[RISC-V] Inconsistent implied/required extension behavior for zabha/zvk*

[patrick-rivos]

Zabha/Zvk currently requires the user to specify Zaamo/{V|Zve}.

Godbolt:
zabha: https://godbolt.org/z/bYrY8fzcv
zvk: https://godbolt.org/z/fqheWoM7d

This behavior doesn't seem to match the RISC-V spec:
Zabha spec: https://github.com/riscv/riscv-isa-manual/blob/d2999932625d2cab25bb2081f7da915411e65ad3/src/zabha.adoc?plain=1#L28-L29
Zvk* spec: https://github.com/riscv/riscv-isa-manual/blob/d2999932625d2cab25bb2081f7da915411e65ad3/src/vector-crypto.adoc?plain=1#L471-L473

"requires" is equivalent to "depends upon": riscvarchive/riscv-v-spec#723 (comment)
"depends upon" seems to be specified behavior: https://github.com/riscv/riscv-isa-manual/blob/399c9a759eb4540a65c60e2cc236164821ff2346/src/naming.tex#L46

This behavior isn't consistent within LLVM:
zcb "depends upon" zca and is implicitly added by llvm: https://godbolt.org/z/aoYGbaWvz

[llvmbot]

@llvm/issue-subscribers-backend-risc-v

Author: Patrick O'Neill (patrick-rivos)

Zabha/Zvk currently requires the user to specify Zaamo/{V|Zve}.

Godbolt:
zabha: https://godbolt.org/z/bYrY8fzcv
zvk: https://godbolt.org/z/fqheWoM7d

This behavior doesn't seem to match the RISC-V spec:
Zabha spec: https://github.com/riscv/riscv-isa-manual/blob/d2999932625d2cab25bb2081f7da915411e65ad3/src/zabha.adoc?plain=1#L28-L29
Zvk* spec: https://github.com/riscv/riscv-isa-manual/blob/d2999932625d2cab25bb2081f7da915411e65ad3/src/vector-crypto.adoc?plain=1#L471-L473

"requires" is equivalent to "depends upon": riscvarchive/riscv-v-spec#723 (comment)
"depends upon" seems to be specified behavior: https://github.com/riscv/riscv-isa-manual/blob/399c9a759eb4540a65c60e2cc236164821ff2346/src/naming.tex#L46

This behavior isn't consistent within LLVM:
zcb "depends upon" zca and is implicitly added by llvm: https://godbolt.org/z/aoYGbaWvz

[lenary]

I think the Zabha issue will be solved by #115694

I'm not sure there's a PR open to fix Zvk* yet.

[topperc]

I think the Zabha issue will be solved by #115694

I'm not sure there's a PR open to fix Zvk* yet.

Zvk was already fixed a week or two ago.

[tclin914]

I think the Zabha issue will be solved by #115694

I'm not sure there's a PR open to fix Zvk* yet.

Zvk was fixed by 94e7d9c and f53889f.

[tclin914]

Zabha was fixed by: #115694