[SYCL] Remove manual diagnostic checking for stmt attrs.

[AaronBallman]

Statement attributes now have automatic diagnostic checking, so let's
use that. This does have a change in behavior in that passing the wrong
number of arguments is now an error instead of a warning, but that
behavior is more consistent with other attributes.

[smanna12]

LGTM

[AaronBallman]

The Jenkins test failure is covered by: #8814

[premanandrao]

Changes look good to me. Looks like there is a conflict in attr.td though.

[premanandrao]

LGTM

[AaronBallman]

FYI, I found one more cleanup that I'm in the process of testing out locally. I thought it was an upstream attribute but it turns out it's one of ours. Sorry for the noise on the review!

[AaronBallman]

Assuming CI is happy, that should be the last of the changes on this patch. I love seeing all that red with so little green. :-)

[premanandrao]

LGTM

[smanna12]

@AaronBallman, we are still using /opt/1 for IVDEP attr and SYCLIntelFPGALoopCoalesce. Do we still need them?

[AaronBallman]

@AaronBallman, we are still using /opt/1 for IVDEP attr. Do we still need this?

I think that's something we should investigate. It looks like we handle the case where no arguments are given (https://github.com/intel/llvm/blob/sycl/clang/lib/Sema/SemaStmtAttr.cpp#L272), but one thing that confuses me is the attribute takes three optional arguments when the semantic handler for it only ever seems to care about two of them. :-/

[AaronBallman]

Btw, as best I can tell, the test failure in check-sycl is unrelated to my changes.

[smanna12]

@AaronBallman, we are still using /opt/1 for IVDEP attr. Do we still need this?

I think that's something we should investigate. It looks like we handle the case where no arguments are given (https://github.com/intel/llvm/blob/sycl/clang/lib/Sema/SemaStmtAttr.cpp#L272), but one thing that confuses me is the attribute takes three optional arguments when the semantic handler for it only ever seems to care about two of them. :-/

Thanks for the explanation. May be we can create a new issue so that we do not forget about this.

[bader]

Btw, as best I can tell, the test failure in check-sycl is unrelated to my changes.

The problem has been addressed, so I restarted validation to confirm.

[bader]

Resolved merge conflict with #3388 - no functional changes.

[premanandrao]

LGTM