Skip to content

Conversation

@haampie
Copy link
Member

@haampie haampie commented Dec 25, 2017

Closes #181. Some code benefits here, like cg, chebyshev and bicgstabl. The only downside is that we're missing threading of openblas, but that helps only marginally on the CPU for BLAS 1.

@haampie haampie changed the title Use . rather than BLAS calls Use broadcasting rather than BLAS calls Dec 25, 2017
@haampie haampie force-pushed the feature-blas-1-as-broadcast branch from 765fb8f to e894d69 Compare December 27, 2017 18:42
@haampie haampie merged commit 7103a27 into JuliaLinearAlgebra:master Dec 27, 2017
@andreasnoack
Copy link
Member

I still think this is fine but I realized one catch here. If you want to try out these methods with a new fancy array type then there is a tax from using dots since supporting broadcasting is much harder than supporting BLAS-1.

@haampie haampie deleted the feature-blas-1-as-broadcast branch December 28, 2017 14:58
@haampie
Copy link
Member Author

haampie commented Dec 28, 2017

True, I had to do it in the test of lsmr and it took me a while to figure out how to do it. Hopefully the docs are clear when Julia 1.0 is released

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants