dpetrisko
commented
1 year ago Major changes
bsg_manycore:
- going through and figuring out where 32b is assumed instead of data_width_p (hopefully not many places)
- minor changes to decoder and ALU to support 32/64b ops
- testbench/, machine/Makefile.include, Makefile changes to support 32/64b. Specifically adding a parameter for data_width/XLEN
- Build a new or multilib gnu-toolchain to support 64b
- updates to software/bsg_manycore_lib to templatize those functions and classes
- not necessary, but probably desired: changes to the address map, which is hardcoded to 32b.
bsg_replicant:
- update to machines/Makefile.include, specifically data_width/XLEN
- updates to bsgmanycore*_packet.h to templatize to 32/64b
There is probably more, but those are the major points. I think @mrutt92 was thinking about this at some point and he's a C++ wizard so he might have more to say on the software side
amithmath
Hi,
Can computing node be RISC-V 64? If yes, what changes are required bsg_manycore, bsg_replicant, etc.?
-Amit