_prefetch() helps on MatRelax_SeqSBAIJ_1() by 10 percent on PETSc
machine. Intel compiler seems to ignore GCC's __builtin_prefetc
added PETSC_Prefetch() for Intel compiler with configure check for _mm_prefetch() helps on MatRelax_SeqSBAIJ_1() by 10 percent on PETSc
machine. Intel compiler seems to ignore GCC's __builtin_prefetch() command
Hg-commit: a2e3967dd3c5eff62c6411454563a0c279d4aab4
show more ...
|
| b2843cf1 | 10-Aug-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
commit for some reason I forgot
Hg-commit: 1c3e42f78e384fba2d4d7edc3762fd0165dc814e |
| a2d7c930 | 07-Aug-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
fixed bugs introduced in last commit to MatRelax_SeqSBAIJ()
Hg-commit: e15ffe0c382f61591fb05db01ca225cad89f7048 |
| 9e2a9236 | 07-Aug-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
attempts at performace optimizations for MatRelax_SeqSBAIJ()
Hg-commit: 119905f2856b76348984a6b906c68ad208b5c75f |
| 64831d72 | 18-Jul-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
cleanup of errors that crept in because people did not fix examples when changing interface
Hg-commit: 09528a08bd96b5ecb599a789bc768fafa2591c75 |
| c60f0209 | 16-Jul-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
moved all the vector fortran kernels to correct location and the block invert kernels used by BAIJ and SBAIJ
Hg-commit: e1f076dca7c745ad92f3e0051127f9b15b56c9dc |
| f09c7c2a | 15-Jul-2009 |
Satish Balay <balay@mcs.anl.gov> |
correct flopcount in MatRelax_SeqSBAIJ()
Hg-commit: a1a36ccd19d24ef2cbd0ed60a675d143dd718e9d |
| be7314b0 | 15-Jul-2009 |
Barry Smith <bsmith@mcs.anl.gov> |
removed unneeded inline spops.h file
removed unused macros from some inline files
moved axpy.h to include/private/petscaxpy.h
moved first Fortran kernels from central location to sys/utils/ftn-kernel
removed unneeded inline spops.h file
removed unused macros from some inline files
moved axpy.h to include/private/petscaxpy.h
moved first Fortran kernels from central location to sys/utils/ftn-kernels where it is used
Hg-commit: c9e7a1a8dc579cf824e1f087165c3df9932aefb9
show more ...
|