Index: lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -504,6 +504,11 @@
     Ctx.emitError("too many user SGPRs used");
   }
 
+  if (MFI->LDSSize > static_cast<unsigned>(STM.getLocalMemorySize())) {
+    LLVMContext &Ctx = MF.getFunction()->getContext();
+    Ctx.emitError("LDS size exceeds device maximum");
+  }
+
   ProgInfo.VGPRBlocks = (ProgInfo.NumVGPR - 1) / 4;
   ProgInfo.SGPRBlocks = (ProgInfo.NumSGPR - 1) / 8;
   // Set the value to initialize FP_ROUND and FP_DENORM parts of the mode
Index: test/CodeGen/AMDGPU/32-bit-local-address-space.ll
===================================================================
--- test/CodeGen/AMDGPU/32-bit-local-address-space.ll
+++ test/CodeGen/AMDGPU/32-bit-local-address-space.ll
@@ -91,12 +91,12 @@
 
 
 @ptr = addrspace(3) global i32 addrspace(3)* undef
-@dst = addrspace(3) global [16384 x i32] undef
+@dst = addrspace(3) global [16383 x i32] undef
 
 ; FUNC-LABEL: {{^}}global_ptr:
 ; SI: ds_write_b32
 define void @global_ptr() nounwind {
-  store i32 addrspace(3)* getelementptr ([16384 x i32], [16384 x i32] addrspace(3)* @dst, i32 0, i32 16), i32 addrspace(3)* addrspace(3)* @ptr
+  store i32 addrspace(3)* getelementptr ([16383 x i32], [16383 x i32] addrspace(3)* @dst, i32 0, i32 16), i32 addrspace(3)* addrspace(3)* @ptr
   ret void
 }
 
Index: test/CodeGen/AMDGPU/over-max-lds-size.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AMDGPU/over-max-lds-size.ll
@@ -0,0 +1,14 @@
+; RUN: not llc -march=amdgcn -mcpu=tahiti < %s 2>&1 | FileCheck -check-prefix=ERROR %s
+; RUN: not llc -march=amdgcn -mcpu=hawaii < %s 2>&1 | FileCheck -check-prefix=ERROR %s
+; RUN: not llc -march=amdgcn -mcpu=fiji < %s 2>&1 | FileCheck -check-prefix=ERROR %s
+
+; ERROR: error: LDS size exceeds device maximum
+
+@huge = internal unnamed_addr addrspace(3) global [100000 x i32] undef, align 4
+
+define void @promote_alloca_size_256() {
+entry:
+  %v0 = getelementptr inbounds [100000 x i32], [100000 x i32] addrspace(3)* @huge, i32 0, i32 0
+  store i32 0, i32 addrspace(3)* %v0
+  ret void
+}