Fix Windows x64 ABI

src/arch/x86_64.rs

@@ -42,8 +42,8 @@ pub unsafe fn init(stack: &Stack, f: unsafe extern "C" fn(usize) -> !) -> StackP
 
 #[inline(always)]
 pub unsafe fn swap(arg: usize, old_sp: &mut StackPointer, new_sp: &StackPointer) -> usize {
-  let ret: usize;
+  macro_rules! swap_body {
-  asm!(
+    () => {
       r#"
         # Save frame pointer explicitly; LLVM doesn't spill it even if it is
         # marked as clobbered.
@@ -71,6 +71,14 @@ pub unsafe fn swap(arg: usize, old_sp: &mut StackPointer, new_sp: &StackPointer)
         jmpq    *%rbx
       2:
       "#
+    }
+  }
+
+  #[cfg(not(windows))]
+  #[inline(always)]
+  unsafe fn swap_impl(arg: usize, old_sp: &mut StackPointer, new_sp: &StackPointer) -> usize {
+    let ret: usize;
+    asm!(swap_body!()
       : "={rdi}" (ret)
       : "{rdi}" (arg)
         "{rsi}" (old_sp)
@@ -90,3 +98,32 @@ pub unsafe fn swap(arg: usize, old_sp: &mut StackPointer, new_sp: &StackPointer)
       : "volatile", "alignstack");
     ret
   }
+
+
+  #[cfg(windows)]
+  #[inline(always)]
+  unsafe fn swap_impl(arg: usize, old_sp: &mut StackPointer, new_sp: &StackPointer) -> usize {
+    let ret: usize;
+    asm!(swap_body!()
+      : "={rcx}" (ret)
+      : "{rcx}" (arg)
+        "{rsi}" (old_sp)
+        "{rdx}" (new_sp)
+      : "rax",   "rbx",   "rcx",   "rdx",   "rsi",   "rdi", //"rbp",   "rsp",
+        "r8",    "r9",    "r10",   "r11",   "r12",   "r13",   "r14",   "r15",
+        "xmm0",  "xmm1",  "xmm2",  "xmm3",  "xmm4",  "xmm5",  "xmm6",  "xmm7",
+        "xmm8",  "xmm9",  "xmm10", "xmm11", "xmm12", "xmm13", "xmm14", "xmm15",
+        "xmm16", "xmm17", "xmm18", "xmm19", "xmm20", "xmm21", "xmm22", "xmm23",
+        "xmm24", "xmm25", "xmm26", "xmm27", "xmm28", "xmm29", "xmm30", "xmm31"
+        "cc", "fpsr", "flags", "memory"
+        // Ideally, we would set the LLVM "noredzone" attribute on this function
+        // (and it would be propagated to the call site). Unfortunately, rustc
+        // provides no such functionality. Fortunately, by a lucky coincidence,
+        // the "alignstack" LLVM inline assembly option does exactly the same
+        // thing on x86_64.
+      : "volatile", "alignstack");
+    ret
+  }
+
+  swap_impl(arg, old_sp, new_sp)
+}