[SYCL] Use checkArgCountAtLeast instead

[premanandrao]

Rather than replicate the code, call existing routine checkArgCountAtLeast
for similar functionality.

Add a similar routine checkArgCountAtMost to check if a call expression's
argument count exceeds the most expected. Replace where equivalent
with calls to this routine.

[premanandrao]

This was a post-commit review comment from @erichkeane

[smanna12]

LGTM.

One question:

// Make sure we don't have too many arguments.
  if (NumArgs > MaxNumArgs)
    return Diag(TheCall->getEndLoc(),
                diag::err_typecheck_call_too_many_args_at_most)
           << 0 /*function call*/ << MaxNumArgs << NumArgs
           << TheCall->getSourceRange();

``

I am wondering whether we can use checkAtMostNumArgs() functionality instead for bottom one.

[premanandrao]

I am wondering whether we can use checkAtMostNumArgs() functionality instead for bottom one.

That particular routine is for checking the number of arguments in an attribute, if I am not mistaken. But I could mimic checkArgCountAtLeast (in SemaChecking.cpp) and create a checkArgCountAtMost for the second check. Since this builtin would be the only one calling it, I wasn't too motivated to do that as part of this change.

[smanna12]

But I could mimic checkArgCountAtLeast (in SemaChecking.cpp) and create a checkArgCountAtMost for the second check. Since this builtin would be the only one calling it, I wasn't too motivated to do that as part of this change.

Agreed.

[premanandrao]

The CUDA failure is unrelated to my changes.

[erichkeane]

Adding a checkArgCountAtMost would be a nice addition, but up to you guys whether it is necessary for htis patch.

[smanna12]

Thank you.

@premanandrao, could you please update PR description?

[premanandrao]

@premanandrao, could you please update PR description?

Thank you, done.

[premanandrao]

No need to review the latest force push; I did that to force testing after a rebase. (The PR is in draft status.)

[premanandrao]

Only one test failure now: SYCL :: Assert/assert_in_simultaneous_kernels.cpp, which is unrelated to my changes.

@smanna12, could you please approve my changes again? There are no changes since you last approved it, except for an empty forced push. Thanks!

[premanandrao]

@intel/llvm-gatekeepers, this is ready for a merge. Thanks!