[llvm-commits] [PATCH] Fix bug in SExt optimsations

[Jose Fonseca]

----- Original Message -----
> >>> Two diffs, one for llvm which fixes the bug and one for nightly
> >>> test-suite
> >>> which tests the bug.
> >> The tests for such sort of thing should be a part of standard
> >> regression testsuite.
> >
> > as these are tests that need to be executed, I asked him to put
> > them in the
> > nightly testsuite.
> Why it's not possible to match the generated assembler?

Yes, it is possible to match generated assembler, but the problem is the in generic codegen code, so it affects multiple targets: we saw it in x86 (that's the one we care), but we also coudl reproduce on arm, and there might be others.

The lli based test has the advantage that it can exercise and verify this bug on multiple targets.  But if you prefer a x86 assembly test we can do that too.


As long it is the final say on this matter, we're fine either way.


Jose