search

lowRISC / sonata-system

A full micro-controller system utilizing the CHERIoT Ibex core, part of the Sunburst project funded by UKRI
Apache License 2.0
20 stars 11 forks source link

Shouldn't the Ibex be configured as RV32E=1'b1? #136

Open ColinSCISemi opened 3 weeks ago

ColinSCISemi commented 3 weeks ago

Hi,

CHERIoT is spec'ed as being 16 GP registers only. However, in Sonata the RV32E is set to 0 which means that there are 32 GP registers. Shouldn't the RV32E parameter be set when the CHERIoTEn parameter is set?

Cheers, Colin

HU90m commented 3 weeks ago

Yes indeed, thanks for opening this

marnovandermaas commented 3 weeks ago

At the moment the CHERI enable bit is configurable at runtime. The hardware essentially turns on RV32E when CHERI is enabled and otherwise is RV32I. RV32E is a parameter and cannot be changed at runtime. If we go down that route, we should re-do the legacy software part of this repo to work on RV32E first.

marnovandermaas commented 1 week ago

Also thanks to @ColinSCISemi for creating a patch for CHERIoT Ibex to fix Verilator lint warnings: https://github.com/microsoft/cheriot-ibex/issues/39

GregAC commented 1 week ago

For sonata-system at least, given it's status as an FPGA based evaluation platform I'd prefer to leave RV32E as 0 because of @marnovandermaas's comments above. Whilst we haven't got everything in place to make this work well we want to ensure the released sonata system is able to boot in both CHERIoT mode and vanilla RV32 mode, which should be RV32I given RV32E is little used.

Advanced users of Sonata who wish to optimize resource usage are of course free to ditch the vanilla RV32 mode and set RV32E to 1 and build their own FPGA image. We should make sure it is simple to do this and make clear in the documentation it's an option.