Index: llvm/trunk/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
===================================================================
--- llvm/trunk/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ llvm/trunk/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -619,6 +619,7 @@
   header.kernarg_segment_byte_size = MFI->ABIArgOffset;
   header.wavefront_sgpr_count = KernelInfo.NumSGPR;
   header.workitem_vgpr_count = KernelInfo.NumVGPR;
+  header.workitem_private_segment_byte_size = KernelInfo.ScratchSize;
 
   AMDGPUTargetStreamer *TS =
       static_cast<AMDGPUTargetStreamer *>(OutStreamer->getTargetStreamer());
Index: llvm/trunk/test/CodeGen/AMDGPU/large-alloca-compute.ll
===================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/large-alloca-compute.ll
+++ llvm/trunk/test/CodeGen/AMDGPU/large-alloca-compute.ll
@@ -34,7 +34,7 @@
 ; GCNHSA: enable_sgpr_grid_workgroup_count_x = 0
 ; GCNHSA: enable_sgpr_grid_workgroup_count_y = 0
 ; GCNHSA: enable_sgpr_grid_workgroup_count_z = 0
-; GCNHSA: workitem_private_segment_byte_size = 0
+; GCNHSA: workitem_private_segment_byte_size = 32772
 ; GCNHSA: private_segment_alignment = 4
 ; GCNHSA: .end_amd_kernel_code_t
 
Index: llvm/trunk/test/CodeGen/AMDGPU/private-memory.ll
===================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/private-memory.ll
+++ llvm/trunk/test/CodeGen/AMDGPU/private-memory.ll
@@ -1,6 +1,7 @@
 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck %s -check-prefix=R600 -check-prefix=FUNC
 ; RUN: llc -show-mc-encoding -mattr=+promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=SI < %s | FileCheck %s -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC
 ; RUN: llc -show-mc-encoding -mattr=-promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=SI < %s | FileCheck %s -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC
+; RUN: llc -show-mc-encoding -mattr=-promote-alloca -verify-machineinstrs -mtriple=amdgcn-amdhsa -mcpu=kaveri < %s | FileCheck %s -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC -check-prefix=HSA-ALLOCA
 ; RUN: llc -show-mc-encoding -mattr=+promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck %s -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC
 ; RUN: llc -show-mc-encoding -mattr=-promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck %s -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC
 
@@ -18,6 +19,12 @@
 ; SI-PROMOTE: ds_read_b32
 ; SI-PROMOTE: ds_read_b32
 
+; HSA-ALLOCA: .amd_kernel_code_t
+; FIXME: Creating the emergency stack slots causes us to over-estimate scratch
+; by 4 bytes.
+; HSA-ALLOCA: workitem_private_segment_byte_size = 24
+; HSA-ALLOCA: .end_amd_kernel_code_t
+
 ; SI-ALLOCA: buffer_store_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen ; encoding: [0x00,0x10,0x70,0xe0
 ; SI-ALLOCA: buffer_store_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offen ; encoding: [0x00,0x10,0x70,0xe0
 define void @mova_same_clause(i32 addrspace(1)* nocapture %out, i32 addrspace(1)* nocapture %in) {