Index: llvm/trunk/lib/Target/AArch64/AArch64InstrInfo.td
===================================================================
--- llvm/trunk/lib/Target/AArch64/AArch64InstrInfo.td
+++ llvm/trunk/lib/Target/AArch64/AArch64InstrInfo.td
@@ -3087,6 +3087,14 @@
 defm CMHS    : SIMDThreeSameVector<1, 0b00111, "cmhs", AArch64cmhs>;
 defm CMTST   : SIMDThreeSameVector<0, 0b10001, "cmtst", AArch64cmtst>;
 defm FABD    : SIMDThreeSameVectorFP<1,1,0b010,"fabd", int_aarch64_neon_fabd>;
+let Predicates = [HasNEON] in {
+foreach VT = [ v2f32, v4f32, v2f64 ] in
+def : Pat<(fabs (fsub VT:$Rn, VT:$Rm)), (!cast<Instruction>("FABD"#VT) VT:$Rn, VT:$Rm)>;
+}
+let Predicates = [HasNEON, HasFullFP16] in {
+foreach VT = [ v4f16, v8f16 ] in
+def : Pat<(fabs (fsub VT:$Rn, VT:$Rm)), (!cast<Instruction>("FABD"#VT) VT:$Rn, VT:$Rm)>;
+}
 defm FACGE   : SIMDThreeSameVectorFPCmp<1,0,0b101,"facge",int_aarch64_neon_facge>;
 defm FACGT   : SIMDThreeSameVectorFPCmp<1,1,0b101,"facgt",int_aarch64_neon_facgt>;
 defm FADDP   : SIMDThreeSameVectorFP<1,0,0b010,"faddp",int_aarch64_neon_addp>;
@@ -3396,6 +3404,11 @@
 defm FABD     : SIMDFPThreeScalar<1, 1, 0b010, "fabd", int_aarch64_sisd_fabd>;
 def : Pat<(v1f64 (int_aarch64_neon_fabd (v1f64 FPR64:$Rn), (v1f64 FPR64:$Rm))),
           (FABD64 FPR64:$Rn, FPR64:$Rm)>;
+let Predicates = [HasFullFP16] in {
+def : Pat<(fabs (fsub f16:$Rn, f16:$Rm)), (FABD16 f16:$Rn, f16:$Rm)>;
+}
+def : Pat<(fabs (fsub f32:$Rn, f32:$Rm)), (FABD32 f32:$Rn, f32:$Rm)>;
+def : Pat<(fabs (fsub f64:$Rn, f64:$Rm)), (FABD64 f64:$Rn, f64:$Rm)>;
 defm FACGE    : SIMDThreeScalarFPCmp<1, 0, 0b101, "facge",
                                      int_aarch64_neon_facge>;
 defm FACGT    : SIMDThreeScalarFPCmp<1, 1, 0b101, "facgt",
Index: llvm/trunk/test/CodeGen/AArch64/arm64-vabs.ll
===================================================================
--- llvm/trunk/test/CodeGen/AArch64/arm64-vabs.ll
+++ llvm/trunk/test/CodeGen/AArch64/arm64-vabs.ll
@@ -219,6 +219,40 @@
 declare <4 x float> @llvm.aarch64.neon.fabd.v4f32(<4 x float>, <4 x float>) nounwind readnone
 declare <2 x double> @llvm.aarch64.neon.fabd.v2f64(<2 x double>, <2 x double>) nounwind readnone
 
+define <2 x float> @fabd_2s_from_fsub_fabs(<2 x float>* %A, <2 x float>* %B) nounwind {
+;CHECK-LABEL: fabd_2s_from_fsub_fabs:
+;CHECK: fabd.2s
+        %tmp1 = load <2 x float>, <2 x float>* %A
+        %tmp2 = load <2 x float>, <2 x float>* %B
+        %sub = fsub <2 x float> %tmp1, %tmp2
+        %abs = call <2 x float> @llvm.fabs.v2f32(<2 x float> %sub)
+        ret <2 x float> %abs
+}
+
+define <4 x float> @fabd_4s_from_fsub_fabs(<4 x float>* %A, <4 x float>* %B) nounwind {
+;CHECK-LABEL: fabd_4s_from_fsub_fabs:
+;CHECK: fabd.4s
+        %tmp1 = load <4 x float>, <4 x float>* %A
+        %tmp2 = load <4 x float>, <4 x float>* %B
+        %sub = fsub <4 x float> %tmp1, %tmp2
+        %abs = call <4 x float> @llvm.fabs.v4f32(<4 x float> %sub)
+        ret <4 x float> %abs
+}
+
+define <2 x double> @fabd_2d_from_fsub_fabs(<2 x double>* %A, <2 x double>* %B) nounwind {
+;CHECK-LABEL: fabd_2d_from_fsub_fabs:
+;CHECK: fabd.2d
+        %tmp1 = load <2 x double>, <2 x double>* %A
+        %tmp2 = load <2 x double>, <2 x double>* %B
+        %sub = fsub <2 x double> %tmp1, %tmp2
+        %abs = call <2 x double> @llvm.fabs.v2f64(<2 x double> %sub)
+        ret <2 x double> %abs
+}
+
+declare <2 x float> @llvm.fabs.v2f32(<2 x float>) nounwind readnone
+declare <4 x float> @llvm.fabs.v4f32(<4 x float>) nounwind readnone
+declare <2 x double> @llvm.fabs.v2f64(<2 x double>) nounwind readnone
+
 define <8 x i8> @sabd_8b(<8 x i8>* %A, <8 x i8>* %B) nounwind {
 ;CHECK-LABEL: sabd_8b:
 ;CHECK: sabd.8b
@@ -829,6 +863,25 @@
 declare double @llvm.aarch64.sisd.fabd.f64(double, double) nounwind readnone
 declare float @llvm.aarch64.sisd.fabd.f32(float, float) nounwind readnone
 
+define float @fabds_from_fsub_fabs(float %a, float %b) nounwind {
+; CHECK-LABEL: fabds_from_fsub_fabs:
+; CHECK: fabd s0, s0, s1
+  %sub = fsub float %a, %b
+  %abs = tail call float @llvm.fabs.f32(float %sub)
+  ret float %abs
+}
+
+define double @fabdd_from_fsub_fabs(double %a, double %b) nounwind {
+; CHECK-LABEL: fabdd_from_fsub_fabs:
+; CHECK: fabd d0, d0, d1
+  %sub = fsub double %a, %b
+  %abs = tail call double @llvm.fabs.f64(double %sub)
+  ret double %abs
+}
+
+declare float @llvm.fabs.f32(float) nounwind readnone
+declare double @llvm.fabs.f64(double) nounwind readnone
+
 define <2 x i64> @uabdl_from_extract_dup(<4 x i32> %lhs, i32 %rhs) {
 ; CHECK-LABEL: uabdl_from_extract_dup:
 ; CHECK-NOT: ext.16b
Index: llvm/trunk/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll
===================================================================
--- llvm/trunk/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll
+++ llvm/trunk/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll
@@ -6,6 +6,7 @@
 declare half @llvm.aarch64.neon.frsqrts.f16(half, half)
 declare half @llvm.aarch64.neon.frecps.f16(half, half)
 declare half @llvm.aarch64.neon.fmulx.f16(half, half)
+declare half @llvm.fabs.f16(half)
 
 define dso_local half @t_vabdh_f16(half %a, half %b) {
 ; CHECK-LABEL: t_vabdh_f16:
@@ -16,6 +17,16 @@
   ret half %vabdh_f16
 }
 
+define dso_local half @t_vabdh_f16_from_fsub_fabs(half %a, half %b) {
+; CHECK-LABEL: t_vabdh_f16_from_fsub_fabs:
+; CHECK:         fabd h0, h0, h1
+; CHECK-NEXT:    ret
+entry:
+  %sub = fsub half %a, %b
+  %abs = tail call half @llvm.fabs.f16(half %sub)
+  ret half %abs
+}
+
 define dso_local i16 @t_vceqh_f16(half %a, half %b) {
 ; CHECK-LABEL: t_vceqh_f16:
 ; CHECK:         fcmp h0, h1
Index: llvm/trunk/test/CodeGen/AArch64/fp16_intrinsic_vector_2op.ll
===================================================================
--- llvm/trunk/test/CodeGen/AArch64/fp16_intrinsic_vector_2op.ll
+++ llvm/trunk/test/CodeGen/AArch64/fp16_intrinsic_vector_2op.ll
@@ -6,6 +6,10 @@
 declare <8 x half> @llvm.aarch64.neon.fminnmp.v8f16(<8 x half>, <8 x half>)
 declare <4 x half> @llvm.aarch64.neon.fmaxnmp.v4f16(<4 x half>, <4 x half>)
 declare <8 x half> @llvm.aarch64.neon.fmaxnmp.v8f16(<8 x half>, <8 x half>)
+declare <4 x half> @llvm.aarch64.neon.fabd.v4f16(<4 x half>, <4 x half>)
+declare <8 x half> @llvm.aarch64.neon.fabd.v8f16(<8 x half>, <8 x half>)
+declare <4 x half> @llvm.fabs.v4f16(<4 x half>)
+declare <8 x half> @llvm.fabs.v8f16(<8 x half>)
 
 define dso_local <4 x half> @t_vdiv_f16(<4 x half> %a, <4 x half> %b) {
 ; CHECK-LABEL: t_vdiv_f16:
@@ -78,3 +82,41 @@
   %vpmaxnm2.i = tail call <8 x half> @llvm.aarch64.neon.fmaxnmp.v8f16(<8 x half> %a, <8 x half> %b)
   ret <8 x half> %vpmaxnm2.i
 }
+
+define dso_local <4 x half> @t_vabd_f16(<4 x half> %a, <4 x half> %b) {
+; CHECK-LABEL: t_vabd_f16:
+; CHECK:         fabd v0.4h, v0.4h, v1.4h
+; CHECK-NEXT:    ret
+entry:
+  %vabdh_f16 = tail call <4 x half> @llvm.aarch64.neon.fabd.v4f16(<4 x half> %a, <4 x half> %b)
+  ret <4 x half> %vabdh_f16
+}
+
+define dso_local <8 x half> @t_vabdq_f16(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: t_vabdq_f16:
+; CHECK:         fabd v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    ret
+entry:
+  %vabdh_f16 = tail call <8 x half> @llvm.aarch64.neon.fabd.v8f16(<8 x half> %a, <8 x half> %b)
+  ret <8 x half> %vabdh_f16
+}
+
+define dso_local <4 x half> @t_vabd_f16_from_fsub_fabs(<4 x half> %a, <4 x half> %b) {
+; CHECK-LABEL: t_vabd_f16_from_fsub_fabs:
+; CHECK:         fabd v0.4h, v0.4h, v1.4h
+; CHECK-NEXT:    ret
+entry:
+  %sub = fsub <4 x half> %a, %b
+  %abs = tail call <4 x half> @llvm.fabs.v4f16(<4 x half> %sub)
+  ret <4 x half> %abs
+}
+
+define dso_local <8 x half> @t_vabdq_f16_from_fsub_fabs(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: t_vabdq_f16_from_fsub_fabs:
+; CHECK:         fabd v0.8h, v0.8h, v1.8h
+; CHECK-NEXT:    ret
+entry:
+  %sub = fsub <8 x half> %a, %b
+  %abs = tail call <8 x half> @llvm.fabs.v8f16(<8 x half> %sub)
+  ret <8 x half> %abs
+}