================
@@ -1446,6 +1452,90 @@ bool SPIRVInstructionSelector::selectRsqrt(Register
ResVReg,
.constrainAllUses(TII, TRI, RBI);
}
+// Select the OpDot instruction for the given float dot
+bool SPIRVInstructionSelector::selectFloatDot(Register ResVReg,
+ const SPIRVType *ResType,
+ MachineInstr &I) const {
+ assert(I.getNumOperands() == 4);
+ assert(I.getOperand(2).isReg());
+ assert(I.getOperand(3).isReg());
+
+ [[maybe_unused]] SPIRVType *VecType =
+ GR.getSPIRVTypeForVReg(I.getOperand(2).getReg());
+
+ assert(GR.getScalarOrVectorComponentCount(VecType) > 1 &&
+ "dot product requires a vector of at least 2 components");
+
----------------
farzonl wrote:
should we also add an assert to make sure its a float vector? I know we have
guarantees about that via `Intrinsic::spv_fdot` but I was concerned about
misuse of the `selectFloatDot` function if say someone else wants to come along
and use it for opencl? Its a nit pick so feel free to ignore.
https://github.com/llvm/llvm-project/pull/104656
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
