diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
index 0824e11574009..f2163373ccaed 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -2662,7 +2662,7 @@ SDValue SelectionDAGLegalize::ExpandPARITY(SDValue Op, const SDLoc &dl) {
 
   // If CTPOP is legal, use it. Otherwise use shifts and xor.
   SDValue Result;
-  if (TLI.isOperationLegal(ISD::CTPOP, VT)) {
+  if (TLI.isOperationLegalOrPromote(ISD::CTPOP, VT)) {
     Result = DAG.getNode(ISD::CTPOP, dl, VT, Op);
   } else {
     Result = Op;
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index fc6bde29b50fc..2a012666b4719 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -431,6 +431,9 @@ X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
     setOperationAction(ISD::PARITY, MVT::i64, Custom);
   if (Subtarget.hasPOPCNT()) {
     setOperationPromotedToType(ISD::CTPOP, MVT::i8, MVT::i32);
+    // popcntw is longer to encode than popcntl and also has a false dependency
+    // on the dest that popcntl hasn't had since Cannon Lake.
+    setOperationPromotedToType(ISD::CTPOP, MVT::i16, MVT::i32);
   } else {
     setOperationAction(ISD::CTPOP          , MVT::i8   , Expand);
     setOperationAction(ISD::CTPOP          , MVT::i16  , Expand);
diff --git a/llvm/test/CodeGen/X86/parity-vec.ll b/llvm/test/CodeGen/X86/parity-vec.ll
index 0d5f8f4fd0c53..e52a32d261ed2 100644
--- a/llvm/test/CodeGen/X86/parity-vec.ll
+++ b/llvm/test/CodeGen/X86/parity-vec.ll
@@ -18,7 +18,7 @@ define i1 @canonical_parity(<16 x i1> %x) {
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    psllw $7, %xmm0
 ; CHECK-NEXT:    pmovmskb %xmm0, %eax
-; CHECK-NEXT:    popcntw %ax, %ax
+; CHECK-NEXT:    popcntl %eax, %eax
 ; CHECK-NEXT:    testb $1, %al
 ; CHECK-NEXT:    setne %al
 ; CHECK-NEXT:    retq
@@ -33,8 +33,8 @@ define i1 @canonical_parity_noncanonical_pred(<16 x i1> %x) {
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    psllw $7, %xmm0
 ; CHECK-NEXT:    pmovmskb %xmm0, %eax
-; CHECK-NEXT:    popcntw %ax, %ax
-; CHECK-NEXT:    # kill: def $al killed $al killed $ax
+; CHECK-NEXT:    popcntl %eax, %eax
+; CHECK-NEXT:    # kill: def $al killed $al killed $eax
 ; CHECK-NEXT:    retq
   %i1 = bitcast <16 x i1> %x to i16
   %i2 = call i16 @llvm.ctpop.i16(i16 %i1)
@@ -62,7 +62,7 @@ define i1 @canonical_nonparity(<16 x i1> %x) {
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    psllw $7, %xmm0
 ; CHECK-NEXT:    pmovmskb %xmm0, %eax
-; CHECK-NEXT:    popcntw %ax, %ax
+; CHECK-NEXT:    popcntl %eax, %eax
 ; CHECK-NEXT:    testb $1, %al
 ; CHECK-NEXT:    sete %al
 ; CHECK-NEXT:    retq
@@ -77,7 +77,7 @@ define i1 @canonical_nonparity_noncanonical_pred(<16 x i1> %x) {
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    psllw $7, %xmm0
 ; CHECK-NEXT:    pmovmskb %xmm0, %eax
-; CHECK-NEXT:    popcntw %ax, %ax
+; CHECK-NEXT:    popcntl %eax, %eax
 ; CHECK-NEXT:    andl $1, %eax
 ; CHECK-NEXT:    xorb $1, %al
 ; CHECK-NEXT:    # kill: def $al killed $al killed $eax
diff --git a/llvm/test/CodeGen/X86/parity.ll b/llvm/test/CodeGen/X86/parity.ll
index ce403c6e8e7fd..5f33db141b542 100644
--- a/llvm/test/CodeGen/X86/parity.ll
+++ b/llvm/test/CodeGen/X86/parity.ll
@@ -59,14 +59,16 @@ define i16 @parity_16(i16 %x) {
 ;
 ; X86-POPCNT-LABEL: parity_16:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    popcntw {{[0-9]+}}(%esp), %ax
+; X86-POPCNT-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-POPCNT-NEXT:    popcntl %eax, %eax
 ; X86-POPCNT-NEXT:    andl $1, %eax
 ; X86-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_16:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    popcntw %di, %ax
+; X64-POPCNT-NEXT:    movzwl %di, %eax
+; X64-POPCNT-NEXT:    popcntl %eax, %eax
 ; X64-POPCNT-NEXT:    andl $1, %eax
 ; X64-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-POPCNT-NEXT:    retq
@@ -98,14 +100,16 @@ define i16 @parity_16_load(i16* %x) {
 ; X86-POPCNT-LABEL: parity_16_load:
 ; X86-POPCNT:       # %bb.0:
 ; X86-POPCNT-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    popcntw (%eax), %ax
+; X86-POPCNT-NEXT:    movzwl (%eax), %eax
+; X86-POPCNT-NEXT:    popcntl %eax, %eax
 ; X86-POPCNT-NEXT:    andl $1, %eax
 ; X86-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_16_load:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    popcntw (%rdi), %ax
+; X64-POPCNT-NEXT:    movzwl (%rdi), %eax
+; X64-POPCNT-NEXT:    popcntl %eax, %eax
 ; X64-POPCNT-NEXT:    andl $1, %eax
 ; X64-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-POPCNT-NEXT:    retq
@@ -420,7 +424,8 @@ define i16 @parity_16_shift(i16 %0) {
 ;
 ; X86-POPCNT-LABEL: parity_16_shift:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    popcntw {{[0-9]+}}(%esp), %ax
+; X86-POPCNT-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-POPCNT-NEXT:    popcntl %eax, %eax
 ; X86-POPCNT-NEXT:    andl $1, %eax
 ; X86-POPCNT-NEXT:    addl %eax, %eax
 ; X86-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
@@ -428,7 +433,8 @@ define i16 @parity_16_shift(i16 %0) {
 ;
 ; X64-POPCNT-LABEL: parity_16_shift:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    popcntw %di, %ax
+; X64-POPCNT-NEXT:    movzwl %di, %eax
+; X64-POPCNT-NEXT:    popcntl %eax, %eax
 ; X64-POPCNT-NEXT:    andl $1, %eax
 ; X64-POPCNT-NEXT:    addl %eax, %eax
 ; X64-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
diff --git a/llvm/test/CodeGen/X86/popcnt.ll b/llvm/test/CodeGen/X86/popcnt.ll
index bc24b41ff3d9f..ad0fb4ccfd85b 100644
--- a/llvm/test/CodeGen/X86/popcnt.ll
+++ b/llvm/test/CodeGen/X86/popcnt.ll
@@ -107,12 +107,16 @@ define i16 @cnt16(i16 %x) nounwind readnone {
 ;
 ; X86-POPCNT-LABEL: cnt16:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    popcntw {{[0-9]+}}(%esp), %ax
+; X86-POPCNT-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-POPCNT-NEXT:    popcntl %eax, %eax
+; X86-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: cnt16:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    popcntw %di, %ax
+; X64-POPCNT-NEXT:    movzwl %di, %eax
+; X64-POPCNT-NEXT:    popcntl %eax, %eax
+; X64-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-POPCNT-NEXT:    retq
   %cnt = tail call i16 @llvm.ctpop.i16(i16 %x)
   ret i16 %cnt
@@ -1564,14 +1568,13 @@ define i32 @popcount_i16_zext(i16 zeroext %x) {
 ;
 ; X86-POPCNT-LABEL: popcount_i16_zext:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    popcntw {{[0-9]+}}(%esp), %ax
-; X86-POPCNT-NEXT:    movzwl %ax, %eax
+; X86-POPCNT-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
+; X86-POPCNT-NEXT:    popcntl %eax, %eax
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: popcount_i16_zext:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    popcntw %di, %ax
-; X64-POPCNT-NEXT:    movzwl %ax, %eax
+; X64-POPCNT-NEXT:    popcntl %edi, %eax
 ; X64-POPCNT-NEXT:    retq
   %cnt = tail call i16 @llvm.ctpop.i16(i16 %x)
   %z = zext i16 %cnt to i32