Index: lib/Target/AArch64/AArch64ISelLowering.cpp
===================================================================
--- lib/Target/AArch64/AArch64ISelLowering.cpp
+++ lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -7510,6 +7510,16 @@
   unsigned Align = cast<ConstantSDNode>(Op.getOperand(2))->getZExtValue();
   EVT VT = Node->getValueType(0);
 
+  if (DAG.getMachineFunction().getFunction().hasFnAttribute(
+          "no-stack-arg-probe")) {
+    SDValue SP = DAG.getCopyFromReg(Chain, dl, AArch64::SP, MVT::i64);
+    Chain = SP.getValue(1);
+    SP = DAG.getNode(ISD::SUB, dl, MVT::i64, SP, Size);
+    Chain = DAG.getCopyToReg(Chain, dl, AArch64::SP, SP);
+    SDValue Ops[2] = {SP, Chain};
+    return DAG.getMergeValues(Ops, dl);
+  }
+
   Chain = DAG.getCALLSEQ_START(Chain, 0, 0, dl);
 
   Chain = LowerWindowsDYNAMIC_STACKALLOC(Op, Chain, Size, DAG);
Index: lib/Target/ARM/ARMISelLowering.cpp
===================================================================
--- lib/Target/ARM/ARMISelLowering.cpp
+++ lib/Target/ARM/ARMISelLowering.cpp
@@ -13950,6 +13950,16 @@
   SDValue Chain = Op.getOperand(0);
   SDValue Size  = Op.getOperand(1);
 
+  if (DAG.getMachineFunction().getFunction().hasFnAttribute(
+          "no-stack-arg-probe")) {
+    SDValue SP = DAG.getCopyFromReg(Chain, DL, ARM::SP, MVT::i32);
+    Chain = SP.getValue(1);
+    SP = DAG.getNode(ISD::SUB, DL, MVT::i32, SP, Size);
+    Chain = DAG.getCopyToReg(Chain, DL, ARM::SP, SP);
+    SDValue Ops[2] = { SP, Chain };
+    return DAG.getMergeValues(Ops, DL);
+  }
+
   SDValue Words = DAG.getNode(ISD::SRL, DL, MVT::i32, Size,
                               DAG.getConstant(2, DL, MVT::i32));
 
Index: test/CodeGen/AArch64/no-stack-arg-probe.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AArch64/no-stack-arg-probe.ll
@@ -0,0 +1,12 @@
+; RUN: llc -mtriple=aarch64-windows -verify-machineinstrs %s -o - \
+; RUN:  | FileCheck %s
+
+define void @check_watermark() "no-stack-arg-probe" {
+entry:
+  %buffer = alloca [4096 x i8], align 1
+  ret void
+}
+
+; CHECK: check_watermark:
+; CHECK: sub sp, sp, #1, lsl #12
+; CHECK-NOT:     bl __chkstk
Index: test/CodeGen/AArch64/win-alloca-no-stack-probe.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AArch64/win-alloca-no-stack-probe.ll
@@ -0,0 +1,21 @@
+; RUN: llc -mtriple aarch64-windows -verify-machineinstrs -filetype asm -o - %s | FileCheck %s
+
+define void @func(i64 %a) "no-stack-arg-probe" {
+entry:
+  %0 = alloca i8, i64 %a, align 16
+  call void @func2(i8* nonnull %0)
+  ret void
+}
+
+declare void @func2(i8*)
+
+; The -O0 version here ends up much less elegant, so just check the
+; details of the optimized form, but check that -O0 at least emits the
+; call to __chkstk.
+
+; CHECK: add [[REG1:x[0-9]+]], x0, #15
+; CHECK-NOT: bl __chkstk
+; CHECK: mov [[REG2:x[0-9]+]], sp
+; CHECK: and [[REG1]], [[REG1]], #0xfffffffffffffff0
+; CHECK: sub [[REG3:x[0-9]+]], [[REG2]], [[REG1]]
+; CHECK: mov sp, [[REG3]]
Index: test/CodeGen/ARM/Windows/alloca-no-stack-arg-probe.ll
===================================================================
--- /dev/null
+++ test/CodeGen/ARM/Windows/alloca-no-stack-arg-probe.ll
@@ -0,0 +1,21 @@
+; RUN: llc -mtriple thumbv7-windows -filetype asm -o - %s | FileCheck %s
+
+declare arm_aapcs_vfpcc i32 @num_entries()
+
+define arm_aapcs_vfpcc void @test___builtin_alloca() "no-stack-arg-probe" {
+entry:
+  %array = alloca i8*, align 4
+  %call = call arm_aapcs_vfpcc i32 @num_entries()
+  %mul = mul i32 4, %call
+  %0 = alloca i8, i32 %mul
+  store i8* %0, i8** %array, align 4
+  ret void
+}
+
+; CHECK: bl num_entries
+; CHECK: movs [[R1:r[0-9]+]], #7
+; CHECK: add.w [[R0:r[0-9]+]], [[R1]], [[R0]], lsl #2
+; CHECK: bic [[R0]], [[R0]], #7
+; CHECK-NOT: bl __chkstk
+; CHECK: sub.w [[R0]], sp, [[R0]]
+; CHECK: mov sp, [[R0]]
Index: test/CodeGen/ARM/Windows/chkstk-no-stack-arg-probe.ll
===================================================================
--- /dev/null
+++ test/CodeGen/ARM/Windows/chkstk-no-stack-arg-probe.ll
@@ -0,0 +1,12 @@
+; RUN: llc -mtriple=thumbv7-windows -verify-machineinstrs %s -o - \
+; RUN:  | FileCheck %s
+
+define arm_aapcs_vfpcc void @check_watermark() "no-stack-arg-probe" {
+entry:
+  %buffer = alloca [4096 x i8], align 1
+  ret void
+}
+
+; CHECK: check_watermark:
+; CHECK-NOT: bl __chkstk
+; CHECK: sub.w sp, sp, #4096