This refactors the vectorizer cost model to call it when everything is
ready, avoiding some fixups. It also fixes cost compute for
SLP reductions.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
2015-05-28 Richard Biener
* tree-vectorizer.h (struct _slp_insta
Hi Guys,
I am applying the patch below to enhance the RX backend so that it
will push and pop multiple groups of registers using the PUSHM and
POPM instructions, thus reducing code size and increasing
performance.
Cheers
Nick
gcc/ChangeLog
2015-05-28 Nick Clifton
* config/r
On Wed, May 27, 2015 at 9:34 PM, Jason Merrill wrote:
> OK, I see the issue. We're calling debug_abstract_function to build debug
> info for the abstract instance of a function that we already built from
> dwarf2out_early_global_decl.
>
> It occurs to me that the early-dwarf work should make
> de
On Thu, May 28, 2015 at 7:11 AM, DJ Delorie wrote:
>
> Copied the way other tests get uint32_t. Ok?
Ok.
> * gcc.c-torture/execute/pr65369.c: Don't assume int is 32 bits.
>
> Index: gcc.c-torture/execute/pr65369.c
> ===
> --
On 27/05/15 20:53, Andreas Tobler wrote:
> Is this ok for trunk?
Excellent, thanks.
Andrew.
This patch changes the implementation of controlled types so that in simple
cases, they are just as efficient as noncontrolled types where initialization
and cleanup is done by hand.
Tested on x86_64-pc-linux-gnu, committed on trunk
2015-05-27 Bob Duff
* exp_ch3.adb (Build_Array_Init_
This patch fixes a spurious visibility error on a selected component in an
instance body, when the type of the prefix of the selected component is an
actual of the instance, and the desired component is inherited through
one or more derivations
The package derived.ads below must compile quietly:
Ontop of this patch you would have to
git grep -l -E "(cleanup-.*-dump|cleanup-saved-temps)" \
| egrep -v "(ChangeLog|/lib/)" \
| sed -e "s|[^/]*$||" | sort | uniq \
| while read d;
do
find $d -type f -exec \
sed -i -e "/cleanup-[^-]*[-]*dump/d;/cleanup-saved-temps/d" {} +
done
---
I
* configure.ac (TARGETLIB_PIC, TARGETLIB_NOASAN): New variables.
* configure: Regenerate.
* maint-tool: Refactor pic/ and noasan/ handling.
* Makefile.in: Likewise. Regenerate dependencies.
---
The below does the same but attempts to be limited to what POSIX
require
On 3 February 2015 at 01:07, Mike Stump wrote:
> On Feb 2, 2015, at 3:22 PM, Bernhard Reutner-Fischer
> wrote:
>> Untested draft patch
>
> I looked it over, seems to slot in nicely.
>
>
> + gfc_error ("% directive does not commence a loop at %C”);
>
> So, don’t like commence here.
Does an
Richard Biener writes:
> * gcc.dg/tree-ssa/ssa-fre-44.c: New testcase.
On ia64:
$ grep -c "return 3.0" ssa-fre-44.c.035t.fre1
3
$ tail ssa-fre-44.c.035t.fre1
_19 = _8->u.x;
_23 = _8->u.y;
f_24 = _19 + _23;
bar (&p);
p ={v} {CLOBBER};
return f_24;
}
Andreas.
--
Andreas S
Hi,
this patch removes a superfluous main function from a test-case.
Committed.
Thanks,
- Tom
Remove superfluous main in kernels-loop-n.c
2015-05-27 Tom de Vries
* c-c++-common/goacc/kernels-loop-n.c (main): Remove.
diff --git a/gcc/testsuite/c-c++-common/goacc/kernels-loop-n.c b/gcc/tes
Hi,
I've committed this patch.
It enables parallelization of the kernels-loops-n.c testcase.
This is now possible, due the commit of the fix for PR65637 to the
gomp-4_0-branch (
https://gcc.gnu.org/ml/gcc-patches/2015-05/msg01569.html ).
Thanks,
- Tom
Enable parallelization of kernels-loop-
Christophe Lyon wrote:
On 26 May 2015 at 18:25, Alan Lawrence wrote:
I don't see this symptom - I am able to execute such subsets with either my,
or Sandra's, advsimd-intrinsics.exp.
I didn't try to run with your patch, I thought it was an oversight of yours.
Sorry, indeed I've just checked
On Wed, May 20, 2015 at 02:58:09PM +0100, Ramana Radhakrishnan wrote:
> Hi,
>
> Someone privately pointed out that the ARM and AArch64 ports do not
> define TARGET_RELAXED_ORDERING given that the architecture(s) mandates a
> weak memory model. This patch fixes it for AArch64, the ARM patch
On Thu, May 21, 2015 at 09:54:19AM +0100, Ramana Radhakrishnan wrote:
> And here's an additional patch for the testsuite which was missed in the
> original posting.
>
> This is a testism that's testing code generation as per
> TARGET_RELAXED_ORDERING being false and therefore needs to be adjuste
On Wed, 27 May 2015, Kyrill Tkachov wrote:
> Hi Richard,
>
> On 26/05/15 14:54, Richard Biener wrote:
> > The following fixes the testcase in PR66142
> >
> > Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
> >
> > Richard.
> >
> > 2015-05-26 Richard Biener
> >
> >
On Wed, 2015-05-27 at 15:56 +0200, mliska wrote:
> gcc/ChangeLog:
>
> 2015-04-30 Martin Liska
>
> * ira-color.c (init_update_cost_records): Use new type-based pool
> allocator.
> (get_update_cost_record): Likewise.
> (free_update_cost_record_list): Likewise.
> (finish_
Hi All,
Here is a simple patch which removes restriction on outer-loop
vectorization - allow references in inner-loop with zero step. This
case was found in one important benchmark.
Bootstrap and regression testing did not show any new failures.
Is it OK for trunk.
ChangeLog:
2015-05-28 Yuri R
Hi!
On Fri, 15 May 2015 16:38:40 +0200, Martin Liška wrote:
> Following patch attempts to rewrite memory reports for GCC's internal
> allocations
> [...]
(Got commtited to trunk in r223748.)
> * hash-map-traits.h: New file.
In that one you added a copyright/licensing header, but...
>
Hi!
On Wed, 20 May 2015 14:30:38 -0600, Jeff Law wrote:
> On 05/20/2015 11:04 AM, Andreas Tobler wrote:
> > the attached patch enables some PIE tests on FreeBSD.
> Wouldn't it be better to remove the target selector and instead add:
>
> /* { dg-require-effective-target pie } */
>
> In each of
For partitioned loops, we're currently calling library functions (in
libgcc) to determine the cardinality of the set of threads a particular
loop is distributed over (given a set of gang/worker/vector toggles),
and the index of the current thread within that set.
This patch reimplements those two
On 28 May 2015 at 12:22, Alan Lawrence wrote:
> Christophe Lyon wrote:
>>
>> On 26 May 2015 at 18:25, Alan Lawrence wrote:
>>>
>>> I don't see this symptom - I am able to execute such subsets with either
>>> my,
>>> or Sandra's, advsimd-intrinsics.exp.
>>
>>
>> I didn't try to run with your patch
On Thu, May 28, 2015 at 1:00 PM, Yuri Rumyantsev wrote:
> Hi All,
>
> Here is a simple patch which removes restriction on outer-loop
> vectorization - allow references in inner-loop with zero step. This
> case was found in one important benchmark.
>
> Bootstrap and regression testing did not show
On Thu, May 21, 2015 at 4:10 AM, Jakub Jelinek wrote:
> On Thu, May 21, 2015 at 01:02:12PM +0200, Thomas Schwinge wrote:
>> Hi!
>>
>> On Thu, 7 May 2015 19:32:26 +0100, Julian Brown
>> wrote:
>> > Here's a new version of the patch [...]
>>
>> > OK for trunk?
>>
>> Makes sense to me (with just a
Hello,
>> #define BACKTRACE_SUPPORTS_THREADS @BACKTRACE_SUPPORTS_THREADS@
>> +
>> +/* BACKTRACE_SUPPORTS_DATA will be #defined'd as 1 if the backtrace library
>> + also handles data symbols, 0 if not. */
>> +
>> +#define BACKTRACE_SUPPORTS_DATA @BACKTRACE_SUPPORTS_DATA@
>
> End users are expec
Unsurprisingly ubsan doesn't like referencing a null pointer.
With this change __array_traits::_S_ref is only used to access an
element, which is invalid for std::array anyway.
Tested powerpc64le-linux, committed to trunk.
commit 0d999cf16b8f6a0d9bbf4bfe96b29e7b73a259e4
Author: Jonathan Wakely
> On 27 May 2015, at 15:36, Jeff Law wrote:
>> +static int
>> +coff_is_symbol (const b_coff_internal_symbol *isym)
>> +{
>> + return isym->type == 0x20 && isym->sec > 0;
>> +}
> You probably want const or enum so that you can have a symbolic name rather
> than 0x20 here. It also seems like th
On Fri, May 22, 2015 at 5:42 PM, Alan Lawrence wrote:
> This example which I wrote to test ifconversion, currently fails to
> if-convert or vectorize:
>
> int foo ()
> {
> for (int i = 0; i < 32 ; i++)
> {
> int m = (a[i] & i) ? 5 : 4;
> b[i] = a[i] * m;
> }
> }
>
> ...becaus
> On 28 May 2015, at 02:26, Ian Lance Taylor wrote:
> The #include will break cross-compilers. It's not OK for
> trunk until that is fixed.
I am confused by this comment, for two reasons:
- I don’t see how that would break cross-compilers. Cross compilers
hosted on windows are not impacted
On May 28, 2015, at 2:02 AM, Bernhard Reutner-Fischer
wrote:
>
> Does anybody have a better suggestion?
>
> directive not at the start of a loop at %C
> directive not followed by a loop at %C
I prefer either of these. I have a slight preference for the first.
> Mike, did you tweak the one or
On 28 May 2015 at 13:32, Christophe Lyon wrote:
> On 28 May 2015 at 12:22, Alan Lawrence wrote:
>> Christophe Lyon wrote:
>>>
>>> On 26 May 2015 at 18:25, Alan Lawrence wrote:
I don't see this symptom - I am able to execute such subsets with either
my,
or Sandra's, advsimd-in
This PR points out that we weren't able to optimize 1 << x == 2 to just
x == 1. This is my attempt to fix that: if we see (CST1 << A) == CST2
and CST2 is a multiple of CST1, use log2 to get rid of the shift, but
only if the result of the shift is a natural number (including zero).
If CST2 is not
On 05/28/2015 01:15 PM, Thomas Schwinge wrote:
Hi!
On Fri, 15 May 2015 16:38:40 +0200, Martin Liška wrote:
Following patch attempts to rewrite memory reports for GCC's internal
allocations
[...]
(Got commtited to trunk in r223748.)
* hash-map-traits.h: New file.
In that one you
On May 20, 2015, at 2:57 PM, Lawrence Velázquez wrote:
> 2015-05-15 Lawrence Velázquez
>
> PR target/63810
> * gcc/config/darwin-c.c (version_components): New global enum.
> (parse_version, version_as_legacy_macro)
> (version_as_modern_macro, macosx_version_as_macro): N
On 13/05/15 14:36 +0100, Jonathan Wakely wrote:
http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4383.html#2455
Voted into the WP in Lenexa.
We already did the right thing for vector, so only basic_string needs
to change.
Tested powerpc64le-linux, committed to trunk.
Also committed t
On 28/05/15 12:53 +0100, Jonathan Wakely wrote:
Unsurprisingly ubsan doesn't like referencing a null pointer.
With this change __array_traits::_S_ref is only used to access an
element, which is invalid for std::array anyway.
Tested powerpc64le-linux, committed to trunk.
And gcc-5-branch.
co
On 01/05/15 16:23 +0100, Jonathan Wakely wrote:
The name of this alias template is a copy&paste error, it should be
_Alloc. Tested powerpc64le-linux, committed to trunk.
Also committed to gcc-5-branch.
commit a772309ec9fc300e57edd750fa32b8320d68004a
Author: Jonathan Wakely
Date: Fri May 1
On Thu, May 28, 2015 at 02:15:45PM +0200, Marek Polacek wrote:
> This PR points out that we weren't able to optimize 1 << x == 2 to just
> x == 1. This is my attempt to fix that: if we see (CST1 << A) == CST2
> and CST2 is a multiple of CST1, use log2 to get rid of the shift, but
> only if the res
On Thu, May 28, 2015 at 2:15 PM, Marek Polacek wrote:
> This PR points out that we weren't able to optimize 1 << x == 2 to just
> x == 1. This is my attempt to fix that: if we see (CST1 << A) == CST2
> and CST2 is a multiple of CST1, use log2 to get rid of the shift, but
> only if the result of t
On 05/27/2015 07:44 PM, Jeff Law wrote:
On 05/27/2015 07:56 AM, mliska wrote:
Hello.
Following patch set attempts to replace old-style pool allocator
to a type-based one. Moreover, as we utilize classes and structs that are used
just by a pool allocator, these types have overwritten ctors and
This patch avoids the use of the secondary stack, and the corresponding cleanup
handlers, in many cases. For example, access discriminants no longer force
functions to return on the secondary stack. This is a speed improvement.
It is particularly relevant to the Ada.Containers.
Tested on x86_64-pc
Hi Matthew,
> > +
> > +/* Implement TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS. */
> > +
> > +static reg_class_t
> > +mips_ira_change_pseudo_allocno_class (int regno, reg_class_t
> > +allocno_class) {
> > + if (FLOAT_MODE_P (PSEUDO_REGNO_MODE (regno)) || allocno_class !=
> > ALL_REGS)
> > +retur
On Thu, 28 May 2015 04:48:58 -0700
"H.J. Lu" wrote:
> On Thu, May 21, 2015 at 4:10 AM, Jakub Jelinek
> wrote:
> > On Thu, May 21, 2015 at 01:02:12PM +0200, Thomas Schwinge wrote:
> >> Hi!
> >>
> >> On Thu, 7 May 2015 19:32:26 +0100, Julian Brown
> >> wrote:
> >> > Here's a new version of the pa
This patch avoids the use of the secondary stack, and the corresponding cleanup
handlers, in many cases. For example, access discriminants no longer force
functions to return on the secondary stack. This is a speed improvement.
It is particularly relevant to the Ada.Containers.
Tested on x86_64-pc
On Thu, 28 May 2015, Jonathan Wakely wrote:
Unsurprisingly ubsan doesn't like referencing a null pointer.
With this change __array_traits::_S_ref is only used to access an
element, which is invalid for std::array anyway.
Should
return *static_cast<_Tp*>(nullptr);
be replaced with
__builtin
The following patch fixes loop vectorization of SLP reduction chains
that involve patterns. The issue here is that pattern recog runs
after reduction detection and this screws things up. Re-ordering
this created interesting side-effects so I didn't explore this
further (for now) but instead fix
Hi,
here is updated version of patch. It makes alias_set_subset_of to be symmetric
for
ptr_type_node and other pointer type and moves the logic of creating subsets
to get_alias_set.
I tested that perlbmk works when built at -O3 x86_64
Bootstrapped/regtested x86_64-linux, OK?
Honza
*
On 28/05/15 14:38 +0100, Jonathan Wakely wrote:
On 28/05/15 15:26 +0200, Marc Glisse wrote:
On Thu, 28 May 2015, Jonathan Wakely wrote:
Unsurprisingly ubsan doesn't like referencing a null pointer.
With this change __array_traits::_S_ref is only used to access an
element, which is invalid for
On 28/05/15 15:26 +0200, Marc Glisse wrote:
On Thu, 28 May 2015, Jonathan Wakely wrote:
Unsurprisingly ubsan doesn't like referencing a null pointer.
With this change __array_traits::_S_ref is only used to access an
element, which is invalid for std::array anyway.
Should
return *static_cast
Christophe Lyon wrote:
So in fact, except for the comment about '-w' it seems you initial
patch was mostly OK, right?
Well, my removing a bunch of that c-torture-init stuff, was what was causing the
"-Og -g" variant to go missing, but apart from that, yes.
--Alan
OK to install?
H.J.
---
Index: gcc-6/changes.html
===
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-6/changes.html,v
retrieving revision 1.10
diff -u -p -r1.10 changes.html
--- gcc-6/changes.html 26 May 2015 10:12:08 - 1.10
+++ gcc-
On Thu, 28 May 2015, Jonathan Wakely wrote:
Would there be a danger of an object compiled with gcc-5.1 that calls
array::data() finding the _S_ref from an object compiled with
gcc-5.2 and hitting the __builtin_unreachable in vali code?
At -O0, maybe. To be safe you would need to give this _S_r
On Thu, 28 May 2015, Jan Hubicka wrote:
> Hi,
> here is updated version of patch. It makes alias_set_subset_of to be
> symmetric for
> ptr_type_node and other pointer type and moves the logic of creating subsets
> to get_alias_set.
>
> I tested that perlbmk works when built at -O3 x86_64
>
>
On Wed, May 27, 2015 at 03:49:24AM +0100, weixiangyu wrote:
> Hi,
Hi,
> The first patch backports the fix of PR
> target/64011(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64011) to the 4.9
> branch from trunk r219717,
I can't approve this patch to be backported, so please do not commit it
witho
Hello!
This patch avoids unnecessary clearing of 4th argument for futex_wake
syscall for 32bit targets.
2015-05-28 Uros Bizjak
* config/linux/x86/futex.h (sys_futex0) [!__x86_64__]: Remove function.
(futex_wait) [!__x86_64__]: Use __asm operand instead of sys_futex0.
(futex_wake)
For NVPTX, it is vitally important that the divergence of threads
within a warp can be controlled: in particular we must be able to
generate code that we know "reconverges" at a particular point.
Unfortunately GCC's middle-end optimisers can cause this property to
be violated, which causes problems
On 27 May 2015 14:20, John David Anglin wrote:
> On 2015-05-27 1:50 PM, Mike Frysinger wrote:
> > since i'm not looped into gcc development normally, which branches are those
> > currently ? naively reading gcc.gnu.org homepage makes me think none since
> > they're labled "regression fixes" and af
> > +alias_set_entry
> > +init_alias_set_entry (alias_set_type set)
> > +{
> > + alias_set_entry ase = ggc_cleared_alloc ();
>
> no need to use cleared_alloc if you also init ->is_pointer to false.
OK, will update the patch.
>
> > + ase->alias_set = set;
> > + ase->children
> > += hash_map
On 28/05/15 15:52 +0200, Marc Glisse wrote:
On Thu, 28 May 2015, Jonathan Wakely wrote:
Would there be a danger of an object compiled with gcc-5.1 that calls
array::data() finding the _S_ref from an object compiled with
gcc-5.2 and hitting the __builtin_unreachable in vali code?
At -O0, maybe
On Thu, May 28, 2015 at 03:06:35PM +0100, Julian Brown wrote:
> For NVPTX, it is vitally important that the divergence of threads
> within a warp can be controlled: in particular we must be able to
> generate code that we know "reconverges" at a particular point.
> Unfortunately GCC's middle-end op
On Thu, May 28, 2015 at 4:06 PM, Julian Brown wrote:
> For NVPTX, it is vitally important that the divergence of threads
> within a warp can be controlled: in particular we must be able to
> generate code that we know "reconverges" at a particular point.
> Unfortunately GCC's middle-end optimisers
On 28/05/15 12:53 +0100, Jonathan Wakely wrote:
Unsurprisingly ubsan doesn't like referencing a null pointer.
With this change __array_traits::_S_ref is only used to access an
element, which is invalid for std::array anyway.
Tested powerpc64le-linux, committed to trunk.
I forgot the debug and
Hi!
On Thu, 28 May 2015 16:20:11 +0200, Jakub Jelinek wrote:
> On Thu, May 28, 2015 at 03:06:35PM +0100, Julian Brown wrote:
> > [...]
> I think the lowering of this already at ompexp time is premature
Yes, we're aware of this "wart". :-|
> I think much better would be to have a function attr
Le 27/05/2015 23:09, Steve Kargl a écrit :
> On Wed, May 27, 2015 at 06:24:25PM +0200, Mikael Morin wrote:
>> Le 27/05/2015 16:07, Andre Vehreschild a ?crit :
>>> Hi Paul, hi Mikael,
>>>
>>> about renaming the identifier emitted: I would like to keep it short.
>>> Remember,
>>> there is always a n
On Thu, 28 May 2015 16:58:44 +0200
Mikael Morin wrote:
> Le 27/05/2015 23:09, Steve Kargl a écrit :
> > On Wed, May 27, 2015 at 06:24:25PM +0200, Mikael Morin wrote:
> >> Le 27/05/2015 16:07, Andre Vehreschild a ?crit :
> >>> Hi Paul, hi Mikael,
> >>>
> >>> about renaming the identifier emitted:
On Thu, May 28, 2015 at 04:49:43PM +0200, Thomas Schwinge wrote:
> > I think much better would be to have a function attribute (or cgraph
> > flag) that would be set for functions you want to compile this way
> > (plus a targetm flag that the targets want to support it that way),
> > plus a flag in
On Thu, May 28, 2015 at 5:01 AM, Tristan Gingold wrote:
>
>> On 28 May 2015, at 02:26, Ian Lance Taylor wrote:
>
>> The #include will break cross-compilers. It's not OK for
>> trunk until that is fixed.
>
> I am confused by this comment, for two reasons:
>
> - I don’t see how that would break c
On Tue, May 26, 2015 at 04:37:46PM +0930, Alan Modra wrote:
> On powerpc64le, modifying the way combine treats function parameters
> and call arguments results in some regressions.
>
> For instance, this testcase from varasm.c
>
> extern int foo3 (void *, ...);
> extern void foo4 (void *, const c
This fixes various cases of a common pattern that would result in an ICE in
the gimplifier because it is trying to create a temporary of variable size.
Tested on x86_64-suse-linux, applied on the mainline.
2015-05-28 Eric Botcazou
* gcc-interface/gigi.h (gnat_stabilize_reference): A
Hi,
this patch fixes PR65419.
Consider this test-case:
...
void
f (void)
{
int i;
#pragma acc data copyout (i)
{
}
}
...
When compiling the oacc data region, the start and end are marked with
GOACC_data_start and GOACC_data_end:
...
.omp_data_arr.1.i = &i;
GOACC_data_start (-1, 1,
Hi Mikael,
thanks for the comments so far.
> I don't understand why one of your previous patches was factoring the
> source expression evaluation to a temporary in gfc_trans_allocate, and
> now with this patch you do the same thing in gfc_resolve_allocate, not
> reusing the part in gfc_trans_allo
And more precisely the block of code at the end responsible for setting the
name of types and creating associated typedefs if necessary.
Tested on x86_64-suse-linux, applied on the mainline.
2015-05-28 Eric Botcazou
* gcc-interface/utils.c (gnat_pushdecl): Reunify the handling of ar
We raise bogus Constraint_Error for the Max_Size_In_Storage_Elements attribute
applied to unconstrained array types and types derived from them. This very
likely was introduced when sizetype was changed to unsigned.
Tested on x86_64-suse-linux, applied on the mainline.
2015-05-28 Eric Botcaz
On Thu, May 28, 2015 at 1:57 AM, Bernhard Reutner-Fischer
wrote:
>
> +stamp-pic-ofiles: $(CFILES:%=$(srcdir)/%)
To the best of my knowledge, in POSIX make variable substitutions of
this form do not recognize % specially. In POSIX make this kind of
substitution can only be used to change the file
> This patch avoids the use of the secondary stack, and the corresponding
> cleanup handlers, in many cases. For example, access discriminants no
> longer force functions to return on the secondary stack. This is a speed
> improvement. It is particularly relevant to the Ada.Containers.
It also unc
Richard,
First of all, I don't think that it is possible to write out test for
outer-loop vectorization with zero-step reference because of possible
loop-carried dependencies and run-time aliasing is not supported for
outer-loop. If there are no such dependencies pre or pdse does
hoisting (sinking
I've tested this on aarch64, aarch64_be, and arm, and in all cases, the same
tests are executed (whether running the whole advsimd-intrinsics.exp, or
manually specifying a single file). AFAICT the loop, explicit runtest_file_p,
and gcc_set_parallelization_enable, all stem from a point where we w
On 05/28/2015 04:42 AM, David Malcolm wrote:
Am I right in thinking that this is a statically-allocated object with a
non-trivial constructor? i.e. that this constructor has to run before
"main" is entered?
Do our coding guidelines allow for this? (I've been burned by this
before, on a buggy
On 05/28/2015 06:29 AM, Martin Liška wrote:
Hello.
Thank you for pointing about missing copyright.
Following patch adds that.
Ready for trunk?
Yes.
jeff
On 05/28/2015 06:49 AM, Martin Liška wrote:
.
This mechanism has been just adapted. I find it quite useful as we have
examples in source code where we
allocate same struct/class types from a various pool. For debugging
purpose, it helps to identify if
release operation is called for a correct po
This patch extends -fno-plt to normal non-PIC calls on x86. -fno-plt
works in 64-bit mode with the existing binutils. For 32-bit, we need
the updated assembler and linker to support "call/jmp *foo@GOT" with
a new relocation different from R_386_GOT32 to indicate that this
relocation applies to in
On Thu, May 28, 2015 at 12:52 AM, Kaz Kojima wrote:
>> * config/linux/sh/futex_bits.h (sys_futex0) Change operands
>> "op" and "val" to int.
>>
>> Untested.
>>
>> OK for mainline?
>
> OK.
>
> Although it looks obvious, I've confirmed that there is no build
> issue/regression on sh4-unknow
On May 28, 2015 7:06:36 PM GMT+02:00, Jeff Law wrote:
>On 05/28/2015 04:42 AM, David Malcolm wrote:
>>
>> Am I right in thinking that this is a statically-allocated object
>with a
>> non-trivial constructor? i.e. that this constructor has to run
>before
>> "main" is entered?
>>
>> Do our coding g
On Thu, May 28, 2015 at 07:57:39PM +0200, Richard Biener wrote:
> But we've been trying to avoid this. And the jit might not be too happy about
> it either.
Yeah, we should certainly try to avoid them, especially if it would affect
many variables having to be constructed.
Jakub
Le 28/05/2015 17:29, Andre Vehreschild a écrit :
> *** resolve_allocate_expr (gfc_expr *e, gfc_
> *** 7103,7112
> --- 7103,7123
> if (!ref2 || ref2->type != REF_ARRAY || ref2->u.ar.type == AR_FULL
> || (dimension && ref2->u.ar.dimen == 0))
> {
> + /* F0
All,
This patch restores bootstrap on i386-*-freebsd*.
The build was failing after the introduction of -std=c++98
configure/build flag. The -std=c++98 enables strict_ansi and on FreeBSD
the libc function atoll is not defined for this.
But the configure always stated atoll available.
A bi
On 28 May 2015 at 17:48, Ian Lance Taylor wrote:
> On Thu, May 28, 2015 at 1:57 AM, Bernhard Reutner-Fischer
> wrote:
>>
>> +stamp-pic-ofiles: $(CFILES:%=$(srcdir)/%)
>
> To the best of my knowledge, in POSIX make variable substitutions of
> this form do not recognize % specially. In POSIX make
I have attached a patch that adds the new attribute "noplt". Please review.
* config/i386/i386.c (avoid_plt_to_call): New function.
(ix86_output_call_insn): Generate indirect call for functions
marked with "noplt" attribute.
(attribute_spec ix86_attribute_): Define new attribute "noplt".
* doc/ex
On Thu, May 28, 2015 at 11:34 AM, Sriraman Tallam wrote:
> I have attached a patch that adds the new attribute "noplt". Please review.
>
> * config/i386/i386.c (avoid_plt_to_call): New function.
> (ix86_output_call_insn): Generate indirect call for functions
> marked with "noplt" attribute.
> (at
On 05/28/2015 08:03 PM, Jakub Jelinek wrote:
On Thu, May 28, 2015 at 07:57:39PM +0200, Richard Biener wrote:
But we've been trying to avoid this. And the jit might not be too happy about
it either.
Yeah, we should certainly try to avoid them, especially if it would affect
many variables havin
On Thu, May 28, 2015 at 11:42 AM, H.J. Lu wrote:
> On Thu, May 28, 2015 at 11:34 AM, Sriraman Tallam wrote:
>> I have attached a patch that adds the new attribute "noplt". Please review.
>>
>> * config/i386/i386.c (avoid_plt_to_call): New function.
>> (ix86_output_call_insn): Generate indirect c
On 05/27/2015 08:39 AM, Jason Merrill wrote:
On 05/20/2015 11:50 AM, Aldy Hernandez wrote:
+ /* Fill in the size of variable-length fields in late dwarf. */
+ if (TREE_ASM_WRITTEN (type)
+ && !early_dwarf_dumping)
+{
+ tree member;
+ for (member = TYPE_FIELDS (type); memb
On Thu, May 28, 2015 at 11:50 AM, Sriraman Tallam wrote:
> On Thu, May 28, 2015 at 11:42 AM, H.J. Lu wrote:
>> On Thu, May 28, 2015 at 11:34 AM, Sriraman Tallam
>> wrote:
>>> I have attached a patch that adds the new attribute "noplt". Please review.
>>>
>>> * config/i386/i386.c (avoid_plt_to_
On 28 May 2015 at 18:45, Alan Lawrence wrote:
> I've tested this on aarch64, aarch64_be, and arm, and in all cases, the same
> tests are executed (whether running the whole advsimd-intrinsics.exp, or
> manually specifying a single file). AFAICT the loop, explicit
> runtest_file_p, and gcc_set_para
On Thu, 28 May 2015, Marek Polacek wrote:
This PR points out that we weren't able to optimize 1 << x == 2 to just
x == 1.
Side note: if we are looking for extra patterns to simplify, llvm has an
almost unlimited supply. Here are a few we don't seem to have (there are
more where those came fr
hello,
only providing you the testcase why I need transitive closure of "contains
pointer" via the extra child I noticed that there is extra symmetry to handle:
struct a {void *ptr;}
char **ptr = (char **)&a.ptr;
ptr = ...
This one doesn't really fly with my extra subset code, beca
On Thu, May 28, 2015 at 08:47:16PM +0200, Martin Liška wrote:
> On 05/28/2015 08:03 PM, Jakub Jelinek wrote:
> >On Thu, May 28, 2015 at 07:57:39PM +0200, Richard Biener wrote:
> >>But we've been trying to avoid this. And the jit might not be too happy
> >>about it either.
> >
> >Yeah, we should ce
OK, thanks.
Jason
On 25/05/2015 20:41, Jonathan Wakely wrote:
On 25/05/15 15:31 +0200, François Dumont wrote:
Hi
This is a patch to clean the debug mode code.
I have introduced a new debug header, assertions.h, so that
headers that only need _GLIBCXX_DEBUG_ASSERT do not have to include
the big debug.h.
1 - 100 of 130 matches
Mail list logo
