Index: llvm/trunk/lib/Target/AMDGPU/SIISelLowering.h
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIISelLowering.h
+++ llvm/trunk/lib/Target/AMDGPU/SIISelLowering.h
@@ -121,6 +121,7 @@
   std::pair<unsigned, const TargetRegisterClass *>
   getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI,
                                StringRef Constraint, MVT VT) const override;
+  ConstraintType getConstraintType(StringRef Constraint) const override;
   SDValue copyToM0(SelectionDAG &DAG, SDValue Chain, SDLoc DL, SDValue V) const;
 };
 
Index: llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp
+++ llvm/trunk/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -2436,13 +2436,41 @@
 SITargetLowering::getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI,
                                                StringRef Constraint,
                                                MVT VT) const {
-  if (Constraint == "r") {
-    switch(VT.SimpleTy) {
-      default: llvm_unreachable("Unhandled type for 'r' inline asm constraint");
-      case MVT::i64:
-        return std::make_pair(0U, &AMDGPU::SGPR_64RegClass);
-      case MVT::i32:
+
+  if (Constraint.size() == 1) {
+    switch (Constraint[0]) {
+    case 's':
+    case 'r':
+      switch (VT.getSizeInBits()) {
+      default:
+        return std::make_pair(0U, nullptr);
+      case 32:
         return std::make_pair(0U, &AMDGPU::SGPR_32RegClass);
+      case 64:
+        return std::make_pair(0U, &AMDGPU::SGPR_64RegClass);
+      case 128:
+        return std::make_pair(0U, &AMDGPU::SReg_128RegClass);
+      case 256:
+        return std::make_pair(0U, &AMDGPU::SReg_256RegClass);
+      }
+
+    case 'v':
+      switch (VT.getSizeInBits()) {
+      default:
+        return std::make_pair(0U, nullptr);
+      case 32:
+        return std::make_pair(0U, &AMDGPU::VGPR_32RegClass);
+      case 64:
+        return std::make_pair(0U, &AMDGPU::VReg_64RegClass);
+      case 96:
+        return std::make_pair(0U, &AMDGPU::VReg_96RegClass);
+      case 128:
+        return std::make_pair(0U, &AMDGPU::VReg_128RegClass);
+      case 256:
+        return std::make_pair(0U, &AMDGPU::VReg_256RegClass);
+      case 512:
+        return std::make_pair(0U, &AMDGPU::VReg_512RegClass);
+      }
     }
   }
 
@@ -2463,3 +2491,16 @@
   }
   return TargetLowering::getRegForInlineAsmConstraint(TRI, Constraint, VT);
 }
+
+SITargetLowering::ConstraintType
+SITargetLowering::getConstraintType(StringRef Constraint) const {
+  if (Constraint.size() == 1) {
+    switch (Constraint[0]) {
+    default: break;
+    case 's':
+    case 'v':
+      return C_RegisterClass;
+    }
+  }
+  return TargetLowering::getConstraintType(Constraint);
+}
Index: llvm/trunk/test/CodeGen/AMDGPU/inline-constraints.ll
===================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/inline-constraints.ll
+++ llvm/trunk/test/CodeGen/AMDGPU/inline-constraints.ll
@@ -0,0 +1,23 @@
+; RUN: llc < %s -march=amdgcn -mcpu=bonaire -verify-machineinstrs | FileCheck --check-prefix=GCN %s
+; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck --check-prefix=GCN %s
+
+; GCN-LABEL: {{^}}inline_reg_constraints:
+; GCN: flat_load_dword v{{[0-9]+}}, v[{{[0-9]+:[0-9]+}}]
+; GCN: flat_load_dwordx2 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}]
+; GCN: flat_load_dwordx4 v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}]
+; GCN: s_load_dword s{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}]
+; GCN: s_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}]
+; GCN: s_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}]
+; GCN: s_load_dwordx8 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}]
+
+define void @inline_reg_constraints(i32 addrspace(1)* %ptr) {
+entry:
+  %v32 = tail call i32 asm sideeffect "flat_load_dword   $0, $1", "=v,v"(i32 addrspace(1)* %ptr)
+  %v64 = tail call <2 x i32> asm sideeffect "flat_load_dwordx2 $0, $1", "=v,v"(i32 addrspace(1)* %ptr)
+  %v128 = tail call <4 x i32> asm sideeffect "flat_load_dwordx4 $0, $1", "=v,v"(i32 addrspace(1)* %ptr)
+  %s32 =  tail call i32 asm sideeffect "s_load_dword $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
+  %s64 =  tail call <2 x i32> asm sideeffect "s_load_dwordx2 $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
+  %s128 =  tail call <4 x i32> asm sideeffect "s_load_dwordx4 $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
+  %s256 =  tail call <8 x i32> asm sideeffect "s_load_dwordx8 $0, $1", "=s,s"(i32 addrspace(1)* %ptr)
+  ret void
+}