diff --git a/lib/Target/X86/X86FixupKinds.h b/lib/Target/X86/X86FixupKinds.h
index c8dac3ce457..a8117d47bf6 100644
--- a/lib/Target/X86/X86FixupKinds.h
+++ b/lib/Target/X86/X86FixupKinds.h
@@ -17,7 +17,8 @@ namespace X86 {
 enum Fixups {
   reloc_pcrel_4byte = FirstTargetFixupKind,  // 32-bit pcrel, e.g. a branch.
   reloc_pcrel_1byte,                         // 8-bit pcrel, e.g. branch_1
-  reloc_riprel_4byte                         // 32-bit rip-relative
+  reloc_riprel_4byte,                        // 32-bit rip-relative
+  reloc_riprel_4byte_movq_load               // 32-bit rip-relative in movq
 };
 }
 }
diff --git a/lib/Target/X86/X86MCCodeEmitter.cpp b/lib/Target/X86/X86MCCodeEmitter.cpp
index 3f18696d855..4547a3a3313 100644
--- a/lib/Target/X86/X86MCCodeEmitter.cpp
+++ b/lib/Target/X86/X86MCCodeEmitter.cpp
@@ -38,14 +38,15 @@ public:
   ~X86MCCodeEmitter() {}
 
   unsigned getNumFixupKinds() const {
-    return 3;
+    return 4;
   }
 
   const MCFixupKindInfo &getFixupKindInfo(MCFixupKind Kind) const {
     const static MCFixupKindInfo Infos[] = {
       { "reloc_pcrel_4byte", 0, 4 * 8 },
       { "reloc_pcrel_1byte", 0, 1 * 8 },
-      { "reloc_riprel_4byte", 0, 4 * 8 }
+      { "reloc_riprel_4byte", 0, 4 * 8 },
+      { "reloc_riprel_4byte_movq_load", 0, 4 * 8 }
     };
     
     if (Kind < FirstTargetFixupKind)
@@ -197,6 +198,14 @@ void X86MCCodeEmitter::EmitMemModRMByte(const MCInst &MI, unsigned Op,
            "Invalid rip-relative address");
     EmitByte(ModRMByte(0, RegOpcodeField, 5), CurByte, OS);
     
+    unsigned FixupKind = X86::reloc_riprel_4byte;
+    
+    // movq loads are handled with a special relocation form which allows the
+    // linker to eliminate some loads for GOT references which end up in the
+    // same linkage unit.
+    if (MI.getOpcode() == X86::MOV64rm_TC)
+      FixupKind = X86::reloc_riprel_4byte_movq_load;
+    
     // rip-relative addressing is actually relative to the *next* instruction.
     // Since an immediate can follow the mod/rm byte for an instruction, this
     // means that we need to bias the immediate field of the instruction with
@@ -204,7 +213,7 @@ void X86MCCodeEmitter::EmitMemModRMByte(const MCInst &MI, unsigned Op,
     // expression to emit.
     int ImmSize = X86II::hasImm(TSFlags) ? X86II::getSizeOfImm(TSFlags) : 0;
     
-    EmitImmediate(Disp, 4, MCFixupKind(X86::reloc_riprel_4byte),
+    EmitImmediate(Disp, 4, MCFixupKind(FixupKind),
                   CurByte, OS, Fixups, -ImmSize);
     return;
   }
diff --git a/test/MC/AsmParser/X86/x86_64-new-encoder.s b/test/MC/AsmParser/X86/x86_64-new-encoder.s
index 797558a81ef..712803e41fe 100644
--- a/test/MC/AsmParser/X86/x86_64-new-encoder.s
+++ b/test/MC/AsmParser/X86/x86_64-new-encoder.s
@@ -25,5 +25,13 @@ movq	$12, foo(%rip)
 // CHECK: encoding: [0x48,0xc7,0x05,A,A,A,A,0x0c,0x00,0x00,0x00]
 // CHECK:    fixup A - offset: 3, value: foo-8, kind: reloc_riprel_4byte
 
-// CHECK: addq	$-424, %rax             # encoding: [0x48,0x05,0x58,0xfe,0xff,0xff]
+// CHECK: addq	$-424, %rax
+// CHECK: encoding: [0x48,0x05,0x58,0xfe,0xff,0xff]
 addq $-424, %rax
+
+
+// CHECK: movq	_foo@GOTPCREL(%rip), %rax
+// CHECK:  encoding: [0x48,0x8b,0x05,A,A,A,A]
+// CHECK:  fixup A - offset: 3, value: _foo@GOTPCREL, kind: reloc_riprel_4byte_movq_load
+movq _foo@GOTPCREL(%rip), %rax
+