Exn #71
Exn #71
Conversation
…that every memory used in an excepting pipeline is Locked (this can be weakened to: any memory which is _written_ to, or has reservations made in the main body of the pipeline must be locked)
…commit and exception blocks
|
|
||
| val new_m = addExnVars(m) | ||
| new_m.name.typ = m.name.typ | ||
| print(new_m.body) |
There was a problem hiding this comment.
Delete this print statement before merging.
There was a problem hiding this comment.
addExnVars should probably just take a Command and return a new Command and only be applied to m.body
This lets us avoid making like 3 new ModuleDefs.
| m.copy(body = CSeq(IStageClear(), convertPrimitives(m.body)), commit_blk = m.commit_blk, except_blk = m.except_blk) | ||
| } | ||
|
|
||
| def convertPrimitives(c: Command): Command = { |
There was a problem hiding this comment.
Maybe change the name or at least add a javadoc style comment - it's not really clear what this is doing from the name. (It's converting exception call arguments and inserting stage kill commands, right?)
| verify(s, pc + 1); | ||
| } | ||
| } | ||
| } |
There was a problem hiding this comment.
adding in stage separators (with nothing between them) can be a way to ensure we have more speculative or excepting threads that need to be killed so that the tests are meaningful.
| } | ||
| } | ||
| --- | ||
| commit: |
There was a problem hiding this comment.
this commit block is kinda early - normally we would want it to be the final stage. I don't think we're testing anything interesting in this example (other than having more code than the little tests).
We should probably have the commit block be only the final stage (so that there are rf.write and dmem.write calls before the commit block that actually need to be aborted)
resolve all comments and fix bugs, add ehr
fix tests
move to ubuntu 22
This is tracking the Exception feature as documented by the
docs/exceptions.mdfile.