Reserve frame word for machine return address

* libguile/frames.h: Add machine return address to diagram.
  (SCM_FRAME_MACHINE_RETURN_ADDRESS):
  (SCM_FRAME_SET_MACHINE_RETURN_ADDRESS): New macros.
  (SCM_FRAME_PREVIOUS_SP):
  (SCM_FRAME_DYNAMIC_LINK):
  (SCM_FRAME_SET_DYNAMIC_LINK): Adapt for new frame size.
* libguile/vm-engine.c (halt): Set frame size to 3.
  (call, call-label): Set mRA to 0.
* libguile/vm.c (push_interrupt_frame, reinstate_continuation_x):
  (scm_call_n): Set frame size to 3.  In push_interrupt_frame, init the
  mRA of the frame.
  (vm_builtin_call_with_values_code, vm_handle_interrupt_code): Allocate
  larger frames.
* module/language/cps/slot-allocation.scm (allocate-slots): Frame size
  is 3.
* module/system/vm/disassembler.scm (define-clobber-parser): Bump frame
  size.

libguile/frames.h

@@ -44,6 +44,8 @@
    | Dynamic link                 |
    +------------------------------+
    | Virtual return address (vRA) |
+   +------------------------------+
+   | Machine return address (mRA) |
    +==============================+ <- fp
    | Local 0                      |
    +------------------------------+
@@ -57,12 +59,12 @@
    The stack grows down.
 
    The calling convention is that a caller prepares a stack frame
-   consisting of the saved FP and the saved virtual return address,
+   consisting of the saved FP, the saved virtual return addres, and the
-   followed by the procedure and then the arguments to the call, in
+   saved machine return address of the calling function, followed by the
-   order.  Thus in the beginning of a call, the procedure being called
+   procedure and then the arguments to the call, in order.  Thus in the
-   is in slot 0, the first argument is in slot 1, and the SP points to
+   beginning of a call, the procedure being called is in slot 0, the
-   the last argument.  The number of arguments, including the procedure,
+   first argument is in slot 1, and the SP points to the last argument.
-   is thus FP - SP.
+   The number of arguments, including the procedure, is thus FP - SP.
 
    After ensuring that the correct number of arguments have been passed,
    a function will set the stack pointer to point to the last local
@@ -103,11 +105,13 @@ union scm_vm_stack_element
   scm_t_bits as_bits;
 };
 
-#define SCM_FRAME_PREVIOUS_SP(fp)	((fp) + 2)
+#define SCM_FRAME_PREVIOUS_SP(fp)	((fp) + 3)
-#define SCM_FRAME_VIRTUAL_RETURN_ADDRESS(fp)    ((fp)[0].as_vcode)
+#define SCM_FRAME_MACHINE_RETURN_ADDRESS(fp)         ((fp)[0].as_mcode)
-#define SCM_FRAME_SET_VIRTUAL_RETURN_ADDRESS(fp, ra) ((fp)[0].as_vcode = (ra))
+#define SCM_FRAME_SET_MACHINE_RETURN_ADDRESS(fp, ra) ((fp)[0].as_mcode = (ra))
-#define SCM_FRAME_DYNAMIC_LINK(fp)      ((fp) + (fp)[1].as_uint)
+#define SCM_FRAME_VIRTUAL_RETURN_ADDRESS(fp)         ((fp)[1].as_vcode)
-#define SCM_FRAME_SET_DYNAMIC_LINK(fp, dl) ((fp)[1].as_uint = ((dl) - (fp)))
+#define SCM_FRAME_SET_VIRTUAL_RETURN_ADDRESS(fp, ra) ((fp)[1].as_vcode = (ra))
+#define SCM_FRAME_DYNAMIC_LINK(fp)      ((fp) + (fp)[2].as_uint)
+#define SCM_FRAME_SET_DYNAMIC_LINK(fp, dl) ((fp)[2].as_uint = ((dl) - (fp)))
 #define SCM_FRAME_SLOT(fp,i)            ((fp) - (i) - 1)
 #define SCM_FRAME_LOCAL(fp,i)           (SCM_FRAME_SLOT (fp, i)->as_scm)
 #define SCM_FRAME_NUM_LOCALS(fp, sp)    ((fp) - (sp))

libguile/vm-engine.c

@@ -321,7 +321,7 @@ VM_NAME (scm_thread *thread)
    */
   VM_DEFINE_OP (0, halt, "halt", OP1 (X32))
     {
-      size_t frame_size = 2;
+      size_t frame_size = 3;
       /* Boot closure, then empty frame, then callee, then values.  */
       size_t first_value = 1 + frame_size + 1;
       uint32_t nvals = FRAME_LOCALS_COUNT_FROM (first_value);
@@ -373,6 +373,7 @@ VM_NAME (scm_thread *thread)
       VP->fp = SCM_FRAME_SLOT (old_fp, proc - 1);
       SCM_FRAME_SET_DYNAMIC_LINK (VP->fp, old_fp);
       SCM_FRAME_SET_VIRTUAL_RETURN_ADDRESS (VP->fp, ip + 2);
+      SCM_FRAME_SET_MACHINE_RETURN_ADDRESS (VP->fp, 0);
 
       RESET_FRAME (nlocals);
 
@@ -416,6 +417,7 @@ VM_NAME (scm_thread *thread)
       VP->fp = SCM_FRAME_SLOT (old_fp, proc - 1);
       SCM_FRAME_SET_DYNAMIC_LINK (VP->fp, old_fp);
       SCM_FRAME_SET_VIRTUAL_RETURN_ADDRESS (VP->fp, ip + 3);
+      SCM_FRAME_SET_MACHINE_RETURN_ADDRESS (VP->fp, 0);
 
       RESET_FRAME (nlocals);
 

libguile/vm.c

@@ -351,11 +351,11 @@ static const uint32_t vm_builtin_abort_to_prompt_code[] = {
 
 static const uint32_t vm_builtin_call_with_values_code[] = {
   SCM_PACK_OP_24 (assert_nargs_ee, 3),
-  SCM_PACK_OP_24 (alloc_frame, 7),
+  SCM_PACK_OP_24 (alloc_frame, 8),
-  SCM_PACK_OP_12_12 (mov, 0, 5),
+  SCM_PACK_OP_12_12 (mov, 0, 6),
-  SCM_PACK_OP_24 (call, 6), SCM_PACK_OP_ARG_8_24 (0, 1),
+  SCM_PACK_OP_24 (call, 7), SCM_PACK_OP_ARG_8_24 (0, 1),
   SCM_PACK_OP_24 (long_fmov, 0), SCM_PACK_OP_ARG_8_24 (0, 2),
-  SCM_PACK_OP_24 (tail_call_shuffle, 7)
+  SCM_PACK_OP_24 (tail_call_shuffle, 8)
 };
 
 static const uint32_t vm_builtin_call_with_current_continuation_code[] = {
@@ -364,9 +364,9 @@ static const uint32_t vm_builtin_call_with_current_continuation_code[] = {
 };
 
 static const uint32_t vm_handle_interrupt_code[] = {
-  SCM_PACK_OP_24 (alloc_frame, 3),
+  SCM_PACK_OP_24 (alloc_frame, 4),
-  SCM_PACK_OP_12_12 (mov, 0, 2),
+  SCM_PACK_OP_12_12 (mov, 0, 3),
-  SCM_PACK_OP_24 (call, 2), SCM_PACK_OP_ARG_8_24 (0, 1),
+  SCM_PACK_OP_24 (call, 3), SCM_PACK_OP_ARG_8_24 (0, 1),
   SCM_PACK_OP_24 (return_from_interrupt, 0)
 };
 
@@ -1014,7 +1014,7 @@ static void
 push_interrupt_frame (scm_thread *thread, uint8_t *mra)
 {
   union scm_vm_stack_element *old_fp;
-  size_t frame_overhead = 2;
+  size_t frame_overhead = 3;
   size_t old_frame_size = frame_locals_count (thread);
   SCM proc = scm_i_async_pop (thread);
 
@@ -1030,6 +1030,7 @@ push_interrupt_frame (scm_thread *thread, uint8_t *mra)
   /* Arrange to return to the same handle-interrupts opcode to handle
      any additional interrupts.  */
   SCM_FRAME_SET_VIRTUAL_RETURN_ADDRESS (thread->vm.fp, thread->vm.ip);
+  SCM_FRAME_SET_MACHINE_RETURN_ADDRESS (thread->vm.fp, mra);
 
   SCM_FRAME_LOCAL (thread->vm.fp, 0) = proc;
 }
@@ -1070,7 +1071,7 @@ reinstate_continuation_x (scm_thread *thread, SCM cont)
   scm_t_contregs *continuation = scm_i_contregs (cont);
   struct scm_vm *vp = &thread->vm;
   struct scm_vm_cont *cp;
-  size_t n, i, frame_overhead = 2;
+  size_t n, i, frame_overhead = 3;
   union scm_vm_stack_element *argv;
   struct return_to_continuation_data data;
 
@@ -1374,7 +1375,7 @@ scm_call_n (SCM proc, SCM *argv, size_t nargs)
      elements and each element is at least 4 bytes, nargs will not be
      greater than INTMAX/2 and therefore we don't have to check for
      overflow here or below.  */
-  size_t return_nlocals = 1, call_nlocals = nargs + 1, frame_size = 2;
+  size_t return_nlocals = 1, call_nlocals = nargs + 1, frame_size = 3;
   ptrdiff_t stack_reserve_words;
   size_t i;
 
@@ -1399,6 +1400,7 @@ scm_call_n (SCM proc, SCM *argv, size_t nargs)
   return_fp = call_fp + frame_size + return_nlocals;
 
   SCM_FRAME_SET_VIRTUAL_RETURN_ADDRESS (return_fp, vp->ip);
+  SCM_FRAME_SET_MACHINE_RETURN_ADDRESS (return_fp, 0);
   SCM_FRAME_SET_DYNAMIC_LINK (return_fp, vp->fp);
   SCM_FRAME_LOCAL (return_fp, 0) = vm_boot_continuation;
 
@@ -1406,6 +1408,7 @@ scm_call_n (SCM proc, SCM *argv, size_t nargs)
   vp->fp = call_fp;
 
   SCM_FRAME_SET_VIRTUAL_RETURN_ADDRESS (call_fp, vp->ip);
+  SCM_FRAME_SET_MACHINE_RETURN_ADDRESS (call_fp, 0);
   SCM_FRAME_SET_DYNAMIC_LINK (call_fp, return_fp);
   SCM_FRAME_LOCAL (call_fp, 0) = proc;
   for (i = 0; i < nargs; i++)

module/language/cps/slot-allocation.scm

@@ -807,7 +807,7 @@ are comparable with eqv?.  A tmp slot may be used."
                                                needs-slot)
                             empty-intset)))
 
-    (define frame-size 2)
+    (define frame-size 3)
 
     (define (empty-live-slots)
       #b0)

module/system/vm/disassembler.scm

@@ -615,7 +615,7 @@ address of that offset."
                                   (lambda ()
                                     (disassemble-one code (/ pos 4)))
                                 (lambda (len elt)
-                                  (define frame-size 2)
+                                  (define frame-size 3)
                                   (match elt
                                     ((_ proc . _)
                                      (let lp ((slot (- proc frame-size)))