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
The AP field tables include the M-bit
ACPERM doesn't AND that permission bit but the rules do include clearing it if X is cleared
GCPERM doesn't return that permission
But the M-bit is encoded with the other architectural permissions, for RV32 and for RV64, in tables 3 and 4.
My proposed fix is to include the M-bit into the definition of ACPERM/GCPERM so we have a consistent definition of AP.
The text was updated successfully, but these errors were encountered:
The M-bit isn't a permission, it can be toggled arbitrarily (providing the cap isn't sealed)
It can be cleared as a result of ACPERM, if X is removed, sort of. In fact if X=0 then it is no longer the M-bit, so it becomes non-existant - so it can change state between existant and non-existant as a result of ACPERM.
I'll tidy up the doc to remove the M-bit from the AP field. The current spec for ACPERM, GCPERM is believed correct.
The AP field tables include the M-bit
ACPERM doesn't AND that permission bit but the rules do include clearing it if X is cleared
GCPERM doesn't return that permission
But the M-bit is encoded with the other architectural permissions, for RV32 and for RV64, in tables 3 and 4.
My proposed fix is to include the M-bit into the definition of ACPERM/GCPERM so we have a consistent definition of AP.
The text was updated successfully, but these errors were encountered: