[X86] Don't use popcnt for parity if only bits 7:0 of the input can be non-zero.

[craig.topper]

Summary

Without popcnt we had a special case for using the parity flag from a single test i8 test instruction if only bits 7:0 could be non-zero. That special case is still useful when we have popcnt.

To reach this special case, we enable custom lowering of parity for i16/i32/i64 even when popcnt is enabled. The check for POPCNT being enabled is now after the special case in LowerPARITY.

Fixes PR52093

[spatel]

llvm/test/CodeGen/X86/parity.ll

2–5

Add shared prefixes for any "X86" and "X64" to reduce the duplicated check lines?

[craig.topper]

Rebase after adding X86 and X64 check prefixes

[spatel]

LGTM - thanks!

llvm/lib/Target/X86/X86ISelLowering.cpp

X86TargetLowering::X86TargetLowering(const X86TargetMachine &TM,
   setLoadExtAction(ISD::EXTLOAD, MVT::f80, MVT::f16, Expand);
   setLoadExtAction(ISD::EXTLOAD, MVT::f128, MVT::f16, Expand);
   setTruncStoreAction(MVT::f32, MVT::f16, Expand);
   setTruncStoreAction(MVT::f64, MVT::f16, Expand);
   setTruncStoreAction(MVT::f80, MVT::f16, Expand);
   setTruncStoreAction(MVT::f128, MVT::f16, Expand);
 
   setOperationAction(ISD::PARITY, MVT::i8, Custom);
+  setOperationAction(ISD::PARITY, MVT::i16, Custom);
+  setOperationAction(ISD::PARITY, MVT::i32, Custom);
+  if (Subtarget.is64Bit())
+    setOperationAction(ISD::PARITY, MVT::i64, Custom);
   if (Subtarget.hasPOPCNT()) {
     setOperationPromotedToType(ISD::CTPOP, MVT::i8, MVT::i32);
   } else {
     setOperationAction(ISD::CTPOP          , MVT::i8   , Expand);
     setOperationAction(ISD::CTPOP          , MVT::i16  , Expand);
     setOperationAction(ISD::CTPOP          , MVT::i32  , Expand);
     if (Subtarget.is64Bit())
       setOperationAction(ISD::CTPOP        , MVT::i64  , Expand);
     else
       setOperationAction(ISD::CTPOP        , MVT::i64  , Custom);
-
-    setOperationAction(ISD::PARITY, MVT::i16, Custom);
-    setOperationAction(ISD::PARITY, MVT::i32, Custom);
-    if (Subtarget.is64Bit())
-      setOperationAction(ISD::PARITY, MVT::i64, Custom);
   }
 
   setOperationAction(ISD::READCYCLECOUNTER , MVT::i64  , Custom);
 
   if (!Subtarget.hasMOVBE())
     setOperationAction(ISD::BSWAP          , MVT::i16  , Expand);
 
   // X86 wants to expand cmov itself.
if (VT == MVT::i8 ||
     SDValue Flags = DAG.getNode(X86ISD::CMP, DL, MVT::i32, X,
                                 DAG.getConstant(0, DL, MVT::i8));
     // Copy the inverse of the parity flag into a register with setcc.
     SDValue Setnp = getSETCC(X86::COND_NP, Flags, DL, DAG);
     // Extend to the original type.
     return DAG.getNode(ISD::ZERO_EXTEND, DL, VT, Setnp);
   }
 
+  // If we have POPCNT, use the default expansion.
+  if (Subtarget.hasPOPCNT())
+    return SDValue();
+
   if (VT == MVT::i64) {
     // Xor the high and low 16-bits together using a 32-bit operation.
     SDValue Hi = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32,
                              DAG.getNode(ISD::SRL, DL, MVT::i64, X,
                                          DAG.getConstant(32, DL, MVT::i8)));
     SDValue Lo = DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, X);
     X = DAG.getNode(ISD::XOR, DL, MVT::i32, Lo, Hi);
   }

llvm/test/CodeGen/X86/parity.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=-popcnt | FileCheck %s --check-prefix=X86-NOPOPCNT
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=-popcnt | FileCheck %s --check-prefix=X64-NOPOPCNT
 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+popcnt | FileCheck %s --check-prefix=X86-POPCNT
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+popcnt | FileCheck %s --check-prefix=X64-POPCNT
 
 define i4 @parity_4(i4 %x) {
 ; X86-NOPOPCNT-LABEL: parity_4:
 ; X86-NOPOPCNT:       # %bb.0:
 ; X86-NOPOPCNT-NEXT:    testb $15, {{[0-9]+}}(%esp)
 ; X86-NOPOPCNT-NEXT:    setnp %al
 ; X86-NOPOPCNT-NEXT:    retl
 ;

 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb %dil, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_16_zexti8:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    popcntw %ax, %ax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    cmpb $0, {{[0-9]+}}(%esp)
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_16_zexti8:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    movzbl %dil, %eax
-; X64-POPCNT-NEXT:    popcntw %ax, %ax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb %dil, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-POPCNT-NEXT:    retq
   %a = zext i8 %x to i16
   %b = tail call i16 @llvm.ctpop.i16(i16 %a)
   %c = and i16 %b, 1
   ret i16 %c
 }
 

 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb %dil, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_16_mask255:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    popcntw %ax, %ax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    cmpb $0, {{[0-9]+}}(%esp)
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_16_mask255:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    movzbl %dil, %eax
-; X64-POPCNT-NEXT:    popcntw %ax, %ax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb %dil, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-POPCNT-NEXT:    retq
   %a = and i16 %x, 255
   %b = tail call i16 @llvm.ctpop.i16(i16 %a)
   %c = and i16 %b, 1
   ret i16 %c
 }
 

 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb $15, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_16_mask15:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movzwl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    andl $15, %eax
-; X86-POPCNT-NEXT:    popcntw %ax, %ax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    movzwl {{[0-9]+}}(%esp), %ecx
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    testb $15, %cl
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_16_mask15:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    andl $15, %edi
-; X64-POPCNT-NEXT:    popcntw %di, %ax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb $15, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    # kill: def $ax killed $ax killed $eax
 ; X64-POPCNT-NEXT:    retq
   %a = and i16 %x, 15
   %b = tail call i16 @llvm.ctpop.i16(i16 %a)
   %c = and i16 %b, 1
   ret i16 %c
 }
 

 ; X64-NOPOPCNT:       # %bb.0:
 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb %dil, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_32_zexti8:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    popcntl %eax, %eax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    cmpb $0, {{[0-9]+}}(%esp)
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_32_zexti8:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    movzbl %dil, %eax
-; X64-POPCNT-NEXT:    popcntl %eax, %eax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb %dil, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    retq
   %a = zext i8 %x to i32
   %b = tail call i32 @llvm.ctpop.i32(i32 %a)
   %c = and i32 %b, 1
   ret i32 %c
 }
 
 define i32 @parity_32_mask255(i32 %x) {
 ; X86-NOPOPCNT-LABEL: parity_32_mask255:
 ; X86-NOPOPCNT:       # %bb.0:
 ; X86-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X86-NOPOPCNT-NEXT:    cmpb $0, {{[0-9]+}}(%esp)
 ; X86-NOPOPCNT-NEXT:    setnp %al
 ; X86-NOPOPCNT-NEXT:    retl
 ;
 ; X64-NOPOPCNT-LABEL: parity_32_mask255:
 ; X64-NOPOPCNT:       # %bb.0:
 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb %dil, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_32_mask255:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    popcntl %eax, %eax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    cmpb $0, {{[0-9]+}}(%esp)
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_32_mask255:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    movzbl %dil, %eax
-; X64-POPCNT-NEXT:    popcntl %eax, %eax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb %dil, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    retq
   %a = and i32 %x, 255
   %b = tail call i32 @llvm.ctpop.i32(i32 %a)
   %c = and i32 %b, 1
   ret i32 %c
 }
 
 define i32 @parity_32_mask15(i32 %x) {
 ; X86-NOPOPCNT-LABEL: parity_32_mask15:
 ; X86-NOPOPCNT:       # %bb.0:
 ; X86-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X86-NOPOPCNT-NEXT:    testb $15, {{[0-9]+}}(%esp)
 ; X86-NOPOPCNT-NEXT:    setnp %al
 ; X86-NOPOPCNT-NEXT:    retl
 ;
 ; X64-NOPOPCNT-LABEL: parity_32_mask15:
 ; X64-NOPOPCNT:       # %bb.0:
 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb $15, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_32_mask15:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    andl $15, %eax
-; X86-POPCNT-NEXT:    popcntl %eax, %eax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    testb $15, {{[0-9]+}}(%esp)
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_32_mask15:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    andl $15, %edi
-; X64-POPCNT-NEXT:    popcntl %edi, %eax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb $15, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    retq
   %a = and i32 %x, 15
   %b = tail call i32 @llvm.ctpop.i32(i32 %a)
   %c = and i32 %b, 1
   ret i32 %c
 }
 
 define i32 @parity_32_shift(i32 %0) {

 ; X64-NOPOPCNT:       # %bb.0:
 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb %dil, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_64_zexti8:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    popcntl %eax, %eax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    cmpb $0, {{[0-9]+}}(%esp)
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    xorl %edx, %edx
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_64_zexti8:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    # kill: def $edi killed $edi def $rdi
-; X64-POPCNT-NEXT:    movzbl %dil, %eax
-; X64-POPCNT-NEXT:    popcntq %rax, %rax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb %dil, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    retq
   %a = zext i8 %x to i64
   %b = tail call i64 @llvm.ctpop.i64(i64 %a)
   %c = and i64 %b, 1
   ret i64 %c
 }
 
 define i64 @parity_64_mask255(i64 %x) {

 ; X64-NOPOPCNT:       # %bb.0:
 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb %dil, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_64_mask255:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    popcntl %eax, %eax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    cmpb $0, {{[0-9]+}}(%esp)
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    xorl %edx, %edx
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_64_mask255:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    movzbl %dil, %eax
-; X64-POPCNT-NEXT:    popcntq %rax, %rax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb %dil, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    retq
   %a = and i64 %x, 255
   %b = tail call i64 @llvm.ctpop.i64(i64 %a)
   %c = and i64 %b, 1
   ret i64 %c
 }
 
 define i64 @parity_64_mask15(i64 %x) {

 ; X64-NOPOPCNT:       # %bb.0:
 ; X64-NOPOPCNT-NEXT:    xorl %eax, %eax
 ; X64-NOPOPCNT-NEXT:    testb $15, %dil
 ; X64-NOPOPCNT-NEXT:    setnp %al
 ; X64-NOPOPCNT-NEXT:    retq
 ;
 ; X86-POPCNT-LABEL: parity_64_mask15:
 ; X86-POPCNT:       # %bb.0:
-; X86-POPCNT-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-POPCNT-NEXT:    andl $15, %eax
-; X86-POPCNT-NEXT:    popcntl %eax, %eax
-; X86-POPCNT-NEXT:    andl $1, %eax
+; X86-POPCNT-NEXT:    xorl %eax, %eax
+; X86-POPCNT-NEXT:    testb $15, {{[0-9]+}}(%esp)
+; X86-POPCNT-NEXT:    setnp %al
 ; X86-POPCNT-NEXT:    xorl %edx, %edx
 ; X86-POPCNT-NEXT:    retl
 ;
 ; X64-POPCNT-LABEL: parity_64_mask15:
 ; X64-POPCNT:       # %bb.0:
-; X64-POPCNT-NEXT:    andl $15, %edi
-; X64-POPCNT-NEXT:    popcntq %rdi, %rax
-; X64-POPCNT-NEXT:    andl $1, %eax
+; X64-POPCNT-NEXT:    xorl %eax, %eax
+; X64-POPCNT-NEXT:    testb $15, %dil
+; X64-POPCNT-NEXT:    setnp %al
 ; X64-POPCNT-NEXT:    retq
   %a = and i64 %x, 15
   %b = tail call i64 @llvm.ctpop.i64(i64 %a)
   %c = and i64 %b, 1
   ret i64 %c
 }
 
 define i64 @parity_64_shift(i64 %0) {