You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
An integration wtih magma would provide state of the art lapack support for cuda and rocm platforms.
This, along with jna bindings to lapack in tech.compute would I think be enough to enable a good portion of scientific computing to happen across multiple backends assuming the basic tensor ops were implemented for the various backends (currently only assignment is done).
A very interesting potential side project. Integrating lapack and magma into tvm I think would be well received by the tvm and magma communities. That could happen independent of tvm-clj and the results would be easily used then from the tvm-clj system.
The text was updated successfully, but these errors were encountered:
An integration wtih magma would provide state of the art lapack support for cuda and rocm platforms.
This, along with jna bindings to lapack in tech.compute would I think be enough to enable a good portion of scientific computing to happen across multiple backends assuming the basic tensor ops were implemented for the various backends (currently only assignment is done).
A very interesting potential side project. Integrating lapack and magma into tvm I think would be well received by the tvm and magma communities. That could happen independent of tvm-clj and the results would be easily used then from the tvm-clj system.
The text was updated successfully, but these errors were encountered: