-
Notifications
You must be signed in to change notification settings - Fork 30
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
Delete MLIR-AIE submodule #639
Conversation
cc5c31d
to
1185ec7
Compare
1185ec7
to
b7cfe18
Compare
501bd33
to
ba87aef
Compare
ba87aef
to
9aa42ab
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
@nirvedhmeshram @newling @Abhishek-Varma you guys want to take a look or you're good with this? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, just a minor question but you can address however and land.
b591cd4
to
550ceda
Compare
This PR delete/removes the MLIR-AIE submodule. The load-bearing change is forking all of the
AIEOps.td
type things. The main reason for doing this now is Xilinx/mlir-aie#1660, which does not fix the parser issue that broke us but makes it permanent.In performing this fork we take the most minimal set of auxiliary/support impls/function/code. Specifically, we take none of their verifiers. The reason is, as usual, 1) we don't intend to continue to have most/many of these ops 2) they're now not an egress boundary i.e., they're purely "compiler state" ops/IR. Naturally, whichever ops we keep or iterate on, we should have verifiers for but I prefer 1) to add them incrementally rather than eagerly 2) add them in a separate PR (or PRs).
A couple things to call out:
i32
->i8
). This is an indication that the fork/port/vendoring is successful;aie-rt
, incurs that cost thatuint8_t
s need to printed withstd::to_string
1;sed
s. I did it this way because the patches would break an isolated build there3. Not the prettiest thing but I'll keep up with it.Note, there's a still bit to do (I want to rename thexilinx::AIE
namespace and then usetypedef
s) but this is ready to review.On second thought let's merge this and then move all the ops/stuff to our
mlir::iree_compiler::AMDAIE
namespace in a follow-up.Fixes #430
Footnotes
Otherwise you print the ASCII character at the position rather the number itself. ↩
Stuff like
getIDInt
when the existing attr getter isgetID
. ↩Which never happens sure but just seems like it sprawls the hackiness rather than containing it to one place. ↩