[llvm] [AMDGPU] Implement moveToVALU for S_CSELECT_B64 (PR #70352)

[Stanislav Mekhanoshin via llvm-commits]

================
@@ -4737,20 +4737,42 @@ MachineBasicBlock *SITargetLowering::EmitInstrWithCustomInserter(
     const auto *CondRC = TRI->getRegClass(AMDGPU::SReg_1_XEXECRegClassID);
     Register SrcCondCopy = MRI.createVirtualRegister(CondRC);
 
+    const TargetRegisterClass *Src0RC = Src0.isReg()
+                                            ? MRI.getRegClass(Src0.getReg())
+                                            : &AMDGPU::VReg_64RegClass;
+    const TargetRegisterClass *Src1RC = Src1.isReg()
+                                            ? MRI.getRegClass(Src1.getReg())
+                                            : &AMDGPU::VReg_64RegClass;
+
+    const TargetRegisterClass *Src0SubRC =
+        TRI->getSubRegisterClass(Src0RC, AMDGPU::sub0);
+    const TargetRegisterClass *Src1SubRC =
+        TRI->getSubRegisterClass(Src1RC, AMDGPU::sub1);
+
+    MachineOperand Src0Sub0 = TII->buildExtractSubRegOrImm(
+        MI, MRI, Src0, Src0RC, AMDGPU::sub0, Src0SubRC);
+    MachineOperand Src1Sub0 = TII->buildExtractSubRegOrImm(
+        MI, MRI, Src1, Src1RC, AMDGPU::sub0, Src1SubRC);
+
+    MachineOperand Src0Sub1 = TII->buildExtractSubRegOrImm(
+        MI, MRI, Src0, Src0RC, AMDGPU::sub1, Src0SubRC);
+    MachineOperand Src1Sub1 = TII->buildExtractSubRegOrImm(
----------------
rampitec wrote:

Wow! I didn't know it does exist. Seems useful.

https://github.com/llvm/llvm-project/pull/70352