make sure when we adjust frame size, stack pointer is 16-bytes aligned

(for calling convention)

src/compiler/backend/arch/x86_64/asm_backend.rs

@@ -633,7 +633,15 @@ impl MachineCode for ASMCode {
         regs_to_remove
     }
 
-    fn patch_frame_size(&mut self, size: usize) {
+    fn patch_frame_size(&mut self, size: usize, size_used: usize) {
+        // calling convention requires stack pointer to be 16 bytes aligned before a call
+        // we make frame size a multipl of 16 bytes
+        let size = if (size + size_used) % 16 == 0 {
+            size
+        } else {
+            ( (size + size_used) / 16 + 1) * 16 - size_used
+        };
+
         let size = size.to_string();
 
         debug_assert!(size.len() <= FRAME_SIZE_PLACEHOLDER_LEN);

src/compiler/backend/reg_alloc/graph_coloring/mod.rs

@@ -95,7 +95,7 @@ impl RegisterAllocation {
             let size_to_patch = total_frame_size - size_for_callee_saved_regs;
 
             trace!("patching the code to grow/shrink size of {} bytes", size_to_patch);
-            coloring.cf.mc_mut().patch_frame_size(size_to_patch);
+            coloring.cf.mc_mut().patch_frame_size(size_to_patch, size_for_callee_saved_regs);
         }
 
         coloring.cf.mc().trace_mc();

src/compiler/machine_code.rs

@@ -192,7 +192,7 @@ pub trait MachineCode {
     /// returns what registers push/pop have been deleted
     fn remove_unnecessary_callee_saved(&mut self, used_callee_saved: Vec<MuID>) -> Vec<MuID>;
     /// patch frame size
-    fn patch_frame_size(&mut self, size: usize);
+    fn patch_frame_size(&mut self, size: usize, size_used: usize);
 
     fn as_any(&self) -> &Any;
 }