[AArch64] Handle extract_subvector(..., 0) in ISel.

[• chatur01]

Summary

Lowering this pattern early to an EXTRACT_SUBREG was making it impossible to match larger patterns in tblgen that use extract_subvector(..., 0) as part of the their input pattern.

It seems like there will exist somewhere a better way of specifying this pattern over all relevant register value types, but I didn't manage to find it.

[• chatur01]

Ping

[jmolloy]

Hi Charlie,

This looks totally reasonable to me. I also know you've thrown all the testing we've got at it, so LGTM.

James

lib/Target/AArch64/AArch64ISelLowering.cpp

if (!VT.isVector())
     return SDValue();
 
   ConstantSDNode *Cst = dyn_cast<ConstantSDNode>(Op.getOperand(1));
   if (!Cst)
     return SDValue();
   unsigned Val = Cst->getZExtValue();
 
   unsigned Size = Op.getValueType().getSizeInBits();
-  if (Val == 0) {
-    switch (Size) {
-    case 8:
-      return DAG.getTargetExtractSubreg(AArch64::bsub, dl, Op.getValueType(),
-                                        Op.getOperand(0));
-    case 16:
-      return DAG.getTargetExtractSubreg(AArch64::hsub, dl, Op.getValueType(),
-                                        Op.getOperand(0));
-    case 32:
-      return DAG.getTargetExtractSubreg(AArch64::ssub, dl, Op.getValueType(),
-                                        Op.getOperand(0));
-    case 64:
-      return DAG.getTargetExtractSubreg(AArch64::dsub, dl, Op.getValueType(),
-                                        Op.getOperand(0));
-    default:
-      llvm_unreachable("Unexpected vector type in extract_subvector!");
-    }
-  }
+
+  // This will get lowered to an appropriate EXTRACT_SUBREG in ISel.
+  if (Val == 0)
+    return Op;
+
   // If this is extracting the upper 64-bits of a 128-bit vector, we match
   // that directly.
   if (Size == 64 && Val * VT.getVectorElementType().getSizeInBits() == 64)
     return Op;
 
   return SDValue();
 }
 

lib/Target/AArch64/AArch64InstrInfo.td

 def : Pat<(v16i8 (bitconvert (v2f64 FPR128:$src))),
                              (v16i8 (REV64v16i8 FPR128:$src))>;
 def : Pat<(v16i8 (bitconvert (v4f32 FPR128:$src))),
                              (v16i8 (REV32v16i8 FPR128:$src))>;
 def : Pat<(v16i8 (bitconvert (v8f16 FPR128:$src))),
                              (v16i8 (REV16v16i8 FPR128:$src))>;
 }
 
+def : Pat<(v4i16 (extract_subvector V128:$Rn, (i64 0))),
+           (EXTRACT_SUBREG V128:$Rn, dsub)>;
+def : Pat<(v8i8 (extract_subvector V128:$Rn, (i64 0))),
+           (EXTRACT_SUBREG V128:$Rn, dsub)>;
+def : Pat<(v2f32 (extract_subvector V128:$Rn, (i64 0))),
+           (EXTRACT_SUBREG V128:$Rn, dsub)>;
+def : Pat<(v4f16 (extract_subvector V128:$Rn, (i64 0))),
+           (EXTRACT_SUBREG V128:$Rn, dsub)>;
+def : Pat<(v2i32 (extract_subvector V128:$Rn, (i64 0))),
+           (EXTRACT_SUBREG V128:$Rn, dsub)>;
+def : Pat<(v1i64 (extract_subvector V128:$Rn, (i64 0))),
+           (EXTRACT_SUBREG V128:$Rn, dsub)>;
+def : Pat<(v1f64 (extract_subvector V128:$Rn, (i64 0))),
+           (EXTRACT_SUBREG V128:$Rn, dsub)>;
+
 def : Pat<(v8i8 (extract_subvector (v16i8 FPR128:$Rn), (i64 1))),
           (EXTRACT_SUBREG (DUPv2i64lane FPR128:$Rn, 1), dsub)>;
 def : Pat<(v4i16 (extract_subvector (v8i16 FPR128:$Rn), (i64 1))),
           (EXTRACT_SUBREG (DUPv2i64lane FPR128:$Rn, 1), dsub)>;
 def : Pat<(v2i32 (extract_subvector (v4i32 FPR128:$Rn), (i64 1))),
           (EXTRACT_SUBREG (DUPv2i64lane FPR128:$Rn, 1), dsub)>;
 def : Pat<(v1i64 (extract_subvector (v2i64 FPR128:$Rn), (i64 1))),
           (EXTRACT_SUBREG (DUPv2i64lane FPR128:$Rn, 1), dsub)>;