Ok, but it would be good to add a comment that this might not work on Windows.

It seems this would result in duplicate definitions of mlir::registerGpuSerializeToCubinPass() on Windows.

Thanks Artem!

Could we fix the documentation to no longer refer to ForwardDataFlowAnalysis?

Thanks for reviewing and sorry for the delay. Just to be clear, although I think it's a minor improvement, I'm not attached to this change. If you prefer to keep things as is, I will just abandon it.

This API is doing too much and splitting 3. out would be the much cleaner solution IMO.

Clean up test TODO.

Nicolas, could you take another look please?

Have you tried to simply use e.g. "//:WORKSPACE" (without Label, without @llvm-raw)?
I think this might work because those rules are provided by @llvm-raw (or whatever the repo is called).

This looks good to me.

Add 'results' member to LoopNest.

Fix comment.

Switch from lambda to template function: uglier formatting, but no uncommon c++ pattern.

I think you can remove the surrounding Label() if the repo is explicitly stated in the target string.

Rename to scalarizeVectorOp.

Thanks Diego.

Nothing.

In D132354#3833295, @bondhugula wrote:

@csigg, the code you've added here yields a warning:

Fix clang tests that relied on empty input or accidentally did not provide the input in the RUN directive.

In D135067#3830815, @probinson wrote:

So what would such a tool see? Would it get an error trying to open stdin, or something else?

A tool reading from stdin would get stdin, not an error and not an empty string (which it did before this change).

Fix bazel config to actually test lit tests.

Add test, fix external shell mode.

FYI, I pushed 5faebb5 fix for Windows for the test added in this change.

Rebase.

Note: pre-merge failures are unrelated, I expect this change can be landed as-is (but I will definitely rerun the pre-merge check once it's fixed).

Update clang/BUILD.bazel as well.

Rebase.

In D133842#3789093, @mehdi_amini wrote:

How are things working right now without this? Does it mean that these may not be useful actually?

Add missing toyc-ch6/7.

Changing this revision to only fix the toy substitutions.

Rebase.

Restore diff.

In D133750#3786735, @mehdi_amini wrote:

LG but I think it'd be better to rebase to get a passing pre-merge build.

Rebase.

In D118126#3786685, @spatel wrote:

Something has gone wrong with the diff.

Rebase.

Rebase.

Bump bazel version to 5.0.

Rebase

Rebase.