Index: lib/Target/AMDGPU/AMDGPUISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -2690,7 +2690,8 @@
 
 uint32_t AMDGPUTargetLowering::getImplicitParameterOffset(
     const AMDGPUMachineFunction *MFI, const ImplicitParameter Param) const {
-  uint64_t ArgOffset = alignTo(MFI->getABIArgOffset(), 4);
+  unsigned Alignment = Subtarget->getAlignmentForImplicitArgPtr();
+  uint64_t ArgOffset = alignTo(MFI->getABIArgOffset(), Alignment);
   switch (Param) {
   case GRID_DIM:
     return ArgOffset;
Index: lib/Target/AMDGPU/AMDGPUSubtarget.h
===================================================================
--- lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -276,6 +276,10 @@
     return isAmdHsaOS() ? 0 : 36;
   }
 
+  unsigned getAlignmentForImplicitArgPtr() const {
+    return isAmdHsaOS() ? 8 : 4;
+  }
+
   unsigned getStackAlignment() const {
     // Scratch is allocated in 256 dword per wave blocks.
     return 4 * 256 / getWavefrontSize();
Index: lib/Target/AMDGPU/GCNSchedStrategy.cpp
===================================================================
--- lib/Target/AMDGPU/GCNSchedStrategy.cpp
+++ lib/Target/AMDGPU/GCNSchedStrategy.cpp
@@ -34,6 +34,7 @@
   const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
   unsigned MinRegOccupancy = std::min(ST.getOccupancyWithNumSGPRs(SGPRs),
                                       ST.getOccupancyWithNumVGPRs(VGPRs));
+  return MinRegOccupancy;
   return std::min(MinRegOccupancy,
                   ST.getOccupancyWithLocalMemSize(MFI->getLDSSize()));
 }
Index: test/CodeGen/AMDGPU/llvm.amdgcn.kernarg.segment.ptr.ll
===================================================================
--- test/CodeGen/AMDGPU/llvm.amdgcn.kernarg.segment.ptr.ll
+++ test/CodeGen/AMDGPU/llvm.amdgcn.kernarg.segment.ptr.ll
@@ -29,7 +29,7 @@
 
 ; ALL-LABEL: {{^}}test_implicit_alignment
 ; MESA: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0xc
-; HSA: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x3
+; HSA: s_load_dword [[VAL:s[0-9]+]], s[{{[0-9]+:[0-9]+}}], 0x4
 ; ALL: v_mov_b32_e32 [[V_VAL:v[0-9]+]], [[VAL]]
 ; MESA: buffer_store_dword [[V_VAL]]
 ; HSA: flat_store_dword v[{{[0-9]+:[0-9]+}}], [[V_VAL]]