Add patterns for the x86 popcnt instruction.

- Also adds a new POPCNT subtarget feature that is currently enabled if the target
  supports SSE4.2 (nehalem) or SSE4A (barcelona).


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120917 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/X86/X86.td

@@ -23,6 +23,9 @@ include "llvm/Target/Target.td"
 def FeatureCMOV    : SubtargetFeature<"cmov","HasCMov", "true",
                                       "Enable conditional move instructions">;
 
+def FeaturePOPCNT   : SubtargetFeature<"popcnt", "HasPOPCNT", "true",
+                                       "Support POPCNT instruction">;
+
 
 def FeatureMMX     : SubtargetFeature<"mmx","X86SSELevel", "MMX",
                                       "Enable MMX instructions">;
@@ -45,7 +48,7 @@ def FeatureSSE41   : SubtargetFeature<"sse41", "X86SSELevel", "SSE41",
                                       [FeatureSSSE3]>;
 def FeatureSSE42   : SubtargetFeature<"sse42", "X86SSELevel", "SSE42",
                                       "Enable SSE 4.2 instructions",
-                                      [FeatureSSE41]>;
+                                      [FeatureSSE41, FeaturePOPCNT]>;
 def Feature3DNow   : SubtargetFeature<"3dnow", "X863DNowLevel", "ThreeDNow",
                                       "Enable 3DNow! instructions">;
 def Feature3DNowA  : SubtargetFeature<"3dnowa", "X863DNowLevel", "ThreeDNowA",
@@ -63,7 +66,8 @@ def FeatureFastUAMem : SubtargetFeature<"fast-unaligned-mem",
                                         "IsUAMemFast", "true",
                                         "Fast unaligned memory access">;
 def FeatureSSE4A   : SubtargetFeature<"sse4a", "HasSSE4A", "true",
-                                      "Support SSE 4a instructions">;
+                                      "Support SSE 4a instructions",
+                                      [FeaturePOPCNT]>;
 
 def FeatureAVX     : SubtargetFeature<"avx", "HasAVX", "true",
                                       "Enable AVX instructions">;

lib/Target/X86/X86ISelLowering.cpp

@@ -285,21 +285,27 @@ X86TargetLowering::X86TargetLowering(X86TargetMachine &TM)
   setOperationAction(ISD::FREM             , MVT::f80  , Expand);
   setOperationAction(ISD::FLT_ROUNDS_      , MVT::i32  , Custom);
 
-  setOperationAction(ISD::CTPOP            , MVT::i8   , Expand);
   setOperationAction(ISD::CTTZ             , MVT::i8   , Custom);
   setOperationAction(ISD::CTLZ             , MVT::i8   , Custom);
-  setOperationAction(ISD::CTPOP            , MVT::i16  , Expand);
   setOperationAction(ISD::CTTZ             , MVT::i16  , Custom);
   setOperationAction(ISD::CTLZ             , MVT::i16  , Custom);
-  setOperationAction(ISD::CTPOP            , MVT::i32  , Expand);
   setOperationAction(ISD::CTTZ             , MVT::i32  , Custom);
   setOperationAction(ISD::CTLZ             , MVT::i32  , Custom);
   if (Subtarget->is64Bit()) {
-    setOperationAction(ISD::CTPOP          , MVT::i64  , Expand);
     setOperationAction(ISD::CTTZ           , MVT::i64  , Custom);
     setOperationAction(ISD::CTLZ           , MVT::i64  , Custom);
   }
 
+  if (Subtarget->hasPOPCNT()) {
+    setOperationAction(ISD::CTPOP          , MVT::i8   , Promote);
+  } 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);
+  }
+
   setOperationAction(ISD::READCYCLECOUNTER , MVT::i64  , Custom);
   setOperationAction(ISD::BSWAP            , MVT::i16  , Expand);
 

lib/Target/X86/X86InstrSSE.td

@@ -4603,22 +4603,25 @@ defm VTESTPDY : avx_bittest<0x0F, "vtestpd", VR256, f256mem, memopv4f64, v4f64>;
 //===----------------------------------------------------------------------===//
 
 def POPCNT16rr : I<0xB8, MRMSrcReg, (outs GR16:$dst), (ins GR16:$src),
-                   "popcnt{w}\t{$src, $dst|$dst, $src}", []>, OpSize, XS;
+                   "popcnt{w}\t{$src, $dst|$dst, $src}",
-let mayLoad = 1 in
+                   [(set GR16:$dst, (ctpop GR16:$src))]>, OpSize, XS;
 def POPCNT16rm : I<0xB8, MRMSrcMem, (outs GR16:$dst), (ins i16mem:$src),
-                   "popcnt{w}\t{$src, $dst|$dst, $src}", []>, OpSize, XS;
+                   "popcnt{w}\t{$src, $dst|$dst, $src}",
+                   [(set GR16:$dst, (ctpop (loadi16 addr:$src)))]>, OpSize, XS;
 
 def POPCNT32rr : I<0xB8, MRMSrcReg, (outs GR32:$dst), (ins GR32:$src),
-                   "popcnt{l}\t{$src, $dst|$dst, $src}", []>, XS;
+                   "popcnt{l}\t{$src, $dst|$dst, $src}",
-let mayLoad = 1 in
+                   [(set GR32:$dst, (ctpop GR32:$src))]>, XS;
 def POPCNT32rm : I<0xB8, MRMSrcMem, (outs GR32:$dst), (ins i32mem:$src),
-                   "popcnt{l}\t{$src, $dst|$dst, $src}", []>, XS;
+                   "popcnt{l}\t{$src, $dst|$dst, $src}",
+                   [(set GR32:$dst, (ctpop (loadi32 addr:$src)))]>, XS;
 
 def POPCNT64rr : RI<0xB8, MRMSrcReg, (outs GR64:$dst), (ins GR64:$src),
-                    "popcnt{q}\t{$src, $dst|$dst, $src}", []>, XS;
+                    "popcnt{q}\t{$src, $dst|$dst, $src}",
-let mayLoad = 1 in
+                    [(set GR64:$dst, (ctpop GR64:$src))]>, XS;
 def POPCNT64rm : RI<0xB8, MRMSrcMem, (outs GR64:$dst), (ins i64mem:$src),
-                    "popcnt{q}\t{$src, $dst|$dst, $src}", []>, XS;
+                    "popcnt{q}\t{$src, $dst|$dst, $src}",
+                    [(set GR64:$dst, (ctpop (loadi64 addr:$src)))]>, XS;
 
 
 

lib/Target/X86/X86Subtarget.h

@@ -65,6 +65,9 @@ protected:
   ///
   bool HasX86_64;
 
+  /// HasPOPCNT - True if the processor supports POPCNT.
+  bool HasPOPCNT;
+
   /// HasSSE4A - True if the processor supports SSE4A instructions.
   bool HasSSE4A;
 
@@ -150,6 +153,7 @@ public:
   bool hasSSE4A() const { return HasSSE4A; }
   bool has3DNow() const { return X863DNowLevel >= ThreeDNow; }
   bool has3DNowA() const { return X863DNowLevel >= ThreeDNowA; }
+  bool hasPOPCNT() const { return HasPOPCNT; }
   bool hasAVX() const { return HasAVX; }
   bool hasAES() const { return HasAES; }
   bool hasCLMUL() const { return HasCLMUL; }

test/CodeGen/X86/popcnt.ll

@@ -0,0 +1,38 @@
+; RUN: llc -march=x86-64 -mattr=+popcnt < %s | FileCheck %s
+
+define i8 @cnt8(i8 %x) nounwind readnone {
+  %cnt = tail call i8 @llvm.ctpop.i8(i8 %x)
+  ret i8 %cnt
+; CHECK: cnt8:
+; CHECK: popcntw
+; CHECK: ret
+}
+
+define i16 @cnt16(i16 %x) nounwind readnone {
+  %cnt = tail call i16 @llvm.ctpop.i16(i16 %x)
+  ret i16 %cnt
+; CHECK: cnt16:
+; CHECK: popcntw
+; CHECK: ret
+}
+
+define i32 @cnt32(i32 %x) nounwind readnone {
+  %cnt = tail call i32 @llvm.ctpop.i32(i32 %x)
+  ret i32 %cnt
+; CHECK: cnt32:
+; CHECK: popcntl
+; CHECK: ret
+}
+
+define i64 @cnt64(i64 %x) nounwind readnone {
+  %cnt = tail call i64 @llvm.ctpop.i64(i64 %x)
+  ret i64 %cnt
+; CHECK: cnt64:
+; CHECK: popcntq
+; CHECK: ret
+}
+
+declare i8 @llvm.ctpop.i8(i8) nounwind readnone
+declare i16 @llvm.ctpop.i16(i16) nounwind readnone
+declare i32 @llvm.ctpop.i32(i32) nounwind readnone
+declare i64 @llvm.ctpop.i64(i64) nounwind readnone