mathiasbourgoin
mathiasbourgoin
Missing feature? It would be nice to add parameters to modify speed/accel of a whole pattern.
Currently sarek modules are hard coded within the _code generator_. A concrete syntax would help complete and detach sarek stdlib from the compiler. It could also provide a way to...
Tail recursive _functions_ used within a kernel could be automatically translated to loops
Everything is in the title... In the same vein, OpenCL backends could benefit from OpenCL SPIR
Latest Cuda (5+) and OpenCL (2.0+) versions allow dynamic parallelism (launch kernels (with different blocks/grids layouts)) from within a kernel. It currently works only on a limited number of compatible...
Currently, local functions are translated to global functions without taking into account free variables..
Adding automatic monomorphization of polymorphic functions used whithin a kernel (that stays monomorphic).
depends on #9 build a new opam package (sarek_stdlib) ?
## Summary by CodeRabbit - **Chores** - Upgraded continuous integration workflows for more reliable builds and tests. - Enhanced dependency installation by streamlining commands and refining cache management. - Simplified...
## Summary by CodeRabbit - **New Features** - Introduced an enhanced GPU computing module with expanded support for CUDA and OpenCL, offering robust error handling, device management, kernel execution, memory...