Index: llvm/trunk/lib/Target/X86/X86.td
===================================================================
--- llvm/trunk/lib/Target/X86/X86.td
+++ llvm/trunk/lib/Target/X86/X86.td
@@ -239,6 +239,11 @@
 def FeatureSoftFloat
     : SubtargetFeature<"soft-float", "UseSoftFloat", "true",
                        "Use software floating point features.">;
+// On at least some AMD processors, there is no performance hazard to writing
+// only the lower parts of a YMM register without clearing the upper part.
+def FeatureFastPartialYMMWrite
+    : SubtargetFeature<"fast-partial-ymm-write", "HasFastPartialYMMWrite",
+                       "true", "Partial writes to YMM registers are fast">;
 
 //===----------------------------------------------------------------------===//
 // X86 processors supported.
@@ -596,7 +601,8 @@
   FeatureXSAVE,
   FeatureXSAVEOPT,
   FeatureSlowSHLD,
-  FeatureLAHFSAHF
+  FeatureLAHFSAHF,
+  FeatureFastPartialYMMWrite
 ]>;
 
 // Bulldozer
Index: llvm/trunk/lib/Target/X86/X86Subtarget.h
===================================================================
--- llvm/trunk/lib/Target/X86/X86Subtarget.h
+++ llvm/trunk/lib/Target/X86/X86Subtarget.h
@@ -189,6 +189,10 @@
   /// the stack pointer. This is an optimization for Intel Atom processors.
   bool UseLeaForSP;
 
+  /// True if there is no performance penalty to writing only the lower parts
+  /// of a YMM register without clearing the upper part.
+  bool HasFastPartialYMMWrite;
+
   /// True if 8-bit divisions are significantly faster than
   /// 32-bit divisions and should be used when possible.
   bool HasSlowDivide32;
@@ -421,6 +425,7 @@
   bool hasSSEUnalignedMem() const { return HasSSEUnalignedMem; }
   bool hasCmpxchg16b() const { return HasCmpxchg16b; }
   bool useLeaForSP() const { return UseLeaForSP; }
+  bool hasFastPartialYMMWrite() const { return HasFastPartialYMMWrite; }
   bool hasSlowDivide32() const { return HasSlowDivide32; }
   bool hasSlowDivide64() const { return HasSlowDivide64; }
   bool padShortFunctions() const { return PadShortFunctions; }
Index: llvm/trunk/lib/Target/X86/X86Subtarget.cpp
===================================================================
--- llvm/trunk/lib/Target/X86/X86Subtarget.cpp
+++ llvm/trunk/lib/Target/X86/X86Subtarget.cpp
@@ -285,6 +285,7 @@
   HasSSEUnalignedMem = false;
   HasCmpxchg16b = false;
   UseLeaForSP = false;
+  HasFastPartialYMMWrite = false;
   HasSlowDivide32 = false;
   HasSlowDivide64 = false;
   PadShortFunctions = false;
Index: llvm/trunk/lib/Target/X86/X86VZeroUpper.cpp
===================================================================
--- llvm/trunk/lib/Target/X86/X86VZeroUpper.cpp
+++ llvm/trunk/lib/Target/X86/X86VZeroUpper.cpp
@@ -248,7 +248,7 @@
 /// vzeroupper instructions before function calls.
 bool VZeroUpperInserter::runOnMachineFunction(MachineFunction &MF) {
   const X86Subtarget &ST = MF.getSubtarget<X86Subtarget>();
-  if (!ST.hasAVX() || ST.hasAVX512())
+  if (!ST.hasAVX() || ST.hasAVX512() || ST.hasFastPartialYMMWrite())
     return false;
   TII = ST.getInstrInfo();
   MachineRegisterInfo &MRI = MF.getRegInfo();
Index: llvm/trunk/test/CodeGen/X86/avx-vzeroupper.ll
===================================================================
--- llvm/trunk/test/CodeGen/X86/avx-vzeroupper.ll
+++ llvm/trunk/test/CodeGen/X86/avx-vzeroupper.ll
@@ -1,4 +1,9 @@
 ; RUN: llc < %s -x86-use-vzeroupper -mtriple=x86_64-apple-darwin -mattr=+avx | FileCheck %s
+; RUN: llc < %s -x86-use-vzeroupper -mtriple=x86_64-apple-darwin -mattr=+avx,+fast-partial-ymm-write | FileCheck --check-prefix=FASTYMM %s
+; RUN: llc < %s -x86-use-vzeroupper -mtriple=x86_64-apple-darwin -mcpu=btver2 | FileCheck --check-prefix=BTVER2 %s
+
+; FASTYMM-NOT: vzeroupper
+; BTVER2-NOT: vzeroupper
 
 declare i32 @foo()
 declare <4 x float> @do_sse(<4 x float>)