diff --git a/src/arch/x86/mod.rs b/src/arch/x86/mod.rs
index 40aedaf..7789839 100644
--- a/src/arch/x86/mod.rs
+++ b/src/arch/x86/mod.rs
@@ -32,10 +32,13 @@ impl Registers {
   }
 
   #[inline(always)]
-  pub unsafe fn swap(&mut self) {
+  pub unsafe fn swap2(out_regs: *mut Registers, in_regs: *const Registers) {
+    let out_espp = &mut (*out_regs).esp;
+    let in_espp = &(*in_regs).esp;
     asm!(include_str!("swap.s")
           :
-          : "{eax}" (&mut self.esp)
+          : "{eax}" (out_espp),
+            "{ebx}" (in_espp)
           : "eax",  "ebx",  "ecx",  "edx",  "esi",  "edi", //"ebp",  "esp",
             "mmx0", "mmx1", "mmx2", "mmx3", "mmx4", "mmx5", "mmx6", "mmx7",
             "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7",
diff --git a/src/arch/x86/swap.s b/src/arch/x86/swap.s
index baca48a..d3310e5 100644
--- a/src/arch/x86/swap.s
+++ b/src/arch/x86/swap.s
@@ -4,7 +4,8 @@
 
 //! switch to a new context
 //! arguments:
-//!  * eax: stack pointer pointer
+//!  * eax: stack pointer out pointer
+//!  * ebx: stack pointer in pointer
 
 // save the frame pointer
 pushl %ebp
@@ -21,11 +22,11 @@ jmp 2f
 
 1:
   // retrieve the new stack pointer
-  movl (%eax), %ebx
+  movl (%eax), %edx
   // save the old stack pointer
-  movl %esp, (%eax)
+  movl %esp, (%ebx)
   // switch to the new stack pointer
-  movl %ebx, %esp
+  movl %edx, %esp
 
   // jump into the new context (return to the call point)
   // doing this instead of a straight `ret` is 8ns slower,
diff --git a/src/arch/x86_64/mod.rs b/src/arch/x86_64/mod.rs
index 66d1a2e..824b2b7 100644
--- a/src/arch/x86_64/mod.rs
+++ b/src/arch/x86_64/mod.rs
@@ -31,10 +31,13 @@ impl Registers {
   }
 
   #[inline(always)]
-  pub unsafe fn swap(&mut self) {
+  pub unsafe fn swap2(out_regs: *mut Registers, in_regs: *const Registers) {
+    let out_rspp = &mut (*out_regs).rsp;
+    let in_rspp = &(*in_regs).rsp;
     asm!(include_str!("swap.s")
           :
-          : "{rdi}" (&mut self.rsp)
+          : "{rdi}" (out_rspp),
+            "{rdi}" (in_rspp)
           : "rax",   "rbx",   "rcx",   "rdx",   "rsi",   "rdi", //"rbp",   "rsp",
             "r8",    "r9",    "r10",   "r11",   "r12",   "r13",   "r14",   "r15",
             "xmm0",  "xmm1",  "xmm2",  "xmm3",  "xmm4",  "xmm5",  "xmm6",  "xmm7",
diff --git a/src/arch/x86_64/swap.s b/src/arch/x86_64/swap.s
index a4270a9..270fdc1 100644
--- a/src/arch/x86_64/swap.s
+++ b/src/arch/x86_64/swap.s
@@ -4,7 +4,8 @@
 
 //! switch to a new context
 //! arguments:
-//!  * rdi: stack pointer pointer
+//!  * rdi: stack pointer out pointer
+//!  * rsi: stack pointer in pointer
 
 // make sure we leave the red zone alone
 sub $$128, %rsp
@@ -27,7 +28,7 @@ jmp 2f
 
 1:
   // retrieve the new stack pointer
-  movq (%rdi), %rax
+  movq (%rsi), %rax
   // save the old stack pointer
   movq %rsp, (%rdi)
   // switch to the new stack pointer
diff --git a/src/context.rs b/src/context.rs
index ab65add..b647bba 100644
--- a/src/context.rs
+++ b/src/context.rs
@@ -46,7 +46,13 @@ impl<'a, Stack> Context<'a, Stack> where Stack: stack::Stack {
   /// Switch to the context, saving the current thread of execution there.
   #[inline(always)]
   pub unsafe fn swap(&mut self) {
-    self.regs.swap()
+    Context::swap2(self, self)
+  }
+
+  /// Switch to in_ctx, saving the current thread of execution to out_ctx.
+  #[inline(always)]
+  pub unsafe fn swap2(out_ctx: *mut Context<'a, Stack>, in_ctx: *const Context<'a, Stack>) {
+    Registers::swap2(&mut (*out_ctx).regs, &(*in_ctx).regs)
   }
 
   /// Unwrap the context, returning the stack it contained.