Vladimir Makarov writes:
> On 04/10/2012 09:35 AM, Richard Sandiford wrote:
>> Hi Vlad,
>>
>> Back in Decemember, when we were still very much in stage 3, I sent
>> an RFC about an alternative implementation of -fsched-pressure.
>> Just wanted to send a reminder now that we're in the proper stage:
On Tue, Apr 17, 2012 at 8:49 AM, Jan Hubicka wrote:
>> Hello,
>>
>> There is code in stmt.c since the initial checkin, that tries to
>> balance a switch tree according to some ascii heuristics. I see a
>> couple of problems with this code:
>>
>> 1. It doesn't seem to help much. With the attached p
> > Note that it would make a lot of sense to teach this heuristics predict.c
> > and properly identify chars.
>
> Indeed this would be the proper place to implement this logic.
TO a degree - switch expansion needs more info than it can obtain from edge
profile. Having
switch
case 1,3,5,7,8,9:
Il 17/04/2012 10:45, Richard Guenther ha scritto:
> > Also it is possble to get an historgrams from profile feedback into
> > switch expansion. I always wanted to do that once switch expansion code
> > is cleaned up and moved to gimple level...
>
> Indeed. At least the parts that expand switch st
On 16/04/12 16:23, Jason Merrill wrote:
> On 04/14/2012 05:43 PM, Tom de Vries wrote:
>> + tree expr = NULL;
>> + append_to_statement_list (*block,&expr);
>> + *block = expr;
Rather than doing this dance here, I think it would be better to enhance
appe
Hi,
On Thu, Apr 12, 2012 at 07:21:12PM +0200, Eric Botcazou wrote:
> > Well, the commit did not add a testcase and when I looked up the patch
> > in the mailing list archive
> > (http://gcc.gnu.org/ml/gcc-patches/2006-11/msg01449.html) it said it
> > was fixing problems not reproducible on trunk s
>>
>> It has nothing but defines for Android. It did not move any existing
>> code to this file.
>>
>
> Adding linux-common.h to i386 backend needs approval from
> i386 backend maintainer. If a patch also adds Android support,
> i386 backend maintainer may not feel comfortable to review it.
> How
On Tue, Apr 17, 2012 at 10:45 AM, Richard Guenther
wrote:
>> Also it is possble to get an historgrams from profile feedback into
>> switch expansion. I always wanted to do that once switch expansion code
>> is cleaned up and moved to gimple level...
>
> Indeed. At least the parts that expand swit
Hi,
in order to avoid this ICE on invalid, I think it makes sense to
explicitly check for try-block in massage_constexpr_body, since 7.1.5/4
explicitly rules out for constexpr constructors such function-body.
Tested x86_64-linux.
Thanks,
Paolo.
///
/cp
2012-04-17 Paolo
Loop bound computation uses undefined behavior when accessing arrays
outside of their domain. Unfortunately while it tries to honor
issues with trailing arrays in allocated storage its implementation
is broken (for one, it does consider a TYPE_DECL after the array
as a sign that the array is not
> On Tue, Apr 17, 2012 at 10:45 AM, Richard Guenther
> wrote:
> >> Also it is possble to get an historgrams from profile feedback into
> >> switch expansion. I always wanted to do that once switch expansion code
> >> is cleaned up and moved to gimple level...
> >
> > Indeed. At least the parts th
On Sat, Apr 14, 2012 at 9:26 AM, Tom de Vries wrote:
> On 27/01/12 21:37, Tom de Vries wrote:
>> On 24/01/12 11:40, Richard Guenther wrote:
>>> On Mon, Jan 23, 2012 at 10:27 PM, Tom de Vries
>>> wrote:
Richard,
Jakub,
the following patch fixes PR51879.
Consider the
Hi,
one-process methodology cannot be used on VMS because fork/pipe/dup2 aren't
fully supported. To avoid a build failure, it is therefore better to build
using two-process methodology.
But, when twoprocess is selected, gcc emits a warning due to a missing
specifier in printf. The patch fixe
On Tue, Apr 17, 2012 at 12:16 PM, Ilya Enkovich wrote:
>>>
>>> It has nothing but defines for Android. It did not move any existing
>>> code to this file.
>>>
>>
>> Adding linux-common.h to i386 backend needs approval from
>> i386 backend maintainer. If a patch also adds Android support,
>> i386
I think build_constexpr_constructor_member_initializers is a better
place for that check, since it's already looking at the tree structure.
Jason
On Tue, Apr 17, 2012 at 3:16 PM, Uros Bizjak wrote:
> On Tue, Apr 17, 2012 at 12:16 PM, Ilya Enkovich
> wrote:
It has nothing but defines for Android. It did not move any existing
code to this file.
>>>
>>> Adding linux-common.h to i386 backend needs approval from
>>> i386 ba
On Tue, 17 Apr 2012, Richard Guenther wrote:
>
> Loop bound computation uses undefined behavior when accessing arrays
> outside of their domain. Unfortunately while it tries to honor
> issues with trailing arrays in allocated storage its implementation
> is broken (for one, it does consider a TY
OK, thanks.
Jason
On Wed, 11 Apr 2012, Richard Guenther wrote:
> On Wed, Apr 11, 2012 at 4:16 PM, Bernd Schmidt
> wrote:
> > The order of calls to sched_rgn_init and sched_init differs between
> > sched-rgn and sel-sched. This caused a scheduler patch I was working on
> > to segfault once sel-sched was enabled.
This fixes PR53011 - EH cleanup needs to cater for loops now
(or avoid some transforms).
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
Index: gcc/tree-eh.c
===
*** gcc/tree-eh.c (revision 186
On 04/17/2012 03:33 PM, Alexander Monakov wrote:
> Bernd, could you elaborate on the segfault you had seen? Perhaps we could
> offer some advice on fixing it then.
It was only seen with another patch which modified the sched-rgn
initialization code.
> In the meanwhile, could you revert your patc
I have various thoughts:
It's odd that we still treat 'return' as starting a function body long
after we removed that extension.
Maybe we shouldn't look for a function body if we already have an
initializer and aren't dealing with a function declarator.
I guess we should set initializer_tok
This patch reverts the change originally done when adding
number-of-iteration analysis uses to VRP, to have a flag
to toggle whether to derive number of iterations from undefined
behavior. To be able to do so one error in VRP has to be fixed - we
have to check for the number of stmt executions,
On Sun, Apr 15, 2012 at 5:11 PM, Richard Sandiford
wrote:
> The comment in alias.c says:
>
> The contents of an ADDRESS is not normally used, the mode of the
> ADDRESS determines whether the ADDRESS is a function argument or some
> other special value. Pointer equality, not rtx_equal_p, det
Hi Tristan,
On Tue, Apr 17, 2012 at 5:57 AM, Tristan Gingold wrote:
> Hi,
>
> one-process methodology cannot be used on VMS[...]
> But, when twoprocess is selected, gcc emits a warning[...]
> Ok for trunk ?
> diff --git a/fixincludes/configure.ac b/fixincludes/configure.ac
> index e7de791..f1fb2
> On Tue, Apr 17, 2012 at 3:16 PM, Uros Bizjak wrote:
>> On Tue, Apr 17, 2012 at 12:16 PM, Ilya Enkovich
>> wrote:
>
> It has nothing but defines for Android. It did not move any existing
> code to this file.
>
Adding linux-common.h to i386 backend needs approval from
>
Hello,
I found the following problem while investigating SMS on x86-64.
When I run gcc with -march=nocona (on pentium-4 with EM64T extension), all
latencies in data dependency graph become zeros. The global pointer
"insn_default_latency" points to insn_default_latency_none, which
returns zero for
>
> Any other inputs?
>
I would suggest to use "snprintf" b/gcc/config/i386/i386-c.c to avoid
possible buffer overrun.
I also have a question regarding AS compatibility. In case one built
GCC using AS with support of HLE then using this GCC on a machine with
old AS will cause fail because of usup
> I also have a question regarding AS compatibility. In case one built
> GCC using AS with support of HLE then using this GCC on a machine with
> old AS will cause fail because of usupported prefix. Can we support it
I don't think that's a supported use case for gcc.
It also doesn't work with .cfi
On Tue, Apr 17, 2012 at 7:35 AM, Roman Zhuykov wrote:
> Hello,
>
> I found the following problem while investigating SMS on x86-64.
> When I run gcc with -march=nocona (on pentium-4 with EM64T extension), all
> latencies in data dependency graph become zeros. The global pointer
> "insn_default_lat
DImode references do not suffer a major performance hit for < 4-byte aligned
access like the float types.
Bootstrap/regtest on powerpc64-linux with no new regressions. Ok for trunk?
-Pat
2012-04-17 Pat Haugen
* config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Remove DImode.
Index:
On 04/13/12 18:22, Boehm, Hans wrote:
-Original Message-
From: Aldy Hernandez [mailto:al...@redhat.com]
Sent: Thursday, April 12, 2012 3:12 PM
To: Richard Guenther
Cc: Andrew MacLeod; Boehm, Hans; gcc-patches; Torvald Riegel
Subject: [PR tree-optimization/52558]: RFC: questions on stor
On Tue, 17 Apr 2012, H.J. Lu wrote:
> On Tue, Apr 17, 2012 at 7:35 AM, Roman Zhuykov wrote:
> > Hello,
> >
> > I found the following problem while investigating SMS on x86-64.
> > When I run gcc with -march=nocona (on pentium-4 with EM64T extension), all
> > latencies in data dependency graph b
This is the first in a series of patches cleaning up rs6000 prologue
and epilogue generating code. This one is just the formatting/style
changes plus renaming two variables to better reflect their usage,
and moving code around.
The patch series has been bootstrapped and regression tested
powerpc-
On Tue, Apr 17, 2012 at 10:55 AM, Pat Haugen
wrote:
> DImode references do not suffer a major performance hit for < 4-byte aligned
> access like the float types.
>
> Bootstrap/regtest on powerpc64-linux with no new regressions. Ok for trunk?
>
> 2012-04-17 Pat Haugen
>
> * config/rs6000/r
This fixes a lot of confusion in rs6000_frame_related call arguments.
At the time rs6000_frame_related first appeared, the prologue only
used sp_reg_rtx (r1) or frame_ptr_rtx (r12) as frame_reg_rtx to access
register save slots. If r12 was used, it was necessary to add a note
that gave the equival
This continues the prologue and epilogue cleanup. Not many user
visible changes here, except for:
- a bugfix to the LR save RTL emitted by rs6000_emit_savres_rtx which
may affect SPE,
- a bugfix for SPE code emitted when using a static chain,
- vector saves will be done using r1 for large frames
This provides some protection against misuse of r0, r11 and r12. I
found it useful when enabling out-of-line saves for large frames. ;-)
* config/rs6000/rs6000.c (START_USE, END_USE, NOT_INUSE): Define.
(rs6000_emit_prologue): Use the above to catch register overlap.
diff -urp g
This enables out-of-line save and restore for large frames, and for
ABI_AIX when using the static chain.
* config/rs6000/rs6000.c (rs6000_savres_strategy): Allow
out-of-line save/restore for large frames. Don't disable
out-of-line saves on ABI_AIX when using static chain r
On Thu, Apr 12, 2012 at 3:16 PM, Richard Guenther
wrote:
> On Thu, Apr 12, 2012 at 1:05 PM, Igor Zamyatin wrote:
>> On Wed, Apr 11, 2012 at 12:39 PM, Richard Guenther
>> wrote:
>>> On Tue, Apr 10, 2012 at 8:43 PM, Igor Zamyatin wrote:
Hi All!
Here is a patch that enables unroll a
The emergency reassociation patch for PR52976 disabled un-distribution
in the presence of repeated factors to avoid ICEs in zero_one_operation.
This patch fixes such cases properly by teaching zero_one_operation
about __builtin_pow* calls.
Bootstrapped with no new regressions on powerpc64-linux.
On Tue, Apr 17, 2012 at 8:04 AM, Alexander Monakov wrote:
>
>
> On Tue, 17 Apr 2012, H.J. Lu wrote:
>
>> On Tue, Apr 17, 2012 at 7:35 AM, Roman Zhuykov wrote:
>> > Hello,
>> >
>> > I found the following problem while investigating SMS on x86-64.
>> > When I run gcc with -march=nocona (on pentium-
Hi,
I think build_constexpr_constructor_member_initializers is a better
place for that check, since it's already looking at the tree structure.
Indeed. I'm finishing testing the below. Ok if it passes?
Thanks,
Paolo.
/cp
2012-04-17 Paolo Carlini
PR c++/5259
On 04/17/2012 03:55 PM, Jason Merrill wrote:
I have various thoughts:
It's odd that we still treat 'return' as starting a function body long
after we removed that extension.
Maybe we shouldn't look for a function body if we already have an
initializer and aren't dealing with a function decla
On 04/17/2012 05:35 PM, Paolo Carlini wrote:
Hi,
I think build_constexpr_constructor_member_initializers is a better
place for that check, since it's already looking at the tree structure.
Indeed. I'm finishing testing the below. Ok if it passes?
... uhm, actually like this seems more correct t
> My goal for GCC 4.8 is to do just that: Move switch expansion to
> GIMPLE and add value profiling for switch expressions.
And the idea is to put all that code in tree-switch-conversion.c. But
there are a few clean-ups I wish to do on that code before that.
First, there is a global pass info stru
Hello,
this patch adds vector subscripting to C++ by reusing the C code.
build_array_ref and cp_build_array_ref could probably share more, but I
don't understand them enough to do it.
(note that I can't commit, so if you like the patch...)
gcc/cp/ChangeLog
2012-04-17 Marc Glisse
On 04/17/2012 04:29 AM, Richard Sandiford wrote:
Vladimir Makarov writes:
On the other hand, I don't think that 1st insn scheduling will be ever
used for x86. And although the SPECFP2000 rate is the same on x86-64 I
saw that some SPECFP2000 tests benefit from your algorithm on x86-64
(one amaz
Found this bug in 4.4 branches, fixed with the following in later
branches:
http://gcc.gnu.org/ml/gcc-patches/2010-06/msg01616.html
But the test case is useful anyway.
tested x86/linux
-benjamin2012-04-16 Benjamin Kosnik
* testsuite/20_util/specialized_algorithms/uninitialized_copy/
80859
OK.
Jason
Hi Mike,
> On Apr 16, 2012, at 8:03 AM, Rainer Orth wrote:
>> I've long noticed that libstdc++.log (unlike libstdc++.sum) doesn't
>> contain log entries for tests run from abi.exp, but hadn't looked
>> closer, getting used to check libstdc++.log.sep instead which contained
>> everything I expected
Diego Novillo writes:
> On 4/16/12 7:32 AM, Rainer Orth wrote:
>
>> Btw., it occured to me that it might be useful to add an option to
>> locate out-of-tree manifests. I often have several source trees
>> (unmodified sources, ones with local patches) and would like to share
>> manifests between
On 04/17/2012 06:52 PM, Benjamin De Kosnik wrote:
Found this bug in 4.4 branches, fixed with the following in later
branches:
http://gcc.gnu.org/ml/gcc-patches/2010-06/msg01616.html
But the test case is useful anyway.
Definitely, thanks! The name of the new testcase seems a bit weird (for
the F
> Definitely, thanks! The name of the new testcase seems a bit weird
> (for the FSF branches): shall we maybe refer to the original (Fedora
> or RHEL Bugzilla, I suppose) PR in a comment and then use either an
> explicative name (our current practice) or just a small number for
> the name of the t
Hello!
Correct name of atomic "or" named pattern is "atomic_orM", not "atomic_iorM".
Attached patch fixes this oversight.
2012-04-17 Uros Bizjak
PR target/53020
* config/i386/sync.md (atomic_): Rename to
atomic_.
Patch was bootstrapped and tested on x86_64-pc-linux-g
The emergency patch for PR52976 manipulated the operand rank system to
force inserted __builtin_powi calls to occur before uses of the call
results. However, this is generally the wrong approach, as it forces
other computations to move unnecessarily, and extends the lifetimes of
other operands.
T
Hello,
this patch expands __builtin_shuffle for V4DF mode in at most 3 insn. It
is simple and works really well, often generates only 2 insn. It is not
very generic, because other modes don't have an instruction equivalent to
vshufpd. For V8SF (and likely V4DI and V8SI with AVX2, but I still n
On Sun, Apr 15, 2012 at 8:11 AM, Richard Sandiford
wrote:
> The comment in alias.c says:
>
> The contents of an ADDRESS is not normally used, the mode of the
> ADDRESS determines whether the ADDRESS is a function argument or some
> other special value. Pointer equality, not rtx_equal_p, det
I have a patch to fix a compile time warning about an unused variable due
to the use being guarded by #ifndef __GCOV_KERNEL__.
Tested with bootstrap. Ok for google-main?
Teresa
2012-04-17 Teresa Johnson
Google ref b/5910724.
* libgcc/libgcov.c (gcov_cur_module_id): Guard def
On Tue, Apr 17, 2012 at 1:48 AM, Jan Hubicka wrote:
>> > Note that it would make a lot of sense to teach this heuristics predict.c
>> > and properly identify chars.
>>
>> Indeed this would be the proper place to implement this logic.
>
> TO a degree - switch expansion needs more info than it can o
ok.
David
On Tue, Apr 17, 2012 at 11:40 AM, Teresa Johnson wrote:
> I have a patch to fix a compile time warning about an unused variable due
> to the use being guarded by #ifndef __GCOV_KERNEL__.
>
> Tested with bootstrap. Ok for google-main?
>
> Teresa
>
> 2012-04-17 Teresa Johnson
>
>
On 18/04/2012, at 2:32 AM, Ilya Enkovich wrote:
>> On Tue, Apr 17, 2012 at 3:16 PM, Uros Bizjak wrote:
...
>>
>> The patch looks OK to me in the sense, that there is no difference for
>> x86 targets.
>>
>> So, OK for x86.
>>
>> Thanks,
>> Uros.
>
> Thanks, Uros!
>
> Maxim, could you please l
"H.J. Lu" writes:
> On Sun, Apr 15, 2012 at 8:11 AM, Richard Sandiford
> wrote:
>> The comment in alias.c says:
>>
>> The contents of an ADDRESS is not normally used, the mode of the
>> ADDRESS determines whether the ADDRESS is a function argument or some
>> other special value. Pointer eq
> On Tue, Apr 17, 2012 at 1:48 AM, Jan Hubicka wrote:
> >> > Note that it would make a lot of sense to teach this heuristics predict.c
> >> > and properly identify chars.
> >>
> >> Indeed this would be the proper place to implement this logic.
> >
> > TO a degree - switch expansion needs more info
Are these patches ready to go in? It looks like they were ack'd.
http://gcc.gnu.org/ml/gcc-patches/2011-10/msg01815.html
http://gcc.gnu.org/ml/gcc-patches/2011-10/msg01817.html
http://gcc.gnu.org/ml/gcc-patches/2011-10/msg01816.html
http://gcc.gnu.org/ml/gcc-patches/2011-10/msg01818.html
http://gc
This patch enables GET_MODE_WIDER_MODE for MODE_PARTIAL_INT (by setting
the wider mode to the one the partial mode is based on), which is useful
for the port I'm working on: I can avoid defining operations on the
partial modes. Also, convert_modes is changed so that unsignedp is taken
into account
Hello,
This is another step towards moving GIMPLE_SWITCH expansion to an
earlier point in the pipeline.
With the attached patch, some of the logic from stmt.c:add_case_node()
is moved to gimplify.c:gimplify_switch_expr(). This includes:
* Code to drop case labels that are out of range for the sw
On Wed, Apr 18, 2012 at 12:04 AM, Steven Bosscher wrote:
> Hello,
>
> This is another step towards moving GIMPLE_SWITCH expansion to an
> earlier point in the pipeline.
>
> With the attached patch, some of the logic from stmt.c:add_case_node()
> is moved to gimplify.c:gimplify_switch_expr(). This
Otherwise, it starts a new paragraph.Tested by inspecting the
resulting html. Committed as obvious.
Cheers,
Manuel.
Index: gcc/doc/invoke.texi
===
--- gcc/doc/invoke.texi (revision 186552)
+++ gcc/doc/invoke.texi (working copy)
@@ -
Hi Dodji,
I was going to commit this as obvious, but I want to make sure that it
doesn't conflict with your new track-macro-expansion patches. It can
also wait until you commit all your patches.
Cheers,
Manuel.
2012-04-18 Manuel López-Ibáñez
* tree-diagnostic.c (maybe_unwind_expande
Hello, gentle maintainer.
This is a message from the Translation Project robot.
A revised PO file for textual domain 'cpplib' has been submitted
by the Vietnamese team of translators. The file is available at:
http://translationproject.org/latest/cpplib/vi.po
(This file, 'cpplib-4.7.0.vi.p
cpplib-4.7.0.vi.po.gz
Description: Binary data
The Translation Project robot, in the
name of your translation coordinator.
On 5/04/2012, at 10:16 AM, Maxim Kuvyrkov wrote:
> Chao,
>
> Let's take discussion of MIPS changes to gcc-patches@. Please follow up here.
>
> --
> Maxim Kuvyrkov
> CodeSourcery / Mentor Graphics
>
> On 5/04/2012, at 10:10 AM, Fu, Chao-Ying wrote:
>
>> For now, two MIPS changes in gnu-user.h
Maxim Kuvyrkov wrote:
> >
> >> For now, two MIPS changes in gnu-user.h and
> unwind-dw2-fde-dip.c can be posted for comment.
> >> (I didn't tested this patch, though.)
>
> You need to test your patches before posting them for review.
> Below are a couple of comments on your current version.
On 18/04/2012, at 1:10 PM, Fu, Chao-Ying wrote:
> Maxim Kuvyrkov wrote:
>
>> Above definitions are OK.
>
> Thanks!
For avoidance of doubt, please wait for the whole patch to be approved before
committing it.
Index: gcc/libgcc/unwind-dw2-fde-dip.c
===
Hi Jason,
I have a new patch for this issue, another SFINAE issue noticed by
Daniel. Compared to the last version, I extended the complain-ization ;)
to a few more functions in typeck.c (I think the set is more consistent
now) and thoroughly double checked that the return values of all the
fu
Steven,
J"orn,
I am looking into fixing performance regression on EEMBC's bitmnp01, and a
version of your combined patch attached to PR38785 still works very well.
Would you mind me getting it through upstream review, or are there any issues
with contributing this patch to GCC mainline?
We (C
To be submitted to the google/integration branch and merged into
google/{main,gcc-4_6,gcc-4_7}.
Add -Xclang-only option (which is ignored).
This is used by certain drivers to pass options selectively to clang. Adding
support to the gcc driver makes it easier to test GCC in the absence of these
d
Fix the test case failed in ARM state.
* gcc.target/arm/thumb1-imm.c: Skip it in non-thumb1 target
Index: gcc/testsuite/gcc.target/arm/thumb1-imm.c
===
--- gcc/testsuite/gcc.target/arm/thumb1-imm.c (revision 186517)
+++ gcc
Hi,
this patch adds dump_symtab and dump_symtab_node for debugging symbol tables.
It also reorganized existing varpool and cgraph dumping code to use the same
format.
I decided to switch from identifier name to assembler names. It is useless to
see
all those ctor/dtor functions wihtout knowing wh
80 matches
Mail list logo
