Wed, Sep 18
Add test verifying that we fall back to SelectionDAG and die when a musttail call is not lowered as a tail call.
Tue, Sep 17
Fri, Sep 13
Update comments, add -global-isel-abort=1 to test
Add missing check for when the return is a tail call.
Thu, Sep 12
I think this seems reasonable.
Wed, Sep 11
Tue, Sep 10
Mon, Sep 9
Fri, Sep 6
Thu, Sep 5
LGTM in any case; I think it would be good to find out why the register class can come out wrong though.
Fixed the issue. Issue was that I never tested returning something other than void. I disabled tail calling calls with non-void return types, and updated the test to reflect that.
@RKSimon Sorry about that. I'll enable EXPENSIVE_CHECKS.
Wed, Sep 4
I'm going to commit this as-is and put up a separate review for remarks.
LGTM with style nit
LGTM with some nits on comments.
Tue, Sep 3
This is just code movement, but still a couple of nits/questions.
I'm pretty sure this should fall back considering that in the past, INSERT_SUBREG (which wasn't checked for) did. So, LGTM.
+1 for a separate pass. If we're going for an equivalent of -instnamer, I think it makes sense to keep the mode of use consistent.
Testcase to verify that we still reject it?