Index: llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
===================================================================
--- llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -547,14 +547,34 @@
 
 static Optional<Instruction *> instCombineSVELast(InstCombiner &IC,
                                                   IntrinsicInst &II) {
+  IRBuilder<> Builder(II.getContext());
+  Builder.SetInsertPoint(&II);
   Value *Pg = II.getArgOperand(0);
   Value *Vec = II.getArgOperand(1);
-  bool IsAfter = II.getIntrinsicID() == Intrinsic::aarch64_sve_lasta;
+  auto IntrinsicID = II.getIntrinsicID();
+  bool IsAfter = IntrinsicID == Intrinsic::aarch64_sve_lasta;
 
   // lastX(splat(X)) --> X
   if (auto *SplatVal = getSplatValue(Vec))
     return IC.replaceInstUsesWith(II, SplatVal);
 
+  // If x and/or y is a splat value then:
+  // lastX (binop (x, y)) --> binop(lastX(x), lastX(y))
+  Value *LHS, *RHS;
+  if (match(Vec, m_OneUse(m_BinOp(m_Value(LHS), m_Value(RHS))))) {
+    if (isSplatValue(LHS) || isSplatValue(RHS)) {
+      auto OldBinOp = cast<BinaryOperator>(Vec);
+      auto OpC = OldBinOp->getOpcode();
+      auto *NewLHS =
+          Builder.CreateIntrinsic(IntrinsicID, {Vec->getType()}, {Pg, LHS});
+      auto *NewRHS =
+          Builder.CreateIntrinsic(IntrinsicID, {Vec->getType()}, {Pg, RHS});
+      auto *NewBinOp = BinaryOperator::CreateWithCopiedFlags(
+          OpC, NewLHS, NewRHS, OldBinOp, OldBinOp->getName(), &II);
+      return IC.replaceInstUsesWith(II, NewBinOp);
+    }
+  }
+
   auto *C = dyn_cast<Constant>(Pg);
   if (IsAfter && C && C->isNullValue()) {
     // The intrinsic is extracting lane 0 so use an extract instead.
Index: llvm/test/Transforms/InstCombine/AArch64/sve-intrinsic-opts-lasta-lastb.ll
===================================================================
--- llvm/test/Transforms/InstCombine/AArch64/sve-intrinsic-opts-lasta-lastb.ll
+++ llvm/test/Transforms/InstCombine/AArch64/sve-intrinsic-opts-lasta-lastb.ll
@@ -163,7 +163,225 @@
   ret i8 %last
 }
 
+; Check that we move the lastb before the binary operation so that the new binary op is scalar.
+define i8 @lastb_binop_RHS_splat_div(<vscale x 16 x i1> %pg, i8 %scalar, <vscale x 16 x i8> %vector) #0 {
+; OPT-LABEL: @lastb_binop_RHS_splat_div(
+; OPT-NEXT:    %1 = call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %vector)
+; OPT-NEXT:    %binop1 = udiv i8 %1, %scalar
+; OPT-NEXT:    ret i8 %binop1
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = udiv <vscale x 16 x i8> %vector, %splat
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+define i8 @lastb_binop_RHS_splat_sdiv_exact(<vscale x 16 x i1> %pg, i8 %scalar, <vscale x 16 x i8> %vector) #0 {
+; OPT-LABEL: @lastb_binop_RHS_splat_sdiv_exact(
+; OPT-NEXT:    %1 = call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %vector)
+; OPT-NEXT:    %binop1 = sdiv exact i8 %1, %scalar
+; OPT-NEXT:    ret i8 %binop1
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = sdiv exact <vscale x 16 x i8> %vector, %splat
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+define float @lastb_binop_RHS_splat_div_float_fast(<vscale x 4 x i1> %pg, float %scalar, <vscale x 4 x float> %vector) #0 {
+; OPT-LABEL: @lastb_binop_RHS_splat_div_float_fast(
+; OPT-NEXT:    %1 = call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %vector)
+; OPT-NEXT:    %binop1 = fdiv fast float %1, %scalar
+; OPT-NEXT:    ret float %binop1
+  %splat_insert = insertelement <vscale x 4 x float> poison, float %scalar, i32 0
+  %splat = shufflevector <vscale x 4 x float> %splat_insert, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+  %binop = fdiv fast <vscale x 4 x float> %vector, %splat
+  %last = tail call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %binop)
+  ret float %last
+}
+
+define float @lastb_binop_RHS_splat_div_float(<vscale x 4 x i1> %pg, float %scalar, <vscale x 4 x float> %vector) #0 {
+; OPT-LABEL: @lastb_binop_RHS_splat_div_float(
+; OPT-NEXT:    %1 = call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %vector)
+; OPT-NEXT:    %binop1 = fdiv float %1, %scalar
+; OPT-NEXT:    ret float %binop1
+  %splat_insert = insertelement <vscale x 4 x float> poison, float %scalar, i32 0
+  %splat = shufflevector <vscale x 4 x float> %splat_insert, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+  %binop = fdiv <vscale x 4 x float> %vector, %splat
+  %last = tail call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %binop)
+  ret float %last
+}
+
+define i8 @lastb_binop_LHS_splat_div(<vscale x 16 x i1> %pg, i8 %scalar, <vscale x 16 x i8> %vector) #0 {
+; OPT-LABEL: @lastb_binop_LHS_splat_div(
+; OPT-NEXT:    %1 = call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %vector)
+; OPT-NEXT:    %binop1 = udiv i8 %scalar, %1
+; OPT-NEXT:    ret i8 %binop1
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = udiv <vscale x 16 x i8> %splat, %vector
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+define i8 @lastb_binop_LHS_splat_div_sdiv_exact(<vscale x 16 x i1> %pg, i8 %scalar, <vscale x 16 x i8> %vector) #0 {
+; OPT-LABEL: @lastb_binop_LHS_splat_div_sdiv_exact(
+; OPT-NEXT:    %1 = call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %vector)
+; OPT-NEXT:    %binop1 = sdiv exact i8 %scalar, %1
+; OPT-NEXT:    ret i8 %binop1
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = sdiv exact <vscale x 16 x i8> %splat, %vector
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+define float @lastb_binop_LHS_splat_div_float_fast(<vscale x 4 x i1> %pg, float %scalar, <vscale x 4 x float> %vector) #0 {
+; OPT-LABEL: @lastb_binop_LHS_splat_div_float_fast(
+; OPT-NEXT:    %1 = call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %vector)
+; OPT-NEXT:    %binop1 = fdiv fast float %scalar, %1
+; OPT-NEXT:    ret float %binop1
+  %splat_insert = insertelement <vscale x 4 x float> poison, float %scalar, i32 0
+  %splat = shufflevector <vscale x 4 x float> %splat_insert, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+  %binop = fdiv fast <vscale x 4 x float> %splat, %vector
+  %last = tail call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %binop)
+  ret float %last
+}
+
+define float @lastb_binop_LHS_splat_div_float(<vscale x 4 x i1> %pg, float %scalar, <vscale x 4 x float> %vector) #0 {
+; OPT-LABEL: @lastb_binop_LHS_splat_div_float(
+; OPT-NEXT:    %1 = call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %vector)
+; OPT-NEXT:    %binop1 = fdiv float %scalar, %1
+; OPT-NEXT:    ret float %binop1
+  %splat_insert = insertelement <vscale x 4 x float> poison, float %scalar, i32 0
+  %splat = shufflevector <vscale x 4 x float> %splat_insert, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+  %binop = fdiv <vscale x 4 x float> %splat, %vector
+  %last = tail call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %binop)
+  ret float %last
+}
+
+define i8 @lastb_binop_LHS_RHS_splat_div(<vscale x 16 x i1> %pg, i8 %scalar1, i8 %scalar2) #0 {
+; OPT-LABEL: @lastb_binop_LHS_RHS_splat_div(
+; OPT-NEXT:    %binop1 = udiv i8 %scalar1, %scalar2
+; OPT-NEXT:    ret i8 %binop1
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar1, i8 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %splat_insert2 = insertelement <vscale x 16 x i8> poison, i8 %scalar2, i8 0
+  %splat2 = shufflevector <vscale x 16 x i8> %splat_insert2, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = udiv <vscale x 16 x i8> %splat, %splat2
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+define i8 @lastb_binop_RHS_splat_add(<vscale x 16 x i1> %pg, i8 %scalar, <vscale x 16 x i8> %vector) #0 {
+; OPT-LABEL: @lastb_binop_RHS_splat_add(
+; OPT-NEXT:    %1 = call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %vector)
+; OPT-NEXT:    %binop1 = add i8 %1, %scalar
+; OPT-NEXT:    ret i8 %binop1
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = add <vscale x 16 x i8> %vector, %splat
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+define float @lastb_binop_RHS_splat_add_float_fast(<vscale x 4 x i1> %pg, float %scalar, <vscale x 4 x float> %vector) #0 {
+; OPT-LABEL: @lastb_binop_RHS_splat_add_float_fast(
+; OPT-NEXT:    %1 = call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %vector)
+; OPT-NEXT:    %binop1 = fadd fast float %1, %scalar
+; OPT-NEXT:    ret float %binop1
+  %splat_insert = insertelement <vscale x 4 x float> poison, float %scalar, i32 0
+  %splat = shufflevector <vscale x 4 x float> %splat_insert, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+  %binop = fadd fast <vscale x 4 x float> %vector, %splat
+  %last = tail call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %binop)
+  ret float %last
+}
+
+define float @lastb_binop_RHS_splat_add_float(<vscale x 4 x i1> %pg, float %scalar, <vscale x 4 x float> %vector) #0 {
+; OPT-LABEL: @lastb_binop_RHS_splat_add_float(
+; OPT-NEXT:    %1 = call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %vector)
+; OPT-NEXT:    %binop1 = fadd float %1, %scalar
+; OPT-NEXT:    ret float %binop1
+  %splat_insert = insertelement <vscale x 4 x float> poison, float %scalar, i32 0
+  %splat = shufflevector <vscale x 4 x float> %splat_insert, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+  %binop = fadd <vscale x 4 x float> %vector, %splat
+  %last = tail call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %binop)
+  ret float %last
+}
+
+define i8 @lastb_binop_LHS_splat_add(<vscale x 16 x i1> %pg, i8 %scalar, <vscale x 16 x i8> %vector) #0 {
+; OPT-LABEL: @lastb_binop_LHS_splat_add(
+; OPT-NEXT:    %1 = call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %vector)
+; OPT-NEXT:    %binop1 = add i8 %1, %scalar
+; OPT-NEXT:    ret i8 %binop1
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = add <vscale x 16 x i8> %splat, %vector
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+define float @lastb_binop_LHS_splat_add_float_fast(<vscale x 4 x i1> %pg, float %scalar, <vscale x 4 x float> %vector) #0 {
+; OPT-LABEL: @lastb_binop_LHS_splat_add_float_fast(
+; OPT-NEXT:    %1 = call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %vector)
+; OPT-NEXT:    %binop1 = fadd fast float %1, %scalar
+; OPT-NEXT:    ret float %binop1
+  %splat_insert = insertelement <vscale x 4 x float> poison, float %scalar, i32 0
+  %splat = shufflevector <vscale x 4 x float> %splat_insert, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+  %binop = fadd fast <vscale x 4 x float> %splat, %vector
+  %last = tail call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %binop)
+  ret float %last
+}
+
+define float @lastb_binop_LHS_splat_add_float(<vscale x 4 x i1> %pg, float %scalar, <vscale x 4 x float> %vector) #0 {
+; OPT-LABEL: @lastb_binop_LHS_splat_add_float(
+; OPT-NEXT:    %1 = call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %vector)
+; OPT-NEXT:    %binop1 = fadd float %1, %scalar
+; OPT-NEXT:    ret float %binop1
+  %splat_insert = insertelement <vscale x 4 x float> poison, float %scalar, i32 0
+  %splat = shufflevector <vscale x 4 x float> %splat_insert, <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer
+  %binop = fadd <vscale x 4 x float> %splat, %vector
+  %last = tail call float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1> %pg, <vscale x 4 x float> %binop)
+  ret float %last
+}
+
+define i8 @lastb_binop_LHS_RHS_splat_add(<vscale x 16 x i1> %pg, i8 %scalar1, i8 %scalar2) #0 {
+; OPT-LABEL: @lastb_binop_LHS_RHS_splat_add(
+; OPT-NEXT:    %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar1, i8 0 
+; OPT-NEXT:    %splat_insert2 = insertelement <vscale x 16 x i8> poison, i8 %scalar2, i8 0
+; OPT-NEXT:    %1 = add <vscale x 16 x i8> %splat_insert, %splat_insert2 
+; OPT-NEXT:    %binop = shufflevector <vscale x 16 x i8> %1, <vscale x 16 x i8> undef, <vscale x 16 x i32> zeroinitializer
+; OPT-NEXT:    %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+; OPT-NEXT:    ret i8 %last
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar1, i8 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %splat_insert2 = insertelement <vscale x 16 x i8> poison, i8 %scalar2, i8 0
+  %splat2 = shufflevector <vscale x 16 x i8> %splat_insert2, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = add <vscale x 16 x i8> %splat, %splat2
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+; Check that we dont do anything as the binary op has multiple uses.
+define i8 @lastb_binop_nochange(<vscale x 16 x i1> %pg, i8 %scalar, <vscale x 16 x i8> %vector) #0 {
+; OPT-LABEL: @lastb_binop_nochange(
+; OPT-NEXT:    %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar, i32 0
+; OPT-NEXT:    %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+; OPT-NEXT:    %binop = udiv <vscale x 16 x i8> %vector, %splat
+; OPT-NEXT:    %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+; OPT-NEXT:    call void @use(<vscale x 16 x i8> %binop)
+; OPT-NEXT:    ret i8 %last
+  %splat_insert = insertelement <vscale x 16 x i8> poison, i8 %scalar, i32 0
+  %splat = shufflevector <vscale x 16 x i8> %splat_insert, <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
+  %binop = udiv <vscale x 16 x i8> %vector, %splat
+  %last = tail call i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1> %pg, <vscale x 16 x i8> %binop)
+  call void @use(<vscale x 16 x i8> %binop)
+  ret i8 %last
+}
+
+declare void @use(<vscale x 16 x i8>)
 declare <vscale x 16 x i1> @llvm.aarch64.sve.ptrue.nxv16i1(i32)
+declare float @llvm.aarch64.sve.lastb.nxv4f32(<vscale x 4 x i1>, <vscale x 4 x float>)
 declare i8 @llvm.aarch64.sve.lasta.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>)
 declare i8 @llvm.aarch64.sve.lastb.nxv16i8(<vscale x 16 x i1>, <vscale x 16 x i8>)