* Random cleanups [2/4]: canonicalize ctor values
@ 2011-03-31 1:32 Michael Matz
2011-03-31 8:13 ` Paolo Bonzini
0 siblings, 1 reply; 8+ messages in thread
From: Michael Matz @ 2011-03-31 1:32 UTC (permalink / raw)
To: gcc-patches
Hi,
this came up when looking into why the static ctors contain useless trees
(like casts). We can simply canonicalize them while varpool analyzes
pending decls. It'll look at initialzers once, where we can "gimplify"
them. This requires making canonicalize_constructor_val be able to be
called outside of functions. And it requires the java frontend not
leaving a dangling function decl as current (for the static ctor function
it generates).
Regstrapped on x86_64-linux with the other three cleanups. Okay for
trunk?
Ciao,
Michael.
--
* cgraphbuild.c (record_reference): Canonicalize constructor
values.
* gimple-fold.c (canonicalize_constructor_val): Accept being called
without function context.
java/
* jcf-parse.c (java_emit_static_constructor): Clear cfun and
current_function_decl.
Index: cgraphbuild.c
===================================================================
--- cgraphbuild.c (revision 171537)
+++ cgraphbuild.c (working copy)
@@ -53,6 +53,8 @@ record_reference (tree *tp, int *walk_su
tree decl;
struct record_reference_ctx *ctx = (struct record_reference_ctx *)data;
+restart:
+
switch (TREE_CODE (t))
{
case VAR_DECL:
@@ -98,6 +100,15 @@ record_reference (tree *tp, int *walk_su
break;
}
+ t = canonicalize_constructor_val (t);
+ if (t && t != *tp)
+ {
+ *tp = t;
+ goto restart;
+ }
+ else
+ t = *tp;
+
if ((unsigned int) TREE_CODE (t) >= LAST_AND_UNUSED_TREE_CODE)
return lang_hooks.callgraph.analyze_expr (tp, walk_subtrees);
break;
Index: gimple-fold.c
===================================================================
--- gimple-fold.c (revision 171537)
+++ gimple-fold.c (working copy)
@@ -106,7 +106,7 @@ can_refer_decl_in_current_unit_p (tree d
return true;
}
-/* CVAL is value taken from DECL_INITIAL of variable. Try to transorm it into
+/* CVAL is value taken from DECL_INITIAL of variable. Try to transform it into
acceptable form for is_gimple_min_invariant. */
tree
@@ -131,7 +131,7 @@ canonicalize_constructor_val (tree cval)
|| TREE_CODE (base) == FUNCTION_DECL)
&& !can_refer_decl_in_current_unit_p (base))
return NULL_TREE;
- if (base && TREE_CODE (base) == VAR_DECL)
+ if (cfun && base && TREE_CODE (base) == VAR_DECL)
add_referenced_var (base);
/* We never have the chance to fixup types in global initializers
during gimplification. Do so here. */
Index: java/jcf-parse.c
===================================================================
--- java/jcf-parse.c.orig 2011-03-26 02:19:03.000000000 +0100
+++ java/jcf-parse.c 2011-03-28 06:16:43.000000000 +0200
@@ -1725,6 +1725,8 @@ java_emit_static_constructor (void)
DECL_STATIC_CONSTRUCTOR (decl) = 1;
java_genericize (decl);
cgraph_finalize_function (decl, false);
+ current_function_decl = NULL;
+ set_cfun (NULL);
}
}
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: Random cleanups [2/4]: canonicalize ctor values
2011-03-31 1:32 Random cleanups [2/4]: canonicalize ctor values Michael Matz
@ 2011-03-31 8:13 ` Paolo Bonzini
2011-03-31 9:35 ` Richard Guenther
0 siblings, 1 reply; 8+ messages in thread
From: Paolo Bonzini @ 2011-03-31 8:13 UTC (permalink / raw)
To: Michael Matz; +Cc: gcc-patches
On 03/31/2011 03:22 AM, Michael Matz wrote:
> this came up when looking into why the static ctors contain useless trees
> (like casts). We can simply canonicalize them while varpool analyzes
> pending decls. It'll look at initialzers once, where we can "gimplify"
> them. This requires making canonicalize_constructor_val be able to be
> called outside of functions. And it requires the java frontend not
> leaving a dangling function decl as current (for the static ctor function
> it generates).
canonicalize_constructor_val may be doing useful things on ADDR_EXPR
too, but you don't call it in that case because you only added your code
in the default case. You only reach it when the ADDR_EXPR is wrapped in
a cast.
Any reason why you didn't add it (possibly wrapped in a loop) _before_
the switch statement?
Paolo
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: Random cleanups [2/4]: canonicalize ctor values
2011-03-31 8:13 ` Paolo Bonzini
@ 2011-03-31 9:35 ` Richard Guenther
2011-03-31 16:01 ` Michael Matz
0 siblings, 1 reply; 8+ messages in thread
From: Richard Guenther @ 2011-03-31 9:35 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: Michael Matz, gcc-patches
On Thu, Mar 31, 2011 at 9:14 AM, Paolo Bonzini <bonzini@gnu.org> wrote:
> On 03/31/2011 03:22 AM, Michael Matz wrote:
>>
>> this came up when looking into why the static ctors contain useless trees
>> (like casts). We can simply canonicalize them while varpool analyzes
>> pending decls. It'll look at initialzers once, where we can "gimplify"
>> them. This requires making canonicalize_constructor_val be able to be
>> called outside of functions. And it requires the java frontend not
>> leaving a dangling function decl as current (for the static ctor function
>> it generates).
>
> canonicalize_constructor_val may be doing useful things on ADDR_EXPR too,
> but you don't call it in that case because you only added your code in the
> default case. You only reach it when the ADDR_EXPR is wrapped in a cast.
>
> Any reason why you didn't add it (possibly wrapped in a loop) _before_ the
> switch statement?
Indeed. This comment needs adjustment, too:
/* We never have the chance to fixup types in global initializers
during gimplification. Do so here. */
if (TREE_TYPE (TREE_TYPE (cval)) != TREE_TYPE (TREE_OPERAND (cval, 0)))
cval = build_fold_addr_expr (TREE_OPERAND (cval, 0));
as we now sort-of do this. In fact it looks like all existing
canonicalize_constructor_val calls are subsumed by the lowering, so,
eventually remove those and move this function to a local place next
to its sole caller.
Richard.
> Paolo
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: Random cleanups [2/4]: canonicalize ctor values
2011-03-31 9:35 ` Richard Guenther
@ 2011-03-31 16:01 ` Michael Matz
2011-03-31 16:04 ` Richard Guenther
0 siblings, 1 reply; 8+ messages in thread
From: Michael Matz @ 2011-03-31 16:01 UTC (permalink / raw)
To: Richard Guenther; +Cc: Paolo Bonzini, gcc-patches
[-- Attachment #1: Type: TEXT/PLAIN, Size: 4919 bytes --]
On Thu, 31 Mar 2011, Richard Guenther wrote:
> > canonicalize_constructor_val may be doing useful things on ADDR_EXPR
> > too, but you don't call it in that case because you only added your
> > code in the default case. Â You only reach it when the ADDR_EXPR is
> > wrapped in a cast.
> >
> > Any reason why you didn't add it (possibly wrapped in a loop) _before_
> > the switch statement?
Nope, probably I didn't want to pay the cost each time, but you're right.
> Indeed. This comment needs adjustment, too:
>
> /* We never have the chance to fixup types in global initializers
> during gimplification. Do so here. */
Done in new patch.
> as we now sort-of do this. In fact it looks like all existing
> canonicalize_constructor_val calls are subsumed by the lowering, so,
> eventually remove those and move this function to a local place next to
> its sole caller.
Yeah. I couldn't immediately convince myself that all simplifications
will also be done on local vars (for which record_reference won't run), so
I'll leave this for later (I'm for instance doubtful about the NULL return
for when the current unit "can't" refer to a symbol). FWIW I'm currently
testing with these asserts which should tell me:
-----------------------------------------------------
@@ -152,7 +151,8 @@ get_symbol_constant_value (tree sym)
tree val = DECL_INITIAL (sym);
if (val)
{
- val = canonicalize_constructor_val (val);
+ tree newval = canonicalize_constructor_val (val);
+ gcc_assert (newval == val);
if (val && is_gimple_min_invariant (val))
return val;
else
@@ -3164,7 +3164,11 @@ fold_ctor_reference (tree type, tree cto
/* We found the field with exact match. */
if (useless_type_conversion_p (type, TREE_TYPE (ctor))
&& !offset)
- return canonicalize_constructor_val (ctor);
+ {
+ tree newctor = canonicalize_constructor_val (ctor);
+ gcc_assert (newctor == ctor);
+ return newctor;
+ }
/* We are at the end of walk, see if we can view convert the
result. */
@@ -3174,6 +3178,7 @@ fold_ctor_reference (tree type, tree cto
TYPE_SIZE (TREE_TYPE (ctor)), 0))
{
ret = canonicalize_constructor_val (ctor);
+ gcc_assert (ret == ctor);
ret = fold_unary (VIEW_CONVERT_EXPR, type, ret);
if (ret)
STRIP_NOPS (ret);
-----------------------------------------------------
In the meanwhile, is the below version okay?
Ciao,
Michael.
--
* cgraphbuild.c (record_reference): Canonicalize constructor
values.
* gimple-fold.c (canonicalize_constructor_val): Accept being called
without function context.
java/
* jcf-parse.c (java_emit_static_constructor): Clear cfun and
current_function_decl.
Index: cgraphbuild.c
===================================================================
--- cgraphbuild.c.orig 2011-03-29 17:08:15.000000000 +0200
+++ cgraphbuild.c 2011-03-31 17:43:29.000000000 +0200
@@ -53,6 +53,12 @@ record_reference (tree *tp, int *walk_su
tree decl;
struct record_reference_ctx *ctx = (struct record_reference_ctx *)data;
+ t = canonicalize_constructor_val (t);
+ if (!t)
+ t = *tp;
+ else if (t != *tp)
+ *tp = t;
+
switch (TREE_CODE (t))
{
case VAR_DECL:
Index: gimple-fold.c
===================================================================
--- gimple-fold.c.orig 2011-03-29 17:08:15.000000000 +0200
+++ gimple-fold.c 2011-03-31 17:42:45.000000000 +0200
@@ -106,7 +106,7 @@ can_refer_decl_in_current_unit_p (tree d
return true;
}
-/* CVAL is value taken from DECL_INITIAL of variable. Try to transorm it into
+/* CVAL is value taken from DECL_INITIAL of variable. Try to transform it into
acceptable form for is_gimple_min_invariant. */
tree
@@ -131,10 +131,9 @@ canonicalize_constructor_val (tree cval)
|| TREE_CODE (base) == FUNCTION_DECL)
&& !can_refer_decl_in_current_unit_p (base))
return NULL_TREE;
- if (base && TREE_CODE (base) == VAR_DECL)
+ if (cfun && base && TREE_CODE (base) == VAR_DECL)
add_referenced_var (base);
- /* We never have the chance to fixup types in global initializers
- during gimplification. Do so here. */
+ /* Fixup types in global initializers. */
if (TREE_TYPE (TREE_TYPE (cval)) != TREE_TYPE (TREE_OPERAND (cval, 0)))
cval = build_fold_addr_expr (TREE_OPERAND (cval, 0));
}
Index: java/jcf-parse.c
===================================================================
--- java/jcf-parse.c.orig 2011-03-29 17:08:15.000000000 +0200
+++ java/jcf-parse.c 2011-03-29 17:12:05.000000000 +0200
@@ -1725,6 +1725,8 @@ java_emit_static_constructor (void)
DECL_STATIC_CONSTRUCTOR (decl) = 1;
java_genericize (decl);
cgraph_finalize_function (decl, false);
+ current_function_decl = NULL;
+ set_cfun (NULL);
}
}
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: Random cleanups [2/4]: canonicalize ctor values
2011-03-31 16:01 ` Michael Matz
@ 2011-03-31 16:04 ` Richard Guenther
2011-04-03 10:14 ` Michael Matz
0 siblings, 1 reply; 8+ messages in thread
From: Richard Guenther @ 2011-03-31 16:04 UTC (permalink / raw)
To: Michael Matz; +Cc: Paolo Bonzini, gcc-patches
On Thu, Mar 31, 2011 at 5:57 PM, Michael Matz <matz@suse.de> wrote:
> On Thu, 31 Mar 2011, Richard Guenther wrote:
>
>> > canonicalize_constructor_val may be doing useful things on ADDR_EXPR
>> > too, but you don't call it in that case because you only added your
>> > code in the default case. You only reach it when the ADDR_EXPR is
>> > wrapped in a cast.
>> >
>> > Any reason why you didn't add it (possibly wrapped in a loop) _before_
>> > the switch statement?
>
> Nope, probably I didn't want to pay the cost each time, but you're right.
>
>> Indeed. This comment needs adjustment, too:
>>
>> /* We never have the chance to fixup types in global initializers
>> during gimplification. Do so here. */
>
> Done in new patch.
>
>> as we now sort-of do this. In fact it looks like all existing
>> canonicalize_constructor_val calls are subsumed by the lowering, so,
>> eventually remove those and move this function to a local place next to
>> its sole caller.
>
> Yeah. I couldn't immediately convince myself that all simplifications
> will also be done on local vars (for which record_reference won't run), so
> I'll leave this for later (I'm for instance doubtful about the NULL return
> for when the current unit "can't" refer to a symbol). FWIW I'm currently
> testing with these asserts which should tell me:
>
> -----------------------------------------------------
> @@ -152,7 +151,8 @@ get_symbol_constant_value (tree sym)
> tree val = DECL_INITIAL (sym);
> if (val)
> {
> - val = canonicalize_constructor_val (val);
> + tree newval = canonicalize_constructor_val (val);
> + gcc_assert (newval == val);
> if (val && is_gimple_min_invariant (val))
> return val;
> else
> @@ -3164,7 +3164,11 @@ fold_ctor_reference (tree type, tree cto
> /* We found the field with exact match. */
> if (useless_type_conversion_p (type, TREE_TYPE (ctor))
> && !offset)
> - return canonicalize_constructor_val (ctor);
> + {
> + tree newctor = canonicalize_constructor_val (ctor);
> + gcc_assert (newctor == ctor);
> + return newctor;
> + }
>
> /* We are at the end of walk, see if we can view convert the
> result. */
> @@ -3174,6 +3178,7 @@ fold_ctor_reference (tree type, tree cto
> TYPE_SIZE (TREE_TYPE (ctor)), 0))
> {
> ret = canonicalize_constructor_val (ctor);
> + gcc_assert (ret == ctor);
> ret = fold_unary (VIEW_CONVERT_EXPR, type, ret);
> if (ret)
> STRIP_NOPS (ret);
> -----------------------------------------------------
>
> In the meanwhile, is the below version okay?
If it bootstraps & tests ok then yes. The java parts look obvious.
Thanks,
Richard.
>
> Ciao,
> Michael.
> --
> * cgraphbuild.c (record_reference): Canonicalize constructor
> values.
> * gimple-fold.c (canonicalize_constructor_val): Accept being called
> without function context.
>
> java/
> * jcf-parse.c (java_emit_static_constructor): Clear cfun and
> current_function_decl.
>
> Index: cgraphbuild.c
> ===================================================================
> --- cgraphbuild.c.orig 2011-03-29 17:08:15.000000000 +0200
> +++ cgraphbuild.c 2011-03-31 17:43:29.000000000 +0200
> @@ -53,6 +53,12 @@ record_reference (tree *tp, int *walk_su
> tree decl;
> struct record_reference_ctx *ctx = (struct record_reference_ctx *)data;
>
> + t = canonicalize_constructor_val (t);
> + if (!t)
> + t = *tp;
> + else if (t != *tp)
> + *tp = t;
> +
> switch (TREE_CODE (t))
> {
> case VAR_DECL:
> Index: gimple-fold.c
> ===================================================================
> --- gimple-fold.c.orig 2011-03-29 17:08:15.000000000 +0200
> +++ gimple-fold.c 2011-03-31 17:42:45.000000000 +0200
> @@ -106,7 +106,7 @@ can_refer_decl_in_current_unit_p (tree d
> return true;
> }
>
> -/* CVAL is value taken from DECL_INITIAL of variable. Try to transorm it into
> +/* CVAL is value taken from DECL_INITIAL of variable. Try to transform it into
> acceptable form for is_gimple_min_invariant. */
>
> tree
> @@ -131,10 +131,9 @@ canonicalize_constructor_val (tree cval)
> || TREE_CODE (base) == FUNCTION_DECL)
> && !can_refer_decl_in_current_unit_p (base))
> return NULL_TREE;
> - if (base && TREE_CODE (base) == VAR_DECL)
> + if (cfun && base && TREE_CODE (base) == VAR_DECL)
> add_referenced_var (base);
> - /* We never have the chance to fixup types in global initializers
> - during gimplification. Do so here. */
> + /* Fixup types in global initializers. */
> if (TREE_TYPE (TREE_TYPE (cval)) != TREE_TYPE (TREE_OPERAND (cval, 0)))
> cval = build_fold_addr_expr (TREE_OPERAND (cval, 0));
> }
> Index: java/jcf-parse.c
> ===================================================================
> --- java/jcf-parse.c.orig 2011-03-29 17:08:15.000000000 +0200
> +++ java/jcf-parse.c 2011-03-29 17:12:05.000000000 +0200
> @@ -1725,6 +1725,8 @@ java_emit_static_constructor (void)
> DECL_STATIC_CONSTRUCTOR (decl) = 1;
> java_genericize (decl);
> cgraph_finalize_function (decl, false);
> + current_function_decl = NULL;
> + set_cfun (NULL);
> }
> }
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: Random cleanups [2/4]: canonicalize ctor values
2011-03-31 16:04 ` Richard Guenther
@ 2011-04-03 10:14 ` Michael Matz
2011-04-03 10:45 ` Richard Guenther
2011-04-04 22:00 ` H.J. Lu
0 siblings, 2 replies; 8+ messages in thread
From: Michael Matz @ 2011-04-03 10:14 UTC (permalink / raw)
To: Richard Guenther; +Cc: Paolo Bonzini, gcc-patches
Hi,
On Thu, 31 Mar 2011, Richard Guenther wrote:
> > In the meanwhile, is the below version okay?
>
> If it bootstraps & tests ok then yes. The java parts look obvious.
So, we indeed can't remove the other calls to
canonicalize_constructor_val, because of local ctors. And fortran has a
similar problem with java. Instead of fixing up all these places of
resetting cfun (where otherwise the frontends don't deal at all with it,
it's mostly just set from the various cgraph routines), I decided to
simply clear this at the appropriate place in
cgraph_finalize_compilation_unit.
Regstrapping in progress again. Still okay if that works?
Ciao,
Michael.
--
* cgraphbuild.c (record_reference): Canonicalize constructor
values.
* gimple-fold.c (canonicalize_constructor_val): Accept being called
without function context.
* cgraphunit.c (cgraph_finalize_compilation_unit): Clear
current_function_decl and cfun.
Index: cgraphbuild.c
===================================================================
--- cgraphbuild.c.orig 2011-04-03 11:28:45.000000000 +0200
+++ cgraphbuild.c 2011-04-03 11:31:21.000000000 +0200
@@ -53,6 +53,12 @@ record_reference (tree *tp, int *walk_su
tree decl;
struct record_reference_ctx *ctx = (struct record_reference_ctx *)data;
+ t = canonicalize_constructor_val (t);
+ if (!t)
+ t = *tp;
+ else if (t != *tp)
+ *tp = t;
+
switch (TREE_CODE (t))
{
case VAR_DECL:
Index: gimple-fold.c
===================================================================
--- gimple-fold.c.orig 2011-04-03 11:28:45.000000000 +0200
+++ gimple-fold.c 2011-04-03 11:31:21.000000000 +0200
@@ -106,7 +106,7 @@ can_refer_decl_in_current_unit_p (tree d
return true;
}
-/* CVAL is value taken from DECL_INITIAL of variable. Try to transorm it into
+/* CVAL is value taken from DECL_INITIAL of variable. Try to transform it into
acceptable form for is_gimple_min_invariant. */
tree
@@ -131,10 +131,9 @@ canonicalize_constructor_val (tree cval)
|| TREE_CODE (base) == FUNCTION_DECL)
&& !can_refer_decl_in_current_unit_p (base))
return NULL_TREE;
- if (base && TREE_CODE (base) == VAR_DECL)
+ if (cfun && base && TREE_CODE (base) == VAR_DECL)
add_referenced_var (base);
- /* We never have the chance to fixup types in global initializers
- during gimplification. Do so here. */
+ /* Fixup types in global initializers. */
if (TREE_TYPE (TREE_TYPE (cval)) != TREE_TYPE (TREE_OPERAND (cval, 0)))
cval = build_fold_addr_expr (TREE_OPERAND (cval, 0));
}
Index: cgraphunit.c
===================================================================
--- cgraphunit.c.orig 2011-04-03 11:41:46.000000000 +0200
+++ cgraphunit.c 2011-04-03 11:50:01.000000000 +0200
@@ -1066,6 +1066,11 @@ cgraph_finalize_compilation_unit (void)
{
timevar_push (TV_CGRAPH);
+ /* If we're here there's no current function anymore. Some frontends
+ are lazy in clearing these. */
+ current_function_decl = NULL;
+ set_cfun (NULL);
+
/* Do not skip analyzing the functions if there were errors, we
miss diagnostics for following functions otherwise. */
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: Random cleanups [2/4]: canonicalize ctor values
2011-04-03 10:14 ` Michael Matz
@ 2011-04-03 10:45 ` Richard Guenther
2011-04-04 22:00 ` H.J. Lu
1 sibling, 0 replies; 8+ messages in thread
From: Richard Guenther @ 2011-04-03 10:45 UTC (permalink / raw)
To: Michael Matz; +Cc: Paolo Bonzini, gcc-patches
On Sun, Apr 3, 2011 at 12:14 PM, Michael Matz <matz@suse.de> wrote:
> Hi,
>
> On Thu, 31 Mar 2011, Richard Guenther wrote:
>
>> > In the meanwhile, is the below version okay?
>>
>> If it bootstraps & tests ok then yes. The java parts look obvious.
>
> So, we indeed can't remove the other calls to
> canonicalize_constructor_val, because of local ctors. And fortran has a
> similar problem with java. Instead of fixing up all these places of
> resetting cfun (where otherwise the frontends don't deal at all with it,
> it's mostly just set from the various cgraph routines), I decided to
> simply clear this at the appropriate place in
> cgraph_finalize_compilation_unit.
>
> Regstrapping in progress again. Still okay if that works?
Ok.
Thanks,
Richard.
>
> Ciao,
> Michael.
> --
> * cgraphbuild.c (record_reference): Canonicalize constructor
> values.
> * gimple-fold.c (canonicalize_constructor_val): Accept being called
> without function context.
> * cgraphunit.c (cgraph_finalize_compilation_unit): Clear
> current_function_decl and cfun.
>
> Index: cgraphbuild.c
> ===================================================================
> --- cgraphbuild.c.orig 2011-04-03 11:28:45.000000000 +0200
> +++ cgraphbuild.c 2011-04-03 11:31:21.000000000 +0200
> @@ -53,6 +53,12 @@ record_reference (tree *tp, int *walk_su
> tree decl;
> struct record_reference_ctx *ctx = (struct record_reference_ctx *)data;
>
> + t = canonicalize_constructor_val (t);
> + if (!t)
> + t = *tp;
> + else if (t != *tp)
> + *tp = t;
> +
> switch (TREE_CODE (t))
> {
> case VAR_DECL:
> Index: gimple-fold.c
> ===================================================================
> --- gimple-fold.c.orig 2011-04-03 11:28:45.000000000 +0200
> +++ gimple-fold.c 2011-04-03 11:31:21.000000000 +0200
> @@ -106,7 +106,7 @@ can_refer_decl_in_current_unit_p (tree d
> return true;
> }
>
> -/* CVAL is value taken from DECL_INITIAL of variable. Try to transorm it into
> +/* CVAL is value taken from DECL_INITIAL of variable. Try to transform it into
> acceptable form for is_gimple_min_invariant. */
>
> tree
> @@ -131,10 +131,9 @@ canonicalize_constructor_val (tree cval)
> || TREE_CODE (base) == FUNCTION_DECL)
> && !can_refer_decl_in_current_unit_p (base))
> return NULL_TREE;
> - if (base && TREE_CODE (base) == VAR_DECL)
> + if (cfun && base && TREE_CODE (base) == VAR_DECL)
> add_referenced_var (base);
> - /* We never have the chance to fixup types in global initializers
> - during gimplification. Do so here. */
> + /* Fixup types in global initializers. */
> if (TREE_TYPE (TREE_TYPE (cval)) != TREE_TYPE (TREE_OPERAND (cval, 0)))
> cval = build_fold_addr_expr (TREE_OPERAND (cval, 0));
> }
> Index: cgraphunit.c
> ===================================================================
> --- cgraphunit.c.orig 2011-04-03 11:41:46.000000000 +0200
> +++ cgraphunit.c 2011-04-03 11:50:01.000000000 +0200
> @@ -1066,6 +1066,11 @@ cgraph_finalize_compilation_unit (void)
> {
> timevar_push (TV_CGRAPH);
>
> + /* If we're here there's no current function anymore. Some frontends
> + are lazy in clearing these. */
> + current_function_decl = NULL;
> + set_cfun (NULL);
> +
> /* Do not skip analyzing the functions if there were errors, we
> miss diagnostics for following functions otherwise. */
>
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: Random cleanups [2/4]: canonicalize ctor values
2011-04-03 10:14 ` Michael Matz
2011-04-03 10:45 ` Richard Guenther
@ 2011-04-04 22:00 ` H.J. Lu
1 sibling, 0 replies; 8+ messages in thread
From: H.J. Lu @ 2011-04-04 22:00 UTC (permalink / raw)
To: Michael Matz; +Cc: Richard Guenther, Paolo Bonzini, gcc-patches
On Sun, Apr 3, 2011 at 3:14 AM, Michael Matz <matz@suse.de> wrote:
> Hi,
>
> On Thu, 31 Mar 2011, Richard Guenther wrote:
>
>> > In the meanwhile, is the below version okay?
>>
>> If it bootstraps & tests ok then yes. The java parts look obvious.
>
> So, we indeed can't remove the other calls to
> canonicalize_constructor_val, because of local ctors. And fortran has a
> similar problem with java. Instead of fixing up all these places of
> resetting cfun (where otherwise the frontends don't deal at all with it,
> it's mostly just set from the various cgraph routines), I decided to
> simply clear this at the appropriate place in
> cgraph_finalize_compilation_unit.
>
> Regstrapping in progress again. Still okay if that works?
>
>
> Ciao,
> Michael.
> --
> * cgraphbuild.c (record_reference): Canonicalize constructor
> values.
>
> Index: cgraphbuild.c
> ===================================================================
> --- cgraphbuild.c.orig 2011-04-03 11:28:45.000000000 +0200
> +++ cgraphbuild.c 2011-04-03 11:31:21.000000000 +0200
> @@ -53,6 +53,12 @@ record_reference (tree *tp, int *walk_su
> tree decl;
> struct record_reference_ctx *ctx = (struct record_reference_ctx *)data;
>
> + t = canonicalize_constructor_val (t);
> + if (!t)
> + t = *tp;
> + else if (t != *tp)
> + *tp = t;
> +
> switch (TREE_CODE (t))
> {
> case VAR_DECL:
This change caused:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48440
--
H.J.
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2011-04-04 22:00 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-03-31 1:32 Random cleanups [2/4]: canonicalize ctor values Michael Matz
2011-03-31 8:13 ` Paolo Bonzini
2011-03-31 9:35 ` Richard Guenther
2011-03-31 16:01 ` Michael Matz
2011-03-31 16:04 ` Richard Guenther
2011-04-03 10:14 ` Michael Matz
2011-04-03 10:45 ` Richard Guenther
2011-04-04 22:00 ` H.J. Lu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).