Add the Zibi extension.

[challenger1024]

Spec: https://github.com/riscv/zibi/releases/
Info: riscv/zibi#2

[jrtc27]

You don't need to sell the point of the extension in the commit message / PR description, just what it actually does

[jrtc27]

As this is a scattered union clause, this could also be caused by using a type defined after the 'scattered union' declaration

Did you read this message?

[challenger1024]

I’ve fixed it . thank you for the hint!

[jordancarlin]

Several (fairly minor) comments. Also make sure to add the extension to the README and Changelog.

[github-actions]

Test Results

2 117 tests  +2   2 117 ✅ +2   23m 35s ⏱️ -6s
    1 suites ±0       0 💤 ±0 
    1 files   ±0       0 ❌ ±0 

Results for commit 2f9b1c7. ± Comparison against base commit 0b4a9f2.

♻️ This comment has been updated with latest results.

[challenger1024]

if use:

mapping clause assembly = BITYPE(imm, cimm, rs1, op)
  <-> bitype_mnemonic(op) ^ spc() ^ reg_name(rs1) ^ sep() ^ hex_bits_signed_5(b_imm5(cimm)) ^ sep() ^ hex_bits_signed_13(imm)

we will get:

Type error:
extensions/Zibi/zibi_insts.sail:48.78-90:
48 |  <-> bitype_mnemonic(op) ^ spc() ^ reg_name(rs1) ^ sep() ^ hex_bits_signed_5(b_imm5(cimm)) ^ sep() ^ hex_bits_signed_13(imm)
| b_imm5 is not a union constructor or mapping in mapping-pattern b_imm5(cimm)

So the code ended up being written this way.
last commit:

• add zibi extension to README.md and doc/ChangeLog.md;
• move currentlyEnabled() function to model/extensions/Zibi/zibi_insts.sail
• modify assembly mapping of zibi

[Timmmm]

@pmundkur We should maybe use this as an example of how to add an extension when it is done since it is so nice and simple.

[jordancarlin]

LGTM. This will be great to merge as our first experimental extension.

[Timmmm]

LGTM. I guess we should probably have some kind of policy on how far along the process extensions need to be before they can be merged into this model?

This page seems to have a fairly clear 6-step progression (Zibi is at 'Stabilization').

It probably also depends on how large and invasive the code is. I think it's @pmundkur's decision but at least in this case the code is small and neat and doesn't affect anything else, so I'd vote merge.

[pmundkur]

Agree that this seems okay to merge. "Stabilization" seems a reasonable stage to merge small self-contained extensions, more invasive ones could be frozen or ratification ready as the case may be?