This avoids sigjmp_buf use which maybe fixes newlib targets.
Tested on x86_64-unknown-linux-gnu, applied.
Richard.
2013-04-24 Richard Biener
PR testsuite/57050
* gcc.c-torture/execute/pr56982.c: Avoid sigjmp_buf use.
Index: gcc/testsuite/gcc.c-torture/execute/pr56982.c
On 23/04/13 17:37, Greta Yorsh wrote:
Ok to backport to gcc4.8?
I'm attaching an updated version - just fixed a spelling error in the
comment.
Thanks,
Greta
gcc/ChangeLog
PR target/56797
* config/arm/arm.c (load_multiple_sequence): Require SP
as base register for load
Practically all of the patterns in rs6000.md having a low-part subreg
use zero as the byte number. That's correct when little-endian, but
not when big-endian, so combine loses optimization opportunities.
This patch duplicates the insns for big-endian. I know Segher has
been working on removing t
Before the target was removed, IRIX defined DWARF_FRAME_RETURN_COLUMN to
be reg 64, which is something other than INCOMING_RETURN_ADDR_RTX (reg
31). This means that when we see a save of reg 31, we discover that
this is really a save of reg 64 and produce dwarf information
accordingly. However, t
This fixes C++ testcases that used size_t when casting a pointer to
integer so that they use intptr_t instead. There's also an testcase
using pointer subtraction where ptrdiff_t is the correct choice, and a
fix to the ptrmem.C testcase to use sizeof on a function pointer rather
than a data pointer.
The pr8781.C testcase appears to have undefined behaviour as far as I
can tell. A function called noop is called with argument pred; this
creates a noop_t object which stores a reference to that argument, and
then the function returns the object. At that point I think the
reference becomes stale. T
This is a first pass at getting powerpc64 little-endian support into
shape. There are no doubt more bugs lurking, but this lets me build
a little-endian C compiler and libgcc to start the little-endian
bootstrap process. Bootstrapped and regression tested powerpc64-linux
to ensure this doesn't br
Hi Guys,
The TARGET_NARROW_VOLATILE_BITFIELD target macro defaults to false:
DEFHOOK
(narrow_volatile_bitfield,
"",
bool, (void),
hook_bool_void_false)
but the documentation currently states that it defaults to
!TARGET_STRICT_ALIGN. The patch below corrects this,
This is a patch series that changes the way vtables are constructed in
the C++ frontend (the first two are small preliminary patches, the meat
will be in 3/3). The problem I was trying to solve was that on the port
I was working on, function pointers are smaller than data pointers, and
size_t is sm
This just removes some unnecessary tests. It's easy to see that if
the sign bit isn't set then xor 0x8000 followed by subtract 0x8000
leaves the value unchanged. Bootstrapped etc. powerpc-linux. OK
for mainline?
* config/rs6000/rs6000.c (rs6000_emit_set_long_const): Tidy.
Index: gcc/co
This is another preliminary piece. It teaches some folding code that
vtable references could contain COMPONENT_REFs and how to deal with that.
Bootstrapped and tested on x86_64-linux, both with and without the final
vtables patch. Ok?
Bernd
commit d1b84c0bbdb1816b7173dca29486bc88124e4641
Author
Committed as obvious.
* config/rs6000/driver-rs6000.c (elf_dcachebsize): Fix comment pasto.
Index: gcc/config/rs6000/driver-rs6000.c
===
--- gcc/config/rs6000/driver-rs6000.c (revision 198174)
+++ gcc/config/rs6000/driver-r
This is the final piece which contains the modifications to the C++
frontend.
Bootstrapped and tested on x86_64-linux, including compat tests against
the installed compiler. I've also built an ia64-hpux (the only target
defining TARGET_VTABLE_DATA_ENTRY_DISTANCE) cc1plus and verified that at
least
Hi,
one of those cases where we error out in a SFINAE context because the
tsubst_flags_t isn't propagated far enough. The straightforward patch
works fine mainline and 4_8-branch.
Tested x86_64-linux.
Thanks,
Paolo.
/cp
2013-04-24 Paolo Carlini
PR c++/56
On Tue, Apr 16, 2013 at 10:17 PM, Kenneth Zadeck
wrote:
> Here is a refreshed version of the rtl changes for wide-int. the only
> change from the previous versions is that the wide-int binary operations
> have been simplified to use the new wide-int binary templates.
Looking for from_rtx calls
I noticed that some functions in SCEV aren't used at all.
They were added in 4.0, but never used since.
Regtested/bootstrapped on x86_64-linux, ok for trunk?
2013-04-24 Marek Polacek
* tree-scalar-evolution.h (analyze_scalar_evolution): Remove.
* tree-scalar-evolution.c (get_e
Joern Rennecke wrote:
> As discussed in the comments to PR38449, this patch replaces
> MD_REDIRECT_BRANCH with TARGET_CAN_FOLLOW_JUMP
>
> bootstrapped / regtested on i686-pc-linux-gnu
>
> cross-built / regtested for i686-pc-linux-gnu X sh-elf
FYI, I've tested the patch with the top level "make
Richard Biener writes:
> Can we in such cases please to a preparatory patch and change the
> CONST_INT/CONST_DOUBLE paths to do an explicit [sz]ext to
> mode precision first?
I'm not sure what you mean here. CONST_INT HWIs are already sign-extended
from mode precision to HWI precision. The 8-bi
On Tue, Apr 23, 2013 at 9:59 PM, Jakub Jelinek wrote:
> On Tue, Apr 23, 2013 at 03:58:06PM -0700, Sriraman Tallam wrote:
>> This patch generates labels for cold function parts that are split when
>> using the option -freorder-blocks-and-partition. The cold label name
>> is generated by suffixin
OK.
Jason
This one is in response to a comment made by Ulrich Weigand. Refer
http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01804.html for why
replace_equiv_address_nv is better than change_address in this reload
related function. Bootstrapped etc. powerpc64-linux.
* config/rs6000/rs6000.c (rs6000_s
This patch improves the messages that get printed by gnatbind when it is
invoked with the -v switch.
Tested on x86_64-pc-linux-gnu, committed on trunk
2013-04-24 Bob Duff
* ali-util.ads (Source_Record): New component Stamp_File
to record from whence the Stamp came.
* a
This patch corrects a failure to issue a message when a Convention
pragma specifying Stdcall applied to a set of homonyms, and other
than the last was a dispatching pragma. The following program
compiles with the indicated warnings:
1. package SCDispatch is
2.type T1 is tagged null r
gnatmake, gnatclean and the GNAT driver were incorrectly putting
directories that are specified with switches -aP after the default
project search directories in the project search path.
This patch ensures that the default project search directories are
always at the end of the project search path.
This implements a new check name Predicate_Check that can be used in
Suppress and Unsuppress pragmas. This allows predicate checks to be
turned on and off for specific sections of code (unlike the effect
of Assertion_Policy which is to enable or disable predicates at the
point where they are define
Any chance of this patch getting a review and commit soon so that it
can go into gcc trunk and gcc-4_8-branch?
Jack
On Sat, Mar 30, 2013 at 03:17:59PM +0100, Magnus Granberg wrote:
> This patch readd TARGET_H that was removed with revision 188166
> IPA_PROP_H is in use by PLUGIN_HEADERS
On Wed, Apr 24, 2013 at 2:44 PM, Richard Sandiford
wrote:
> Richard Biener writes:
>> Can we in such cases please to a preparatory patch and change the
>> CONST_INT/CONST_DOUBLE paths to do an explicit [sz]ext to
>> mode precision first?
>
> I'm not sure what you mean here. CONST_INT HWIs are al
On 04/22/2013 03:19 PM, Jason Merrill wrote:
The only thing missing from our implementation is support for
list-initialization as well as = initialization; I'll add that soon.
These patches add that and parenthesized initializers, and also conform
to the proposal that init-captures be nameable
On Mon, Apr 22, 2013 at 11:16 AM, Jan Hubicka wrote:
> Hi,
> sorry for getting back to this late.
>> >> That's a larger error than I had expected from the merging, although
>> >> as you note it is an approximation so there is going to be some amount
>> >> of error. If the error is that large then
On Wed, Apr 24, 2013 at 2:18 PM, Marek Polacek wrote:
> I noticed that some functions in SCEV aren't used at all.
> They were added in 4.0, but never used since.
>
> Regtested/bootstrapped on x86_64-linux, ok for trunk?
Ok.
Thanks,
Richard.
> 2013-04-24 Marek Polacek
>
> * tree-scala
On Tue, Apr 23, 2013 at 1:51 PM, Sriraman Tallam wrote:
> Hi,
>
> This patch brings the following to the linker function reordering plugin
> present in gcc-4_7
>
> * Node profiles: Callgraph node profiles from the compiler are passed to the
> linker plugin. The node profiles are passed as bb ent
On Wed, Apr 24, 2013 at 7:04 AM, Alan Modra wrote:
> This just removes some unnecessary tests. It's easy to see that if
> the sign bit isn't set then xor 0x8000 followed by subtract 0x8000
> leaves the value unchanged. Bootstrapped etc. powerpc-linux. OK
> for mainline?
>
> * config/rs6
On Wed, Apr 24, 2013 at 9:07 AM, Alan Modra wrote:
> This one is in response to a comment made by Ulrich Weigand. Refer
> http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01804.html for why
> replace_equiv_address_nv is better than change_address in this reload
> related function. Bootstrapped etc.
Richard Biener writes:
> On Wed, Apr 24, 2013 at 2:44 PM, Richard Sandiford
> wrote:
>> Richard Biener writes:
>>> Can we in such cases please to a preparatory patch and change the
>>> CONST_INT/CONST_DOUBLE paths to do an explicit [sz]ext to
>>> mode precision first?
>>
>> I'm not sure what you
On Wed, Apr 24, 2013 at 6:47 AM, Alan Modra wrote:
> This is a first pass at getting powerpc64 little-endian support into
> shape. There are no doubt more bugs lurking, but this lets me build
> a little-endian C compiler and libgcc to start the little-endian
> bootstrap process. Bootstrapped and
On Wed, Apr 24, 2013 at 5:01 AM, Alan Modra wrote:
> Practically all of the patterns in rs6000.md having a low-part subreg
> use zero as the byte number. That's correct when little-endian, but
> not when big-endian, so combine loses optimization opportunities.
>
> This patch duplicates the insns
On Wed, Apr 24, 2013 at 4:03 PM, Richard Sandiford
wrote:
> Richard Biener writes:
>> On Wed, Apr 24, 2013 at 2:44 PM, Richard Sandiford
>> wrote:
>>> Richard Biener writes:
Can we in such cases please to a preparatory patch and change the
CONST_INT/CONST_DOUBLE paths to do an explici
In some cases, the value of global variable Unit_Casing could be read before
being initialized. This is now fixed.
Tested on x86_64-pc-linux-gnu, committed on trunk
2013-04-24 Yannick Moy
* repinfo.adb (List_Rep_Info): Set the value of Unit_Casing before
calling subprograms wh
This patch reimplements the expansion of attribute 'Loop_Entry. Instead of
performing the attribute and related loop transformation post-order, the same
actions are now carried out in-order. This allows for the proper type of the
constant to be propagated to the related context of the attribute.
T
This patch corrects the placement of an error message concerning a redundant
comparison to True. The patch also add machinery to explain the nature of the
redundant True.
-- Source --
-- main.adb
procedure Main is
type Rec (Discr : Boolean) is null record;
funct
On Windows, when gnatfind is called with a pattern and a source file name
that includes capital letters, as in "gnatfind Put_Line:A-textio.ads -r",
gnatfind does not give the same output that it would have if the file
was not capitalized ("gnatfind Put_Line:a-textio.ads -r").
This is corrected by t
This patch suppresses the creation of routine _Postconditions when the related
context lacks invariants or predicates and all postcindition aspect / pragmas
are disabled.
-- Source --
-- main.adb
procedure Main is
X : Integer := 0;
procedure P with
Post =>
Richard Biener writes:
> I suppose the above should use immed_double_int_const (v, mode), too,
In practice it doesn't matter, because...
> which oddly only ever truncates to mode for modes <= HOST_BITS_PER_WIDE_INT
> via gen_int_mode.
...right. That's because there's not really any proper supp
A new switch -aPdir is added to gnatls. When gnatls is called with one
or several switches -aPdir and also switch -v, the directories specified
in the -aP switches are displayed immediately after the current directory
in the Project Search Path.
A new warning is issued by gnatls for switches that a
We do not need checks if we are not generating code, and in SPARK mode, we
specifically don't want the frontend to expand these checks, which are dealt
with directly in the formal verification backend. This is similar to what is
already done for other checks.
Tested on x86_64-pc-linux-gnu, committ
Hi,
The vld1_* patterns in arm_neon.h did not correctly describe
their register/memory constraints. This could lead to incorrect
code generation where they were used.
This patch fixes the bug by giving the patterns the correct
register constraints.
Regression tested on aarch64-none-elf without r
On 04/24/2013 09:36 AM, Richard Biener wrote:
On Wed, Apr 24, 2013 at 2:44 PM, Richard Sandiford
wrote:
Richard Biener writes:
Can we in such cases please to a preparatory patch and change the
CONST_INT/CONST_DOUBLE paths to do an explicit [sz]ext to
mode precision first?
I'm not sure what y
Contract nodes are meant to carry information for subprogram spec entities,
not subprogram body entities (for bodies that are completions of specs).
So we remove the contract node when a spec entity is changed to a body entity.
Depending on how refined contracts on bodies are handled, this may need
The following patch makes us create and preserves loops from the
point we build the CFG. As-is the patch passes bootstrap
on x86_64-unknown-linux-gnu for all languages including Ada
and has no ICEs running the testsuite. There are a few testcases
that fail which I still have to investigate.
I t
On Wed, Apr 24, 2013 at 4:29 PM, Richard Sandiford
wrote:
> Richard Biener writes:
>> I suppose the above should use immed_double_int_const (v, mode), too,
>
> In practice it doesn't matter, because...
>
>> which oddly only ever truncates to mode for modes <= HOST_BITS_PER_WIDE_INT
>> via gen_int
Hi,
I believe this is all we need in order to get the ball rolling in the
library for -std=c++1y.
If we think it's conceptually clearer (no difference in practice,
because cxx11 == cxx0x), for the legacy C++0x macro we could also do:
if (cxx_dialect >= cxx0x && cxx_dialect < cxx1y)
This patch corrects the parser to detect an illegal placement of the aspect
specification list.
-- Source --
-- semantics.adb:
procedure Semantics is
function Error (Formal : Integer) return Boolean
with Pre => Formal > 1234 is (Formal = 5678);
begin
null;
e
In Ada 2012 it is possible to complete an incomplete type with a record one
of whose components is an anonymous access to task initialized with an
allocator. The analysis of the allocator in the init_proc for the type creates
an anonymous access that requires a master id. This must be obtained from
On Wed, Apr 24, 2013 at 4:59 AM, Bernd Schmidt wrote:
> This fixes C++ testcases that used size_t when casting a pointer to
> integer so that they use intptr_t instead. There's also an testcase
> using pointer subtraction where ptrdiff_t is the correct choice, and a
> fix to the ptrmem.C testcase
On 04/24/2013 10:42 AM, Richard Biener wrote:
On Wed, Apr 24, 2013 at 4:29 PM, Richard Sandiford
wrote:
Richard Biener writes:
I suppose the above should use immed_double_int_const (v, mode), too,
In practice it doesn't matter, because...
which oddly only ever truncates to mode for modes <
On Wed, Apr 24, 2013 at 6:37 AM, Teresa Johnson wrote:
> On Mon, Apr 22, 2013 at 11:16 AM, Jan Hubicka wrote:
>> Hi,
>> sorry for getting back to this late.
>>> >> That's a larger error than I had expected from the merging, although
>>> >> as you note it is an approximation so there is going to b
This patch disables the generation of internal procedure _Postconditions when
invariants and preconditions are disabled.
-- Source --
-- main.adb
procedure Main is
X : Integer := 0;
type R is new Integer with Predicate => X > 0;
package Pack is
type T is
This patch fixes a small missunderstanding in the implementation of fall-back
termination handlers. Previously, a fall-back termination handler set by a
given task would apply for itself. However, it has been now corrected because
it applies only to dependent tasks (see ARM C.7.3 par. 9/2).
The fo
On Wed, Apr 24, 2013 at 9:43 AM, Paolo Carlini wrote:
> Hi,
>
> I believe this is all we need in order to get the ball rolling in the
> library for -std=c++1y.
>
> If we think it's conceptually clearer (no difference in practice, because
> cxx11 == cxx0x), for the legacy C++0x macro we could also
Richard Biener writes:
> On Wed, Apr 24, 2013 at 4:29 PM, Richard Sandiford
> wrote:
>> In other words, one of the reasons wide_int can't be exactly 1:1
>> in practice is because it is clearing out these mistakes (GEN_INT
>> rather than gen_int_mode) and missing features (non-power-of-2 widths).
On Wed, Apr 24, 2013 at 4:35 PM, Kenneth Zadeck
wrote:
> On 04/24/2013 09:36 AM, Richard Biener wrote:
>>
>> On Wed, Apr 24, 2013 at 2:44 PM, Richard Sandiford
>> wrote:
>>>
>>> Richard Biener writes:
Can we in such cases please to a preparatory patch and change the
CONST_INT/CONS
This patch enables Redundant Extension Elimination pass for AArch64.
Testing shows no regressions on linux and bare-metal.
In terms of performance impact, it reduces code-size for some benchmarks
and makes no difference on others.
OK to commit to trunk?
Cheers,
Ian
2013-04-24 Ian Bolton
On 24/04/13 15:34, James Greenhalgh wrote:
Hi,
The vld1_* patterns in arm_neon.h did not correctly describe
their register/memory constraints. This could lead to incorrect
code generation where they were used.
This patch fixes the bug by giving the patterns the correct
register constraints.
Re
> this isn't true, it would be helpful to know. Given the size_t/uintptr_t
> testsuite patch I just submitted I'm thinking they don't exist, but I'm
> kind of wondering about m32c, so Cc'ing DJ.
For m32c-elf with -mcpu=m32c, size_t is 16 bits but void* is 24 bits.
On 24/04/13 16:09, Marcus Shawcroft wrote:
On 24/04/13 15:34, James Greenhalgh wrote:
Hi,
The vld1_* patterns in arm_neon.h did not correctly describe
their register/memory constraints. This could lead to incorrect
code generation where they were used.
This patch fixes the bug by giving the pa
On 24/04/13 16:06, Ian Bolton wrote:
This patch enables Redundant Extension Elimination pass for AArch64.
Testing shows no regressions on linux and bare-metal.
In terms of performance impact, it reduces code-size for some benchmarks
and makes no difference on others.
OK to commit to trunk?
Ch
On Wed, Apr 24, 2013 at 5:00 PM, Richard Sandiford
wrote:
> Richard Biener writes:
>> On Wed, Apr 24, 2013 at 4:29 PM, Richard Sandiford
>> wrote:
>>> In other words, one of the reasons wide_int can't be exactly 1:1
>>> in practice is because it is clearing out these mistakes (GEN_INT
>>> rather
On Wed, 3 Apr 2013, Senthil Kumar Selvaraj wrote:
> 2013-04-03Senthil Kumar Selvaraj
>
> * c-common.c (check_user_alignment): Emit error for negative values
>
> * gcc.dg/c1x-align-3.c: Add test for negative power of 2
OK (but note there should be a "." at the end of each C
On Fri, 5 Apr 2013, Paolo Carlini wrote:
> Hi,
>
> in the audit trail of c++/56815 Manuel noticed that, inconsistently with the
> documentation, a LangEnabledBy was missing for -Wpointer-arith vs -Wpedantic.
>
> Then I noticed that a clean up was possible in the actual pedwarn calls,
> which, in
Richard Biener writes:
> On Wed, Apr 24, 2013 at 4:35 PM, Kenneth Zadeck
> wrote:
>> On 04/24/2013 09:36 AM, Richard Biener wrote:
>>>
>>> On Wed, Apr 24, 2013 at 2:44 PM, Richard Sandiford
>>> wrote:
Richard Biener writes:
>
> Can we in such cases please to a preparatory patc
On 12/04/13 15:19, Kyrylo Tkachov wrote:
Hi all,
This patch adds testsuite options to check for a neon-fp16
effective target and add appropriate options.
These are needed to test the half-precision NEON intrinsics that adding
with patch [1/2] of this set.
Tested on qemu with the tests added in p
On 19/04/13 22:40, Mike Frysinger wrote:
The bpabi.h header already sets up defines to automatically use the
--fix-v4bx flag with the assembler & linker as needed, and creates a
default assembly & linker spec which uses those. Unfortunately, the
linux-eabi.h header clobbers the LINK_SPEC define
The following patch incorporates some LRA changes on lra and mike-lra
branches. I am committing them to get a better testing of them for
x86/x86-64 on trunk.
The patch was successfully tested and bootstrapped on x86/x86-64.
Committed as rev. 198252.
2013-04-24 Vladimir Makarov
*
On 2013-04-19 07:30 , Richard Biener wrote:
On Tue, 16 Apr 2013, Diego Novillo wrote:
Thanks for the feedback, folks.
I've removed the builder type and added some overloads to simplify the
creation of gimple assignments. I have only added exactly the functions I
needed to simplify a bit of gc
Richard Biener writes:
> On Wed, Apr 24, 2013 at 5:00 PM, Richard Sandiford
> wrote:
>> Richard Biener writes:
>>> On Wed, Apr 24, 2013 at 4:29 PM, Richard Sandiford
>>> wrote:
In other words, one of the reasons wide_int can't be exactly 1:1
in practice is because it is clearing out t
I would really rather avoid introducing another macro to be removed
again later. Instead, let's use a value of __cplusplus greater than
201103L, perhaps 201300?
Jason
We were already trying to support the rule that *this doesn't need to be
complete for class member access, but we were comparing the wrong thing
to current_class_ref.
Tested x86_64-pc-linux-gnu, applying to trunk.
commit 460a34462cd0dd2787e1ffddf4b08d1c36ff5557
Author: Jason Merrill
Date: We
On Wed, Apr 24, 2013 at 11:26 AM, Jason Merrill wrote:
> I would really rather avoid introducing another macro to be removed again
> later. Instead, let's use a value of __cplusplus greater than 201103L,
> perhaps 201300?
>
> Jason
yes, that makes sense, and even a better path forward.
Hopefully
On 04/24/2013 05:10 PM, DJ Delorie wrote:
>
>> this isn't true, it would be helpful to know. Given the size_t/uintptr_t
>> testsuite patch I just submitted I'm thinking they don't exist, but I'm
>> kind of wondering about m32c, so Cc'ing DJ.
>
> For m32c-elf with -mcpu=m32c, size_t is 16 bits but
On 04/24/2013 12:48 PM, Gabriel Dos Reis wrote:
On Wed, Apr 24, 2013 at 11:26 AM, Jason Merrill wrote:
I would really rather avoid introducing another macro to be removed again
later. Instead, let's use a value of __cplusplus greater than 201103L,
perhaps 201300?
yes, that makes sense, and e
Hi,
On 04/24/2013 06:55 PM, Jason Merrill wrote:
On 04/24/2013 12:48 PM, Gabriel Dos Reis wrote:
On Wed, Apr 24, 2013 at 11:26 AM, Jason Merrill
wrote:
I would really rather avoid introducing another macro to be removed
again
later. Instead, let's use a value of __cplusplus greater than 201
ARM build (on chrome) is broken. This patch fixed the problem. Tested
by building arm cross compiler successfully.
* gcc/config/config.cc: Removed duplicated header files which
causes error in generating gtyp-input.list.
* gcc/config/arm.md: Most of define_c_enum "unspec"
All
Currently using the match_scratch predicate on a destination operand will
trigger the warning
"warning: destination operand 0 allows non-lvalue",
This happends because add_predicate_code() will set the
pred->allows_non_lvalue when
it encounters a SCRATCH rtx code wh
Hi again ;)
thus I'm finishing testing (past g++.dg/dg.exp) the below. I added
proper GNUCXX1Y and CXX1Y modes and then the rest seems largely
straightforward.
Ok if testing passes?
Thanks,
Paolo.
///
/libcpp
2013-04-24 Paolo Carlini
* include/cpplib.h (e
On Wed, Apr 24, 2013 at 11:55 AM, Jason Merrill wrote:
> On 04/24/2013 12:48 PM, Gabriel Dos Reis wrote:
>>
>> On Wed, Apr 24, 2013 at 11:26 AM, Jason Merrill wrote:
>>>
>>> I would really rather avoid introducing another macro to be removed again
>>> later. Instead, let's use a value of __cplus
On 04/24/2013 02:02 PM, Paolo Carlini wrote:
+#if __cplusplus < 201300L
Don't test for this value. Use <= 201103L instead.
OK with that change.
Jason
Hi,
On Fri, Apr 19, 2013 at 09:27:28AM -0600, Jeff Law wrote:
> On 04/18/2013 05:08 PM, Martin Jambor wrote:
> >On Fri, Apr 19, 2013 at 12:37:58AM +0200, Steven Bosscher wrote:
> >>On Fri, Apr 19, 2013 at 12:09 AM, Martin Jambor wrote:
> >>>I also have not tried scheduling the hard register copy p
When GCC is configured for x86_64-elf for embedded target, --x32 isn't
passed to assembler for "gcc -mx32". This patch adds -mx32 support to
GCC assembler driver. OK for trunk?
Thanks.
H.J.
---
2013-04-24 H.J. Lu
* config/i386/x86-64.h (ASM_SPEC): Support -mx32.
diff --git a/gcc/c
This patch is a consolodation of the hash_table patches to the
cxx-conversion branch for files under gcc/config.
Recipients:
config/arm/arm.c - ni...@redhat.com, ramana.radhakrish...@arm.com
config/ia64/ia64.c - wil...@tuliptree.org, sell...@mips.com
config/mips/mips.c - rdsandif...@googlemail.com
> 24 bits stored as three bytes, or four? How does this affect vtable
> layout? I would have expected the C++ frontend and libsupc++ to
> currently be inconsistent with each other given such a setup.
In memory, four, I think. The address registers really are three
bytes though. They're PSImode
> When GCC is configured for x86_64-elf for embedded target, --x32 isn't
> passed to assembler for "gcc -mx32". This patch adds -mx32 support to
> GCC assembler driver. OK for trunk?
OK,
Thanks
Honza
On 04/24/2013 12:24 PM, Martin Jambor wrote:
Here they are. First, I simply looked at how many instructions would
be changed by a second run of the pass in its current position during
C and C++ bootstrap:
| | Insns changed | % |
|-
I'm seeing a lot of failures with these changes in make check.
The first two that I noticed on a build that did not use --with-cpu=power7:
1) c-c++-common/dfp/call-by-value.c (and others in the directory) fails with -O0
for all targets before power7 because it can't spill SDmode. Note, in the
ea
Hi,
On 04/24/2013 05:49 PM, Vladimir Makarov wrote:
The following patch incorporates some LRA changes on lra and mike-lra
branches. I am committing them to get a better testing of them for
x86/x86-64 on trunk.
Just in case nobody reported it already, when the patch went in these
regressions a
Hi,
the attached patch hopefully fixes the PR.
Regtested on x86_64-unknown-linux-gnu, Ok for trunk?
2013-04-24 Janne Blomqvist
PR bootstrap/57028
* Make-lang.in (f951): Link in ZLIB.
(CFLAGS-module.o): Add zlib include directory.
--
Janne Blomqvist
fortzlib.diff
Description:
The following patch fixes
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57046
The patch was successfully bootstrapped and tested on x86/x86-64.
Committed as rev. 198263.
2013-04-24 Vladimir Makarov
PR rtl-optimizations/57046
* lra-constraints (split_reg): Set up lra_ris
On 04/24/2013 11:41 AM, Graham Stott wrote:
All
Currently using the match_scratch predicate on a destination operand will
trigger the warning
"warning: destination operand 0 allows non-lvalue",
This happends because add_predicate_code() will set the
pred->allows_non_lva
Comments inline below.
>
> This patch brings the following to the linker function reordering plugin
> present in gcc-4_7
>
> * Node profiles: Callgraph node profiles from the compiler are passed to the
> linker plugin. The node profiles are passed as bb entry count and max count
> of the corresp
Hi!
This patch brings the C FE roughly on feature parity with what
has been done earlier to the C++ FE only.
Thus, #pragma omp simd should work with C FE now, etc.
2013-04-24 Jakub Jelinek
c/
* c-parser.c (c_parser_compound_statement,
c_parser_statement): Adjust comments for O
The C++0x draft defined std::copy_exception but it was renamed to
std::make_exception_ptr in the final C++11 standard. This changes the
library to use the new name and deprecates the old one, so we can
remove it one day.
PR libstdc++/56905
* libsupc++/exception_ptr.h (copy_excepti
1 - 100 of 113 matches
Mail list logo
