You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There's no rotation instruction in the base vector spec at the moment, so I'd suggest adding:
vror.vv vd, vs2, vs1 , vm # Vector-vector
vror.vx vd, vs2, rs1 , vm # Vector-scalar
vrol.vv vd, vs2, vs1 , vm # Vector-vector
vrol.vx vd, vs2, rs1 , vm # Vector-scalar
vrori.vi vd, vs2, uimm, vm # Vector-immediate
These mirror the format of the existing Vector Single-Width Bit Shift Instructions, and replicate the same set of functionalities that the scalar Bitmanip instructions provide.
For the vector Crypto extension, SEW=64 is the critical case for SHA3. I'm not sure about other element widths?
It might make sense for the base vector spec to define this instruction, and then for the vector crypto extension to require only the SEW values that it needs. Arguably, crypto might only need the vector-vector variants?
The text was updated successfully, but these errors were encountered:
From a mailing list discussion.
These mirror the format of the existing Vector Single-Width Bit Shift Instructions, and replicate the same set of functionalities that the scalar Bitmanip instructions provide.
For the vector Crypto extension, SEW=64 is the critical case for SHA3. I'm not sure about other element widths?
It might make sense for the base vector spec to define this instruction, and then for the vector crypto extension to require only the SEW values that it needs. Arguably, crypto might only need the vector-vector variants?
The text was updated successfully, but these errors were encountered: