rtl

Type of issue: documentation

EH1 has a lot of examples where a sequent CFunc contains only a handful ot statements (often 1 or 2) and is called only once. This causes a performance penalty when --output-split puts these in a different function than _eval and hence the compiler can't inline them, so we should inline these ourselve when it's obviously the right thing to do.

Type of issue: feature request

Impact: new rtl

Development Phase: request

Other information
As the vector extension proposal is near stable and the necessary things is defined in the rocket chip repo I think now the vector e

We have axi_cut and axi_multicut for pipelining (to be unified), but we currently don't have a module for buffering AXI beats in FIFOs. This gap can be filled by a new axi_fifo module.

Implementing thi