We have been developing the LLVM backend for AAP, a experimental 16-bit
Harvard architecture with features representative of a vast array of
deeply embedded microprocessors used today. The architecture
specification can be found at
Application Notes – Embecosm.
We discussed this backend as part of our poster at EuroLLVM 2015 (LLVM
for Deeply Embedded Systems), and we wish to upstream this for the
benefit of the development and maintenance of other out-of-tree deeply
embedded systems which utilize LLVM as their primary compiler.
Currently, this backend can generate code that can execute on a
simulator and Verilog implementation of this architecture. Currently, it
is available on GitHub (Embecosm · GitHub).
We feel this backend is in a state where it is suitable for upstreaming.
What is the best approach to contribute our efforts?
Edward Jones, Simon Cook