[CIR][CIRGen][Builtin][Neon] Lower neon_vshl_n_v and neon_vshlq_n_v

[ghehg]

As title, but important step in this PR is to allow CIR ShiftOp to take vector of int type as input type. As result, I added a verifier to ShiftOp with 2 constraints

1. Input type either all vector or int type. This is consistent with LLVM::ShlOp, vector shift amount is expected.
2. In the spirit of C99 6.5.7.3, shift amount type must be the same as result type, the if vector type is used. (This is enforced in LLVM lowering for scalar int type).

[ghehg]

Read for review, but keep it draft as I anticipate manual merging will happen once other pending PR's accepted

[bcardosolopes]

Thanks! Some minor changes needed

[dkolsen-pgi]

It was my mistake that I forgot to implement the shift operators for GNU vector types as part of issue #284. I even have a test program for that in my test area. But I must have assumed that BinOp included all the binary operators that I needed to worry about and didn't go looking for other binary operators that were implemented with different CIR ops.

Changing ShiftOp to also handle vector types is the right thing to do.

If this change fixes vector shift operations in user code (it looks like it might, but I'm not certain), then please update the PR description to say that, and please add shift operations to clang/test/CIR/CodeGen/vectype.cpp (and maybe also to clang/test/CIR/Lowering/vectype.cpp).

If this change doesn't fix vector shift operations, then leave this change alone and open an issue about this. Someone, probably me, will finish fixing that in a later PR.

[ghehg]

It was my mistake that I forgot to implement the shift operators for GNU vector types as part of issue #284. I even have a test program for that in my test area. But I must have assumed that BinOp included all the binary operators that I needed to worry about and didn't go looking for other binary operators that were implemented with different CIR ops.

Changing ShiftOp to also handle vector types is the right thing to do.

If this change fixes vector shift operations in user code (it looks like it might, but I'm not certain), then please update the PR description to say that, and please add shift operations to clang/test/CIR/CodeGen/vectype.cpp (and maybe also to clang/test/CIR/Lowering/vectype.cpp).

If this change doesn't fix vector shift operations, then leave this change alone and open an issue about this. Someone, probably me, will finish fixing that in a later PR.

The change might support user code, I'll try to find user test code to see if this is the case, if this change alone is enough to support user code, I'll just update clang/test/CIR/CodeGen/vectype.cpp and clang/test/CIR/Lowering/vectype.cpp.
Otherwise, I'll open an issue and either you or I can fix it in a separate PR.

[dkolsen-pgi]

The test case that I wrote (and then forgot about and didn't use) for vector shift is:

typedef int vi4 __attribute__((vector_size(16)));
int main() {
  vi4 a = { 1, 2, 3, 4 };
  vi4 b = { 5, 6, 7, 8 };
  vi4 c = a << b;
  vi4 d = a >> b;
  return c[1] + d[1];
}

You can start with that for your tests of user code.

[ghehg]

user code support is confirmed. Added user code test cases covering left, right(logic and arith) shifts

[bcardosolopes]

Sounds like we have a buggy " right" with an extra space, let me create an issue.

[ghehg]

Haha, I noticed that yesterday but not sure that extra space was on purpose as we had existing test case with that