| 88b60a73 | 29-Jun-2010 |
Victor Minden <victorminden@gmail.com> |
changed arch-cuda to use the -x flag for CUDA 3.1 Also added event logging for copying vectors up and down to GPU and began creating subclass for SeqCUDA to move code from Seq class.
Hg-commit: c1a
changed arch-cuda to use the -x flag for CUDA 3.1 Also added event logging for copying vectors up and down to GPU and began creating subclass for SeqCUDA to move code from Seq class.
Hg-commit: c1aa6f2e29f76ec64ce854bf916a7edd016eb261
show more ...
|
| 7b641abf | 29-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
fixes for totalview viewing of vecs and mats
H fo: Enter commit message. Lines beginning with 'HG:' are removed.
Hg-commit: 02a6848ed1c46ea72f35212aa069342446a4123f |
| 2082002f | 29-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
display basic information about Mat and Vec object for Totalview debugger
:w!
:w!
Hg-commit: d5bf9a455370c2b92b8a8df0eab291460186a475 |
| 99279cb2 | 25-Jun-2010 |
Victor Minden <victorminden@gmail.com> |
Switched Vec functions to use CUSP BLAS instead of CUBLAS. Also fixed some CUDA copy up/down problems.
Hg-commit: 53f17d50e2f2ecfa147f24dd7b6916b2f5eaec14 |
| fc5669d5 | 25-Jun-2010 |
Matthew G Knepley <knepley@gmail.com> |
Who the FUCK called VecScale collective?
- I gave this as a homework and everyone got it wrong because the docs are messed up
Hg-commit: f057bdb14bdc6abf6c539b0a18832c73d17f5906 |
| 1ff3db77 | 24-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
changed custom VecLoadIntoVector() for Fortran to custom VecLoad() since need custom to handle special viewer cases
Hg-commit: 64bd2758c540e6fa5b39a870aa71ef6027c7791a |
| 8d35a466 | 23-Jun-2010 |
Victor Minden <victorminden@gmail.com> |
made changes to make VecView use CUDA copy up, a few Vec_Seq functions use correct form of VecGetArray. Added function to 'allocate' GPU memory without copying down for Vec_seq
Hg-commit: 90375c35f
made changes to make VecView use CUDA copy up, a few Vec_Seq functions use correct form of VecGetArray. Added function to 'allocate' GPU memory without copying down for Vec_seq
Hg-commit: 90375c35fa92e90e4031e89ec6ff43c897caa6d6
show more ...
|
| d1574f38 | 23-Jun-2010 |
Victor Minden <victorminden@gmail.com> |
commit after merge
Hg-commit: 5acee20bde6ff44083e9e137280f70145913eb2b |
| 8f9db167 | 23-Jun-2010 |
Victor Minden <victorminden@gmail.com> |
edited the new example for ease of use
Hg-commit: 4cd6ddbc026e6dda37e3349ec73b009058e0d28e |
| 44773932 | 20-Jun-2010 |
Jed Brown <jed@59A2.org> |
ISComplement unit test
Hg-commit: 117faa30a99e970c52ef868590bb3cd1ae02f296 |
| 5a586d82 | 20-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
removed many unused variables
small amount of reformating
added use of all arguments in MPIUNI call to prevent complaints about unused values
Hg-commit: f1e85e77153d4207c8ab2522ae8e9ec72b7358ec |
| 40db8491 | 18-Jun-2010 |
Victor Minden <victorminden@gmail.com> |
switched CUDA implementation to use Thrust library vectors instead of CUBLAS arrays.
Hg-commit: 605c99c545b2505b9466d8eae8309049af2a02ce |
| 8fb81238 | 17-Jun-2010 |
Shri Abhyankar <abhyshr@mcs.anl.gov> |
Added MPIAIJ implementation for new MatLoad interface.
Small bug fix in VecLoad interface.
Hg-commit: 468674f220b2f370b25aa95c5be79f344669779f |
| 90e7e275 | 16-Jun-2010 |
Victor Minden <victorminden@gmail.com> |
commit after merge
Hg-commit: e0219ecda17af748235879d268f48db3d35db835 |
| b4fc8ab4 | 16-Jun-2010 |
Victor Minden <victorminden@gmail.com> |
made several changes to vector functions to fix cuda functionality (GPU memory was not being allocated properly due to flags being set from unallocated to CPU without any action). Configured make fil
made several changes to vector functions to fix cuda functionality (GPU memory was not being allocated properly due to flags being set from unallocated to CPU without any action). Configured make files to accept .cu for use with cubuilder.py (the new, slow method for compiling the library for use with cuda)
Hg-commit: b1f0a82d5b711f64d8bb8c44f5515261ca1a1729
show more ...
|
| 8b7d6567 | 16-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
forgot what I changed
Hg-commit: 202bb90902021012a4123f61a73b8733e0b15681 |
| 3649974f | 16-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
changed many VecGetArray() to VecGetArrayRead()
Hg-commit: d7f64a5632750c1ca905eea86db48747b7367a8a |
| 91ae87fc | 15-Jun-2010 |
Jed Brown <jed@59A2.org> |
Remove custom VecLoad (zveciof.c, was already removed from build)
Hg-commit: 0971c213b1d8fa1551ec7768bbdac380cc793622
|
| a3424433 | 15-Jun-2010 |
Jed Brown <jed@59A2.org> |
Fix some misuse of SETERRQ, Barry was naughty and didn't compile
Hg-commit: c5b7ef1f9e2d57041449029ea1c5cd28de430ecf |
| f1d8bc90 | 15-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
added VecGetArrayRead() and VecRestoreArrayRead() and used them for mdot and maxpy implementation in vec seq
Hg-commit: 0ca29eba344a09cf693b32a9ba8d421e39ff7295 |
| 31f78ffa | 15-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
introduced VecGetArrayPrivate() and VecRestoreArrayPrivate() for use in vector implementations. The do not increase the state of the vector since that this handled by the calling routine.
Hg-commit:
introduced VecGetArrayPrivate() and VecRestoreArrayPrivate() for use in vector implementations. The do not increase the state of the vector since that this handled by the calling routine.
Hg-commit: c0e05d280f3dadbc710d6c1a35abe589db5561b2
show more ...
|
| 09e5b063 | 15-Jun-2010 |
Barry Smith <bsmith@mcs.anl.gov> |
removed special code for slow blas; removed unused variable
Hg-commit: 01b00ca28265816eec642cec72a0305bc9e84436 |
| b6bd2e71 | 15-Jun-2010 |
Shri Abhyankar <abhyshr@mcs.anl.gov> |
Seperated MPI I/O code from binary for VecLoad
Hg-commit: e7533b5838c91ee1022ff2189a3efb11fbc8a3c1 |
| 8638a4bb | 14-Jun-2010 |
Shri Abhyankar <abhyshr@mcs.anl.gov> |
Fixed checking vector type in VecLoad().
Since vec->type_name is not set in VecSetType() checking the vector type using vec->type_name is incorrect.
Hg-commit: 2d1785e2f9cd3d01f0f2fa783ea185b77fb923
Fixed checking vector type in VecLoad().
Since vec->type_name is not set in VecSetType() checking the vector type using vec->type_name is incorrect.
Hg-commit: 2d1785e2f9cd3d01f0f2fa783ea185b77fb92317
show more ...
|
| fa52e05e | 14-Jun-2010 |
Shri Abhyankar <abhyshr@mcs.anl.gov> |
Automatically generate fortran stub for vecload
Hg-commit: f3c70632fe2a6a5344b27e8b6875ade33d858397 |