-
Notifications
You must be signed in to change notification settings - Fork 61
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Transform enhancement squashed rebased #135
Transform enhancement squashed rebased #135
Conversation
@@ -360,6 +360,40 @@ clsparseDCsrMatrixfromFile( clsparseCsrMatrix* csrMatx, const char* filePath, cl | |||
const cldenseVector* y, | |||
const clsparseControl control ); | |||
|
|||
// functions for tests of scan transformation, as simple as possible |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
See my comments on PR #133
I have not benchmarked your refactoring of coo2csr, csr2coo, dense2csr, csr2dense, but a phenomenal job simplifying the code and eliminating kernel passes! |
@jpola |
1. Added more information to message when failed to read matrix from file. 2. Reorganized functions related to computing rowBlocks they are now in csr-meta.hpp / cpp files instead of clsparse-coo2csr.cpp, rest of functions in clsparse-coo2csr were not used therfore name clsparse-coo2csr-GPU changed to proper one. 3. Csr_matrix_environment allocates matrix in double precision in the first place which is then casted to single precision. 4. Sanity checks in clsparseD/SCsrMatrixFromFile. 5. Added Inclusive and exclusive scans operations + tests. 6. Rewritten reduce by key operation + tests. 7. Rewriten coo2csr and csr2coo which no longer need the use of radix sort. I.e we now using 10 kernel calls instead of 34, simple tests showed 6x speedup. 8. Rewritten dense2csr and csr2dense in more clean way. Improved performance eliminating unnecessary copies. Minor: 1. When reading matix in coo format directly the data need to be also sorted by (row, col). Otherwise we will have column major format which is default for mtx storage.
…duce by key functions and corresponding tests
Hi, sorry for delay. 2015-09-01 6:32 GMT+02:00 Kent Knox notifications@github.com:
|
Transform enhancement squashed rebased 👍
No description provided.