Summary of Privileged Architecture version 1.13 public review feedback

[Greg Favor]

Public review of "Priv 1.13" - namely the Machine-Level ISA version 1.13 (aka Sm1p13) and the Supervisor-Level ISA version 1.13 (aka Ss1p13) - is complete.  Below is a summary of the feedback received and the actions taken in response.

Robert Chyla pointed out that the PDF's DRAFT watermark is misleading, as the document contains specifications of extensions in various stages of development, including frozen and ratified ones.  The watermark describes the state of the document rather than the state of the extensions that it represents, so there is not actually a contradiction, but we agree that this distinction is not always clear. This matter falls under the umbrella of the current ISA spec unification work, an ongoing effort that is outside the scope of this public review.

Ved Shanbhogue raised an issue pertaining to the integration of the Svadu spec with v1.13 of the Supervisor ISA: The constraints on G-stage PTE D-bit handling are unduly strict, complicating the implementation of hardware page-table walkers.  In response to this observation, the ARC discussed and approved a relaxation that allows G-stage D bits to be speculatively set when initiating VS-stage page-table walks.  This relaxed scheme is also employed by contemporary ISAs.  Implementations that conform to the previous stricter specification remain compatible with the latest specification that has been updated to reflect this relaxation.

Greg and Andrew