[AArch64] Add intrinsic to count trailing zero elements

This patch introduces an experimental intrinsic for counting the
trailing zero elements in a vector. The intrinsic has generic expansion
in SelectionDAGBuilder, and for AArch64 there is a pattern which matches
to brkb & cntp instructions where SVE is enabled.

The intrinsic has a second operand, is_zero_poison, similar to the
existing cttz intrinsic.

These changes have been split out from D158291.

Author: kmclaughlin-arm

llvm/docs/LangRef.rst

@@ -18497,6 +18497,45 @@ Arguments:
 Both arguments must be vectors of the same type whereby their logical
 concatenation matches the result type.
 
+'``llvm.experimental.cttz.elts``' Intrinsic
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Syntax:
+"""""""
+
+This is an overloaded intrinsic. You can use ```llvm.experimental.cttz.elts```
+on any vector of integer elements, both fixed width and scalable.
+
+::
+
+      declare i8 @llvm.experimental.cttz.elts.i8.v8i1(<8 x i1> <src>, i1 <is_zero_poison>)
+
+Overview:
+"""""""""
+
+The '``llvm.experimental.cttz.elts``' intrinsic counts the number of trailing
+zero elements of a vector.
+
+Arguments:
+""""""""""
+
+The first argument is the vector to be counted. This argument must be a vector
+with integer element type. The return type must also be an integer type which is
+wide enough to hold the maximum number of elements of the source vector. The
+behaviour of this intrinsic is undefined if the return type is not wide enough
+for the number of elements in the input vector.
+
+The second argument is a constant flag that indicates whether the intrinsic
+returns a valid result if the first argument is all zero. If the first argument
+is all zero and the second argument is true, the result is poison.
+
+Semantics:
+""""""""""
+
+The '``llvm.experimental.cttz.elts``' intrinsic counts the trailing (least
+significant) zero elements in a vector. If ``src == 0`` the result is the
+number of elements in the input vector.
+
 '``llvm.experimental.vector.splice``' Intrinsic
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 

llvm/include/llvm/CodeGen/TargetLowering.h

@@ -465,6 +465,10 @@ class TargetLoweringBase {
     return true;
   }
 
+  /// Return true if the @llvm.experimental.cttz.elts intrinsic should be
+  /// expanded using generic code in SelectionDAGBuilder.
+  virtual bool shouldExpandCttzElements(EVT VT) const { return true; }
+
   // Return true if op(vecreduce(x), vecreduce(y)) should be reassociated to
   // vecreduce(op(x, y)) for the reduction opcode RedOpc.
   virtual bool shouldReassociateReduction(unsigned RedOpc, EVT VT) const {

llvm/include/llvm/IR/Intrinsics.td

@@ -2182,6 +2182,11 @@ def int_experimental_get_vector_length:
                         [IntrNoMem, IntrNoSync, IntrWillReturn,
                          ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<2>>]>;
 
+def int_experimental_cttz_elts:
+  DefaultAttrsIntrinsic<[llvm_anyint_ty],
+            [llvm_anyvector_ty, llvm_i1_ty],
+            [IntrNoMem, IntrNoSync, IntrWillReturn, ImmArg<ArgIndex<1>>]>;
+
 def int_experimental_vp_splice:
   DefaultAttrsIntrinsic<[llvm_anyvector_ty],
             [LLVMMatchType<0>,

llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

@@ -7514,6 +7514,62 @@ void SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I,
     setValue(&I, Trunc);
     return;
   }
+  case Intrinsic::experimental_cttz_elts: {
+    auto DL = getCurSDLoc();
+    SDValue Op = getValue(I.getOperand(0));
+    EVT OpVT = Op.getValueType();
+
+    if (!TLI.shouldExpandCttzElements(OpVT)) {
+      visitTargetIntrinsic(I, Intrinsic);
+      return;
+    }
+
+    if (OpVT.getScalarType() != MVT::i1) {
+      // Compare the input vector elements to zero & use to count trailing zeros
+      SDValue AllZero = DAG.getConstant(0, DL, OpVT);
+      OpVT = EVT::getVectorVT(*DAG.getContext(), MVT::i1,
+                              OpVT.getVectorElementCount());
+      Op = DAG.getSetCC(DL, OpVT, Op, AllZero, ISD::SETNE);
+    }
+
+    // Find the smallest "sensible" element type to use for the expansion.
+    ConstantRange CR(
+        APInt(64, OpVT.getVectorElementCount().getKnownMinValue()));
+    if (OpVT.isScalableVT())
+      CR = CR.umul_sat(getVScaleRange(I.getCaller(), 64));
+
+    // If the zero-is-poison flag is set, we can assume the upper limit
+    // of the result is VF-1.
+    if (!cast<ConstantSDNode>(getValue(I.getOperand(1)))->isZero())
+      CR = CR.subtract(APInt(64, 1));
+
+    unsigned EltWidth = I.getType()->getScalarSizeInBits();
+    EltWidth = std::min(EltWidth, (unsigned)CR.getActiveBits());
+    EltWidth = std::max(llvm::bit_ceil(EltWidth), (unsigned)8);
+
+    MVT NewEltTy = MVT::getIntegerVT(EltWidth);
+
+    // Create the new vector type & get the vector length
+    EVT NewVT = EVT::getVectorVT(*DAG.getContext(), NewEltTy,
+                                 OpVT.getVectorElementCount());
+
+    SDValue VL =
+        DAG.getElementCount(DL, NewEltTy, OpVT.getVectorElementCount());
+
+    SDValue StepVec = DAG.getStepVector(DL, NewVT);
+    SDValue SplatVL = DAG.getSplat(NewVT, DL, VL);
+    SDValue StepVL = DAG.getNode(ISD::SUB, DL, NewVT, SplatVL, StepVec);
+    SDValue Ext = DAG.getNode(ISD::SIGN_EXTEND, DL, NewVT, Op);
+    SDValue And = DAG.getNode(ISD::AND, DL, NewVT, StepVL, Ext);
+    SDValue Max = DAG.getNode(ISD::VECREDUCE_UMAX, DL, NewEltTy, And);
+    SDValue Sub = DAG.getNode(ISD::SUB, DL, NewEltTy, VL, Max);
+
+    EVT RetTy = TLI.getValueType(DAG.getDataLayout(), I.getType());
+    SDValue Ret = DAG.getZExtOrTrunc(Sub, DL, RetTy);
+
+    setValue(&I, Ret);
+    return;
+  }
   case Intrinsic::vector_insert: {
     SDValue Vec = getValue(I.getOperand(0));
     SDValue SubVec = getValue(I.getOperand(1));

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -1791,6 +1791,10 @@ bool AArch64TargetLowering::shouldExpandGetActiveLaneMask(EVT ResVT,
   return false;
 }
 
+bool AArch64TargetLowering::shouldExpandCttzElements(EVT VT) const {
+  return !Subtarget->hasSVEorSME() || VT != MVT::nxv16i1;
+}
+
 void AArch64TargetLowering::addTypeForFixedLengthSVE(MVT VT,
                                                      bool StreamingSVE) {
   assert(VT.isFixedLengthVector() && "Expected fixed length vector type!");
@@ -2634,6 +2638,7 @@ const char *AArch64TargetLowering::getTargetNodeName(unsigned Opcode) const {
     MAKE_CASE(AArch64ISD::MRRS)
     MAKE_CASE(AArch64ISD::MSRR)
     MAKE_CASE(AArch64ISD::RSHRNB_I)
+    MAKE_CASE(AArch64ISD::CTTZ_ELTS)
   }
 #undef MAKE_CASE
   return nullptr;
@@ -5338,6 +5343,12 @@ SDValue AArch64TargetLowering::LowerINTRINSIC_WO_CHAIN(SDValue Op,
     }
     return SDValue();
   }
+  case Intrinsic::experimental_cttz_elts: {
+    SDValue NewCttzElts =
+        DAG.getNode(AArch64ISD::CTTZ_ELTS, dl, MVT::i64, Op.getOperand(1));
+
+    return DAG.getZExtOrTrunc(NewCttzElts, dl, Op.getValueType());
+  }
   }
 }
 

llvm/lib/Target/AArch64/AArch64ISelLowering.h

@@ -335,6 +335,8 @@ enum NodeType : unsigned {
   PTEST_ANY,
   PTRUE,
 
+  CTTZ_ELTS,
+
   BITREVERSE_MERGE_PASSTHRU,
   BSWAP_MERGE_PASSTHRU,
   REVH_MERGE_PASSTHRU,
@@ -927,6 +929,8 @@ class AArch64TargetLowering : public TargetLowering {
 
   bool shouldExpandGetActiveLaneMask(EVT VT, EVT OpVT) const override;
 
+  bool shouldExpandCttzElements(EVT VT) const override;
+
   /// If a change in streaming mode is required on entry to/return from a
   /// function call it emits and returns the corresponding SMSTART or SMSTOP node.
   /// \p Entry tells whether this is before/after the Call, which is necessary

llvm/lib/Target/AArch64/AArch64InstrInfo.td

@@ -842,6 +842,9 @@ def AArch64rshrnb_pf : PatFrags<(ops node:$rs, node:$i),
                             [(AArch64rshrnb node:$rs, node:$i),
                             (int_aarch64_sve_rshrnb node:$rs, node:$i)]>;
 
+def AArch64CttzElts : SDNode<"AArch64ISD::CTTZ_ELTS", SDTypeProfile<1, 1,
+                             [SDTCisInt<0>, SDTCisVec<1>]>, []>;
+
 // Match add node and also treat an 'or' node is as an 'add' if the or'ed operands
 // have no common bits.
 def add_and_or_is_add : PatFrags<(ops node:$lhs, node:$rhs),

llvm/lib/Target/AArch64/AArch64SVEInstrInfo.td

@@ -1964,6 +1964,11 @@ let Predicates = [HasSVEorSME] in {
   defm CNTW_XPiI : sve_int_count<0b100, "cntw", int_aarch64_sve_cntw>;
   defm CNTD_XPiI : sve_int_count<0b110, "cntd", int_aarch64_sve_cntd>;
   defm CNTP_XPP : sve_int_pcount_pred<0b0000, "cntp", int_aarch64_sve_cntp>;
+
+  def : Pat<(i64 (AArch64CttzElts nxv16i1:$Op1)),
+                 (i64 (!cast<Instruction>(CNTP_XPP_B)
+                        (nxv16i1 (!cast<Instruction>(BRKB_PPzP) (PTRUE_B 31), nxv16i1:$Op1)),
+                        (nxv16i1 (!cast<Instruction>(BRKB_PPzP) (PTRUE_B 31), nxv16i1:$Op1))))>;
 }
 
   defm INCB_XPiI : sve_int_pred_pattern_a<0b000, "incb", add, int_aarch64_sve_cntb>;
@@ -2049,6 +2054,17 @@ let Predicates = [HasSVEorSME] in {
   defm INCP_ZP     : sve_int_count_v<0b10000, "incp">;
   defm DECP_ZP     : sve_int_count_v<0b10100, "decp">;
 
+  def : Pat<(i64 (add GPR64:$Op1, (i64 (AArch64CttzElts nxv16i1:$Op2)))),
+                 (i64 (!cast<Instruction>(INCP_XP_B)
+                        (nxv16i1 (!cast<Instruction>(BRKB_PPzP) (PTRUE_B 31), nxv16i1:$Op2)),
+                        GPR64:$Op1))>;
+
+  def : Pat<(i32 (add GPR32:$Op1, (trunc (i64 (AArch64CttzElts nxv16i1:$Op2))))),
+                 (i32 (EXTRACT_SUBREG (i64 (!cast<Instruction>(INCP_XP_B)
+                        (nxv16i1 (!cast<Instruction>(BRKB_PPzP) (PTRUE_B 31), nxv16i1:$Op2)),
+                        (INSERT_SUBREG (i64 (IMPLICIT_DEF)), GPR32:$Op1, sub_32))),
+                      sub_32))>;
+
   defm INDEX_RR : sve_int_index_rr<"index", AArch64mul_p_oneuse>;
   defm INDEX_IR : sve_int_index_ir<"index", AArch64mul_p, AArch64mul_p_oneuse>;
   defm INDEX_RI : sve_int_index_ri<"index">;