Index: llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -1092,6 +1092,7 @@
 
   // kernarg_segment_alignment is specified as log of the alignment.
   // The minimum alignment is 16.
+  // FIXME: The metadata treats the minimum as 4?
   Out.kernarg_segment_alignment = Log2(std::max(Align(16), MaxKernArgAlign));
 }
 
Index: llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
@@ -201,10 +201,11 @@
   Align MaxKernArgAlign;
   HSACodeProps.mKernargSegmentSize = STM.getKernArgSegmentSize(F,
                                                                MaxKernArgAlign);
+  HSACodeProps.mKernargSegmentAlign =
+    std::max(MaxKernArgAlign, Align(4)).value();
+
   HSACodeProps.mGroupSegmentFixedSize = ProgramInfo.LDSSize;
   HSACodeProps.mPrivateSegmentFixedSize = ProgramInfo.ScratchSize;
-  HSACodeProps.mKernargSegmentAlign =
-      std::max(MaxKernArgAlign, Align(4)).value();
   HSACodeProps.mWavefrontSize = STM.getWavefrontSize();
   HSACodeProps.mNumSGPRs = ProgramInfo.NumSGPR;
   HSACodeProps.mNumVGPRs = ProgramInfo.NumVGPR;
@@ -868,6 +869,8 @@
       Kern.getDocument()->getNode(ProgramInfo.LDSSize);
   Kern[".private_segment_fixed_size"] =
       Kern.getDocument()->getNode(ProgramInfo.ScratchSize);
+
+  // FIXME: The metadata treats the minimum as 16?
   Kern[".kernarg_segment_align"] =
       Kern.getDocument()->getNode(std::max(Align(4), MaxKernArgAlign).value());
   Kern[".wavefront_size"] =
Index: llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
+++ llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
@@ -697,6 +697,7 @@
   if (ImplicitBytes != 0) {
     const Align Alignment = getAlignmentForImplicitArgPtr();
     TotalSize = alignTo(ExplicitArgBytes, Alignment) + ImplicitBytes;
+    MaxAlign = std::max(MaxAlign, Alignment);
   }
 
   // Being able to dereference past the end is useful for emitting scalar loads.
Index: llvm/test/CodeGen/AMDGPU/llvm.amdgcn.implicitarg.ptr.ll
===================================================================
--- llvm/test/CodeGen/AMDGPU/llvm.amdgcn.implicitarg.ptr.ll
+++ llvm/test/CodeGen/AMDGPU/llvm.amdgcn.implicitarg.ptr.ll
@@ -272,7 +272,7 @@
 ; HSA-LABEL: - Name:            kernel_implicitarg_ptr_empty
 ; HSA: CodeProps:
 ; HSA: KernargSegmentSize: 56
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL: - Name:            kernel_implicitarg_ptr_empty_0implicit
 ; HSA: KernargSegmentSize: 0
@@ -280,19 +280,19 @@
 
 ; HSA-LABEL: - Name:            opencl_kernel_implicitarg_ptr_empty
 ; HSA: KernargSegmentSize: 48
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL: - Name:            kernel_implicitarg_ptr
 ; HSA: KernargSegmentSize: 168
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL: - Name:            opencl_kernel_implicitarg_ptr
 ; HSA: KernargSegmentSize: 160
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL: - Name:            kernel_call_implicitarg_ptr_func_empty
 ; HSA: KernargSegmentSize: 56
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL: - Name:            kernel_call_implicitarg_ptr_func_empty_implicit0
 ; HSA: KernargSegmentSize: 0
@@ -300,19 +300,19 @@
 
 ; HSA-LABEL:  - Name:            opencl_kernel_call_implicitarg_ptr_func_empty
 ; HSA: KernargSegmentSize: 48
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL:  - Name:            kernel_call_implicitarg_ptr_func
 ; HSA: KernargSegmentSize: 168
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL:  - Name:            opencl_kernel_call_implicitarg_ptr_func
 ; HSA: KernargSegmentSize: 160
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL: - Name:            kernel_call_kernarg_implicitarg_ptr_func
 ; HSA: KernargSegmentSize: 168
-; HSA: KernargSegmentAlign: 4
+; HSA: KernargSegmentAlign: 8
 
 ; HSA-LABEL: - Name:            kernel_implicitarg_no_struct_align_padding
 ; HSA: KernargSegmentSize: 120