[RISCV][GISel] Add initial support for rvv intrinsics

[jacquesguan]

This pr removes the falling back to SDISel of rvv intrinsics and marks them legalized in the legalize pass. Another pr would be created for regbankselect pass to make vf form intriniscs have the right scalar register bank.

[github-actions]

⚠️ undef deprecator found issues in your code. ⚠️

You can test this locally with the following command:

git diff -U0 --pickaxe-regex -S '([^a-zA-Z0-9#_-]undef[^a-zA-Z0-9_-]|UndefValue::get)' 'HEAD~1' HEAD llvm/test/CodeGen/RISCV/GlobalISel/rvv/vfadd.ll llvm/lib/Target/RISCV/GISel/RISCVLegalizerInfo.cpp llvm/lib/Target/RISCV/RISCVISelLowering.cpp llvm/test/CodeGen/RISCV/GlobalISel/irtranslator/fallback.ll

The following files introduce new uses of undef:

• llvm/test/CodeGen/RISCV/GlobalISel/rvv/vfadd.ll

Undef is now deprecated and should only be used in the rare cases where no replacement is possible. For example, a load of uninitialized memory yields undef. You should use poison values for placeholders instead.

In tests, avoid using undef and having tests that trigger undefined behavior. If you need an operand with some unimportant value, you can add a new argument to the function and use that instead.

For example, this is considered a bad practice:

define void @fn() {
  ...
  br i1 undef, ...
}

Please use the following instead:

define void @fn(i1 %cond) {
  ...
  br i1 %cond, ...
}

Please refer to the Undefined Behavior Manual for more information.

[jroelofs]

If II is otherwise unused, this should be:

Suggested change

	const RISCVVIntrinsicsTable::RISCVVIntrinsicInfo *II =
	RISCVVIntrinsicsTable::getRISCVVIntrinsicInfo(IntrinsicID);
	if (II) {
	return true;
	}
	if (RISCVVIntrinsicsTable::getRISCVVIntrinsicInfo(IntrinsicID))
	return true;

[jacquesguan]

Addressed, I was thinking to use the intrinsic info to identify which rvv intrinsics need specific legalization, since the change is not in this pr, I changed this to the suggested form.

[jroelofs]

Suggested change

	if (Op == Instruction::Call) {
	const CallInst &CI = cast<CallInst>(Inst);
	const Function *F = CI.getCalledFunction();
	Intrinsic::ID ID = F ? F->getIntrinsicID() : Intrinsic::not_intrinsic;
	const RISCVVIntrinsicsTable::RISCVVIntrinsicInfo *II =
	RISCVVIntrinsicsTable::getRISCVVIntrinsicInfo(ID);
	// Mark RVV intrinsic is supported.
	if (II)
	return false;
	}
	if (auto *II = dyn_cast<IntrinsicInst>(&Inst)) {
	// Mark RVV intrinsic as supported.
	if (RISCVVIntrinsicsTable::getRISCVVIntrinsicInfo(II->getIntrinsicID()))
	return false;
	}

[jacquesguan]

Addressed.

[topperc]

Another pr would be created for regbankselect pass to make vf form intriniscs have the right scalar register bank.

What about i64 scalar operands on RV32?

[mshockwave]

we no longer need to declare intrinsics

[jacquesguan]

Thanks to point this, this test case is just copied from the SDISel case, maybe we need to refactor them all.

[mshockwave]

nit: for most of the intrinsics here, their operands should have the same type, so a function like intrinsic_vfadd_vv_nxv1f16 should suffice

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[jacquesguan]

Another pr would be created for regbankselect pass to make vf form intriniscs have the right scalar register bank.

What about i64 scalar operands on RV32?

Not only the i64 scalar operands on RV32, all integer scalars that is not XLen bits should be legalized. I plan to implement these in the next pr to support vx and vf form intrinsics.

[jroelofs]

nit: llvm style prefers simple stmt bodies don't have {}s

Suggested change

	if (RISCVVIntrinsicsTable::getRISCVVIntrinsicInfo(IntrinsicID)) {
	return true;
	}
	if (RISCVVIntrinsicsTable::getRISCVVIntrinsicInfo(IntrinsicID))
	return true;

[jroelofs]

okay to keep them here though, because of the comment.