David Biancolin

> Given this, it seems like I should switch this PR to have init-submodules rather than build-toolchains delete the scripts? That's definitely better, but if a user doesn't run the...

Yeah that's tricky. I think what i might do if i were you initially, is create another package under generators/chipyard and instead of modifying chipyard sources, import/extend/copy in them in...

THis is a definite improvement, but does it address Tim's comment that the manager itself should return non-zero if any one of the builds fails?

i have vcs metasimulators running with some change to how the mm_t and mmio_t pointers are handled, but i'm not confident that this problem is resolved yet.

Yeah, that's understandable. Another question is whether this field should also be exposed in the `ini` / `yaml` so maybe we should just punt on it for now. If the...

I think switching to spot instances and being parsimonious with FPGA-based CI deployments is sufficient in the short term to bring our spending down. @abejgonzalez you might as well address...

What's the state of this PR?

I'm very receptive to the idea of making FireSIm more modular and friendly to use in different environments. For instance, some degree of this is required to integrate it more...

Firstly to avoid confusion i will refer to the amazon-provided shell as the "SH region" or SH. I will refer to the fpga-shell's notion of a shell as a "diplo-shell"....

> I heard that AWS F1 is alveo U200? I wonder how could aws achieve this locking? For example the PCIe under SLR0 and one of DDR4 under SLR1 is...