Tal Ben-Nun
Tal Ben-Nun
Reduction as library nodes is so delicate these days... Thanks!
that would be nice. If assign was an access node instead, then RedundantArray would kick in anyway.
I can only think of setting a constant, but that is not handled in assign.
@alexnick83 should be fixed now with views?
Some of the half-precision issues have already been solved in #186 (not all though)
* fun1: This happens because the block size is set to `max(1, N)`, due to the fact that the reduce node happens, by default, on shared memory. I'm taking care...
You are correct, if the SDFG was generated from one of the frontends (Python for example) it retains the order of the arguments. However, if the SDFG API is used,...
> @tbennun, is there any public discussion forum for the project where we can discuss some ideas and decide their feasibility? https://github.com/spcl/dace/discussions
@bartokon Seems like CUDA is failing because its version is too old to support GCC 9 or newer. The official recommendation from NVIDIA is to install the CUDA toolkit only...
Thank you for the detailed example! Any ideas about the FPGA issues @definelicht @TizianoDeMatteis ?