diff --git a/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.cpp b/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.cpp
index f98d5e331fe..7cdd5b1b20f 100644
--- a/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.cpp
+++ b/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.cpp
@@ -107,6 +107,74 @@ bool X86_MC::GetCpuIDAndInfo(unsigned value, unsigned *rEAX,
   return true;
 }
 
+/// GetCpuIDAndInfoEx - Execute the specified cpuid with subleaf and return the
+/// 4 values in the specified arguments.  If we can't run cpuid on the host,
+/// return true.
+bool X86_MC::GetCpuIDAndInfoEx(unsigned value, unsigned subleaf, unsigned *rEAX,
+                               unsigned *rEBX, unsigned *rECX, unsigned *rEDX) {
+#if defined(__x86_64__) || defined(_M_AMD64) || defined (_M_X64)
+  #if defined(__GNUC__)
+    // gcc desn't know cpuid would clobber ebx/rbx. Preseve it manually.
+    asm ("movq\t%%rbx, %%rsi\n\t"
+         "cpuid\n\t"
+         "xchgq\t%%rbx, %%rsi\n\t"
+         : "=a" (*rEAX),
+           "=S" (*rEBX),
+           "=c" (*rECX),
+           "=d" (*rEDX)
+         :  "a" (value),
+            "c" (subleaf));
+    return false;
+  #elif defined(_MSC_VER)
+    // can't use __cpuidex because it isn't available in all supported versions
+    // of MSC
+    __asm {
+      mov   eax,value
+      mov   ecx,subleaf
+      cpuid
+      mov   rsi,rEAX
+      mov   dword ptr [rsi],eax
+      mov   rsi,rEBX
+      mov   dword ptr [rsi],ebx
+      mov   rsi,rECX
+      mov   dword ptr [rsi],ecx
+      mov   rsi,rEDX
+      mov   dword ptr [rsi],edx
+    }
+    return false;
+  #endif
+#elif defined(i386) || defined(__i386__) || defined(__x86__) || defined(_M_IX86)
+  #if defined(__GNUC__)
+    asm ("movl\t%%ebx, %%esi\n\t"
+         "cpuid\n\t"
+         "xchgl\t%%ebx, %%esi\n\t"
+         : "=a" (*rEAX),
+           "=S" (*rEBX),
+           "=c" (*rECX),
+           "=d" (*rEDX)
+         :  "a" (value),
+            "c" (subleaf));
+    return false;
+  #elif defined(_MSC_VER)
+    __asm {
+      mov   eax,value
+      mov   ecx,subleaf
+      cpuid
+      mov   esi,rEAX
+      mov   dword ptr [esi],eax
+      mov   esi,rEBX
+      mov   dword ptr [esi],ebx
+      mov   esi,rECX
+      mov   dword ptr [esi],ecx
+      mov   esi,rEDX
+      mov   dword ptr [esi],edx
+    }
+    return false;
+  #endif
+#endif
+  return true;
+}
+
 void X86_MC::DetectFamilyModel(unsigned EAX, unsigned &Family,
                                unsigned &Model) {
   Family = (EAX >> 8) & 0xf; // Bits 8 - 11
diff --git a/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.h b/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.h
index c144c513de1..a4e0b5af9e1 100644
--- a/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.h
+++ b/lib/Target/X86/MCTargetDesc/X86MCTargetDesc.h
@@ -54,6 +54,11 @@ namespace X86_MC {
   /// the specified arguments.  If we can't run cpuid on the host, return true.
   bool GetCpuIDAndInfo(unsigned value, unsigned *rEAX,
                        unsigned *rEBX, unsigned *rECX, unsigned *rEDX);
+  /// GetCpuIDAndInfoEx - Execute the specified cpuid with subleaf and return
+  /// the 4 values in the specified arguments.  If we can't run cpuid on the
+  /// host, return true.
+  bool GetCpuIDAndInfoEx(unsigned value, unsigned subleaf, unsigned *rEAX,
+                       unsigned *rEBX, unsigned *rECX, unsigned *rEDX);
 
   void DetectFamilyModel(unsigned EAX, unsigned &Family, unsigned &Model);
 
diff --git a/lib/Target/X86/X86Subtarget.cpp b/lib/Target/X86/X86Subtarget.cpp
index 7064dd06fa3..670073a07bb 100644
--- a/lib/Target/X86/X86Subtarget.cpp
+++ b/lib/Target/X86/X86Subtarget.cpp
@@ -177,16 +177,18 @@ unsigned X86Subtarget::getSpecialAddressLatency() const {
 
 void X86Subtarget::AutoDetectSubtargetFeatures() {
   unsigned EAX = 0, EBX = 0, ECX = 0, EDX = 0;
+  unsigned MaxLevel;
   union {
     unsigned u[3];
     char     c[12];
   } text;
-  
-  if (X86_MC::GetCpuIDAndInfo(0, &EAX, text.u+0, text.u+2, text.u+1))
+
+  if (X86_MC::GetCpuIDAndInfo(0, &MaxLevel, text.u+0, text.u+2, text.u+1) ||
+      MaxLevel < 1)
     return;
 
   X86_MC::GetCpuIDAndInfo(0x1, &EAX, &EBX, &ECX, &EDX);
-  
+
   if ((EDX >> 15) & 1) { HasCMov = true;      ToggleFeature(X86::FeatureCMOV); }
   if ((EDX >> 23) & 1) { X86SSELevel = MMX;   ToggleFeature(X86::FeatureMMX);  }
   if ((EDX >> 25) & 1) { X86SSELevel = SSE1;  ToggleFeature(X86::FeatureSSE1); }
@@ -245,27 +247,41 @@ void X86Subtarget::AutoDetectSubtargetFeatures() {
       ToggleFeature(X86::FeatureSlowBTMem);
     }
     // If it's Nehalem, unaligned memory access is fast.
+    // FIXME: Nehalem is family 6. Also include Westmere and later processors?
     if (Family == 15 && Model == 26) {
       IsUAMemFast = true;
       ToggleFeature(X86::FeatureFastUAMem);
     }
 
-    X86_MC::GetCpuIDAndInfo(0x80000001, &EAX, &EBX, &ECX, &EDX);
-    if ((EDX >> 29) & 0x1) {
-      HasX86_64 = true;
-      ToggleFeature(X86::Feature64Bit);
+    unsigned MaxExtLevel;
+    X86_MC::GetCpuIDAndInfo(0x80000000, &MaxExtLevel, &EBX, &ECX, &EDX);
+
+    if (MaxExtLevel >= 0x80000001) {
+      X86_MC::GetCpuIDAndInfo(0x80000001, &EAX, &EBX, &ECX, &EDX);
+      if ((EDX >> 29) & 0x1) {
+        HasX86_64 = true;
+        ToggleFeature(X86::Feature64Bit);
+      }
+      if ((ECX >> 5) & 0x1) {
+        HasLZCNT = true;
+        ToggleFeature(X86::FeatureLZCNT);
+      }
+      if (IsAMD && ((ECX >> 6) & 0x1)) {
+        HasSSE4A = true;
+        ToggleFeature(X86::FeatureSSE4A);
+      }
+      if (IsAMD && ((ECX >> 16) & 0x1)) {
+        HasFMA4 = true;
+        ToggleFeature(X86::FeatureFMA4);
+      }
     }
-    if ((ECX >> 5) & 0x1) {
-      HasLZCNT = true;
-      ToggleFeature(X86::FeatureLZCNT);
-    }
-    if (IsAMD && ((ECX >> 6) & 0x1)) {
-      HasSSE4A = true;
-      ToggleFeature(X86::FeatureSSE4A);
-    }
-    if (IsAMD && ((ECX >> 16) & 0x1)) {
-      HasFMA4 = true;
-      ToggleFeature(X86::FeatureFMA4);
+  }
+
+  if (IsIntel && MaxLevel >= 7) {
+    X86_MC::GetCpuIDAndInfoEx(0x7, 0x0, &EAX, &EBX, &ECX, &EDX);
+    if ((EBX >> 3) & 0x1) {
+      HasBMI = true;
+      ToggleFeature(X86::FeatureBMI);
     }
   }
 }