Add mcontrol6.#538
Conversation
Mcontrol6 is like mcontrol but without a maskmax field and with a sizehi field on RV32. A recipe is provided for determining the max mask when mcontrol6.match=1 (NAPOT).
ernie-sifive
left a comment
ernie-sifive
left a comment
There was a problem hiding this comment.
Looks pretty good... just a couple of comments.
Co-authored-by: Ernie Edgar <43148441+ernie-sifive@users.noreply.github.com>
|
Good suggestions. I intuitively expected the maskmax6 discovery to behave as you said but I described it wrong. I also like the description of tdata2 WARL behavior since it describes required hardware behaviors that debuggers can rely on rather than describing expected debugger behavior and making hardware figure out how it should behave. |
|
Note that I had built the PDF as part of this PR but now that will need to be rebuilt (which I'll leave up to Tim or whoever). |
ernie-sifive
left a comment
ernie-sifive
left a comment
There was a problem hiding this comment.
Thanks, Paul. Looks good to me!
|
Note that I just added verbiage (both to mcontrol and mcontrol6) about the limitation on data value matching > XLEN. I tried to craft it to make the recommendations reasonably equivalent to my understanding of the original recommendations (which I understand to not recommend that address and data value triggers of all supported sizes should be implemented, just that if you're going to support address or data value triggers then all supported sizes should be implemented). |
timsifive
left a comment
timsifive
left a comment
There was a problem hiding this comment.
This change looks good. Thank you. I have a few style comments below, and one quesiton.
- don't duplicate timing table. mcontrol6 is authoritative. - use register/field macros - unify size field in mcontrol6
timsifive
left a comment
timsifive
left a comment
There was a problem hiding this comment.
Looks good to me. @ernie-sifive, do you have anything to add?
If not, I'll merge this on Friday.
|
Oh, one more comment. Can you remove riscv-debug-draft.pdf from this PR? Because git doesn't do a great job handling binary files, I never commit it along with PRs. Instead, I manually rebuild the PDF once a month so it's always reasonably up-to-date. |
OK. I saw your previous comment so I didn't include the new PDF for my recent changes (though I did build it locally to make sure there were no syntax errors and that it looked OK). But I forgot to remove the existing one from the PR. I just did that. |
3 participants
Mcontrol6 is like mcontrol but without a maskmax field and with a sizehi field on RV32.
A recipe is provided for determining the max mask when mcontrol6.match=1 (NAPOT).