extend sparse map[!]/broadcast[!] to structured matrices #19926
Conversation
|
@nanosoldier |
|
tests should probably verify the output type? pretty small amount of additional code, that's encouraging. |
Good call. Output type checks added to the |
|
On that note, in the sparse |
|
don't think we have one, but it would make sense to check for testing |
|
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels |
kshyatt
added
sparse
Sacha0
changed the title
|
Added a commit extending sparse |
|
Rebased out the commits from #19724. Absent objections, requests for time, or someone else merging this PR prior, I plan to merge on Wednesday morning PST. Best! |
|
With the additional commit: @nanosoldier |
|
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels |
…tructured matrices to sparse.
|
Fixed rebase conflict resolution mistakes. Not certain whether those mistakes could have caused the strange nanosoldier results. (The benchmarks that regressed / improved cover unrelated functionality.) Will rerun nanosoldier if / when Travis and AppVeyor approve. Best! |
…ed matrices to sparse.
…d matrices check return type.
…atrices to sparse.
|
@nanosoldier |
|
Your benchmark job has completed - possible performance regressions were detected. A full report can be found here. cc @jrevels |
(Requires #19724, to which the first four commits belong.) This pull request extends sparse
broadcast[!]to structured matrices by promoting structured matrices to sparse where appropriate. (For better future performance, handling structured matrices directly in the underlying sparsebroadcast[!]routines might be possible.) Best!