> >
> > Comdat locals are now used by ipa-comdats, for thunks and for decloned
> > ctors.
> > We probably need to figure out bit more precise limitation of Solaris and
> > either
> > fix or add way for target to say what kind of comdat locals are not
> > supported.
>
> Right. I'll start reghun
Rainer Orth writes:
> Hi Honza,
>
>>> Unfortunately, AIX isn't the only target massively affected by your
>>> recent patches. This all started with r210597
>>>
>>> 2014-05-17 Jan Hubicka
>>>
>>> * tree-pass.h (make_pass_ipa_comdats): New pass.
>>> * timevar.def (TV_IPA_COMDATS):
Hi Honza,
>> Unfortunately, AIX isn't the only target massively affected by your
>> recent patches. This all started with r210597
>>
>> 2014-05-17 Jan Hubicka
>>
>> * tree-pass.h (make_pass_ipa_comdats): New pass.
>> * timevar.def (TV_IPA_COMDATS): New timevar.
>> * pass
>
> Unfortunately, AIX isn't the only target massively affected by your
> recent patches. This all started with r210597
>
> 2014-05-17 Jan Hubicka
>
> * tree-pass.h (make_pass_ipa_comdats): New pass.
> * timevar.def (TV_IPA_COMDATS): New timevar.
> * passes.def (pass_ip
Jan Hubicka writes:
>> Honza,
>>
>> I am not sure that the problem is caused only by aliases and thunks.
>> The large increase in AIX linker warnings about branches not followed
>> by nop also worry me.
>>
>> Your patch was about visibility. How does the more aggressive
>
> ipa-visibility is a
Deavid,
this is patch to disable the vtable rewriting on AIX. I did not make much
progress
on figuring out why it fails today - just observed that if I produce .o file
for the
test and link it against libstdc++ built by older GCC it works. Also objects
compiled
w/o optimization works, so I am b
> Honza,
>
> I am not sure that the problem is caused only by aliases and thunks.
> The large increase in AIX linker warnings about branches not followed
> by nop also worry me.
>
> Your patch was about visibility. How does the more aggressive
ipa-visibility is a pass that basically bring extern
On Tue, Jun 10, 2014 at 2:02 PM, Jan Hubicka wrote:
>> Honza,
>>
>> Thanks for this patch which improves some of the G++ testsuite
>> failures, but most of the libstdc++ testsuite continues to fail on
>> AIX.
>
> Yep, I am still looking into this. Just made new alias verifier that
> catches quite
> Honza,
>
> Thanks for this patch which improves some of the G++ testsuite
> failures, but most of the libstdc++ testsuite continues to fail on
> AIX.
Yep, I am still looking into this. Just made new alias verifier that
catches quite few nonsenses in how C++ builds thunks and same body
aliases.
On Sun, Jun 8, 2014 at 12:58 PM, Jan Hubicka wrote:
> Hi,
> this is the last part. It makes DECL_VIRTUAL to be copied when creating
> aliases. This is
> needed to make sanity check in gimple-fold happy (it checks that vtables are
> DECL_VIRTUAL).
> It also resets initializers of aliases to sav
On Sun, Jun 8, 2014 at 6:44 PM, Jan Hubicka wrote:
>> Honza,
>>
>> I finally was able to bootstrap GCC on AIX yesterday. The bootstrap
>> works, but the testsuite results are in extremely bad shape: 250
>> failures in G++ testsuite and 470 failures in libstdc++ testsuite.
>
> David, this is first
Hi,
this is the last part. It makes DECL_VIRTUAL to be copied when creating
aliases. This is
needed to make sanity check in gimple-fold happy (it checks that vtables are
DECL_VIRTUAL).
It also resets initializers of aliases to save memory.
Bootstrapped/regtested rs6000-aix, comitted.
Index: s
Hi,
this is third part - it fixes handling of aliases in the anchor sections
by recursing in place_block_symbol as discussed and also fixing one case
where we lookup attribute instead of using symbol table
* varasm.c (use_blocks_for_decl_p): Check symbol table instead of alias
attr
Hi,
this is second part of fix - it just makes ipa-visibility to work less hard when
initializer already contains local aliases. I noticed this when debugging the
original
issue.
Bootstrapped/regtested rs6000-aix, comitted.
Honza
* ipa-visibility.c: Include varasm.h
(can_replac
> Honza,
>
> I finally was able to bootstrap GCC on AIX yesterday. The bootstrap
> works, but the testsuite results are in extremely bad shape: 250
> failures in G++ testsuite and 470 failures in libstdc++ testsuite.
David, this is first of the AIX fixes.
It makes aliases to be output after thunk
On Fri, Jun 6, 2014 at 3:09 AM, Jan Hubicka wrote:
> sorry for taking time to get back to it. I went through the code and see no
> obvious flaws
> except that I am somewhat concerned what happens with local aliases into the
> anchors.
> The problem however does not seem to be caused by it. I
> Honza,
>
> How can we make further progress with the large regression on AIX?
David,
sorry for taking time to get back to it. I went through the code and see no
obvious flaws
except that I am somewhat concerned what happens with local aliases into the
anchors.
The problem however does not see
Honza,
How can we make further progress with the large regression on AIX?
Thanks, David
On Fri, May 30, 2014 at 1:24 PM, David Edelsohn wrote:
> Honza,
>
> For example g++.dg/abi/vcall1.C fails at a call in a "localalias"
> function, which jumps to a bad location:
>
>
> (gdb) up
> #1 0x14c
This leads to bootstrap errors like:
In file included from
/tmp/20140531/powerpc-ibm-aix7.1.0.0/libstdc++-v3/include/backward/strstream:54:0,
from
/nasfarm/edelsohn/src/src/libstdc++-v3/src/c++98/strstream.cc:44:
/tmp/20140531/powerpc-ibm-aix7.1.0.0/libstdc++-v3/include/istream: I
David Edelsohn writes:
> Honza,
>
> With the patch, I cannot bootstrap.
I think it was meant to be:
place_block_symbol (target);
rather than:
place_block_symbol (symbol);
I.e. we need to place "target" before copying its offset.
Thanks,
Richard
Honza,
With the patch, I cannot bootstrap.
I am concerned about the much greater frequency of AIX linker warnings
about branch not followed by no-op after your patch.
GCC may bootstrap only because it links libstdc++ statically. If your
patch is too aggressive in its effort to remove PLT calls,
> Honza,
>
> For example g++.dg/abi/vcall1.C fails at a call in a "localalias"
> function, which jumps to a bad location:
>
>
> (gdb) up
> #1 0x14c0 in B::B() [clone .localalias.2] ()
> (gdb) x/16i $pc-32
>0x14a0 <_ZN1BC2Ev+156>: add r10,r10,r8
>0x14a4 <_ZN1BC2Ev+160>:
Honza,
For example g++.dg/abi/vcall1.C fails at a call in a "localalias"
function, which jumps to a bad location:
(gdb) up
#1 0x14c0 in B::B() [clone .localalias.2] ()
(gdb) x/16i $pc-32
0x14a0 <_ZN1BC2Ev+156>: add r10,r10,r8
0x14a4 <_ZN1BC2Ev+160>: mr r3,r10
0x1
On Thu, May 29, 2014 at 3:17 PM, Yufeng Zhang wrote:
> Hi Honza,
>
> I can confirm that with your commit r211045 the arm-none-linux-gnueabi{hf}
> builds are OK now. Thanks for the fix.
Testsuite regressions like
FAIL: g++.dg/ipa/devirt-21.C -std=gnu++1y (test for excess errors)
Excess errors:
Honza,
I finally was able to bootstrap GCC on AIX yesterday. The bootstrap
works, but the testsuite results are in extremely bad shape: 250
failures in G++ testsuite and 470 failures in libstdc++ testsuite.
G++ previously had about 65 failures, mainly TLS and init. libstdc++
previously hovered ar
Jan Hubicka writes:
>> Jan Hubicka writes:
>> >> Richard Sandiford wrote the original section anchors implementation,
>> >> so he would be a good person to comment about the interaction between
>> >> aliases and section anchors.
>> >
>> > Thanks! Richard, does this patch seem sane?
>>
>> Looks g
> Jan Hubicka writes:
> >> Richard Sandiford wrote the original section anchors implementation,
> >> so he would be a good person to comment about the interaction between
> >> aliases and section anchors.
> >
> > Thanks! Richard, does this patch seem sane?
>
> Looks good to me in principle, but w
Hi Honza,
I can confirm that with your commit r211045 the
arm-none-linux-gnueabi{hf} builds are OK now. Thanks for the fix.
Yufeng
On 05/28/14 22:56, Jan Hubicka wrote:
Any update?
I've managed to generate a simple test case from
libstdc++-v3/src/c++98/strstream.cc which reproduces the iss
Jan Hubicka writes:
>> Richard Sandiford wrote the original section anchors implementation,
>> so he would be a good person to comment about the interaction between
>> aliases and section anchors.
>
> Thanks! Richard, does this patch seem sane?
Looks good to me in principle, but with:
> + s
> Richard Sandiford wrote the original section anchors implementation,
> so he would be a good person to comment about the interaction between
> aliases and section anchors.
Thanks! Richard, does this patch seem sane? AIX gcc now builds for me, so I
will test it on x86_64-linux it and commit if i
On Wed, May 28, 2014 at 6:31 PM, Jan Hubicka wrote:
>> Honza,
>>
>> I'm glad that you're making progress.
>>
>> > David, this looks like a bug in the AIX target output macros. I get:
>> > .set
>> > _ZTCSt14basic_ifstreamIcSt11char_traitsIcEE0_Si.localalias.69,_ZTCSt14basic_ifstreamIcSt11c
> Honza,
>
> I'm glad that you're making progress.
>
> > David, this looks like a bug in the AIX target output macros. I get:
> > .set
> > _ZTCSt14basic_ifstreamIcSt11char_traitsIcEE0_Si.localalias.69,_ZTCSt14basic_ifstreamIcSt11char_traitsIcEE0_Si
>
> > (this is correct since localalia
> Any update?
>
> I've managed to generate a simple test case from
> libstdc++-v3/src/c++98/strstream.cc which reproduces the issue on
> ARM that Ramana has reported previously:
>
>
> template struct char_traits;
>
> template
> class basic_ios
> {
> };
>
> template >
> class basic_istream : v
Honza,
I'm glad that you're making progress.
> David, this looks like a bug in the AIX target output macros. I get:
> .set
> _ZTCSt14basic_ifstreamIcSt11char_traitsIcEE0_Si.localalias.69,_ZTCSt14basic_ifstreamIcSt11char_traitsIcEE0_Si
> (this is correct since localalias is really an ali
> On 05/27/14 23:20, Jan Hubicka wrote:
> >>
> >>here we have decl and its local alias:
> >>(gdb) p debug_tree (sect->named.decl)
> >> >> type >> type >> 700f57e0>
> >> unsigned SI
> >> size
> >> unit size
> >> align 32 symtab 45 alias s
On 05/27/14 23:20, Jan Hubicka wrote:
here we have decl and its local alias:
(gdb) p debug_tree (sect->named.decl)
unsigned SI
size
unit size
align 32 symtab 45 alias set 3 canonical type 700f5840
pointer_to_this>
BLK
>
> here we have decl and its local alias:
> (gdb) p debug_tree (sect->named.decl)
> type type 700f57e0>
> unsigned SI
> size
> unit size
> align 32 symtab 45 alias set 3 canonical type 700f5840
> pointer_to_this >
>
>
> Sadly this doesn't appear to have fixed the issue.
I have now reproduced it on AIX. We die at:
/* Sanity check user variables for flag changes. */
if (sect->named.decl != NULL
&& DECL_P (sect->named.decl)
&& decl != sect->named.decl)
On 05/25/2014 07:54 AM, Jan Hubicka wrote:
Hi,
this patch adds code to rerite references in vtable initializers to local
aliases
when doing so is a win.
Bootstrapped/regtested x86_64-linux, comitted.
Honza
* ipa-visibility.c (can_replace_by_local_alias_in_vtable): New function.
On Mon, May 26, 2014 at 2:04 AM, Jan Hubicka wrote:
>> > On Sun, May 25, 2014 at 6:54 AM, Jan Hubicka wrote:
>> > > Hi,
>> > > this patch adds code to rerite references in vtable initializers to
>> > > local aliases
>> > > when doing so is a win.
>> > >
>> > > Bootstrapped/regtested x86_64-linux
> > On Sun, May 25, 2014 at 6:54 AM, Jan Hubicka wrote:
> > > Hi,
> > > this patch adds code to rerite references in vtable initializers to local
> > > aliases
> > > when doing so is a win.
> > >
> > > Bootstrapped/regtested x86_64-linux, comitted.
> >
> > This is the most likely patch to have c
> On Sun, May 25, 2014 at 6:54 AM, Jan Hubicka wrote:
> > Hi,
> > this patch adds code to rerite references in vtable initializers to local
> > aliases
> > when doing so is a win.
> >
> > Bootstrapped/regtested x86_64-linux, comitted.
>
> This is the most likely patch to have caused build failur
On Sun, May 25, 2014 at 6:54 AM, Jan Hubicka wrote:
> Hi,
> this patch adds code to rerite references in vtable initializers to local
> aliases
> when doing so is a win.
>
> Bootstrapped/regtested x86_64-linux, comitted.
This is the most likely patch to have caused build failures on
arm-linux-gn
Hi,
this patch adds code to rerite references in vtable initializers to local
aliases
when doing so is a win.
Bootstrapped/regtested x86_64-linux, comitted.
Honza
* ipa-visibility.c (can_replace_by_local_alias_in_vtable): New function.
(update_vtable_references): New function.
44 matches
Mail list logo
