* [PATCH] Fix PR ipa/65665
@ 2015-04-03 8:45 Martin Liška
2015-04-03 17:19 ` Jan Hubicka
0 siblings, 1 reply; 3+ messages in thread
From: Martin Liška @ 2015-04-03 8:45 UTC (permalink / raw)
To: GCC Patches; +Cc: Jan Hubicka
[-- Attachment #1: Type: text/plain, Size: 217 bytes --]
Hello.
Following patch fixes a new issue, ICE has disappeared on aarch64-linux machine and I was able to boostrap on
the machine. Moreover, no new regression seen on x86_64-linux-pc.
Ready for trunk?
Thanks,
Martin
[-- Attachment #2: 0001-Fix-PR-ipa-65665.patch --]
[-- Type: text/x-patch, Size: 1896 bytes --]
From b8463b8a7b2f6b652d6d9c3a2ece814ec5554619 Mon Sep 17 00:00:00 2001
From: mliska <mliska@suse.cz>
Date: Fri, 3 Apr 2015 09:30:50 +0200
Subject: [PATCH] Fix PR ipa/65665
gcc/ChangeLog:
2015-04-03 Martin Liska <mliska@suse.cz>
PR ipa/65665
* ipa-icf.c (sem_function::equals_wpa): Verify that IPA CP
has computed data structure.
(sem_item_optimizer::update_hash_by_addr_refs): Likewise.
---
gcc/ipa-icf.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/gcc/ipa-icf.c b/gcc/ipa-icf.c
index 8626730..8f8a0cf 100644
--- a/gcc/ipa-icf.c
+++ b/gcc/ipa-icf.c
@@ -535,7 +535,8 @@ sem_function::equals_wpa (sem_item *item,
&& (TREE_CODE (TREE_TYPE (decl)) == METHOD_TYPE
|| TREE_CODE (TREE_TYPE (item->decl)) == METHOD_TYPE)
&& (ipa_node_params_sum == NULL
- || ipa_is_param_used (IPA_NODE_REF (dyn_cast <cgraph_node *>(node)),
+ || IPA_NODE_REF (get_node ())->descriptors.is_empty ()
+ || ipa_is_param_used (IPA_NODE_REF (get_node ()),
0))
&& compare_polymorphic_p ())
{
@@ -2501,14 +2502,15 @@ sem_item_optimizer::update_hash_by_addr_refs ()
m_items[i]->update_hash_by_addr_refs (m_symtab_node_map);
if (m_items[i]->type == FUNC)
{
+ cgraph_node *cnode = dyn_cast <cgraph_node *> (m_items[i]->node);
+
if (TREE_CODE (TREE_TYPE (m_items[i]->decl)) == METHOD_TYPE
&& contains_polymorphic_type_p
(method_class_type (TREE_TYPE (m_items[i]->decl)))
&& (DECL_CXX_CONSTRUCTOR_P (m_items[i]->decl)
|| ((ipa_node_params_sum == NULL
- || ipa_is_param_used (
- IPA_NODE_REF
- (dyn_cast <cgraph_node *>(m_items[i]->node)), 0))
+ || IPA_NODE_REF (cnode)->descriptors.is_empty ()
+ || ipa_is_param_used (IPA_NODE_REF (cnode), 0))
&& static_cast<sem_function *> (m_items[i])
->compare_polymorphic_p ())))
{
--
2.1.4
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Fix PR ipa/65665
2015-04-03 8:45 [PATCH] Fix PR ipa/65665 Martin Liška
@ 2015-04-03 17:19 ` Jan Hubicka
2015-04-07 20:25 ` Martin Liška
0 siblings, 1 reply; 3+ messages in thread
From: Jan Hubicka @ 2015-04-03 17:19 UTC (permalink / raw)
To: Martin Liška; +Cc: GCC Patches, Jan Hubicka
> Hello.
>
> Following patch fixes a new issue, ICE has disappeared on aarch64-linux machine and I was able to boostrap on
> the machine. Moreover, no new regression seen on x86_64-linux-pc.
>
> Ready for trunk?
> Thanks,
> Martin
> >From b8463b8a7b2f6b652d6d9c3a2ece814ec5554619 Mon Sep 17 00:00:00 2001
> From: mliska <mliska@suse.cz>
> Date: Fri, 3 Apr 2015 09:30:50 +0200
> Subject: [PATCH] Fix PR ipa/65665
>
> gcc/ChangeLog:
>
> 2015-04-03 Martin Liska <mliska@suse.cz>
>
> PR ipa/65665
> * ipa-icf.c (sem_function::equals_wpa): Verify that IPA CP
> has computed data structure.
> (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
OK,
we should clean this up next stage1 ideally making passmanager to do the ipa-prop
analysis and making them trigger for all ipa passes consuming them
(ipa-icf/ipa-cp/inliner). Currently the way they are initialized either by inliner
or ipa-cp is quite ugly.
Honza
> ---
> gcc/ipa-icf.c | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/gcc/ipa-icf.c b/gcc/ipa-icf.c
> index 8626730..8f8a0cf 100644
> --- a/gcc/ipa-icf.c
> +++ b/gcc/ipa-icf.c
> @@ -535,7 +535,8 @@ sem_function::equals_wpa (sem_item *item,
> && (TREE_CODE (TREE_TYPE (decl)) == METHOD_TYPE
> || TREE_CODE (TREE_TYPE (item->decl)) == METHOD_TYPE)
> && (ipa_node_params_sum == NULL
> - || ipa_is_param_used (IPA_NODE_REF (dyn_cast <cgraph_node *>(node)),
> + || IPA_NODE_REF (get_node ())->descriptors.is_empty ()
> + || ipa_is_param_used (IPA_NODE_REF (get_node ()),
> 0))
> && compare_polymorphic_p ())
> {
> @@ -2501,14 +2502,15 @@ sem_item_optimizer::update_hash_by_addr_refs ()
> m_items[i]->update_hash_by_addr_refs (m_symtab_node_map);
> if (m_items[i]->type == FUNC)
> {
> + cgraph_node *cnode = dyn_cast <cgraph_node *> (m_items[i]->node);
> +
> if (TREE_CODE (TREE_TYPE (m_items[i]->decl)) == METHOD_TYPE
> && contains_polymorphic_type_p
> (method_class_type (TREE_TYPE (m_items[i]->decl)))
> && (DECL_CXX_CONSTRUCTOR_P (m_items[i]->decl)
> || ((ipa_node_params_sum == NULL
> - || ipa_is_param_used (
> - IPA_NODE_REF
> - (dyn_cast <cgraph_node *>(m_items[i]->node)), 0))
> + || IPA_NODE_REF (cnode)->descriptors.is_empty ()
> + || ipa_is_param_used (IPA_NODE_REF (cnode), 0))
> && static_cast<sem_function *> (m_items[i])
> ->compare_polymorphic_p ())))
> {
> --
> 2.1.4
>
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Fix PR ipa/65665
2015-04-03 17:19 ` Jan Hubicka
@ 2015-04-07 20:25 ` Martin Liška
0 siblings, 0 replies; 3+ messages in thread
From: Martin Liška @ 2015-04-07 20:25 UTC (permalink / raw)
To: gcc-patches
On 04/03/2015 07:19 PM, Jan Hubicka wrote:
>> Hello.
>>
>> Following patch fixes a new issue, ICE has disappeared on aarch64-linux machine and I was able to boostrap on
>> the machine. Moreover, no new regression seen on x86_64-linux-pc.
>>
>> Ready for trunk?
>> Thanks,
>> Martin
>
>> >From b8463b8a7b2f6b652d6d9c3a2ece814ec5554619 Mon Sep 17 00:00:00 2001
>> From: mliska <mliska@suse.cz>
>> Date: Fri, 3 Apr 2015 09:30:50 +0200
>> Subject: [PATCH] Fix PR ipa/65665
>>
>> gcc/ChangeLog:
>>
>> 2015-04-03 Martin Liska <mliska@suse.cz>
>>
>> PR ipa/65665
>> * ipa-icf.c (sem_function::equals_wpa): Verify that IPA CP
>> has computed data structure.
>> (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
>
> OK,
> we should clean this up next stage1 ideally making passmanager to do the ipa-prop
> analysis and making them trigger for all ipa passes consuming them
> (ipa-icf/ipa-cp/inliner). Currently the way they are initialized either by inliner
> or ipa-cp is quite ugly.
Hi.
Agree with you, this problem shows how closely are individual passes connected
and how hard is it to split a pass to LGEN, WPA and LTRANS phase :)
Martin
>
> Honza
>> ---
>> gcc/ipa-icf.c | 10 ++++++----
>> 1 file changed, 6 insertions(+), 4 deletions(-)
>>
>> diff --git a/gcc/ipa-icf.c b/gcc/ipa-icf.c
>> index 8626730..8f8a0cf 100644
>> --- a/gcc/ipa-icf.c
>> +++ b/gcc/ipa-icf.c
>> @@ -535,7 +535,8 @@ sem_function::equals_wpa (sem_item *item,
>> && (TREE_CODE (TREE_TYPE (decl)) == METHOD_TYPE
>> || TREE_CODE (TREE_TYPE (item->decl)) == METHOD_TYPE)
>> && (ipa_node_params_sum == NULL
>> - || ipa_is_param_used (IPA_NODE_REF (dyn_cast <cgraph_node *>(node)),
>> + || IPA_NODE_REF (get_node ())->descriptors.is_empty ()
>> + || ipa_is_param_used (IPA_NODE_REF (get_node ()),
>> 0))
>> && compare_polymorphic_p ())
>> {
>> @@ -2501,14 +2502,15 @@ sem_item_optimizer::update_hash_by_addr_refs ()
>> m_items[i]->update_hash_by_addr_refs (m_symtab_node_map);
>> if (m_items[i]->type == FUNC)
>> {
>> + cgraph_node *cnode = dyn_cast <cgraph_node *> (m_items[i]->node);
>> +
>> if (TREE_CODE (TREE_TYPE (m_items[i]->decl)) == METHOD_TYPE
>> && contains_polymorphic_type_p
>> (method_class_type (TREE_TYPE (m_items[i]->decl)))
>> && (DECL_CXX_CONSTRUCTOR_P (m_items[i]->decl)
>> || ((ipa_node_params_sum == NULL
>> - || ipa_is_param_used (
>> - IPA_NODE_REF
>> - (dyn_cast <cgraph_node *>(m_items[i]->node)), 0))
>> + || IPA_NODE_REF (cnode)->descriptors.is_empty ()
>> + || ipa_is_param_used (IPA_NODE_REF (cnode), 0))
>> && static_cast<sem_function *> (m_items[i])
>> ->compare_polymorphic_p ())))
>> {
>> --
>> 2.1.4
>>
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2015-04-07 20:25 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-04-03 8:45 [PATCH] Fix PR ipa/65665 Martin Liška
2015-04-03 17:19 ` Jan Hubicka
2015-04-07 20:25 ` Martin Liška
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).