In D158107#4641560, @mehdi_amini wrote:

It depends on whether it's intended behavior for ConversionPatternRewriter::replaceAllUsesWith and ConversionPatternRewriter::eraseOp to be unusable in conversion passes

I'm confused, aren't these widely used? At least eraseOp is pretty standard here and I see quite a large number of hit in the conversion passes in-tree (replaceAllUsesWith maybe less common because replaceOpWithNewOp).

In D158107#4641519, @mehdi_amini wrote:

Did we get to the end of this issue?

Make changes to zero9178's comments

one takes a StringRef * that parses any keyword.

Ah yes, now I see why parseKeyword isn't sufficient. I would like to have the bare_identifier be arbitrary, similar to how AffineMap allows for an arbitrary variable name.

In D156356#4540606, @rriddle wrote:

I'm not really sure we need to expose this, why does parseKeyword not work for you?

Add optional variant

Remove caret token def

Still getting the hang of the arc CLI tool, sorry for the noise

Revert ParserTest.cpp formatting change

Removed the parseCaret after determining I can't make it work

What exactly is the followup here? I see the extra test added that ensures ctlz is not lowered when the flag is disabled. Is another test helpful?

I can fix it Monday morning. A rollback is fine if it's urgent.

I'm not able to get to my computer til Monday, but perhaps your issue is
that this feature is disabled by default (see comments above), and you need
to explicitly enable it. The test files in this patch show the invocation.

@vzakhari ping :) thanks for your review and advice!

(btw, I don't have commit access, so someone else will need to commit it for me)

This time only Bazel failed, so ok to merge :)

Try syncing with head

break -> continue, plus move mlir-cpu-runner

Fixed a typo/bad variable name choice from the last update

Sorry for the delay, I was in Korea at a conference.

Address reviewer comments

Added default-disabled convert-ctlz option

The backend we're ultimately planning to support does not have a ctlz builtin, so we have no choice but to emulate it in software. This is why lowering to llvm's builtin doesn't help us (and IIUC CIRCT also doesn't support llvm's ctlz)

Added mlir-cpu-runner tests for ctlz