diff --git a/llvm/lib/Target/AArch64/SVEInstrFormats.td b/llvm/lib/Target/AArch64/SVEInstrFormats.td
--- a/llvm/lib/Target/AArch64/SVEInstrFormats.td
+++ b/llvm/lib/Target/AArch64/SVEInstrFormats.td
@@ -771,7 +771,7 @@
   def : Pat<(i64 (op GPR64:$Rn, (nxv2i1 PPRAny:$Pg))),
             (!cast<Instruction>(NAME # _D) PPRAny:$Pg, $Rn)>;
 
-  // Combine cntp with combine_op
+  // combine_op(x, cntp(all_active, p)) ==> inst p, x
   def : Pat<(i64 (combine_op GPR64:$Rn, (int_aarch64_sve_cntp_oneuse (nxv16i1 (SVEAllActive)), (nxv16i1 PPRAny:$pred)))),
             (!cast<Instruction>(NAME # _B) PPRAny:$pred, $Rn)>;
   def : Pat<(i64 (combine_op GPR64:$Rn, (int_aarch64_sve_cntp_oneuse (nxv8i1 (SVEAllActive)), (nxv8i1 PPRAny:$pred)))),
@@ -780,6 +780,16 @@
             (!cast<Instruction>(NAME # _S) PPRAny:$pred, $Rn)>;
   def : Pat<(i64 (combine_op GPR64:$Rn, (int_aarch64_sve_cntp_oneuse (nxv2i1 (SVEAllActive)), (nxv2i1 PPRAny:$pred)))),
             (!cast<Instruction>(NAME # _D) PPRAny:$pred, $Rn)>;
+
+  // combine_op(x, cntp(p, p)) ==> inst p, x
+  def : Pat<(i64 (combine_op GPR64:$Rn, (int_aarch64_sve_cntp_oneuse (nxv16i1 PPRAny:$pred), (nxv16i1 PPRAny:$pred)))),
+            (!cast<Instruction>(NAME # _B) PPRAny:$pred, $Rn)>;
+  def : Pat<(i64 (combine_op GPR64:$Rn, (int_aarch64_sve_cntp_oneuse (nxv8i1 PPRAny:$pred), (nxv8i1 PPRAny:$pred)))),
+            (!cast<Instruction>(NAME # _H) PPRAny:$pred, $Rn)>;
+  def : Pat<(i64 (combine_op GPR64:$Rn, (int_aarch64_sve_cntp_oneuse (nxv4i1 PPRAny:$pred), (nxv4i1 PPRAny:$pred)))),
+            (!cast<Instruction>(NAME # _S) PPRAny:$pred, $Rn)>;
+  def : Pat<(i64 (combine_op GPR64:$Rn, (int_aarch64_sve_cntp_oneuse (nxv2i1 PPRAny:$pred), (nxv2i1 PPRAny:$pred)))),
+            (!cast<Instruction>(NAME # _D) PPRAny:$pred, $Rn)>;
 }
 
 class sve_int_count_v<bits<2> sz8_64, bits<5> opc, string asm,
diff --git a/llvm/test/CodeGen/AArch64/sve-cntp-combine.ll b/llvm/test/CodeGen/AArch64/sve-cntp-combine.ll
--- a/llvm/test/CodeGen/AArch64/sve-cntp-combine.ll
+++ b/llvm/test/CodeGen/AArch64/sve-cntp-combine.ll
@@ -5,8 +5,8 @@
 
 ; INCP
 
-define i64 @cntp_add_nxv16i1(i64 %x, <vscale x 16 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_add_nxv16i1:
+define i64 @cntp_add_all_active_nxv16i1(i64 %x, <vscale x 16 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_all_active_nxv16i1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    incp x0, p0.b
 ; CHECK-NEXT:    ret
@@ -16,8 +16,8 @@
   ret i64 %add
 }
 
-define i64 @cntp_add_nxv8i1(i64 %x, <vscale x 8 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_add_nxv8i1:
+define i64 @cntp_add_all_active_nxv8i1(i64 %x, <vscale x 8 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_all_active_nxv8i1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    incp x0, p0.h
 ; CHECK-NEXT:    ret
@@ -27,8 +27,8 @@
   ret i64 %add
 }
 
-define i64 @cntp_add_nxv4i1(i64 %x, <vscale x 4 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_add_nxv4i1:
+define i64 @cntp_add_all_active_nxv4i1(i64 %x, <vscale x 4 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_all_active_nxv4i1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    incp x0, p0.s
 ; CHECK-NEXT:    ret
@@ -38,8 +38,8 @@
   ret i64 %add
 }
 
-define i64 @cntp_add_nxv2i1(i64 %x, <vscale x 2 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_add_nxv2i1:
+define i64 @cntp_add_all_active_nxv2i1(i64 %x, <vscale x 2 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_all_active_nxv2i1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    incp x0, p0.d
 ; CHECK-NEXT:    ret
@@ -49,8 +49,8 @@
   ret i64 %add
 }
 
-define i64 @cntp_add_all_active_nxv8i1(i64 %x, <vscale x 8 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_add_all_active_nxv8i1:
+define i64 @cntp_add_all_active_nxv8i1_via_cast(i64 %x, <vscale x 8 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_all_active_nxv8i1_via_cast:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    incp x0, p0.h
 ; CHECK-NEXT:    ret
@@ -61,8 +61,8 @@
   ret i64 %add
 }
 
-define i64 @cntp_add_nxv2i1_oneuse(i64 %x, <vscale x 2 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_add_nxv2i1_oneuse:
+define i64 @cntp_add_all_active_nxv2i1_oneuse(i64 %x, <vscale x 2 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_all_active_nxv2i1_oneuse:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    cntp x8, p1, p0.d
@@ -77,10 +77,64 @@
   ret i64 %res
 }
 
+define i64 @cntp_add_same_active_nxv16i1(i64 %x, <vscale x 16 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_same_active_nxv16i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    incp x0, p0.b
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv16i1(<vscale x 16 x i1> %pg, <vscale x 16 x i1> %pg)
+  %add = add i64 %1, %x
+  ret i64 %add
+}
+
+define i64 @cntp_add_same_active_nxv8i1(i64 %x, <vscale x 8 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_same_active_nxv8i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    incp x0, p0.h
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv8i1(<vscale x 8 x i1> %pg, <vscale x 8 x i1> %pg)
+  %add = add i64 %1, %x
+  ret i64 %add
+}
+
+define i64 @cntp_add_same_active_nxv4i1(i64 %x, <vscale x 4 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_same_active_nxv4i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    incp x0, p0.s
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv4i1(<vscale x 4 x i1> %pg, <vscale x 4 x i1> %pg)
+  %add = add i64 %1, %x
+  ret i64 %add
+}
+
+define i64 @cntp_add_same_active_nxv2i1(i64 %x, <vscale x 2 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_same_active_nxv2i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    incp x0, p0.d
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv2i1(<vscale x 2 x i1> %pg, <vscale x 2 x i1> %pg)
+  %add = add i64 %1, %x
+  ret i64 %add
+}
+
+define i64 @cntp_add_same_active_nxv2i1_oneuse(i64 %x, <vscale x 2 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_add_same_active_nxv2i1_oneuse:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    cntp x8, p0, p0.d
+; CHECK-NEXT:    add x9, x8, x0
+; CHECK-NEXT:    madd x0, x8, x0, x9
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv2i1(<vscale x 2 x i1> %pg, <vscale x 2 x i1> %pg)
+  %add = add i64 %1, %x
+  %mul = mul i64 %1, %x
+  %res = add i64 %add, %mul
+  ret i64 %res
+}
+
 ; DECP
 
-define i64 @cntp_sub_nxv16i1(i64 %x, <vscale x 16 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_sub_nxv16i1:
+define i64 @cntp_sub_all_active_nxv16i1(i64 %x, <vscale x 16 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_all_active_nxv16i1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    decp x0, p0.b
 ; CHECK-NEXT:    ret
@@ -90,8 +144,8 @@
   ret i64 %sub
 }
 
-define i64 @cntp_sub_nxv8i1(i64 %x, <vscale x 8 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_sub_nxv8i1:
+define i64 @cntp_sub_all_active_nxv8i1(i64 %x, <vscale x 8 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_all_active_nxv8i1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    decp x0, p0.h
 ; CHECK-NEXT:    ret
@@ -101,8 +155,8 @@
   ret i64 %sub
 }
 
-define i64 @cntp_sub_nxv4i1(i64 %x, <vscale x 4 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_sub_nxv4i1:
+define i64 @cntp_sub_all_active_nxv4i1(i64 %x, <vscale x 4 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_all_active_nxv4i1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    decp x0, p0.s
 ; CHECK-NEXT:    ret
@@ -112,8 +166,8 @@
   ret i64 %sub
 }
 
-define i64 @cntp_sub_nxv2i1(i64 %x, <vscale x 2 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_sub_nxv2i1:
+define i64 @cntp_sub_all_active_nxv2i1(i64 %x, <vscale x 2 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_all_active_nxv2i1:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    decp x0, p0.d
 ; CHECK-NEXT:    ret
@@ -123,8 +177,8 @@
   ret i64 %sub
 }
 
-define i64 @cntp_sub_all_active_nxv8i1(i64 %x, <vscale x 8 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_sub_all_active_nxv8i1:
+define i64 @cntp_sub_all_active_nxv8i1_via_cast(i64 %x, <vscale x 8 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_all_active_nxv8i1_via_cast:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    decp x0, p0.h
 ; CHECK-NEXT:    ret
@@ -135,8 +189,8 @@
   ret i64 %sub
 }
 
-define i64 @cntp_sub_nxv2i1_multiuse(i64 %x, <vscale x 2 x i1> %pg) #0 {
-; CHECK-LABEL: cntp_sub_nxv2i1_multiuse:
+define i64 @cntp_sub_all_active_nxv2i1_multiuse(i64 %x, <vscale x 2 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_all_active_nxv2i1_multiuse:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ptrue p1.d
 ; CHECK-NEXT:    cntp x8, p1, p0.d
@@ -151,6 +205,59 @@
   ret i64 %res
 }
 
+define i64 @cntp_sub_same_active_nxv16i1(i64 %x, <vscale x 16 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_same_active_nxv16i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    decp x0, p0.b
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv16i1(<vscale x 16 x i1> %pg, <vscale x 16 x i1> %pg)
+  %sub = sub i64 %x, %1
+  ret i64 %sub
+}
+
+define i64 @cntp_sub_same_active_nxv8i1(i64 %x, <vscale x 8 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_same_active_nxv8i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    decp x0, p0.h
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv8i1(<vscale x 8 x i1> %pg, <vscale x 8 x i1> %pg)
+  %sub = sub i64 %x, %1
+  ret i64 %sub
+}
+
+define i64 @cntp_sub_same_active_nxv4i1(i64 %x, <vscale x 4 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_same_active_nxv4i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    decp x0, p0.s
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv4i1(<vscale x 4 x i1> %pg, <vscale x 4 x i1> %pg)
+  %sub = sub i64 %x, %1
+  ret i64 %sub
+}
+
+define i64 @cntp_sub_same_active_nxv2i1(i64 %x, <vscale x 2 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_same_active_nxv2i1:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    decp x0, p0.d
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv2i1(<vscale x 2 x i1> %pg, <vscale x 2 x i1> %pg)
+  %sub = sub i64 %x, %1
+  ret i64 %sub
+}
+
+define i64 @cntp_sub_same_active_nxv2i1_multiuse(i64 %x, <vscale x 2 x i1> %pg) #0 {
+; CHECK-LABEL: cntp_sub_same_active_nxv2i1_multiuse:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    cntp x8, p0, p0.d
+; CHECK-NEXT:    sub x9, x8, x0
+; CHECK-NEXT:    madd x0, x8, x0, x9
+; CHECK-NEXT:    ret
+  %1 = tail call i64 @llvm.aarch64.sve.cntp.nxv2i1(<vscale x 2 x i1> %pg, <vscale x 2 x i1> %pg)
+  %add = sub i64 %1, %x
+  %mul = mul i64 %1, %x
+  %res = add i64 %add, %mul
+  ret i64 %res
+}
 
 declare <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1>)
 declare <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1>)