Add a target hook for sibcall epilogues

Epilogues for sibling calls are generated using the
sibcall_epilogue pattern.  One disadvantage of this approach
is that the target doesn't know which call the epilogue is for,
even though the code that generates the pattern has the call
to hand.

Although call instructions are currently rtxes, and so could be
passed as an operand to the pattern, the main point of introducing
rtx_insn was to move towards separating the rtx and insn types
(a good thing IMO).  There also isn't an existing practice of
passing genuine instructions (as opposed to labels) to
instruction patterns.

This patch therefore adds a hook that can be defined as an
alternative to sibcall_epilogue.  The advantage is that it
can be passed the call; the disadvantage is that it can't
use .md conveniences like generating instructions from
textual patterns (although most epilogues are too complex
to benefit much from that anyway).

gcc/
	* doc/tm.texi.in: Add TARGET_EMIT_EPILOGUE_FOR_SIBCALL.
	* doc/tm.texi: Regenerate.
	* target.def (emit_epilogue_for_sibcall): New hook.
	* calls.cc (can_implement_as_sibling_call_p): Use it.
	* function.cc (thread_prologue_and_epilogue_insns): Likewise.
	(reposition_prologue_and_epilogue_notes): Likewise.
	* config/aarch64/aarch64-protos.h (aarch64_expand_epilogue): Take
	an rtx_call_insn * rather than a bool.
	* config/aarch64/aarch64.cc (aarch64_expand_epilogue): Likewise.
	(TARGET_EMIT_EPILOGUE_FOR_SIBCALL): Define.
	* config/aarch64/aarch64.md (epilogue): Update call.
	(sibcall_epilogue): Delete.

gcc/calls.cc

@@ -2502,7 +2502,8 @@ can_implement_as_sibling_call_p (tree exp,
 				 tree addr,
 				 const args_size &args_size)
 {
-  if (!targetm.have_sibcall_epilogue ())
+  if (!targetm.have_sibcall_epilogue ()
+      && !targetm.emit_epilogue_for_sibcall)
     {
       maybe_complain_about_tail_call
 	(exp,

gcc/config/aarch64/aarch64-protos.h

@@ -888,7 +888,7 @@ const char * aarch64_gen_far_branch (rtx *, int, const char *, const char *);
 const char * aarch64_output_probe_stack_range (rtx, rtx);
 const char * aarch64_output_probe_sve_stack_clash (rtx, rtx, rtx, rtx);
 void aarch64_err_no_fpadvsimd (machine_mode);
-void aarch64_expand_epilogue (bool);
+void aarch64_expand_epilogue (rtx_call_insn *);
 rtx aarch64_ptrue_all (unsigned int);
 opt_machine_mode aarch64_ptrue_all_mode (rtx);
 rtx aarch64_convert_sve_data_to_pred (rtx, machine_mode, rtx);

gcc/config/aarch64/aarch64.cc

@@ -7971,7 +7971,7 @@ aarch64_use_return_insn_p (void)
    from a deallocated stack, and we optimize the unwind records by
    emitting them all together if possible.  */
 void
-aarch64_expand_epilogue (bool for_sibcall)
+aarch64_expand_epilogue (rtx_call_insn *sibcall)
 {
   aarch64_frame &frame = cfun->machine->frame;
   poly_int64 initial_adjust = frame.initial_adjust;
@@ -8102,7 +8102,7 @@ aarch64_expand_epilogue (bool for_sibcall)
     }
 
   /* Stack adjustment for exception handler.  */
-  if (crtl->calls_eh_return && !for_sibcall)
+  if (crtl->calls_eh_return && !sibcall)
     {
       /* If the EH_RETURN_TAKEN_RTX flag is set then we need
 	 to unwind the stack and jump to the handler, otherwise
@@ -8137,7 +8137,7 @@ aarch64_expand_epilogue (bool for_sibcall)
 	   explicitly authenticate.
     */
   if (aarch64_return_address_signing_enabled ()
-      && (for_sibcall || !TARGET_ARMV8_3))
+      && (sibcall || !TARGET_ARMV8_3))
     {
       switch (aarch_ra_sign_key)
 	{
@@ -8155,7 +8155,7 @@ aarch64_expand_epilogue (bool for_sibcall)
     }
 
   emit_use (gen_rtx_REG (DImode, LR_REGNUM));
-  if (!for_sibcall)
+  if (!sibcall)
     emit_jump_insn (ret_rtx);
 }
 
@@ -26427,6 +26427,9 @@ aarch64_libgcc_floating_mode_supported_p
 #undef TARGET_CONST_ANCHOR
 #define TARGET_CONST_ANCHOR 0x1000000
 
+#undef TARGET_EMIT_EPILOGUE_FOR_SIBCALL
+#define TARGET_EMIT_EPILOGUE_FOR_SIBCALL aarch64_expand_epilogue
+
 struct gcc_target targetm = TARGET_INITIALIZER;
 
 #include "gt-aarch64.h"

gcc/config/aarch64/aarch64.md

@@ -863,16 +863,7 @@
   [(clobber (const_int 0))]
   ""
   "
-  aarch64_expand_epilogue (false);
-  DONE;
-  "
-)
-
-(define_expand "sibcall_epilogue"
-  [(clobber (const_int 0))]
-  ""
-  "
-  aarch64_expand_epilogue (true);
+  aarch64_expand_epilogue (nullptr);
   DONE;
   "
 )

gcc/doc/tm.texi

@@ -11898,6 +11898,14 @@ the hook might return true if the prologue and epilogue need to switch
 between instruction sets.
 @end deftypefn
 
+@deftypefn {Target Hook} void TARGET_EMIT_EPILOGUE_FOR_SIBCALL (rtx_call_insn *@var{call})
+If defined, this hook emits an epilogue sequence for sibling (tail)
+call instruction @var{call}.  Another way of providing epilogues
+for sibling calls is to define the @code{sibcall_epilogue} instruction
+pattern; the main advantage of this hook over the pattern is that it
+has access to the call instruction.
+@end deftypefn
+
 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
 If non-null, this hook performs a target-specific pass over the
 instruction stream.  The compiler will run it at all optimization levels,

gcc/doc/tm.texi.in

@@ -7786,6 +7786,8 @@ to by @var{ce_info}.
 
 @hook TARGET_USE_LATE_PROLOGUE_EPILOGUE
 
+@hook TARGET_EMIT_EPILOGUE_FOR_SIBCALL
+
 @hook TARGET_MACHINE_DEPENDENT_REORG
 
 @hook TARGET_INIT_BUILTINS

gcc/function.cc

@@ -6208,7 +6208,17 @@ thread_prologue_and_epilogue_insns (void)
       if (!(CALL_P (insn) && SIBLING_CALL_P (insn)))
 	continue;
 
-      if (rtx_insn *ep_seq = targetm.gen_sibcall_epilogue ())
+      rtx_insn *ep_seq;
+      if (targetm.emit_epilogue_for_sibcall)
+	{
+	  start_sequence ();
+	  targetm.emit_epilogue_for_sibcall (as_a<rtx_call_insn *> (insn));
+	  ep_seq = get_insns ();
+	  end_sequence ();
+	}
+      else
+	ep_seq = targetm.gen_sibcall_epilogue ();
+      if (ep_seq)
 	{
 	  start_sequence ();
 	  emit_note (NOTE_INSN_EPILOGUE_BEG);
@@ -6268,7 +6278,8 @@ reposition_prologue_and_epilogue_notes (void)
 {
   if (!targetm.have_prologue ()
       && !targetm.have_epilogue ()
-      && !targetm.have_sibcall_epilogue ())
+      && !targetm.have_sibcall_epilogue ()
+      && !targetm.emit_epilogue_for_sibcall)
     return;
 
   /* Since the hash table is created on demand, the fact that it is

gcc/target.def

@@ -4174,6 +4174,15 @@ between instruction sets.",
  bool, (),
  hook_bool_void_false)
 
+DEFHOOK
+(emit_epilogue_for_sibcall,
+ "If defined, this hook emits an epilogue sequence for sibling (tail)\n\
+call instruction @var{call}.  Another way of providing epilogues\n\
+for sibling calls is to define the @code{sibcall_epilogue} instruction\n\
+pattern; the main advantage of this hook over the pattern is that it\n\
+has access to the call instruction.",
+ void, (rtx_call_insn *call), NULL)
+
 /* Do machine-dependent code transformations.  Called just before
      delayed-branch scheduling.  */
 DEFHOOK