woruyu wrote:
> @woruyu Are you able to follow up on this?
I want to ask some quetion before writing code:
```c++
// curent - llvm
@llvm.masked.load (ptr, mask, passthru) --- three args
// curent - cir
mlir::Value createMaskedLoad(mlir::Location loc, mlir::Type ty,
mlir::Value ptr, llvm::Align alignment,
mlir::Value mask, mlir::Value passThru) {
....
mlir::Value ops[] = {ptr, this->getUInt32(int32_t(alignment.value()), loc),
mask, passThru};
return cir::LLVMIntrinsicCallOp::create(
*this, loc, getStringAttr("masked.load"), ty, ops)
.getResult();
}
--- ops = (ptr, align, mask, passThru) --- four args
case X86::BI__builtin_ia32_loaddqudi512_mask:
return emitX86MaskedLoad(builder, ops, llvm::Align(1),
getLoc(expr->getExprLoc()));
```
you can see that the extra align operand is default align(1), which is useless
as argument, so Is just deleting align arg better choice than redefining mask
op in cir and writing lowering code for it.
https://github.com/llvm/llvm-project/pull/169464
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
