test possible conflicts

Default device for the machine (typically a single cpu core)

---------- float ----------

---------- float ----------

---------- double ----------

Make our own __half_raw definition that is similar to CUDA's.

---------- double ----------

---------- float ----------

---------- float ----------

---------- float ----------

---------- double ----------

just a helper to store a triple of K,M,N sizes for matrix product

Apply Sqrt to all output elements.

Enhanced std::string */

* Kernel for determining basic Eigen compile-time information * (i.e. the cuda/hip arch) */

The type used to identify a dense storage. */