Merge branch 'jed/fix-mat-sbaij-sor-prefetch'* jed/fix-mat-sbaij-sor-prefetch: Mat SBAIJ: avoid invalid memory read in MatSOR_SBAIJ
Mat SBAIJ: avoid invalid memory read in MatSOR_SBAIJIn the back-solve, the software prefetching code computes the previousline length and issues prefetch instructions before relaxing a row. Fort
Mat SBAIJ: avoid invalid memory read in MatSOR_SBAIJIn the back-solve, the software prefetching code computes the previousline length and issues prefetch instructions before relaxing a row. Forthe first row, the "previous row" does not exist so the indexing wasinvalid. Now just make sure the index is non-negative.Reported-by: Patrick Lacasse <patrick.m.lacasse@gmail.com>
show more ...
Bad LogFlops in MatSOR_SeqSBAIJ.
refactored residuals by adding a 'residual' method to Mat
Merge branch 'barry/improve-memory-logging'Conflicts: src/ksp/pc/impls/redundant/redundant.c src/mat/interface/matnull.c
logging memory now credits to all ancestors
Use classname in object for object name when calling PetscObjectPrintClassNamePrefixType() instead of some string with less precise name passed in
PetscStackCallBLAS: new macro instead of PetscStackCall (skip CHKMEMQ)When CHKMEMQ occurs on every call to BLAS or Lapack, we sometimes seeunacceptable performance degradation (orders of magnitude
PetscStackCallBLAS: new macro instead of PetscStackCall (skip CHKMEMQ)When CHKMEMQ occurs on every call to BLAS or Lapack, we sometimes seeunacceptable performance degradation (orders of magnitude).PetscStackCallBLAS is currently identical to PetscStackCall except forskipping CHKMEMQ, but such checking could be made optional in thefuture.
removed string version of function name for XXXRegister(), PetscFunctionListAdd() and PetscObjectComposeFunction()
removed excessive use of CHKMEMQ;
Warnings: clean compile with gcc-4.8 at -Wextra -Wno-unused-parameter* Mostly fills out some ops struct declarations. Reformatted so Emacs can help with alignment. Squelch whitespace in diff to s
Warnings: clean compile with gcc-4.8 at -Wextra -Wno-unused-parameter* Mostly fills out some ops struct declarations. Reformatted so Emacs can help with alignment. Squelch whitespace in diff to see useful changes.* Use struct instead of casting to data pointer (void*) when serializing function pointers.* strcasecmp is in strings.h
moved "macro" kernel interfaces to petsc-private/kernelsHg-commit: a5f14392b251a7db2c640b6be7e611f6c23ef9ea
Change all PETSC_EXTERN_C to PETSC_EXTERNHg-commit: 8d2ebbb193fb583bccc64015e35640c4e08c3426
merge, terrible manual process with many conflicts with Jed's PETSC_INTERNALHg-commit: 84df07d03c6e55bd0f27bd5ee8c1738562bd529d
visibility: use PETSC_INTERN when possibleHg-commit: 3b4bfb5c23843293f1f08790b48437080a32eb42
finished all EXTERN_C_BEGIN/END fixes up through Mat including all fortran interface filesHg-commit: a763120ba50023930be975663a3f80c58e415743
completed removing unneeded EXTERN_C_BEGIN/END from Mat directories and converting to PETSC_EXTERN_C for constructorstested with and without dynamic with and without C++Hg-commit: 7d27d7f4d9ea3bfe
completed removing unneeded EXTERN_C_BEGIN/END from Mat directories and converting to PETSC_EXTERN_C for constructorstested with and without dynamic with and without C++Hg-commit: 7d27d7f4d9ea3bfe6616fafdfb32d046b5db53a1
changed use of PetscObjectComposeFunctionDynamic() to PetscObjectComposeFunction() to allow use of static for functions for standard use in PETScPetscObjectComposeFunctionDynamic() is still availabl
changed use of PetscObjectComposeFunctionDynamic() to PetscObjectComposeFunction() to allow use of static for functions for standard use in PETScPetscObjectComposeFunctionDynamic() is still available for use if neededalso fixed calls to PetscObjectComposeFunction() to not wrap lines (per PETSc coding style)Hg-commit: 822f9ddaac95a8ff6c2a9ad77fbf07d02d2c20d9
added petscdrawtypes.h removed the typedef _xx* xx definitions from petscdraw.h and removed petscdraw.h from petscsys.hHg-commit: 01be7b4c0ac1e993b540535c319de45af820a255
added PetscObjectComm() and used it to replace (((PetscObject)obj)->comm)Hg-commit: 3da37c458124ad48ae939f4e9823e4430ee0b8be
removed PETSC_NULL from C and Fortran (except declaration in C for backward compatibility). Kept PETSC_NULL_xxx for FortranFixed a few bugs where PETSC_NULL had been used incorrectly.Hg-commit: 05
removed PETSC_NULL from C and Fortran (except declaration in C for backward compatibility). Kept PETSC_NULL_xxx for FortranFixed a few bugs where PETSC_NULL had been used incorrectly.Hg-commit: 054705a517d7f4388a8a084415d7478cbe95dff4
added back in /*I "petscvec.h" I*/ in several files I mistakenly removed it fromwrapped all the BLAS calls in PetscStackCall()Hg-commit: 18be25a7ff539aa5230355449bf40d90e959b06a
Uncrustified src/mat/*.Hg-commit: 5c6f04286a6cfcd98361b2479b884c0041d95b73
Reformatted: "Do not put a blank line immediately before PetscFunctionReturn;"Hg-commit: c8a41c455986e732d0ea0513d7162a54daffaece
Reformatted: "There should be a blank line before PetscFunctionBegin;"Hg-commit: 05bda4416aaceb2dbd9d85b5720629adc88eef98
1...<<11121314151617181920>>...53