-
Notifications
You must be signed in to change notification settings - Fork 1
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merging into master for 0.8 release #99
Merged
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
A 2D mesh of board routers (one per board) now connects together the 2D NoCs on each board. This means that message can pass over a board without travesing the NoC. The per-board router is not yet programmable, but that is direction we are heading.
Support routing-key-based send operations, but programmable routers don't yet interpret routing keys.
Need to support inflight request info for programmable routers now, in addition to caches. Therefore, generalise the info field in a DRAMResp. Need to check that the optimiser does what I expect in this case, i.e. throws away unused bits in the info field.
(But doesn't compile yet)
(Compiles but untested)
(Works, but only tests one type of routing record so far)
These work. Still need to test RR records, and multi-beat records.
Working fine but needed to clarify the docs: the size field in a beat is the number of records, not the number of chunks.
(From 0.6.3)
(From 0.6.3)
This is to distinguish (existing) thread-level routing tables from board-level routing tables (to come).
(Builds but untested)
(Compiles, but untested)
The 'active' condition in the fetcher was inverted, and also referred to the wrong state variable.
Missing 'return' statement, and missing use of 'offset' when uploading routing tables. The POLite heat grid now appears to work in simulation. Ready to try things on FPGA.
Fast mapping version of POLite
I was trying to save area in the ProgRouter crossbar by sharing output queues between inter and intra FPGA links, but this introduces deadlock. Every trick I've tried to reduce area in the ProgRouter has met the same fate. So here's the fully parallel that's working very nicely so far, if a bit greedy in terms of FPGA area.
Optional extra send slot
POLite now allows empty pins, so fan outs of some vertices may be zero
Multiple boxes in the Y direction only has stopped working. Looking back at the first commit the in 0.8 branch, it looks like I accidently changed the behviour of inter-board routing by considering the host bit before knowing that we're on the correct Y coordinate. Hopefully, this is the fix...
Fix multi box issue
If POLITE_DUMP_STATS is defined then measuring time from the host isn't very accurate because it will include the time to transfer a large number of stats over the slow UART. To help users become aware of this, we now don't display the time if POLITE_DUMP_STATS is enabled.
This is faster than METIS and gives similar results for a few graphs that I've tried. It works by picking an unvisited vertex, and then doing a size-bounded BFS from that vertex with the size-bound equal to the partition size. This process is repeated until the partition is full, at which point we move to a new parition and repeat the process until all vertices have been visited. It is just a few lines of code, and could be easily parallelised in future.
Compiles but untested...
coralmw
pushed a commit
that referenced
this pull request
Apr 26, 2022
Merging into master for 0.8 release
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.