On 26/04/12 12:20, Richard Guenther wrote:
On Wed, Apr 25, 2012 at 11:56 PM, Tom de Vries tom_devr...@mentor.com wrote:
On 25/04/12 11:57, Richard Guenther wrote:
SNIP
Hmm. I'm not sure we can conclude that they have the same value!
+int bar (int);
+void baz (int);
+void bla (int);
+
Are you sure that tree-level unrollers are turned on at O2? My
impression was that they work only at O3 or with f[unroll,peel]-loops
flags.
On Tue, Apr 24, 2012 at 6:13 PM, Andi Kleen a...@firstfloor.org wrote:
tejohn...@google.com (Teresa Johnson) writes:
This patch adds heuristics to limit
On Fri, Apr 27, 2012 at 12:07 AM, Igor Zamyatin izamya...@gmail.com wrote:
Are you sure that tree-level unrollers are turned on at O2? My
impression was that they work only at O3 or with f[unroll,peel]-loops
flags.
yes they are on but only have effect on tiny loops with very small
trip count.
On Thu, Apr 26, 2012 at 7:48 PM, Steven Bosscher stevenb@gmail.com wrote:
Hello,
The attached patch re-organizes some code in tree-switch-conversion.c.
All information about a GIMPLE_SWITCH is now collected by one
function, so that my switch lowering code can use the same
On Thu, Apr 26, 2012 at 8:43 PM, Marc Glisse marc.gli...@inria.fr wrote:
On Thu, 26 Apr 2012, Richard Guenther wrote:
On Wed, Apr 25, 2012 at 3:58 PM, Marc Glisse marc.gli...@inria.fr wrote:
Here is take 2 on this patch, which seems cleaner. Bootstrapped and
regression tested.
On 25 April 2012 18:08, Manuel López-Ibáñez lopeziba...@gmail.com wrote:
On 25 April 2012 16:46, H.J. Lu hjl.to...@gmail.com wrote:
On Sat, Apr 21, 2012 at 4:58 AM, Manuel López-Ibáñez
lopeziba...@gmail.com wrote:
This patch improves missing initializers diagnostics. From:
pr36446.c:13:3:
On Fri, Apr 27, 2012 at 12:42:41PM +0800, Thomas Schwinge wrote:
GET_MODE_BITSIZE (lmode)« (8 bits). (With the current sources, lmode is
VOIDmode.)
Is emmitting »BIT_FIELD_REF *common, 32, 0 255« wrong in this case,
or should a later optimization pass be able to figure out that
On Fri, 27 Apr 2012, Richard Guenther wrote:
Do you have a copyright assignment on file?
Yes.
--
Marc Glisse
Ping.
On 10/04/12 14:00, Andrew Stubbs wrote:
Ping.
On 30/03/12 12:15, Andrew Stubbs wrote:
On 28/02/12 16:20, Andrew Stubbs wrote:
Hi all,
This patch implements 64-bit immediate constant loads in NEON.
The current state is that you can load const_vector, but not const_int.
This is clearly
On Thu, Apr 26, 2012 at 7:53 PM, Andrew MacLeod amacl...@redhat.com wrote:
On 04/05/2012 05:14 AM, Richard Guenther wrote:
Ok. Remember that you should use non-tree things if you can in GIMPLE
land. This probably means that both the size and the memmodel operands
should be
+ struct
On Thu, Apr 26, 2012 at 10:07 PM, Andrew MacLeod amacl...@redhat.com wrote:
On 04/05/2012 05:14 AM, Richard Guenther wrote:
+ static inline bool
+ gimple_atomic_has_fail_order (const_gimple gs)
+ {
+ return gimple_atomic_kind (gs) == GIMPLE_ATOMIC_COMPARE_EXCHANGE;
+ }
btw, these kind
On Fri, Apr 27, 2012 at 8:20 AM, Tom de Vries tom_devr...@mentor.com wrote:
On 26/04/12 12:20, Richard Guenther wrote:
On Wed, Apr 25, 2012 at 11:56 PM, Tom de Vries tom_devr...@mentor.com
wrote:
On 25/04/12 11:57, Richard Guenther wrote:
SNIP
Hmm. I'm not sure we can conclude that they
In the GCC FAQ under Dynamic linker is unable to find GCC libraries, one
suggestion is to add '-R' or '-rpath' linker option to the *link or *lib
specs so that the GCC libraries can be found.
E.G. the following line is added to the DRIVER_DEFINES when building gcc
via pkgsrc
Hello,
It makes no sense to mark DF solutions dirty on the gcc_unreachable()
path but not on the return path.
Bootstrappedtested on x86_64-unknown-linux-gnu and
powerpc64-unknown-linux-gnu. I'll this, as obvious, some time late
next week unless I hear objections.
Ciao!
Steven
* cfg.c
Il 27/04/2012 11:18, Steven Bosscher ha scritto:
Hello,
It makes no sense to mark DF solutions dirty on the gcc_unreachable()
path but not on the return path.
Bootstrappedtested on x86_64-unknown-linux-gnu and
powerpc64-unknown-linux-gnu. I'll this, as obvious, some time late
next week
If the value at zero is outside the range [0, GET_MODE_BITSIZE (mode)),
A != 0 and A == 0 can be compiled to clz/ctz followed by a subtraction
or one's complement (only for A != 0) and a shift. This trick can be
used effectively on PPC (though there are other problems in the machine
description
(conditional_replacement): Replace PHIs
whose arguments are -1 and 0, by negating the result of the
conditional.
2012-04-27 Paolo Bonzini bonz...@gnu.org
* gcc.c-torture/execute/20120427-2.c: New testcase.
* gcc.dg/tree-ssa/phi-opt-10.c: New testcase.
Index: tree-ssa
On Fri, Apr 27, 2012 at 11:49 AM, Paolo Bonzini bonz...@gnu.org wrote:
The testcase is miscompiled to this:
...
movl (%rsi), %ecx
cmpl %ecx, (%rdi)
sbbl %edx, %edx
cmovbe %edx, %eax
ret
but sbbl only preserves the carry flag, not the
To followup myself - bootstrap with just the 2nd patch is still
broken:
/abuild/rguenther/obj2/./gcc/xgcc -B/abuild/rguenther/obj2/./gcc/
-B/usr/local/x86_64-unknown-linux-gnu/bin/
-B/usr/local/x86_64-unknown-linux-gnu/lib/ -isystem
/usr/local/x86_64-unknown-linux-gnu/include -isystem
Ah, and all ACATS fails and
-FAIL: gnat.dg/loop_optimization3.adb (test for excess errors)
-FAIL: gnat.dg/loop_optimization3.adb execution test
-FAIL: gnat.dg/test_8bitlong_overflow.adb (test for excess errors)
-FAIL: gnat.dg/test_8bitlong_overflow.adb execution test
are fixed by for
Index: config/rs6000/rs6000.md
===
--- config/rs6000/rs6000.md (revisione 186859)
+++ config/rs6000/rs6000.md (copia locale)
@@ -2129,7 +2129,7 @@
(define_expand abssi2
[(set (match_operand:SI 0 gpc_reg_operand )
Le 06/04/2012 17:27, Pascal Obry a écrit :
Back on this! It turn out that this breaks the shared Ada runtime.
Indeed, exported variables on Ada packages in a DLLs are only accessible
when linking against DLL (thanks to runtime pseudo reloc).
With the patch applied it is not possible to
-04-27 Paolo Bonzini bonz...@gnu.org
* tree-ssa-phiopt.c (conditional_replacement): Replace PHIs
whose arguments are -1 and 0, by negating the result of the
conditional.
2012-04-27 Paolo Bonzini bonz...@gnu.org
* gcc.c-torture/execute/20120427-2.c: New
We do not depend on the block structure any more when dealing with
stack layout for variables in GCC 4.7.0 and above. I am not saying
your patch is incorrect or not needed. Just it will not have an
effect on variable stack layout.
It might be worth backporting to the 4.6 branch though,
http://gcc.gnu.org/viewcvs?view=revisionrevision=186899
Applied as approved in
http://gcc.gnu.org/ml/gcc/2012-04/msg00843.html
Johann
PR c/51527
* convert.c (convert_to_integer): Avoid infinite recursion for
target-defined built-in types.
Index: convert.c
On Fri, Apr 27, 2012 at 12:50 PM, Eric Botcazou ebotca...@adacore.com wrote:
We do not depend on the block structure any more when dealing with
stack layout for variables in GCC 4.7.0 and above. I am not saying
your patch is incorrect or not needed. Just it will not have an
effect on
On Fri, 27 Apr 2012, Paolo Bonzini wrote:
If the value at zero is outside the range [0, GET_MODE_BITSIZE (mode)),
A != 0 and A == 0 can be compiled to clz/ctz followed by a subtraction
or one's complement (only for A != 0) and a shift. This trick can be
used effectively on PPC (though there
This removes is_hidden_global_store in favor of two functions
with more clear semantics.
Bootstrapped and tested on x86_64-unknown-linux-gnu.
Richard.
2012-04-27 Richard Guenther rguent...@suse.de
* tree-flow.h (is_hidden_global_store): Remove.
* tree-ssa-sink.c
Hi,
I'd like to drag some attention to this bug again, it is the only ICE
when LTO building Firefox with debug info and the problem also happens
with the 4.7 so it would be nice to have this fixed for 4.7.1.
On Mon, Mar 12, 2012 at 11:51:05AM +0100, Richard Guenther wrote:
On Thu, Mar 8, 2012
Hi,
On Tue, Apr 24, 2012 at 12:31:38PM +0200, Martin Jambor wrote:
Hi,
On Mon, Apr 23, 2012 at 03:30:19PM +0200, Richard Guenther wrote:
On Mon, 23 Apr 2012, Martin Jambor wrote:
Hi,
On Mon, Apr 23, 2012 at 12:50:51PM +0200, Richard Guenther wrote:
On Fri, 20 Apr 2012,
On Fri, Apr 27, 2012 at 12:50 PM, Eric Botcazou ebotca...@adacore.com wrote:
We do not depend on the block structure any more when dealing with
stack layout for variables in GCC 4.7.0 and above. I am not saying
your patch is incorrect or not needed. Just it will not have an
effect on
On 04/27/2012 04:37 AM, Richard Guenther wrote:
Since integral atomics are always of an unsigned type , I could switch over
and use 'unsigned size' instead of 'tree fntype' for them (I will rename
it), but then things may be more complicated when dealing with generic
atomics... those can be
On Wed, Apr 25, 2012 at 6:41 PM, Richard Guenther
richard.guent...@gmail.com wrote:
On Wed, Apr 25, 2012 at 4:32 PM, Igor Zamyatin izamya...@gmail.com wrote:
On Wed, Apr 25, 2012 at 1:14 PM, Richard Guenther
richard.guent...@gmail.com wrote:
On Wed, Apr 25, 2012 at 10:56 AM, Igor Zamyatin
Il 27/04/2012 13:16, Richard Guenther ha scritto:
In optabs.c we compare the CTZ_DEFINED_VALUE_AT_ZERO against two,
is != 0 really what you want here? The docs suggest to me
that as you are using the optab below you should compare against two, too.
Interesting, first time I hear about this...
Hello guys,
After conversation in IRC with Richard, I've slightly updated the patch.
1. According to Richards suggestion I moved PTA_HLE to `generic` march.
2. Applied and updated Andi's patch (see [1]).
3. Updated tests to use proper memory model combintations
4. Added 1-sentense description to
tzcnt is encoded as rep;bsf and unlike lzcnt is a drop-in replacement
if we don't care about the flags (it has the same semantics for non-zero
values).
Since bsf is usually slower, just emit tzcnt unconditionally. However,
write it as rep;bsf unless -mbmi is in use, to cater for old assemblers.
Hi,
The DWARF spec says (since version 4) that DW_AT_high_pc can be
represented by a constant form.
If the value of the DW_AT_high_pc is of class address, it is the
relocated address of the first location past the last
instruction associated with the entity; if it is of
On Thu, Apr 26, 2012 at 10:08 PM, Sriraman Tallam tmsri...@google.com wrote:
Hi,
I have made the following changes in this new patch which is attached:
* Use target attribute itself to create function versions.
* Handle any number of ISA names and arch= args to target attribute,
On 26 April 2012 12:12, Dodji Seketeli do...@seketeli.org wrote:
Manuel López-Ibáñez lopeziba...@gmail.com a écrit:
On 21 April 2012 14:56, Jason Merrill ja...@redhat.com wrote:
It seems like we'll do this for every line in the header, which could lead
to a lot of leaked memory. Instead, we
On Wed, Apr 25, 2012 at 6:23 AM, Ulrich Weigand uweig...@de.ibm.com wrote:
Andreas Krebbel wrote:
2011-11-17 Andreas Krebbel andreas.kreb...@de.ibm.com
* reload.c (find_reloads): Change the loop nesting when trying an
alternative with swapped operands.
This is OK.
This
Jakub == Jakub Jelinek ja...@redhat.com writes:
Jakub On Thu, Apr 26, 2012 at 01:52:31PM -0400, DJ Delorie wrote:
I will not oppose adding more unrelated stuff to libiberty, but
neither will I approve it. I will let one of the other maintainers or
a global maintainer approve it.
Jakub The
On Thu, Apr 26, 2012 at 8:35 AM, Tom Tromey tro...@redhat.com wrote:
This is ok with this change.
Thanks. Updated and submitted to trunk.
Ollie
On Fri, Apr 27, 2012 at 6:38 AM, H.J. Lu hjl.to...@gmail.com wrote:
On Thu, Apr 26, 2012 at 10:08 PM, Sriraman Tallam tmsri...@google.com wrote:
Hi,
I have made the following changes in this new patch which is attached:
* Use target attribute itself to create function versions.
* Handle
On Fri, Apr 27, 2012 at 7:35 AM, Sriraman Tallam tmsri...@google.com wrote:
On Fri, Apr 27, 2012 at 6:38 AM, H.J. Lu hjl.to...@gmail.com wrote:
On Thu, Apr 26, 2012 at 10:08 PM, Sriraman Tallam tmsri...@google.com
wrote:
Hi,
I have made the following changes in this new patch which is
On Fri, Apr 27, 2012 at 7:38 AM, H.J. Lu hjl.to...@gmail.com wrote:
On Fri, Apr 27, 2012 at 7:35 AM, Sriraman Tallam tmsri...@google.com wrote:
On Fri, Apr 27, 2012 at 6:38 AM, H.J. Lu hjl.to...@gmail.com wrote:
On Thu, Apr 26, 2012 at 10:08 PM, Sriraman Tallam tmsri...@google.com
wrote:
Hi,
Gabriel Dos Reis g...@integrable-solutions.net writes:
On Wed, Apr 25, 2012 at 10:20 AM, Dodji Seketeli do...@redhat.com wrote:
Gabriel Dos Reis g...@integrable-solutions.net writes:
On Wed, Apr 25, 2012 at 9:04 AM, Dodji Seketeli do...@redhat.com wrote:
In gcc/testsuite/gcc.dg/pr30457.c,
Hello Gabriel,
Following your request[1], please find below the implementation for the
-Wvarargs option, as well as its introductory text. It applies on top
the changes to enable -ftrack-macro-expansion by default[2].
[1]: http://gcc.gnu.org/ml/gcc-patches/2012-04/msg01604.html
[2]:
On 30/03/12 12:15, Andrew Stubbs wrote:
On 28/02/12 16:20, Andrew Stubbs wrote:
Hi all,
This patch implements 64-bit immediate constant loads in NEON.
The current state is that you can load const_vector, but not const_int.
This is clearly not ideal. The result is a constant pool entry when
Manuel López-Ibáñez lopeziba...@gmail.com a écrit:
Another drawback I didn't realize until now is that in this way the
canonicalize every path, instead of only touching those that belong to
system headers.
Ah. Good catch.
I guess file-dir-sysp should tell us if we are in a system directory,
On Fri, Apr 27, 2012 at 7:53 AM, Sriraman Tallam tmsri...@google.com wrote:
On Fri, Apr 27, 2012 at 7:38 AM, H.J. Lu hjl.to...@gmail.com wrote:
On Fri, Apr 27, 2012 at 7:35 AM, Sriraman Tallam tmsri...@google.com wrote:
On Fri, Apr 27, 2012 at 6:38 AM, H.J. Lu hjl.to...@gmail.com wrote:
On
On Fri, Apr 27, 2012 at 8:36 AM, H.J. Lu hjl.to...@gmail.com wrote:
On Fri, Apr 27, 2012 at 7:53 AM, Sriraman Tallam tmsri...@google.com wrote:
On Fri, Apr 27, 2012 at 7:38 AM, H.J. Lu hjl.to...@gmail.com wrote:
On Fri, Apr 27, 2012 at 7:35 AM, Sriraman Tallam tmsri...@google.com
wrote:
On
, by negating the result of the
conditional.
2012-04-27 Paolo Bonzini bonz...@gnu.org
* gcc.c-torture/execute/20120427-2.c: New testcase.
* gcc.dg/tree-ssa/phi-opt-10.c: New testcase.
Index: tree-ssa-phiopt.c
On 04/27/12 05:49, Kirill Yukhin wrote:
+ if (targetm.memmodel_check)
+val = targetm.memmodel_check (val);
+ else if (val ~MEMMODEL_MASK)
+
+{
Incorrect vertical whitespace.
+ if ( (failure MEMMODEL_MASK) == MEMMODEL_RELEASE
+ || (failure MEMMODEL_MASK) ==
On Fri, Apr 27, 2012 at 7:11 AM, Tom Tromey tro...@redhat.com wrote:
Jakub == Jakub Jelinek ja...@redhat.com writes:
Jakub On Thu, Apr 26, 2012 at 01:52:31PM -0400, DJ Delorie wrote:
I will not oppose adding more unrelated stuff to libiberty, but
neither will I approve it. I will let one of
On Fri, Apr 27, 2012 at 9:01 AM, H.J. Lu hjl.to...@gmail.com wrote:
On Fri, Apr 27, 2012 at 7:11 AM, Tom Tromey tro...@redhat.com wrote:
Jakub == Jakub Jelinek ja...@redhat.com writes:
Jakub On Thu, Apr 26, 2012 at 01:52:31PM -0400, DJ Delorie wrote:
I will not oppose adding more unrelated
HJ You should add extern C for C++ on those functions moved to
HJ libiberty.
Yeah, sorry about that.
I'm testing the fix.
Tom
Bernd Schmidt wrote:
We're creating new pseudos, and while we're resizing some data
structures, we aren't doing it for everything.
@@ -3983,7 +3983,8 @@ find_moveable_pseudos (void)
last_moveable_pseudo = max_reg_num ();
- fix_reg_equiv_init();
+ fix_reg_equiv_init ();
+
This patch to libgo corrects the implementation of syscall.Setenv for
systems that do not have the setenv library call, but only have putenv.
Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu and
i386-sun-solaris2.9. Committed to mainline and 4.7 branch.
Ian
diff -r 1ed95295e00b
The Solaris 9 ldexp function has a bug: ldexp(-1, -1075) returns
positive zero when it should return negative zero. This patch works
around this bug in the Go math package. This fixes part of PR 52358.
Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu and
i386-sun-solaris2.9.
This patch to libgo provides the long double math functions if they are
not in the system libm. This is needed on i386 because the Go frontend
automatically converts calls to these functions from float64 (aka C
double) to C long double, as guided by express_precision_type. This is
done so that
Tom == Tom Tromey tro...@redhat.com writes:
HJ You should add extern C for C++ on those functions moved to
HJ libiberty.
Tom Yeah, sorry about that.
Tom I'm testing the fix.
Here is what I am checking in.
Tom
ChangeLog:
2012-04-27 Tom Tromey tro...@redhat.com
* dwarf2.h: Wrap
On Fri, Apr 27, 2012 at 5:06 AM, Jan Hubicka hubi...@ucw.cz wrote:
On Fri, Apr 27, 2012 at 12:50 PM, Eric Botcazou ebotca...@adacore.com
wrote:
We do not depend on the block structure any more when dealing with
stack layout for variables in GCC 4.7.0 and above. I am not saying
your patch
On Fri, 2012-04-27 at 15:43 +0200, Jakub Jelinek wrote:
On Fri, Apr 27, 2012 at 03:36:56PM +0200, Mark Wielaard wrote:
But even without this, I think the patch is worth it just to get rid of
all the relocations necessary otherwise.
IMHO we should defer applying this by a few months, given
This tree is also available from
git://repo.or.cz/gcc/rth.git rth/libatomic
Changes since v1:
* I believe I've addressed all of Torvald's feedback especially
re barrier problems. I have not changed the lock hash function,
as there were no concrete suggestions, and certainly that
This caused:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53144
Looks like a PPRE bug.
Paolo
The attached change fixes a problem compiling the linux dynamic loader
for the PA target.
Putting function labels (plabels) in the constant pool results
in a section flag conflict compiling rtld.c. Other targets don't
do this, and testing indicates that it is not necessary.
Tested on
On 04/27/2012 07:56 AM, Martin Jambor wrote:
PR lto/53138
* dwarf2out.c (dwarf2out_decl): Only lookup die representing context
of a variable when the contect is a function.
OK.
Jason
We can perform a single-copy atomic load with an ldrexd insn.
If the load is all we care about, we need not pair this with
a strexd.
Ok?
r~
* config/arm/arm.md (UNSPEC_LL): New.
* config/arm/sync.md (atomic_loaddi, atomic_loaddi_1): New.
diff --git a/gcc/config/arm/arm.md
Dodji Seketeli do...@redhat.com writes:
Tested on x86_64-unknown-linux-gnu against trunk. Bootstrap is still
running ...
* builtins.c (fold_builtin_next_arg): Unwinds to the first
location in real source code.
---
gcc/builtins.c | 23 +++
1 files
Dodji Seketeli do...@redhat.com writes:
Tested on x86_64-unknown-linux-gnu against trunk. Bootstrap for all
languages is still underway.
gcc/c-family/
* c.opt (Wvarargs): Define new option.
gcc/
builtins.c (fold_builtin_next_arg): Use OPT_Wvarargs as an
argument
This patch to the Go frontend changes the representation of array/slice
literals to use less memory when the literal uses index keys.
Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu.
Committed to mainline and 4.7 branch.
Ian
diff -r ebdbe2ad3ef6 go/expressions.cc
---
Hi,
I'm having a look to this PR filed by Daniel, which is again about
SFINAE for combined delete / new expressions, eg a simple example could
be (Daniel provided tens)
templateclass T, class = decltype(::delete ::new T())
auto g(int) - char;
templateclass
auto g(...) - char()[2];
On 27/04/2012, at 1:17 AM, Richard Guenther wrote:
On Wed, Apr 25, 2012 at 8:06 AM, Maxim Kuvyrkov ma...@codesourcery.com
wrote:
...
+ppre_n_insert_for_speed_p (pre_expr expr, basic_block block,
+ unsigned int inserts_needed)
+{
+ /* The more expensive EXPR is,
On 04/27/2012 09:42 PM, Paolo Carlini wrote:
In particular about the exact meaning of the FIXME? More generally, is
the issue here matter of compile-time optimization? Like we want to save
work when we actually *know* the type even in template context? (then
looks like
This patch to the Go frontend fixes a few cases where the compiler was
crashing on invalid code. Bootstrapped and ran Go testsuite on
x86_64-unknown-linux-gnu. Committed to mainline and 4.7 branch.
Ian
diff -r cc46c1bb0278 go/expressions.cc
--- a/go/expressions.cc Fri Apr 27 17:28:13 2012
75 matches
Mail list logo