Re: Tree inlining for the C front end (part 1 of 3)

[Alexandre Oliva <aoliva at redhat dot com>]

On Sep 24, 2001, Alexandre Oliva <aoliva@redhat.com> wrote:

> The plan remains, with patches 2 and 3 to follow this one in a matter
> of seconds, but I decided to put in a bit of addition stuff into patch
                                                       ^al
> #1 (LANG_DISREGARD_INLINE_LIMITS) and name the hooks and hook types
> more uniformly.

Except that the change to inlinable_function_p that took advantage of
and justified LANG_DISREGARD_INLINE_LIMITS somehow didn't make it to
the patch I posted.  Here's a patch that adds applies after
tree-inline-1.patch to get the desired result, and a patch file that
fixes tree-inline-2.patch so that it applies cleanly afterwards,
preserving the fix.

--- gcc/cp/optimize.c 2001/09/24 19:50:30
+++ gcc/cp/optimize.c 2001/09/25 10:59:02
@@ -632,2 +632,4 @@
-     function to eat up half of our budget.  */
-  else if (DECL_NUM_STMTS (fn) * INSNS_PER_STMT > MAX_INLINE_INSNS / 2)
+     function to eat up half of our budget.  Make special allowance
+     for extern inline functions, though.  */
+  else if (! LANG_DISREGARD_INLINE_LIMITS (fn)
+	   && DECL_NUM_STMTS (fn) * INSNS_PER_STMT > MAX_INLINE_INSNS / 2)
@@ -649,2 +651,3 @@
-  if ((DECL_NUM_STMTS (fn) + id->inlined_stmts) * INSNS_PER_STMT
-      > MAX_INLINE_INSNS
+  if (! LANG_DISREGARD_INLINE_LIMITS (fn)
+      && ((DECL_NUM_STMTS (fn) + id->inlined_stmts) * INSNS_PER_STMT
+	  > MAX_INLINE_INSNS)
39,40c38,39
< --- gcc/tree-inline.c Mon Sep 24 12:49:47 2001
< +++ gcc/tree-inline.c Mon Sep 24 12:55:48 2001
---
> --- gcc/tree-inline.c Tue Sep 25 03:59:00 2001
> +++ gcc/tree-inline.c Tue Sep 25 03:56:29 2001
56c55
< @@ -42,0 +57,1303 @@
---
> @@ -43,0 +58,1306 @@
649,650c648,651
< +     function to eat up half of our budget.  */
< +  else if (DECL_NUM_STMTS (fn) * INSNS_PER_STMT > MAX_INLINE_INSNS / 2)
---
> +     function to eat up half of our budget.  Make special allowance
> +     for extern inline functions, though.  */
> +  else if (! LANG_DISREGARD_INLINE_LIMITS (fn)
> +	   && DECL_NUM_STMTS (fn) * INSNS_PER_STMT > MAX_INLINE_INSNS / 2)
666,667c667,669
< +  if ((DECL_NUM_STMTS (fn) + id->inlined_stmts) * INSNS_PER_STMT
< +      > MAX_INLINE_INSNS
---
> +  if (! LANG_DISREGARD_INLINE_LIMITS (fn)
> +      && ((DECL_NUM_STMTS (fn) + id->inlined_stmts) * INSNS_PER_STMT
> +	  > MAX_INLINE_INSNS)
1360,1361c1362,1363
< --- gcc/cp/optimize.c 2001/09/24 19:50:30
< +++ gcc/cp/optimize.c 2001/09/24 19:56:19
---
> --- gcc/cp/optimize.c 2001/09/25 10:59:02
> +++ gcc/cp/optimize.c 2001/09/25 10:56:31
1427c1429
< @@ -107,870 +43,0 @@
---
> @@ -107,873 +43,0 @@
1953,1954c1955,1958
< -     function to eat up half of our budget.  */
< -  else if (DECL_NUM_STMTS (fn) * INSNS_PER_STMT > MAX_INLINE_INSNS / 2)
---
> -     function to eat up half of our budget.  Make special allowance
> -     for extern inline functions, though.  */
> -  else if (! LANG_DISREGARD_INLINE_LIMITS (fn)
> -	   && DECL_NUM_STMTS (fn) * INSNS_PER_STMT > MAX_INLINE_INSNS / 2)
1970,1971c1974,1976
< -  if ((DECL_NUM_STMTS (fn) + id->inlined_stmts) * INSNS_PER_STMT
< -      > MAX_INLINE_INSNS
---
> -  if (! LANG_DISREGARD_INLINE_LIMITS (fn)
> -      && ((DECL_NUM_STMTS (fn) + id->inlined_stmts) * INSNS_PER_STMT
> -	  > MAX_INLINE_INSNS)
2298c2303
< @@ -1067,32 +133,0 @@
---
> @@ -1070,32 +133,0 @@

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                  aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist    *Please* write to mailing lists, not to me