I have released v0.4 of the Co-dfns compiler. This release focuses on overcoming the performance bottlenecks that were plaguing earlier versions by using the DWA infrastructure. You can now expect good performance on scalar computation on both the CPU and the GPU, though you may have to do some tweaking on the GPU to limit data transfer overheads still.

Further improvements in performance are slated for v0.6.

https://github.com/arcfide/Co-dfns/releases/tag/v0.4