Nathaniel Wesley Filardo

Results 237 comments of Nathaniel Wesley Filardo

OK, here's a spin with the registers back to being registers and not a vector thereof (thanks @jordancarlin!), but they're still exposed to the core of the model via an...

It almost surely still needs review and work, but I suppose it's moved from "a quick sketch" to "maybe this is something we might consider merging", so sure.

I'm certainly not opposed to someone doing the dynamic I/E design -- probably it means that the `encdec_reg` `mapping` becomes dependent on `misa.[IE]`, and I think that makes instruction decoding...

Please do. I apologize for my computer's misbehavior this morning... apparently Zoom and Firefox just _do not_ get along at the moment. The question I was attempting to ask in...

I guess another approach is "one `configfile` per `target`" and have each target's `on_config()` run `target:set("configvar", "thevar", --[[ ... --]])` and use a single template file that's just `${thevar}`.

AFAIK, `add_configfiles` always takes an input file. I'm not sure that's "wrong", but it'd be nice if it could just manage writing a string to a file, so that I...

AFAIK, `configfiles` always come from another file (usually named something `.in`): https://github.com/xmake-io/xmake/blob/7301a531f1581e5b22897c13c66d2704c493dbd9/xmake/actions/config/configfiles.lua#L297 which is then modified by variable substitutions in various forms. (These transformations do not have access to much...

> Why can't we do it with config.h.in and configvars? I thought it could do anything, even generating from a string as you said. Yes, that's the "have a shared...

Removing `regidx_offset` might make [certain future affordances](https://github.com/riscv/sail-riscv/pull/646/commits/fd0d6a25c02dcad83f8b1c8cfddf560f2e6d066d) more of a nuisance, requiring sign-extension at call-sites. That might be fine, but just to say.

This seems related to https://github.com/microsoft/snmalloc/blob/7b597335aedc2c60b0174f6e67a7f7560f841760/src/snmalloc/ds_core/mitigations.h#L243-L246 which I could have sworn had an issue associated with it but if so the association has been lost to time. :(