diff --git a/lib/compiler-singlepass/src/emitter_x64.rs b/lib/compiler-singlepass/src/emitter_x64.rs
index a4c0ceecdda..17e1818d6c1 100644
--- a/lib/compiler-singlepass/src/emitter_x64.rs
+++ b/lib/compiler-singlepass/src/emitter_x64.rs
@@ -117,6 +117,7 @@ pub trait Emitter {
     fn emit_xchg(&mut self, sz: Size, src: Location, dst: Location);
     fn emit_lock_xadd(&mut self, sz: Size, src: Location, dst: Location);
     fn emit_lock_cmpxchg(&mut self, sz: Size, src: Location, dst: Location);
+    fn emit_rep_stosq(&mut self);
 
     fn emit_btc_gpr_imm8_32(&mut self, src: u8, dst: GPR);
     fn emit_btc_gpr_imm8_64(&mut self, src: u8, dst: GPR);
@@ -1176,6 +1177,9 @@ impl Emitter for Assembler {
         }
     }
 
+    fn emit_rep_stosq(&mut self) {
+        dynasm!(self ; rep stosq);
+    }
     fn emit_btc_gpr_imm8_32(&mut self, src: u8, dst: GPR) {
         dynasm!(self ; btc Rd(dst as u8), BYTE src as i8);
     }
diff --git a/lib/compiler-singlepass/src/machine.rs b/lib/compiler-singlepass/src/machine.rs
index 17374f8184b..7f446744cd7 100644
--- a/lib/compiler-singlepass/src/machine.rs
+++ b/lib/compiler-singlepass/src/machine.rs
@@ -3,9 +3,12 @@ use crate::emitter_x64::*;
 use crate::x64_decl::{new_machine_state, X64Register};
 use smallvec::smallvec;
 use smallvec::SmallVec;
+use std::cmp;
 use std::collections::HashSet;
 use wasmer_compiler::wasmparser::Type as WpType;
 
+const NATIVE_PAGE_SIZE: usize = 4096;
+
 struct MachineStackOffset(usize);
 
 pub struct Machine {
@@ -447,11 +450,6 @@ impl Machine {
             }
         }
 
-        // Initialize all normal locals to zero.
-        for i in n_params..n {
-            a.emit_mov(Size::S64, Location::Imm32(0), locations[i]);
-        }
-
         // Load vmctx into R15.
         a.emit_mov(
             Size::S64,
@@ -459,6 +457,41 @@ impl Machine {
             Location::GPR(GPR::R15),
         );
 
+        // Stack probe.
+        //
+        // `rep stosq` writes data from low address to high address and may skip the stack guard page.
+        // so here we probe it explicitly when needed.
+        for i in (n_params..n).step_by(NATIVE_PAGE_SIZE / 8).skip(1) {
+            a.emit_mov(Size::S64, Location::Imm32(0), locations[i]);
+        }
+
+        // Initialize all normal locals to zero.
+        let mut init_stack_loc_cnt = 0;
+        let mut last_stack_loc = Location::Memory(GPR::RBP, i32::MAX);
+        for i in n_params..n {
+            match locations[i] {
+                Location::Memory(_, _) => {
+                    init_stack_loc_cnt += 1;
+                    last_stack_loc = cmp::min(last_stack_loc, locations[i]);
+                }
+                Location::GPR(_) => {
+                    a.emit_mov(Size::S64, Location::Imm32(0), locations[i]);
+                }
+                _ => unreachable!(),
+            }
+        }
+        if init_stack_loc_cnt > 0 {
+            // Since these assemblies take up to 24 bytes, if more than 2 slots are initialized, then they are smaller.
+            a.emit_mov(
+                Size::S64,
+                Location::Imm64(init_stack_loc_cnt as u64),
+                Location::GPR(GPR::RCX),
+            );
+            a.emit_xor(Size::S64, Location::GPR(GPR::RAX), Location::GPR(GPR::RAX));
+            a.emit_lea(Size::S64, last_stack_loc, Location::GPR(GPR::RDI));
+            a.emit_rep_stosq();
+        }
+
         // Add the size of all locals allocated to stack.
         self.stack_offset.0 += static_area_size - callee_saved_regs_size;