How to Design a Coprocessor #2518
Conversation
|
✔️ successful run, report available here. |
|
Great contribution ! Very useful for newcomers. To become contributor, you have to sign the Eclipse agreement, can you ? |
| // 1 RISCV instruction class for our Coprocessor | ||
| parameter int unsigned NbInstr = 1; | ||
| parameter copro_issue_resp_t CoproInstr[NbInstr] = '{ | ||
| '{ | ||
| instr: 32'b0000000_00000_00000_000_00000_0101011, // custom1 opcode | ||
| mask: 32'b1111111_00000_00000_110_00000_1111111, // bits that must match | ||
| resp : '{ | ||
| accept : 1'b1, // We will process this instruction | ||
| writeback : 1'b1, // We will write a general-purpose register | ||
| dualwrite : 1'b0, | ||
| dualread : 1'b0, | ||
| loadstore : 1'b0, // No memory operation (Not supported in pipeline) | ||
| exc : 1'b0 // Cannot throw exception | ||
| } | ||
| } | ||
| }; |
There was a problem hiding this comment.
This looks based on a deprecated version of the cvxif_instr_pkg.sv
There was a problem hiding this comment.
I must have been working from the latest tagged release. (v5.1.0 also reportedly doesn't build, sadly, due to a problem with dual-issue parameterisation which was fixed soon after.). I'll update the example to be based on master.
@JeanRochCoulon Apologies, this took us a bit to digest. I believe I have now signed the Eclipse agreement. |
@JeanRochCoulon, can you provide guidance to relaunch this PR as the ECA is now signed? A new commit? @joncapltd, a prerequisite to make the connection on the ECA is to have the same email address in GitHub and the Eclipse account. |
|
I relaunched the CI, but ECA always fails. Can you double check the email address as suggested by @jquevremont Can you rebase your PR please ? |
|
Hi joncapltd, any update on this? |
|
Hi @joncapltd Any news ? |
|
I've found time to update the example to align with the current master and test. Apologies for the delay! I now need to figure out what's going with email addresses and ECA... |
|
✔️ successful run, report available here. |
1 similar comment
|
✔️ successful run, report available here. |
|
When looking at https://api.eclipse.org/git/eca/status/gh/openhwgroup/cva6/2518, I think early commits have been "polluted" and do not pass the ECA. I would recommend to start with a new PR. |
|
@jquevremont Thank you for the pointer to the report; I was uncertain whether I had any commits that passed, and was worried that the check was case-sensitive, as the ECA did not allow capital letters, but my github email has capitals. I'll see what I can do to maybe flatten the commits. |
|
Take this with a pinch of salt. I am not a specialist of the GitHub / ECA integration. I also see some recommendation on the report. |
Add a tutorial for creating a custom coprocessor and testing it. Add an introduction to the tutorial section. Spelling fixes. Update example to reflect the current master branch example coprocessor.
joncapltd
force-pushed
the
how_to_design_a_coprocessor
branch
from
ffa1815
to
ca37fa9
Compare
|
I think I might be passing the ECA agreement now, after help with flattening the commits and force-pushing as the new user. Thanks to Samuel Stark here at the lab for the Git handling. |
|
✔️ successful run, report available here. |
|
✔️ successful run, report available here. |
This adds a tutorial on how to customise the example coprocessor with your own instructions and test them.