Index: lib/Target/AMDGPU/AMDGPUSubtarget.h
===================================================================
--- lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -283,6 +283,13 @@
   bool enableSubRegLiveness() const override {
     return true;
   }
+
+  /// \brief Returns the offset in bytes from the start of the input buffer
+  ///        of the first explicit kernel argument.
+  unsigned getExplicitKernelArgOffset() const {
+    return isAmdHsaOS() ? 0 : 36;
+  }
+
 };
 
 } // End namespace llvm
Index: lib/Target/AMDGPU/SIISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/SIISelLowering.cpp
+++ lib/Target/AMDGPU/SIISelLowering.cpp
@@ -583,7 +583,8 @@
     if (VA.isMemLoc()) {
       VT = Ins[i].VT;
       EVT MemVT = Splits[i].VT;
-      const unsigned Offset = 36 + VA.getLocMemOffset();
+      const unsigned Offset = Subtarget->getExplicitKernelArgOffset() +
+                              VA.getLocMemOffset();
       // The first 36 bytes of the input buffer contains information about
       // thread group and global sizes.
       SDValue Arg = LowerParameter(DAG, VT, MemVT,  DL, DAG.getRoot(),
Index: test/CodeGen/AMDGPU/hsa.ll
===================================================================
--- test/CodeGen/AMDGPU/hsa.ll
+++ test/CodeGen/AMDGPU/hsa.ll
@@ -20,6 +20,7 @@
 ; HSA-NEXT: .ascii  "HSA Code Unit:0.0:AMD:0.1:GFX8.1:0"
 ; Test that the amd_kernel_code_t object is emitted
 ; HSA: .asciz
+; HSA: s_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[0:1], 0x0
 ; Make sure we are setting the ATC bit:
 ; HSA: s_mov_b32 s[[HI:[0-9]]], 0x100f000
 ; HSA: buffer_store_dword v{{[0-9]+}}, s[0:[[HI]]], 0