merge axi and l15 ports into noc port #1267
Conversation
Signed-off-by: Jean-Roch Coulon <jean-roch.coulon@thalesgroup.com>
There was a problem hiding this comment.
LGTM @JeanRochCoulon. I am approving this PR, but please do not merge based only on my approval - I am just the verification geek and we should obtain an approval from more qualified team members.
|
❌ failed run, report available here. |
|
@Jbalkind As I have no way to test the modification in openpiton/ariane_verilog_wrap.sv, could you have a look and maybe try it? Thanks |
|
I just cleaned up some syntax errors but the design will not build and pass tests. I will have to spend some time to dig into this and figure out what's broken. The core makes its first request and receives the response but then it gets stuck. |
|
CVA6 does not come with a NoC (i.e. a complex interconnect). I suggest to use the "socket" word instead to qualify this interface, as it can be used to plug CVA6 on a bus or a NoC. |
| .l15_req_o ( l15_req_o ), | ||
| .l15_rtrn_i ( l15_rtrn_i ) | ||
| .l15_req_o ( noc_req_o ), | ||
| .l15_rtrn_i ( noc_rtrn_i ) |
There was a problem hiding this comment.
This is a typo - should be noc_resp_i instead of noc_rtrn_i. If you fix this, our simulation seems to be passing once again
|
I just made a PR to the ThalesSiliconSecurity/cva6 fork for fixes that make this PR work with openpiton. I don't think any of the fixes involve anything used outside of openpiton so it should be trivial to merge |
We discussed this when we had the meeting proposing this change and agreed on noc. I'm not a huge fan of socket as it's used for other purposes in other cva6 users' projects (e.g. ESP, DECADES). Perhaps "memintf" or "l1intf" or similar? |
Signed-off-by: Jean-Roch Coulon jean-roch.coulon@thalesgroup.com