[AArch64][SVE] Change the asserts in LowerToPredicatedOp to check for legal types

[david-arm]

Summary

When building the LLVM test suite with SVE I discovered a crash
when compiling some Halide tests, which occurs because we try to
use SVE to lower 64-bit vector multiplies and there is no
vscale_range attribute on the function. In this case the min SVE
vector bits was 0, which caused an assert in LowerToPredicatedOp
to fire. I have amended the asserts in this function to check that the
fixed-width type is legal. If the fixed-width type is larger than NEON
and is legal then it must be because we've set the min SVE vector
bits to something > 128. Or if the min SVE bits is 0, then the only
legal types allowed are 128 bit types - for any other types the assert
will fire.

Tests added here:

CodeGen/AArch64/sve-fixed-length-no-vscale-range.ll

[david-arm]

• I have tried an alternative approach with this latest patch. Instead of changing the default min SVE vector bits to 128, I have amended the asserts in LowerToPredicatedOp to check that the fixed-width type is legal.

[paulwalker-arm]

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

20034

This is a target lowering function so you should be able to call isTypeLegal() directly.

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

 SDValue AArch64TargetLowering::LowerToPredicatedOp(SDValue Op,
                                                    SelectionDAG &DAG,
                                                    unsigned NewOp) const {
   EVT VT = Op.getValueType();
   SDLoc DL(Op);
   auto Pg = getPredicateForVector(DAG, DL, VT);
 
   if (VT.isFixedLengthVector()) {
-    assert(VT.getFixedSizeInBits() <= Subtarget->getMinSVEVectorSizeInBits() &&
-           "Cannot use SVE to lower fixed length predicated op!");
+    assert(isTypeLegal(VT) && "Expected only legal fixed-width types");
     EVT ContainerVT = getContainerForFixedLengthVector(DAG, VT);
 
     // Create list of operands by converting existing ones to scalable types.
     SmallVector<SDValue, 4> Operands = {Pg};
     for (const SDValue &V : Op->op_values()) {
       if (isa<CondCodeSDNode>(V)) {
         Operands.push_back(V);
         continue;
       }
 
       if (const VTSDNode *VTNode = dyn_cast<VTSDNode>(V)) {
         EVT VTArg = VTNode->getVT().getVectorElementType();
         EVT NewVTArg = ContainerVT.changeVectorElementType(VTArg);
         Operands.push_back(DAG.getValueType(NewVTArg));
         continue;
       }
 
-      assert(V.getValueType().getFixedSizeInBits() <=
-                 Subtarget->getMinSVEVectorSizeInBits() &&
-             "Only fixed length vectors are supported!");
+      assert(isTypeLegal(V.getValueType()) &&
+             "Expected only legal fixed-width types");
       Operands.push_back(convertToScalableVector(DAG, ContainerVT, V));
     }
 
     if (isMergePassthruOpcode(NewOp))
       Operands.push_back(DAG.getUNDEF(ContainerVT));
 
     auto ScalableRes = DAG.getNode(NewOp, DL, ContainerVT, Operands);
     return convertFromScalableVector(DAG, VT, ScalableRes);

llvm/test/CodeGen/AArch64/sve-fixed-length-no-vscale-range.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s | FileCheck %s
+
+target triple = "aarch64-unknown-linux-gnu"
+
+define <2 x i64> @mul_v2i64(<2 x i64> %op1, <2 x i64> %op2) #0 {
+; CHECK-LABEL: mul_v2i64:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
+; CHECK-NEXT:    ptrue p0.d, vl2
+; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    mul z0.d, p0/m, z0.d, z1.d
+; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
+; CHECK-NEXT:    ret
+  %res = mul <2 x i64> %op1, %op2
+  ret <2 x i64> %res
+}
+
+define <4 x i32> @sdiv_v4i32(<4 x i32> %op1, <4 x i32> %op2) #0 {
+; CHECK-LABEL: sdiv_v4i32:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    // kill: def $q0 killed $q0 def $z0
+; CHECK-NEXT:    ptrue p0.s, vl4
+; CHECK-NEXT:    // kill: def $q1 killed $q1 def $z1
+; CHECK-NEXT:    sdiv z0.s, p0/m, z0.s, z1.s
+; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $z0
+; CHECK-NEXT:    ret
+  %res = sdiv <4 x i32> %op1, %op2
+  ret <4 x i32> %res
+}
+
+attributes #0 = { "target-features"="+sve" }