On Aug 17 2017, Martin Sebor wrote:
> -static void (*resolve_memcpy (void)) (void)
> +static void* (*resolve_memcpy (void))(void *, const void *, size_t)
Please use consistent spacing.
Andreas.
--
Andreas Schwab, sch...@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5
This patch to the go tool in libgo changes it to use -funwind-tables
when compiling C code to support the cgo tool. Using -funwind-tables
is necessary to permit Go code to correctly throw a panic through C
code. This hasn't been necessary in the past as -funwind-tables is
the default on x86. How
Joseph, while looking into implementing enhancement your request
pr81824 I noticed that GCC silently accepts incompatible alias
declarations (pr81854) so as sort of a proof-concept for the
former I enhanced the checking already done for other kinds of
incompatibilities to also detect those mention
On Thu, 2017-08-17 at 00:28 -0500, Segher Boessenkool wrote:
> Hi!
>
> On Wed, Aug 16, 2017 at 03:35:40PM -0500, Steven Munroe wrote:
> > +extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__,
> > __artificial__))
> > +_mm_add_ss (__m128 __A, __m128 __B)
> > +{
> > +#ifdef _ARC
PR target/80210 exposes a problem in rs6000_set_current_function() where
is fails to correctly clear the rs6000_previous_fndecl cache correctly.
With the test case, we notice that rs6000_previous_fndecl is set (when it
shouldn't be) and we end up restoring options from it. In this case,
we end up
This patch deletes some of the debug switches that I've added over the years,
and now don't use any more.
I did bootstrap builds and make check runs on a little endian power8 system.
There were no regressions. Can I check this into the trunk?
2017-08-17 Michael Meissner
* config/rs60
OK.
On Thu, Aug 17, 2017 at 7:50 AM, David Malcolm wrote:
> Ping
>
> On Thu, 2017-07-27 at 17:36 -0400, David Malcolm wrote:
>> PR c++/81514 reports a problem where
>> g++.dg/lookup/missing-std-include-2.C
>> fails on Solaris, offering the suggestion:
>>
>> error: 'string' is not a member of
On Thu, Aug 17, 2017 at 7:56 AM, H.J. Lu wrote:
> On Thu, Aug 17, 2017 at 6:52 AM, Joseph Myers wrote:
>> On Sat, 8 Jul 2017, H.J. Lu wrote:
>>
>>> +@item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
>>> +@opindex Wpacked-not-aligned
>>> +@opindex Wno-packed-not-aligned
>
On Thu, Aug 17, 2017 at 12:00 PM, Nathan Sidwell wrote:
> This patch concludes with simplifying TYPE_TEMPLATE_INFO. It was too
> tricky, given the time available, to make TYPE_TEMPLATE_INFO require a
> templatable _TYPE node, so it remains returning NULL_TREE if you give it a
> random _TYPE node.
Looks like the following patch falled through the cracks
https://gcc.gnu.org/ml/gcc-patches/2012-12/msg01397.html
https://bugs.gentoo.org/show_bug.cgi?id=434180#c16
Thanks!
--
Sergei
pgpUpEUwrmHQi.pgp
Description: Цифровая подпись OpenPGP
On 08/17/2017 01:26 PM, Alexander Monakov wrote:
On Thu, 17 Aug 2017, Martin Sebor wrote:
returns a pointer to the selected implementation function. The
implementation functions' declarations must match the API of the
-function being implemented, the resolver's declaration is be a
-function
On Thu, 17 Aug 2017, Martin Sebor wrote:
> returns a pointer to the selected implementation function. The
> implementation functions' declarations must match the API of the
> -function being implemented, the resolver's declaration is be a
> -function returning pointer to void function returning
This patch concludes with simplifying TYPE_TEMPLATE_INFO. It was too
tricky, given the time available, to make TYPE_TEMPLATE_INFO require a
templatable _TYPE node, so it remains returning NULL_TREE if you give it
a random _TYPE node. That'd be nice to clean up though.
I can't recall the unde
On 8/16/17 5:49 PM, Segher Boessenkool wrote:
>> - (TI "TARGET_VSX_TIMODE")])
>> + (TI "TARGET_VSX")])
>
> You can completely remove the condition here as far as I see.
[snip]
> Okay for trunk with that simplification. Thanks!
Good catch. I do
On 8/16/17 5:56 PM, Peter Bergner wrote:
> On 8/16/17 5:30 PM, Segher Boessenkool wrote:
> I'll make the above changes and commit after another quick test cycle.
Testing the changes came up clean, so I committed it. Thanks.
Peter
The attached patch adjusts the documentation of attribute ifunc
to avoid warnings (in C) and errors (in C++) due to the resolver
returning a pointer to a function of incompatible type.
Martin
PR c/81882 - attribute ifunc documentation uses invalid code
gcc/ChangeLog:
PR c/81882
* doc/extend.t
On 08/16/2017 02:55 PM, David Malcolm wrote:
On Wed, 2017-08-16 at 10:12 -0600, Martin Sebor wrote:
PR c/81859 - [8 Regression] valgrind error from
warn_about_normalization
gcc/ChangeLog:
PR c/81859
* pretty-print.c (pp_format): Use strnlen in %.*s to avoid
reading
past
> It seems with Ada one can have cycles in types via pointer types
> and calling variably_modified_type_p on such type recurses indefinitely.
> The following is an attempt to fix that (albeit I'm not 100% sure
> there's no caller of the function using TREE_VISITED itself).
>
> Cures c38102a.adb wi
Attached is an updated patch with just the minor editorial
tweaks for the issues pointed out by Marek (stray comments
and spaces), and with the C++ and libstdc++ bits removed
and posted separately. I also added some text to the manual
to clarify the const/pure effects.
I thought quite a bit more
Now that template info is on TYPE_LANG_SLOT_1, B_T_T_P_Types don't need
type_lang_specific.
Excised thusly.
nathan
--
Nathan Sidwell
2017-08-17 Nathan Sidwell
* lex.c (maybe_add_lang_type_raw): BOUND_TEMPLATE_TEMPLATE_PARMs
don't need lang_type.
(cxx_make_type): Use maybe_add_lang_type_r
On Thu, 17 Aug 2017, Wilco Dijkstra wrote:
> This patch simplifies pow (C, x) into exp (x * C1) if C > 0, C1 = log (C).
Note this changes the outcome for C == +Inf, x == 0 (pow is specified to
return 1.0 in that case, but x * C1 == NaN). There's another existing
transform with the same issue, 'p
It seems with Ada one can have cycles in types via pointer types
and calling variably_modified_type_p on such type recurses indefinitely.
The following is an attempt to fix that (albeit I'm not 100% sure
there's no caller of the function using TREE_VISITED itself).
Cures c38102a.adb with LTO earl
On Thu, Aug 17, 2017 at 6:52 AM, Joseph Myers wrote:
> On Sat, 8 Jul 2017, H.J. Lu wrote:
>
>> +@item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
>> +@opindex Wpacked-not-aligned
>> +@opindex Wno-packed-not-aligned
>> +Warn if a structure field with explicitly specified a
On Thu, Aug 17, 2017 at 04:36:02PM +0200, Richard Biener wrote:
> On Thu, 17 Aug 2017, Marek Polacek wrote:
>
> > On Thu, Aug 17, 2017 at 04:17:54PM +0200, Richard Biener wrote:
> > > On Thu, 17 Aug 2017, Marek Polacek wrote:
> > >
> > > > This PR is about wrong-code and has gone undetected for o
Ping
On Thu, 2017-07-27 at 17:36 -0400, David Malcolm wrote:
> PR c++/81514 reports a problem where
> g++.dg/lookup/missing-std-include-2.C
> fails on Solaris, offering the suggestion:
>
> error: 'string' is not a member of 'std'
> note: suggested alternative: 'sprintf'
>
> instead of the
On Thu, Aug 17, 2017 at 3:43 PM, Martin Liška wrote:
> Hello.
>
> As described in https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81864#c5 we
> should not
> introduce a new static hash_table variable in order to be sure
> mem_alloc_descriptors
> are initialized earlier.
>
>
> Patch can bootstrap on p
On Thu, Aug 17, 2017 at 1:06 PM, Richard Sandiford
wrote
> Richard Biener writes:
>> On Thu, Aug 17, 2017 at 11:49 AM, Richard Sandiford
>> wrote:
>>> Internal functions that map directly to an optab can only throw an
>>> exception for -fnon-call-exceptions. This patch handles that in
>>> inter
On Thu, 17 Aug 2017, Marek Polacek wrote:
> On Thu, Aug 17, 2017 at 04:17:54PM +0200, Richard Biener wrote:
> > On Thu, 17 Aug 2017, Marek Polacek wrote:
> >
> > > This PR is about wrong-code and has gone undetected for over 10 years (!).
> > > The issue is that e.g. the following
> > >
> > >
On Thu, Aug 17, 2017 at 04:17:54PM +0200, Richard Biener wrote:
> On Thu, 17 Aug 2017, Marek Polacek wrote:
>
> > This PR is about wrong-code and has gone undetected for over 10 years (!).
> > The issue is that e.g. the following
> >
> > (signed char) x == 0 ? (unsigned long long) x : 0
> >
>
Hi,
[Patch, rs6000] testcase coverage for vec_perm built-ins
Add some Testcase coverage for the vector permute intrinsics.
Tested across power platforms. OK for trunk?
Thanks,
-Will
[gcc/testsuite]
2017-08-17 Will Schmidt
* gcc.target/powerpc/fold-vec-perm-char.c: New.
*
Hi,
[PATCH, rs6000] testcase coverage for vec_perm built-ins
Add testcase coverage for the vec_ld intrinsic builtins.
Tested across power platforms (p6 and newer). OK for trunk?
Thanks,
-Will
[gcc/testsuite]
2017-08-17 Will Schmidt
* gcc.target/powerpc/fold-vec-ld-char.c: New.
Hi,
[Patch, rs6000] Add testcase coverage for vec_sums built-in
Add some testcase coverage for the vec_sums() built-in.
Tested across power platforms (p6 and newer). OK for trunk?
Thanks,
-Will
[gcc/testsuite]
2017-08-17 Will Schmidt
* gcc.target/powerpc/fold-vec-sums-int.c: New.
On Thu, 17 Aug 2017, Marek Polacek wrote:
> This PR is about wrong-code and has gone undetected for over 10 years (!).
> The issue is that e.g. the following
>
> (signed char) x == 0 ? (unsigned long long) x : 0
>
> was wrongly folded to 0, because fold_cond_expr_with_comparison will fold
> A
On Thu, 17 Aug 2017, Marek Polacek wrote:
> I've been itching to remove this code for some time now. The comment suggests
> that the code is actually unused, so I replaced the body of that "else if"
> with
> gcc_unreachable (); and ran regtest/bootstrap and nothing broke, so I propose
> to do aw
This patch simplifies pow (C, x) into exp (x * C1) if C > 0, C1 = log (C).
Do this only for fast-math as accuracy is reduced. This is much faster
since pow is more complex than exp - with current GLIBC the speedup is
more than 7 times for this transformation.
The worst-case ULP error of the trans
On Sat, 8 Jul 2017, H.J. Lu wrote:
> +@item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
> +@opindex Wpacked-not-aligned
> +@opindex Wno-packed-not-aligned
> +Warn if a structure field with explicitly specified alignment in a
> +packed struct or union is misaligned. For e
Hello.
As described in https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81864#c5 we should
not
introduce a new static hash_table variable in order to be sure
mem_alloc_descriptors
are initialized earlier.
Patch can bootstrap on ppc64le-redhat-linux and survives regression tests.
Ready to be inst
Currently non-bootstrap of --enable-languages=ada fails, the following
fixes this.
Bootstrapped and non-bootstrapped on x86_64-unknown-linux-gnu.
Approved by Eric in the PR, applied.
Richard.
2017-08-17 Richard Biener
PR ada/81878
* Makefile.in (CXX_LFLAGS): Remove.
On 16/08/17 16:19, Jackson Woodruff wrote:
> Hi Richard,
>
> I have changed the condition as you suggest below. OK for trunk?
>
> Jackson.
>
I renamed the testcase to vect_str_zero.c, as that seems to more closely
match the naming style, and checked this in.
Thanks for the patch.
R.
> On 08/
With yesterday's patch to move TYPE_BINFO onto TYPE_MAX_VALUES_RAW, I
freed up TYPE_LANG_SLOT_1 for class types.
This patch moves template_info from lang_type to that field for both
RECORD_TYPESs and BOUND_TEMPLATE_TEMPLATE_PARM_TYPEs.
now, you'll notice that TYPE_LANG_SLOT_1 was where ENUMER
On Wed, Aug 9, 2017 at 3:39 AM, H.J. Lu wrote:
> On Tue, Aug 8, 2017 at 10:36 PM, Richard Biener
> wrote:
>> On August 9, 2017 12:18:41 AM GMT+02:00, "H.J. Lu"
>> wrote:
>>>This patch adds -static-pie to GCC driver to create static PIE. A
>>>static
>>>position independent executable (PIE) is s
This PR is about wrong-code and has gone undetected for over 10 years (!).
The issue is that e.g. the following
(signed char) x == 0 ? (unsigned long long) x : 0
was wrongly folded to 0, because fold_cond_expr_with_comparison will fold
A != 0 ? A : 0 to 0. But for x = 0x0100 this is wrong:
On 8 August 2017 at 09:50, Prathamesh Kulkarni
wrote:
> On 31 July 2017 at 23:53, Prathamesh Kulkarni
> wrote:
>> On 23 May 2017 at 19:10, Prathamesh Kulkarni
>> wrote:
>>> On 19 May 2017 at 19:02, Jan Hubicka wrote:
>
> * LTO and memory management
> This is a general question about
On 8 August 2017 at 09:51, Prathamesh Kulkarni
wrote:
> On 1 August 2017 at 00:10, Prathamesh Kulkarni
> wrote:
>> On 11 July 2017 at 17:59, Prathamesh Kulkarni
>> wrote:
>>> On 13 June 2017 at 01:47, Joseph Myers wrote:
This is OK with one fix:
> +C ObjC Var(warn_enum_conversion)
On Thu, Aug 17, 2017 at 12:35 PM, Richard Sandiford
wrote:
> "Bin.Cheng" writes:
>> On Wed, Aug 16, 2017 at 6:50 PM, Richard Sandiford
>> wrote:
>>> "Bin.Cheng" writes:
On Wed, Aug 16, 2017 at 5:00 PM, Richard Sandiford
wrote:
> "Bin.Cheng" writes:
>> On Wed, Aug 16, 2017 at
On Thu, 17 Aug 2017, Richard Biener wrote:
> > Without global analysis it seems best to move constants/negates to the
> > toplevel if they can't be trivially removed in a subexpression. Eg. -x
> > / (a * b * -c).
>
> Sure. So both patterns are canonicalization which is fine for match.pd.
> T
Hi,
this patch requires effective target nonlocal_goto for ipa/pr81696.c.
Committed.
Thanks,
- Tom
2017-08-17 Tom de Vries
* gcc.dg/ipa/pr81696.c: Require effective target nonlocal_goto.
---
gcc/testsuite/gcc.dg/ipa/pr81696.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/gc
On Thu, Aug 17, 2017 at 1:37 PM, Maxim Ostapenko
wrote:
> Hi,
>
> as discussed in PR, currently we are missing some SSP related options when
> streaming target specific options in LTO mode. This leads to incorrect code
> and segfaults (e.g. in ASan's pr64820.c testcase).
> This patch fixes this by
Hi,
as discussed in PR, currently we are missing some SSP related options
when streaming target specific options in LTO mode. This leads to
incorrect code and segfaults (e.g. in ASan's pr64820.c testcase).
This patch fixes this by moving options saving machinery down to setting
ix86_stack_prot
"Bin.Cheng" writes:
> On Wed, Aug 16, 2017 at 6:50 PM, Richard Sandiford
> wrote:
>> "Bin.Cheng" writes:
>>> On Wed, Aug 16, 2017 at 5:00 PM, Richard Sandiford
>>> wrote:
"Bin.Cheng" writes:
> On Wed, Aug 16, 2017 at 2:38 PM, Richard Sandiford
> wrote:
>> The first loop in th
Richard Biener writes:
> On Thu, Aug 17, 2017 at 11:49 AM, Richard Sandiford
> wrote:
>> Internal functions that map directly to an optab can only throw an
>> exception for -fnon-call-exceptions. This patch handles that in
>> internal_fn_flags, in a similar way to ATTR_*NOTHROW in builtins.def.
> -Original Message-
> From: Richard Biener [mailto:rguent...@suse.de]
> Sent: 08 June 2017 19:23
> To: Joseph Myers
> Cc: Tamar Christina; Christophe Lyon; Markus Trippelsdorf; Jeff Law; GCC
> Patches; Wilco Dijkstra; Michael Meissner; nd
> Subject: Re: [PATCH][GCC][PATCHv3] Improve fpcl
On Wed, Aug 16, 2017 at 6:50 PM, Richard Sandiford
wrote:
> "Bin.Cheng" writes:
>> On Wed, Aug 16, 2017 at 5:00 PM, Richard Sandiford
>> wrote:
>>> "Bin.Cheng" writes:
On Wed, Aug 16, 2017 at 2:38 PM, Richard Sandiford
wrote:
> The first loop in the testcase regressed after my re
Richard Biener writes:
> On Thu, Aug 17, 2017 at 11:47 AM, Richard Sandiford
> wrote:
>> This patch adds missing ECF_NOTHROW flags to the vectorisable
>> integer internal functions. I noticed it while doing some SVE
>> work but don't have a testcase that's useful now.
>>
>> Tested on aarch64-lin
On Thu, Aug 17, 2017 at 11:49 AM, Richard Sandiford
wrote:
> Internal functions that map directly to an optab can only throw an
> exception for -fnon-call-exceptions. This patch handles that in
> internal_fn_flags, in a similar way to ATTR_*NOTHROW in builtins.def.
>
> (Functions that don't throw
I've been itching to remove this code for some time now. The comment suggests
that the code is actually unused, so I replaced the body of that "else if" with
gcc_unreachable (); and ran regtest/bootstrap and nothing broke, so I propose
to do away with it.
Bootstrapped/regtested on x86_64-linux, o
On Thu, Aug 17, 2017 at 11:47 AM, Richard Sandiford
wrote:
> This patch adds missing ECF_NOTHROW flags to the vectorisable
> integer internal functions. I noticed it while doing some SVE
> work but don't have a testcase that's useful now.
>
> Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK
On Thu, Aug 17, 2017 at 11:55 AM, Wilco Dijkstra wrote:
> Richard Biener wrote:
>> On Tue, Aug 15, 2017 at 4:11 PM, Wilco Dijkstra
>> wrote:
>> > Richard Biener wrote:
> We also change the association of
>
> x / (y * C) -> (x / C) / y
>
> If C is a constant.
>>
Richard Biener wrote:
> On Tue, Aug 15, 2017 at 4:11 PM, Wilco Dijkstra
> wrote:
> > Richard Biener wrote:
>>> > We also change the association of
>>> >
>>> > x / (y * C) -> (x / C) / y
>>> >
>>> > If C is a constant.
>>>
>>> Why's that profitable?
>>
>> It enables (x * C1) / (y * C2) -> (x
Internal functions that map directly to an optab can only throw an
exception for -fnon-call-exceptions. This patch handles that in
internal_fn_flags, in a similar way to ATTR_*NOTHROW in builtins.def.
(Functions that don't throw even for flag_non_call_exceptions should be
explicitly marked ECF_NO
This patch adds missing ECF_NOTHROW flags to the vectorisable
integer internal functions. I noticed it while doing some SVE
work but don't have a testcase that's useful now.
Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
Richard
2017-08-17 Richard Sandiford
gcc/
On Tue, Aug 15, 2017 at 4:11 PM, Wilco Dijkstra wrote:
> Richard Biener wrote:
>> > We also change the association of
>> >
>> > x / (y * C) -> (x / C) / y
>> >
>> > If C is a constant.
>>
>> Why's that profitable?
>
> It enables (x * C1) / (y * C2) -> (x * C1/C2) / y for example.
> Also 1/y i
On Thu, Aug 17, 2017 at 10:57 AM, Marek Polacek wrote:
> Ping.
Ok.
Richard.
> On Thu, Aug 10, 2017 at 08:50:44PM +0200, Marek Polacek wrote:
>> My new warning triggered here and said "bitwise comparison always evaluates
>> to
>> true" and I believe it's right: GOVD_WRITTEN = 131072 which is 10
The following patch resulted from PR81827 analysis where I was on
the wrong track initially. It still results in a small speedup
because we avoid useless duplicate work in case nodes are unified
and reduce the number of graph edges.
Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
Ping.
On Thu, Aug 10, 2017 at 08:50:44PM +0200, Marek Polacek wrote:
> My new warning triggered here and said "bitwise comparison always evaluates to
> true" and I believe it's right: GOVD_WRITTEN = 131072 which is 10... in
> binary, so n->value & GOVD_WRITTEN can never be 1; I suppose 0 was m
I am currently testing the following patch to improve bitmap density
during points-to analysis. This improves the testcase in PR81827 from
> /usr/bin/time /space/rguenther/install/gcc-7.2/bin/gfortran -S t.f90 -O
1139.91user 1.99system 19:02.37elapsed 99%CPU (0avgtext+0avgdata
8035048maxreside
I was notifed I broke proper handling of undefined overflow in
multiplicative ops handling. The following resurrects previous
behavior (and adds a testcase).
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
2017-08-17 Richard Biener
* tree-vrp.c (vrp_
67 matches
Mail list logo
