On Tue, Aug 21, 2012 at 03:08:43PM -0700, Mike Stump wrote:
> On Aug 11, 2012, at 10:39 AM, Senthil Kumar Selvaraj wrote:
> > This patch allows cflags set in board config files using
> > "set_board_info cflags" to be used in the selectors of
> > dg-skip-if and other dejagnu commands that use the c
Hi Sébastien,
After applying your patch, I get an assertion failure in
emit_return_into_block, at function.c:5600 when trying to compile
anything (line is "gcc_assert (ANY_RETURN_P (pat));").
Any ideas?
Doh - I should have checked further. The problem is the return_internal
pattern. emit_r
Hello,
This adapts SH's logical right shift patterns to work/look the same way
as left shift patterns. It mainly fixes a few issues with dynamic shift
insn selection.
Tested on rev 190580 with
make -k check RUNTESTFLAGS="--target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a/-mb,-m4/-ml,-m4/-mb,-m4a/-ml,-m
On Tue, Aug 21, 2012 at 9:26 PM, Steven Bosscher wrote:
> Hello,
>
> Two more bitmap obstacks, this time in tree-ssa-coalesce.c.
>
> The advantage isn't so much in having the bitmaps on the non-default
> obstack, but more in that the bitmaps can be free'ed all at once by
> simply releasing the obs
On Tue, Aug 21, 2012 at 10:20 PM, Sandra Loosemore
wrote:
> This patch is a followup to the addition of support for
> -fstrict-volatile-bitfields (required by the ARM EABI); see this thread
>
> http://gcc.gnu.org/ml/gcc-patches/2010-10/msg01889.html
>
> for discussion of the original patch.
>
> Th
On Tue, Aug 21, 2012 at 10:33 PM, Jakub Jelinek wrote:
> Hi!
>
> The recent change in find_assert_locations from XCNEWVEC to XNEWVEC
> caused a valgrind warning, because bb_rpo[ENTRY_BLOCK] used to
> be accessed, but was never initialized.
>
> Fixed by ignoring edges from ENTRY_BLOCK altogether.
>
On Tue, Aug 21, 2012 at 10:33 PM, Jakub Jelinek wrote:
> Hi!
>
> The recent change in find_assert_locations from XCNEWVEC to XNEWVEC
> caused a valgrind warning, because bb_rpo[ENTRY_BLOCK] used to
> be accessed, but was never initialized.
Sorry for this breakage. I looked at each of the changes
Hello,
On 22/08/2012 07:56, Tobias Burnus wrote:
> Committed as Rev. 190586 after successful regtesting.
>
> That's the version I also had attached to
> http://gcc.gnu.org/ml/fortran/2012-08/msg00118.html; as written there:
I have one minor comment about it. See below.
>
> "The patch is incomple
Hi!
Another leak and another uninitialized var.
Testing in progress, ok for trunk if it passes?
2012-08-22 Jakub Jelinek
* tree-vect-loop.c (vect_transform_loop): Initialize
check_profitability to false.
* tree-predcom.c (try_combine_chains): Free the worklist vector
Hello,
This both speeds up and improves RTL alias analysis by propagating the
alias chains information in a visit in topological order of the CFG.
Perhaps unsurprisingly, this is motivated again by PR
middle-end/54146, where I noticed that the loop in init_alias_analysis
terminated because the it
On Wed, Aug 22, 2012 at 11:38 AM, Jakub Jelinek wrote:
> Hi!
>
> Another leak and another uninitialized var.
> Testing in progress, ok for trunk if it passes?
Ok.
Thanks,
Richard.
> 2012-08-22 Jakub Jelinek
>
> * tree-vect-loop.c (vect_transform_loop): Initialize
> check_prof
On 22/08/12 02:20, Terry Guo wrote:
> Hi,
>
> Due to the impact of ARM UAL, the Thumb1 and Thumb2 mode use LSRS
> instruction while the ARM mode uses MOVS instruction. So the following case
> is updated accordingly. Is it OK to trunk?
>
> BR,
> Terry
>
> 2012-08-21 Terry Guo
>
> * gc
On Wed, Aug 22, 2012 at 11:53 AM, Steven Bosscher wrote:
> Hello,
>
> This both speeds up and improves RTL alias analysis by propagating the
> alias chains information in a visit in topological order of the CFG.
>
> Perhaps unsurprisingly, this is motivated again by PR
> middle-end/54146, where I
> >
> > Due to the impact of ARM UAL, the Thumb1 and Thumb2 mode use LSRS
> > instruction while the ARM mode uses MOVS instruction. So the
> following case
> > is updated accordingly. Is it OK to trunk?
> >
> > BR,
> > Terry
> >
> > 2012-08-21 Terry Guo
> >
> > * gcc.target/arm/combine-
Hello,
when I adapted VRP PLUS_EXPR handling for __int128, I missed one place
where double_int can overflow. Note that I have no idea if that helps for
bug 54317, but that's where I noticed the issue.
2012-08-21 Marc Glisse
PR tree-optimization/54317
gcc/
* tree-vrp.c (ex
This finally limits the number of alias queries we do when looking
for redundant loads/stores from inside SCCVN (thus FRE and PRE).
Previously the number was bound roughly by O(n_loads * n_stores) while
now it is bound by O(n_loads) with the constant factor
--param sccvn-max-alias-queries-per-acc
On Wed, Aug 22, 2012 at 1:55 PM, Marc Glisse wrote:
> Hello,
>
> when I adapted VRP PLUS_EXPR handling for __int128, I missed one place where
> double_int can overflow. Note that I have no idea if that helps for bug
> 54317, but that's where I noticed the issue.
Ok.
Thanks,
Richard.
> 2012-08-2
On Wed, Aug 22, 2012 at 02:19:19PM +0200, Richard Guenther wrote:
> > 2012-08-21 Marc Glisse
> >
> > PR tree-optimization/54317
> >
> > gcc/
> > * tree-vrp.c (extract_range_from_binary_expr_1): Test for
> > double_int overflow.
> > Remove dead tests.
> >
> > gcc/t
On Wed, 22 Aug 2012, Jakub Jelinek wrote:
2012-08-21 Marc Glisse
+ n <<= (8 * sizeof (NT) - 1);
Better use __CHAR_BIT__ instead of 8 here...
Ok, I committed the __CHAR_BIT__ version (I just compiled that one file
manually with old and new compilers, I didn't restart the testsuite, I
ho
On Wed, Aug 22, 2012 at 02:34:01PM +0200, Marc Glisse wrote:
> On Wed, 22 Aug 2012, Jakub Jelinek wrote:
>
> >>>2012-08-21 Marc Glisse
> >>>+ n <<= (8 * sizeof (NT) - 1);
> >
> >Better use __CHAR_BIT__ instead of 8 here...
>
> Ok, I committed the __CHAR_BIT__ version (I just compiled that one
Hi,
yesterday I had a look to this old PR and noticed that we are almost
doing already the right thing for the original testcase: we are for
classes, but we ICE (something recent) for enums. The latter issue seems
easy to fix: handle specially enums at the beginning of
supplement_binding_1 on
On Wed, 22 Aug 2012, Jakub Jelinek wrote:
On Wed, Aug 22, 2012 at 02:34:01PM +0200, Marc Glisse wrote:
Ok, I committed the __CHAR_BIT__ version (I just compiled that one
file manually with old and new compilers, I didn't restart the
testsuite, I hope the manual test is enough to detect any typo
Richard Guenther writes:
> Doesn't that belong in system.h instead? And removed from rtl.h?
Fine with me. The following patch does just that, moving the #undefs
as far down in system.h as seemed reasonable.
Bootstrapped without regressions on i386-pc-solaris2.10 and
x86_64-unknown-linux-gnu;
While we should already be in loop-closed SSA form for virtual
operands most of the time (because we have a virtual use at
the return statement) and loop-closed SSA form for virtuals
is cheap (we only have a single virtual operand now) the following
makes sure that a loop-closed PHI node for virtu
Hi,
On Tue, Aug 21, 2012 at 01:30:47PM +0200, Richard Guenther wrote:
> On Tue, Aug 21, 2012 at 1:27 PM, Martin Jambor wrote:
> > On Wed, Aug 15, 2012 at 05:21:04PM +0200, Martin Jambor wrote:
> >> Hi,
> >>
> >> On Fri, Aug 10, 2012 at 04:57:41PM +0200, Eric Botcazou wrote:
> >> > > - ada/gcc-int
Hi!
init_function_start is these days supposed to be called only at the start of
RTL expansion, it shouldn't be called much earlier, because then we leak
e.g. the memory allocated by init_emit.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
2012-08-22 Jakub Jelinek
For PR46590 we spend a lot of time doing incremental SSA update.
The following makes sure that time isn't increased by overeager
asserts.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
2012-08-22 Richard Guenther
* tree-ssa-loop-ch.c (copy_loop_headers): Remo
On 08/22/2012 03:03 PM, Jakub Jelinek wrote:
init_function_start is these days supposed to be called only at the start of
RTL expansion, it shouldn't be called much earlier, because then we leak
e.g. the memory allocated by init_emit.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for
On Wed, Aug 22, 2012 at 3:03 PM, Jakub Jelinek wrote:
> Hi!
>
> init_function_start is these days supposed to be called only at the start of
> RTL expansion, it shouldn't be called much earlier, because then we leak
> e.g. the memory allocated by init_emit.
>
> Bootstrapped/regtested on x86_64-lin
On Wed, Aug 22, 2012 at 3:04 PM, Martin Jambor wrote:
> Hi,
>
> On Tue, Aug 21, 2012 at 01:30:47PM +0200, Richard Guenther wrote:
>> On Tue, Aug 21, 2012 at 1:27 PM, Martin Jambor wrote:
>> > On Wed, Aug 15, 2012 at 05:21:04PM +0200, Martin Jambor wrote:
>> >> Hi,
>> >>
>> >> On Fri, Aug 10, 2012
On Wed, Aug 22, 2012 at 12:57 AM, Steven Bosscher wrote:
> Hello,
>
> This patch back-ports most of the changes made to resolve the worst of
> PR54146 on trunk to GCC 4.7. This PR is basically an accumulation of
> various compiler speed and memory usage regressions, and all fixes are
> almost triv
The discard member function of the mersenne_twister_engine class is
unnecessarily inefficient. It currently discard elements one-by-one.
It is possible to discard with higher granularity by discarding the
entire internal buffer. The attached patch implements this. To avoid
duplication a new in
On Wed, Aug 22, 2012 at 03:41:08PM +0200, Richard Guenther wrote:
> On Wed, Aug 22, 2012 at 12:57 AM, Steven Bosscher
> wrote:
> > Hello,
> >
> > This patch back-ports most of the changes made to resolve the worst of
> > PR54146 on trunk to GCC 4.7. This PR is basically an accumulation of
> > var
Hello,
While working on something else, I noticed that debug_tree (vec), when
vec is a TREE_VEC, was crashing because TREE_NOTHROW was asserting that
its argument is not a TREE_VEC, so print_node would crash.
It turned out that TREE_NOTHROW was accidentally modified by this
change set:
commit 87
On 08/22/2012 08:43 AM, Paolo Carlini wrote:
yesterday I had a look to this old PR and noticed that we are almost
doing already the right thing for the original testcase: we are for
classes, but we ICE (something recent) for enums. The latter issue seems
easy to fix: handle specially enums at the
On 08/22/2012 03:43 PM, Ulrich Drepper wrote:
The discard member function of the mersenne_twister_engine class is
unnecessarily inefficient. It currently discard elements one-by-one.
It is possible to discard with higher granularity by discarding the
entire internal buffer. The attached patc
Hi,
On 08/22/2012 04:09 PM, Jason Merrill wrote:
On 08/22/2012 08:43 AM, Paolo Carlini wrote:
yesterday I had a look to this old PR and noticed that we are almost
doing already the right thing for the original testcase: we are for
classes, but we ICE (something recent) for enums. The latter iss
On 22/08/12 12:16, Terry Guo wrote:
>
>>>
>>> Due to the impact of ARM UAL, the Thumb1 and Thumb2 mode use LSRS
>>> instruction while the ARM mode uses MOVS instruction. So the
>> following case
>>> is updated accordingly. Is it OK to trunk?
>>>
>>> BR,
>>> Terry
>>>
>>> 2012-08-21 Terry Guo
>>
Hi.
On 08/22/2012 04:14 PM, Paolo Carlini wrote:
Hi,
On 08/22/2012 04:09 PM, Jason Merrill wrote:
On 08/22/2012 08:43 AM, Paolo Carlini wrote:
yesterday I had a look to this old PR and noticed that we are almost
doing already the right thing for the original testcase: we are for
classes, but
On 08/19/2012 11:22 AM, Richard Sandiford wrote:
Not sure whether a peephole is the right choice here. In practice,
I'd imagine these opportunities would only come from a DImode move of
$0 into a doubleword register, so we could simply emit the pattern in
mips_split_doubleword_move.
That would
Hi,
Long double is the same as double in Bionic. This patch
1. Add -mlong-double-80:
80-bit long double
Enabled for Linux by default.
2. Add -mlong-double-64:
64-bit long double
Predefine a new C/C++ macro, __LONG_DOUBLE_64__.
Enabled for Android by defaul
. thus, in short, what is happening is that, for this testcase:
class B
{
protected:
enum E { E1, E2, E3 };
};
class D : private B
{
public:
using B::E;
private:
enum E { };
};
we parse the new declaration enum E { }; and we reach
supplement_binding_1 before setting the underlying type
Inside the vectorizer they are not necessary (stmt update will make
sure they are marked for renaming - sth I'm going to improve with
a follwup). Also BB vectorization doesn't need to update SSA form
after vectorizing each BB.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richar
On 08/22/2012 10:55 AM, Paolo Carlini wrote:
. thus, in short, what is happening is that, for this testcase:
class B
{
protected:
enum E { E1, E2, E3 };
};
class D : private B
{
public:
using B::E;
private:
enum E { };
};
we parse the new declaration enum E { }; and we reach
suppleme
Hi again,
On 08/22/2012 05:13 PM, Jason Merrill wrote:
On 08/22/2012 10:55 AM, Paolo Carlini wrote:
. thus, in short, what is happening is that, for this testcase:
class B
{
protected:
enum E { E1, E2, E3 };
};
class D : private B
{
public:
using B::E;
private:
enum E { };
};
we pa
Dear all,
first, a question to Mikael (and others knowing the scalarizer): How to
properly fix the following:
implicit none
REAL qss(3)
REAL, ALLOCATABLE :: qj(:,:)
INTEGER :: qcount
qss(:)=qj(:,qcount)
end
For that one calls gfc_cleanup_loop (&loop) - and in gfc_free_ss:
cas
On Wed, Aug 22, 2012 at 3:44 PM, Jakub Jelinek wrote:
> To me the patch looks way too long, changing too many things, to be suitable
> for the release branch.
Most of the changes are only to allocate bitmaps on dedicated
obstacks. It makes the patch long, but the changes are not really
significan
On Wed, Aug 22, 2012 at 3:01 PM, Richard Guenther wrote:
>
> While we should already be in loop-closed SSA form for virtual
> operands most of the time (because we have a virtual use at
> the return statement) and loop-closed SSA form for virtuals
> is cheap (we only have a single virtual operand
On Wed, Aug 22, 2012 at 4:41 PM, H.J. Lu wrote:
> Long double is the same as double in Bionic. This patch
>
> 1. Add -mlong-double-80:
> 80-bit long double
> Enabled for Linux by default.
> 2. Add -mlong-double-64:
> 64-bit long double
> Predefine a new C/C++ macr
On Tue, Aug 21, 2012 at 5:05 PM, H.J. Lu wrote:
> long double may not be 80-bit on i386. We can't use
> REAL_VALUE_TO_TARGET_LONG_DOUBLE for XFmode. This patch replaces
> REAL_VALUE_TO_TARGET_LONG_DOUBLE with real_to_target. OK to install?
>
> Thanks.
>
> H.J.
> ---
> 2012-08-21 H.J. Lu
>
>
On Tue, Aug 21, 2012 at 1:34 AM, Oleg Endo wrote:
> On Sat, 2012-08-18 at 10:23 +0200, Uros Bizjak wrote:
>> On Sat, Aug 18, 2012 at 10:14 AM, Uros Bizjak wrote:
>>
>> > After discussion with Oleg, it looks that it is enough to prevent
>> > wrong registers in the output of the (multi-output) insn
On 22/08/2012 19:19, Tobias Burnus wrote:
> Dear all,
>
> first, a question to Mikael (and others knowing the scalarizer): How to
> properly fix the following:
>
> implicit none
> REAL qss(3)
> REAL, ALLOCATABLE :: qj(:,:)
> INTEGER :: qcount
> qss(:)=qj(:,qcount)
> end
>
> For that
Hi,
-dm hasn't worked for a long time, at least dating back to GCC 3.4.
This patch removes -dm and puts back -da, which was removed by accident.
OK to install?
Thanks.
H.J.
---
2012-08-22 H.J. Lu
PR driver/54335
* doc/invoke.texi: Add -da and remove -dm.
diff --git a/gcc/do
Am 22.08.2012 11:18, schrieb Tobias Burnus:
Regarding -Wcompare-real, I wonder whether it makes sense to either
ignore comparisions against zero or to put them into a different flag
(-Wcompare-real-zero);
Here is a patch to not warn for comparisons against zero.
Regression-tested, also tested
Hi.
I've been having to make this small change to the 'configure' script
when building on sparc-sun-solaris2.10 to accomodate the shell executing
the script. Without the change, I get an error message like so:
configure: test: unknown operator ==
With the double equals being replaced by a singl
> + bool packedp = false;
> +
> + if (TREE_CODE(to) == COMPONENT_REF
> + && (TYPE_PACKED (TREE_TYPE (TREE_OPERAND (to, 0)))
> + || (TREE_CODE (TREE_OPERAND (to, 1)) == FIELD_DECL
> + && DECL_PACKED (TREE_OPERAND (to
Oleg Endo wrote:
> This adapts SH's logical right shift patterns to work/look the same way
> as left shift patterns. It mainly fixes a few issues with dynamic shift
> insn selection.
> Tested on rev 190580 with
> make -k check RUNTESTFLAGS="--target_board=sh-sim
> \{-m2/-ml,-m2/-mb,-m2a/-mb,-m4/-
Hi again,
On 08/22/2012 05:13 PM, Jason Merrill wrote:
On 08/22/2012 10:55 AM, Paolo Carlini wrote:
. thus, in short, what is happening is that, for this testcase:
class B
{
protected:
enum E { E1, E2, E3 };
};
class D : private B
{
public:
using B::E;
private:
enum E { };
};
we pa
Hello Bill,
This patch plugs a leak in rs6000.c:rs6000_density_test(). You have to
free the array that get_loop_body returns. Noticed while going over
all uses of get_loop_body (it's a common mistake to leak the return
array).
Patch is completely untested because I don't know when/how this
functi
Hi Richard,
You never responded to this. Is there something wrong with this fix?
Can you address whether it's sufficient for align_loops > align_labels
and such cases that Julian Brown raised?
A patch against the current trunk is below.
Thanks,
Roland
On Wed, Aug 1, 2012 at 2:43 PM, Roland M
On Thu, 2012-08-23 at 00:53 +0200, Steven Bosscher wrote:
> Hello Bill,
>
> This patch plugs a leak in rs6000.c:rs6000_density_test(). You have to
> free the array that get_loop_body returns. Noticed while going over
> all uses of get_loop_body (it's a common mistake to leak the return
> array).
>
Hello Everyone,
I have ten patches which are approved or obvious but waiting on commit,
each of which is attached to this email. Feel free to consider this a
ping, HOWEVER, they are rebased onto the latest trunk so they're no
longer stale. Additionally, I updated the commit messages and with
This patch is for the google/gcc-4_7 branch.
If a function contains a local typedef of an anonymous structure, GCC
will generate a typedef DIE in the function, where the typedef DIE points
to a structure_type DIE at the top level. That structure_type DIE, if
it's a non-POD, can contain ctor and d
The Go frontend had some old code for special handling of unsafe.Pointer
for type assertions. That is not part of the current language, and
would actually break some otherwise valid programs. This patch removes
the special handling, and fixes one piece of gccgo-specific code in
libgo that used it
On 08/22/12 17:05, rbmj wrote:
Hello Everyone,
I have ten patches which are approved or obvious but waiting on commit
The include fixing stuff looks fine to me.
However I think it might be simpler to tweak mkfixinc.sh to
sed '/if test -s .{MACRO_LIST}/s/$/ && false/' \
${srcdir}/fixin
I hope this can get looked at, thanks.
http://gcc.gnu.org/ml/gcc-patches/2012-06/msg01208.html
On Wed, Aug 22, 2012 at 5:11 PM, Cary Coutant wrote:
> This patch is for the google/gcc-4_7 branch.
Approved for google/gcc-4_7 branch.
Thanks,
--
Paul Pluzhnikov
On Tue, Aug 21, 2012 at 11:18 PM, Jan Hubicka wrote:
>> On Tue, Aug 21, 2012 at 6:56 PM, Jan Hubicka wrote:
>> >> > I can go ahead with the histogram approach. There is some roundoff
>> >> > error from the working set scaling approach that can affect different
>> >> > merging orders as you note,
On 08/21/2012 02:23 PM, Richard Sandiford wrote:
Would be nice to add a compile test for -mabi=64 just to make sure
that Pmode == DImode works. A copy of an existing test like
code-readable-1.c would be fine.
I'm having problems with this part -- it seems like every combination of
options wi
On 8/22/2012 8:52 PM, Bruce Korb wrote:
On 08/22/12 17:05, rbmj wrote:
Hello Everyone,
I have ten patches which are approved or obvious but waiting on commit
The include fixing stuff looks fine to me.
However I think it might be simpler to tweak mkfixinc.sh to
sed '/if test -s .{MACRO_LIS
>
> I'm concerned about inaccuracies arising when multiple runs have
> different sizes and therefore counters are in different buckets in the
> corresponding histograms. We would end up in a situation where the
> merged histogram has more "counters" in it than there are actual
> counters in the pr
On Wed, Aug 22, 2012 at 10:47:34PM +0200, Thomas Koenig wrote:
> Am 22.08.2012 11:18, schrieb Tobias Burnus:
>
> >Regarding -Wcompare-real, I wonder whether it makes sense to either
> >ignore comparisions against zero or to put them into a different flag
> >(-Wcompare-real-zero);
>
> Here is a pa
On 08/15/2012 03:43 AM, Dodji Seketeli wrote:
Or we could just require people to put the attribute in the right
place (or one of the right places) if they want it to apply to the
decl. That is, either at the beginning of the declaration statement
or after the declarator-id.
Just to make sure I
On 08/22/2012 12:01 PM, Paolo Carlini wrote:
At this point, let me know, I could either add to the testcase a
templated variant like the above (see attached), or rework the code to
explicitly check the underlying type (I would add locals hosting the
TREE_TYPEs to shorten a bit things, etc).
The
On 08/22/2012 03:27 PM, Eric Botcazou wrote:
+ bool packedp = false;
+
+ if (TREE_CODE(to) == COMPONENT_REF
+&& (TYPE_PACKED (TREE_TYPE (TREE_OPERAND (to, 0)))
+ || (TREE_CODE (TREE_OPERAND (to, 1)) == FIELD_DECL
+&& DECL_PACKED (TREE_OPERAND (to, 1))
The Go language was changed slightly so that comparisons return an
untyped boolean value rather than the named type "bool". This patch
implements that change in gccgo. One of the tests in the testsuite
changed as well. Bootstrapped and ran Go testsuite on
x86_64-unknown-linux-gnu. Committed to
* PING *
On August 19, 2012, Tobias Burnus wrote:
Dear all,
attached is a slightly updated patch:
* Call finalizers of nonallocatable, nonpointer components
* Generate FINAL wrapper for abstract types which have a finalizer.
(The allocatable components are deallocated in the first type
(abst
Sandra Loosemore writes:
> 2012-08-22 Julian Brown
> Sandra Loosemore
>
> gcc/
> * config/mips/mips.md
> (UNSPEC_CASESI_DISPATCH): New.
> (MIPS16_T_REGNUM): New constant.
> (tablejump): Don't use for MIPS16_SHORT_JUMP_TABLES.
> (casesi): New.
>
78 matches
Mail list logo
