Index: lib/Target/AMDGPU/AMDGPUISelLowering.h
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -76,6 +76,7 @@
   SDValue performLoadCombine(SDNode *N, DAGCombinerInfo &DCI) const;
   SDValue performStoreCombine(SDNode *N, DAGCombinerInfo &DCI) const;
   SDValue performClampCombine(SDNode *N, DAGCombinerInfo &DCI) const;
+  SDValue performAssertSZExtCombine(SDNode *N, DAGCombinerInfo &DCI) const;
 
   SDValue splitBinaryBitConstantOpImpl(DAGCombinerInfo &DCI, const SDLoc &SL,
                                        unsigned Opc, SDValue LHS,
Index: lib/Target/AMDGPU/AMDGPUISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -573,6 +573,8 @@
   setTargetDAGCombine(ISD::FSUB);
   setTargetDAGCombine(ISD::FNEG);
   setTargetDAGCombine(ISD::FABS);
+  setTargetDAGCombine(ISD::AssertZext);
+  setTargetDAGCombine(ISD::AssertSext);
 }
 
 //===----------------------------------------------------------------------===//
@@ -2591,6 +2593,31 @@
   return SDValue(CSrc, 0);
 }
 
+// FIXME: This should go in generic DAG combiner with an isTruncateFree check,
+// but isTruncateFree is inaccurate for i16 now because of SALU vs. VALU
+// issues.
+SDValue AMDGPUTargetLowering::performAssertSZExtCombine(SDNode *N,
+                                                        DAGCombinerInfo &DCI) const {
+  SelectionDAG &DAG = DCI.DAG;
+  SDValue N0 = N->getOperand(0);
+
+  // (vt2 (assertzext (truncate vt0:x), vt1)) ->
+  //     (vt2 (truncate (assertzext vt0:x, vt1)))
+  if (N0.getOpcode() == ISD::TRUNCATE) {
+    SDValue N1 = N->getOperand(1);
+    EVT ExtVT = cast<VTSDNode>(N1)->getVT();
+    SDLoc SL(N);
+
+    SDValue Src = N0.getOperand(0);
+    EVT SrcVT = Src.getValueType();
+    if (SrcVT.bitsGE(ExtVT)) {
+      SDValue NewInReg = DAG.getNode(N->getOpcode(), SL, SrcVT, Src, N1);
+      return DAG.getNode(ISD::TRUNCATE, SL, N->getValueType(0), NewInReg);
+    }
+  }
+
+  return SDValue();
+}
 /// Split the 64-bit value \p LHS into two 32-bit components, and perform the
 /// binary operation \p Opc to it with the corresponding constant operands.
 SDValue AMDGPUTargetLowering::splitBinaryBitConstantOpImpl(
@@ -3521,6 +3548,9 @@
 
     break;
   }
+  case ISD::AssertZext:
+  case ISD::AssertSext:
+    return performAssertSZExtCombine(N, DCI);
   }
   return SDValue();
 }
Index: lib/Target/AMDGPU/SIISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/SIISelLowering.cpp
+++ lib/Target/AMDGPU/SIISelLowering.cpp
@@ -1380,10 +1380,37 @@
 
     unsigned Reg = VA.getLocReg();
     const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg, VT);
+    EVT ValVT = VA.getValVT();
 
     Reg = MF.addLiveIn(Reg, RC);
     SDValue Val = DAG.getCopyFromReg(Chain, DL, Reg, VT);
 
+    // If this is an 8 or 16-bit value, it is really passed promoted
+    // to 32 bits. Insert an assert[sz]ext to capture this, then
+    // truncate to the right size.
+    switch (VA.getLocInfo()) {
+    case CCValAssign::Full:
+      break;
+    case CCValAssign::BCvt:
+      Val = DAG.getNode(ISD::BITCAST, DL, ValVT, Val);
+      break;
+    case CCValAssign::SExt:
+      Val = DAG.getNode(ISD::AssertSext, DL, VT, Val,
+                        DAG.getValueType(ValVT));
+      Val = DAG.getNode(ISD::TRUNCATE, DL, ValVT, Val);
+      break;
+    case CCValAssign::ZExt:
+      Val = DAG.getNode(ISD::AssertZext, DL, VT, Val,
+                        DAG.getValueType(ValVT));
+      Val = DAG.getNode(ISD::TRUNCATE, DL, ValVT, Val);
+      break;
+    case CCValAssign::AExt:
+      Val = DAG.getNode(ISD::TRUNCATE, DL, ValVT, Val);
+      break;
+    default:
+      llvm_unreachable("Unknown loc info!");
+    }
+
     if (IsShader && Arg.VT.isVector()) {
       // Build a vector from the registers
       Type *ParamType = FType->getParamType(Arg.getOrigArgIndex());
Index: test/CodeGen/AMDGPU/function-args.ll
===================================================================
--- test/CodeGen/AMDGPU/function-args.ll
+++ test/CodeGen/AMDGPU/function-args.ll
@@ -34,6 +34,22 @@
   ret void
 }
 
+; GCN-LABEL: {{^}}i1_arg_i1_use:
+; GCN: v_and_b32_e32 v0, 1, v0
+; GCN: v_cmp_eq_u32_e32 vcc, 1, v0
+; GCN: s_xor_b64 s{{\[[0-9]+:[0-9]+\]}}, vcc, -1
+define void @i1_arg_i1_use(i1 %arg) #0 {
+bb:
+  br i1 %arg, label %bb2, label %bb1
+
+bb1:
+  store volatile i32 0, i32 addrspace(1)* undef
+  br label %bb2
+
+bb2:
+  ret void
+}
+
 ; GCN-LABEL: {{^}}void_func_i8:
 ; GCN-NOT: v0
 ; GCN: buffer_store_byte v0, off