On Tue, 1 May 2018, Jan Hubicka wrote:
> Richard,
> I have updated my patch for incremental linking into LTO bytecode, but it now
> breaks in
> dwarf2out:
> lto1: internal compiler error: in dwarf2out_die_ref_for_decl, at
> dwarf2out.c:5859
> 0x5d8a37 dwarf2out_die_ref_for_decl
> ../../g
On Tue, 1 May 2018, Tom de Vries wrote:
> Hi,
>
> if lto-wrapper can't find mkoffload for an acceleration target it shows this
> error:
> ...
> $ gcc -fopenacc test.c
> lto-wrapper: fatal error: problem with building target image for nvptx-none
>
> compilation terminated.
> /usr/bin/ld: error: l
On Fri, Jan 26, 2018 at 3:16 AM, Martin Sebor wrote:
> PR tree-optimization/83776 - [6/7/8 Regression] missing
> -Warray-bounds indexing past the end of a string literal,
> identified a not-so-recent improvement to constant propagation
> as the reason for GCC no longer being able to detect out-of-
On Tue, May 1, 2018 at 8:34 PM, Richard Sandiford
wrote:
> The problem in this PR was that we didn't consider aliases between
> writes in the same strided group. After tightening the early exit
> we get the expected abs(step) >= 2 versioning check.
>
> Tested on aarch64-linux-gnu and x86_64-linux
On 03/29/2018 11:16 AM, Tom de Vries wrote:
On 03/29/2018 11:11 AM, Tom de Vries wrote:
Hi,
Consider an lto multi-source test-case main.c and foo.c:
..
$ cat main.c
extern int foo (void);
int
main ()
{
return foo () + 1;
}
$ cat foo.c
int __attribute__((noinline, noclone))
foo (void)
{
Hi!
operands[2] is an input operand in these define_insn_and_split patterns,
so when the masking needs to be still performed, we shouldn't modify that
input register; all the patterns are guarded with can_create_pseudo_p,
so we can create new pseudos there.
Bootstrapped/regtested on x86_64-linux
On 04/03/2018 07:49 PM, Bernhard Reutner-Fischer wrote:
This patch adds scan-ltrans-tree-dump.
Please check all error calls to talk about the correct function -- at least
scan-ltrans-tree-dump-times is wrong.
Fixed in this updated patch.
OK for trunk?
Thanks,
- Tom
[testsuite] Add scan-lt
On Wed, May 2, 2018 at 9:42 AM, Jakub Jelinek wrote:
> Hi!
>
> operands[2] is an input operand in these define_insn_and_split patterns,
> so when the masking needs to be still performed, we shouldn't modify that
> input register; all the patterns are guarded with can_create_pseudo_p,
> so we can c
The following fixes into-SSA gimplification for SAVE_EXPRs introduced
late during GIMPLE. In that case we can reasonably expect to have
a situation which gimplifies into sth without control flow so allow
SSA temps for its destination.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied
On Wed, 2 May 2018, Tom de Vries wrote:
> On 03/29/2018 11:16 AM, Tom de Vries wrote:
> > On 03/29/2018 11:11 AM, Tom de Vries wrote:
> > > Hi,
> > >
> > > Consider an lto multi-source test-case main.c and foo.c:
> > > ..
> > > $ cat main.c
> > > extern int foo (void);
> > >
> > > int
> > > main
On Wed, May 02, 2018 at 09:50:07AM +0200, Uros Bizjak wrote:
> > 2018-05-02 Jakub Jelinek
> >
> > PR target/85582
> > * config/i386/i386.md (*ashl3_doubleword_mask,
> > *ashl3_doubleword_mask_1, *3_doubleword_mask,
> > *3_doubleword_mask_1): If and[sq]i3 is needed
On Wed, May 2, 2018 at 10:02 AM, Jakub Jelinek wrote:
> On Wed, May 02, 2018 at 09:50:07AM +0200, Uros Bizjak wrote:
>> > 2018-05-02 Jakub Jelinek
>> >
>> > PR target/85582
>> > * config/i386/i386.md (*ashl3_doubleword_mask,
>> > *ashl3_doubleword_mask_1, *3_doubleword_m
On Mon, 30 Apr 2018, Kyrill Tkachov wrote:
> Hi Richard,
>
>
> On 27/04/18 09:24, Richard Biener wrote:
> > On Thu, 26 Apr 2018, Richard Biener wrote:
> >
> > >
> > > The following makes loop-header copying stop after the first exit test
> > > it copied. The reports rightfully complain about t
On 02/05/18 07:11 +0200, François Dumont wrote:
Hi
Here is the patch I eventually would like to commit.
Tested under Linux x86_64, ok for trunk ?
Ah yes, this is the one I remember thinking would be OK for stage 1.
OK for trunk, thanks.
On Tue, Apr 10, 2018 at 2:52 AM, Kugan Vivekanandarajah
wrote:
> I would like to queue this patch for stage1 review.
>
> In DSE, while in dse_classify_store, as soon as we see a PHI use
> statement that is part of the loop, we are immediately giving up.
>
> As far as I understand, this can be impr
This patch fixes false dependencies for vmovss, vmovsd, vrcpss, vrsqrtss,
vsqrtss and vsqrtsd instructions.
Tested on x86-64/Linux, no new test fails, some SPEC 2006/2017 performance
gains.
Please let me know if something is wrong here and should be changed.
--
Alexander Nesterovskiy
falsedep
On Mon, Apr 30, 2018 at 7:16 PM, Jeff Law wrote:
> On 01/09/2018 02:41 PM, Martin Sebor wrote:
>> I found a few problems in the previous revision:
>>
>> 1) It didn't handle the simple case of member arrays of const
>>struct objects (only member arrays of const arrays of structs
>>were han
>
> Hmm, if that's from ltrans stage then you probably need
>
> Index: gcc/dwarf2out.c
> ===
> --- gcc/dwarf2out.c (revision 259764)
> +++ gcc/dwarf2out.c (working copy)
> @@ -5835,7 +5835,7 @@ dwarf2out_die_ref_for_decl (tre
On Mon, Apr 30, 2018 at 7:41 PM, Kyrill Tkachov
wrote:
> Hi all,
>
> We can improve the performance of complex floating-point multiplications by
> inlining the expansion a bit more aggressively.
> We can inline complex x = a * b as:
> x = (ar*br - ai*bi) + i(ar*bi + br*ai);
> if (isunordered (__r
The following fixes the LTO part of the -f[no-]checking miscompare issue.
I introduce a compare-lto script similar to compare-debug where I strip
the LTO option section and re-compare. I have no easy way to test
the nonplugin path and at least for targets using simple-object to
wrap all LTO secti
On Mon, Apr 30, 2018 at 8:49 PM, Marc Glisse wrote:
> On Fri, 12 Jan 2018, Wilco Dijkstra wrote:
>
>> Hi,
>>
>> Here is the updated version:
>>
>> This patch implements some of the optimizations discussed in
>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71026.
>>
>> Simplify (C / x >= 0.0) into
On Mon, Apr 30, 2018 at 05:05:26PM +, Segher Boessenkool wrote:
> This removes paired single (used on the 750CL and friends). It was
> deprecated in GCC 8. Removing it means we only have one vector model
> to deal with (VMX+VSX, 16-byte vectors).
>
> Tested on powerpc64-linux {-m32,-m64} and
On Wed, May 2, 2018 at 12:44 AM, Jeff Law wrote:
>
> Time to come back to this now that the trunk is open again...
>
>
> Forwarded Message
> Subject: [GCC 9][RFC][PATCH] Optimize PHIs with constant arguments better
> Date: Thu, 30 Nov 2017 16:24:33 -0700
> From: Jeff Law
> To: g
On Wed, 2 May 2018, Tom de Vries wrote:
> On 04/03/2018 07:49 PM, Bernhard Reutner-Fischer wrote:
> > > This patch adds scan-ltrans-tree-dump.
> >
> > Please check all error calls to talk about the correct function -- at least
> > scan-ltrans-tree-dump-times is wrong.
> >
>
> Fixed in this upda
On 01/05/18 21:56 +0200, François Dumont wrote:
Hi
If not told otherwise I'll commit attached patch tomorrow.
Please do not commit it, see below.
Already discussed here:
https://gcc.gnu.org/ml/libstdc++/2017-10/msg00053.html
There's no changelog entry with the patch, so to recap, the chan
The following fixes PR85597 in an easy backportable way, the proper
fix involves refactoring of the vect_get_vec_defs interface.
Bootstrap and regtest running on x86_64-unknown-linux-gnu.
Richard.
2018-05-02 Richard Biener
PR tree-optimization/85597
* tree-vect-stmts.c (vect
Committed to CVS.
Index: htdocs/gcc-8/changes.html
===
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-8/changes.html,v
retrieving revision 1.73
diff -u -r1.73 changes.html
--- htdocs/gcc-8/changes.html 2 May 2018 10:21:15 - 1.73
+++ htdo
On Wed, May 02, 2018 at 10:09:22AM +0200, Uros Bizjak wrote:
> On Wed, May 2, 2018 at 10:02 AM, Jakub Jelinek wrote:
> > On Wed, May 02, 2018 at 09:50:07AM +0200, Uros Bizjak wrote:
> >> > 2018-05-02 Jakub Jelinek
> >> >
> >> > PR target/85582
> >> > * config/i386/i386.md (*ashl
Hi Kelvin,
On Tue, Apr 24, 2018 at 09:12:52AM -0500, Kelvin Nilsen wrote:
> * doc/extend.texi: Tidy documentation of PowerPC built-in functions.
You should note the sections you renamed:
(PowerPC Built-in Functions): Rename to ...
(Low-Level PowerPC Built-in Functions): ... th
Hi,
On Tue, 1 May 2018, Jeff Law wrote:
> The very nature of a traditional asm implies that it can read or write
> anything visible to compiler. We can't realistically peek inside to see
> what's happening and the user hasn't provided appropriate dataflow
> information. One could make the ar
On 05/02/2018 07:45 AM, Michael Matz wrote:
> Hi,
>
> On Tue, 1 May 2018, Jeff Law wrote:
>
>> The very nature of a traditional asm implies that it can read or write
>> anything visible to compiler. We can't realistically peek inside to see
>> what's happening and the user hasn't provided appr
Hi!
On Wed, Apr 25, 2018 at 11:28:44AM -0500, Kelvin Nilsen wrote:
> >>> 4. Remove descriptions of built-in function that do not belong in this
> >>> section because the
> >>>built-in functions are generic (not specific to PowerPC):
> >>> __builtin_fabsq,
> >>>__builtin_copysignq, __builti
Hi Richard,
On 02/05/18 11:15, Richard Biener wrote:
> On Mon, Apr 30, 2018 at 7:41 PM, Kyrill Tkachov
> wrote:
>> Hi all,
>>
>> We can improve the performance of complex floating-point multiplications by
>> inlining the expansion a bit more aggressively.
>> We can inline complex x = a * b as:
On 05/02/2018 03:27 AM, Richard Biener wrote:
> On Tue, Apr 10, 2018 at 2:52 AM, Kugan Vivekanandarajah
> wrote:
>> I would like to queue this patch for stage1 review.
>>
>> In DSE, while in dse_classify_store, as soon as we see a PHI use
>> statement that is part of the loop, we are immediately g
In libstdc++ the deprecated char* streams are non-copyable, as was
required pre-C++11.
Since C++11 the standard implies that those streams should be copyable,
but doesn't specify the effects of copying them. This is surely a
defect, so for consistency with other implementations this change makes
On 02/05/18 17:25 +0100, Jonathan Wakely wrote:
+int
+main()
+{
+ test01();
+ test02();
+ test03();
+ test04();
+ test05();
+ test05();
After committing this I noticed test05() is run twice. Fixed by this
patch.
commit 81d3cafc6bd00d44413adca3420f75c52870b1eb
Author: Jonathan Wakely
Dat
On 06/04/18 11:23 +0100, Jonathan Wakely wrote:
On 6 April 2018 at 09:50, Marc Glisse wrote:
On Fri, 6 Apr 2018, Jonathan Wakely wrote:
This attempts to solve some of the problems when mixing std::valarray
operations and 'auto', by storing nested closure objects as values
instead of references
Richard Biener wrote:
> why use BUILT_IN_ISUNORDERED but not a GIMPLE_COND with
> UNORDERED_EXPR? Note again that might trap/throw with -fsignalling-nans
> so better avoid this transform for flag_signalling_nans as well...
Both currently trap on signalling NaNs due to the implementation of the C
Hello, gentle maintainer.
This is a message from the Translation Project robot.
A revised PO file for textual domain 'gcc' has been submitted
by the Spanish team of translators. The file is available at:
http://translationproject.org/latest/gcc/es.po
(This file, 'gcc-8.1-b20180401.es.po',
On 02/16/2018 05:43 AM, Nick Clifton wrote:
> Hi David,
>
> Attached is a revised version of the patch which I hope addresses all
> of your (very helpful) comments on the v3 patch.
>
> OK to apply once the sources are back on stage 1 ?
>
> Cheers
> Nick
>
> gcc/ChangeLog
> 2018-02-09
On Wed, May 02, 2018 at 11:00:20AM -0600, Jeff Law wrote:
> + escaped = (char *) xmalloc (len * 2 + 1);
> I believe that generally we want a slower growth pattern. You'll find
> that we regularly do something like len * 3 / 2 + 1. Seems wise here too.
We don't want to use xmalloc in
We have an unused overload of _Rb_tree::erase which takes a pair of
pointers (not iterators, pointers) to keys and calls erase for every
key in the range.
It's not actually doing any harm, as the function is only on the
internal _Rb_tree type. It's not exposed as a member of std::map,
std::set e
On 03/13/2018 03:56 PM, dave.pa...@oracle.com wrote:
> This patch fixes ICE when statement expression used in old-style
> function declaration parameter list. See
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=30552
>
> In c_parser_postfix_expression (), after seeing a left paren followed by
>
On 03/14/2018 11:40 AM, Sudakshina Das wrote:
> Hi
>
> This patch is another partial fix for PR 84521. This is adding a
> definition to one of the target hooks used in the SJLJ implemetation so
> that AArch64 defines the hard_frame_pointer_rtx as the
> TARGET_BUILTIN_SETJMP_FRAME_VALUE. As pointed
Hi all, Jason,
On 24/04/2018 01:00, Paolo Carlini wrote:
On 23/04/2018 23:43, Jason Merrill wrote:
On Mon, Apr 23, 2018 at 5:17 PM, Paolo Carlini
wrote:
Hi,
this issue is by and large fixed in trunk, thus I wanted to resolve
it as
such, marking it as fixed for 8.1.0. However the text of th
On May 2, 2018 6:17:50 PM GMT+02:00, Jeff Law wrote:
>On 05/02/2018 03:27 AM, Richard Biener wrote:
>> On Tue, Apr 10, 2018 at 2:52 AM, Kugan Vivekanandarajah
>> wrote:
>>> I would like to queue this patch for stage1 review.
>>>
>>> In DSE, while in dse_classify_store, as soon as we see a PHI use
On 05/02/2018 11:36 AM, Richard Biener wrote:
> On May 2, 2018 6:17:50 PM GMT+02:00, Jeff Law wrote:
>> On 05/02/2018 03:27 AM, Richard Biener wrote:
>>> On Tue, Apr 10, 2018 at 2:52 AM, Kugan Vivekanandarajah
>>> wrote:
I would like to queue this patch for stage1 review.
In DSE, w
OK.
On Wed, May 2, 2018 at 1:33 PM, Paolo Carlini wrote:
> Hi all, Jason,
>
> On 24/04/2018 01:00, Paolo Carlini wrote:
>>
>> On 23/04/2018 23:43, Jason Merrill wrote:
>>>
>>> On Mon, Apr 23, 2018 at 5:17 PM, Paolo Carlini
>>> wrote:
Hi,
this issue is by and large fixed in tr
The patch below includes a description of a new feature in gfortran for
inclusion in the GCC 8 changes.html file. I don’t have commit rights. Could
someone approve and apply this? Thanks.
Damian
cvs diff: Diffing .
Index: changes.html
The suggested resolution of LWG 3083 is to make invalid indices
undefined, but we can fairly easily check for them and treat them as
errors in the same way as allocation failure. This avoids a segfault or
worse, setting an error flag on the stream instead.
PR libstdc++/68197
* inc
This is a new version of a patch I sent out last year to stop gcc from
trying to do a link when creating precompiled headers and a linker
flag is also given.
When I build and test GCC I also build glibc and then I run the GCC tests
with -Wl,-rpath and -Wl,--dynamic-linker so that I don't have to i
Libgomp's usage of struct gomp_coalesce_buf is a little confusing. The
member chunks is an array where the even elements represent the starting
address of the chunk cache line and the odd elements represent the
corresponding ending addresses. This patch clarifies the usage of the
chunks member by i
[ was: [PATCH, 0/2] Add scan-ltrans-tree-dump and scan-wpa-ipa-dump ]
Likewise, this patch adds scan-offload-tree-dump, and uses it in
testsuite/libgomp.oacc-c/vec.c, fixing the todo there.
[ The discussion on how to address to todo started here (
https://gcc.gnu.org/ml/gcc-patches/2017-06/m
This patch by Than McIntosh breaks the dependence of go-unwind.c on
unwind-pe.h, by adding the required definitions and code directly to
go-unwind.c. go-unwind.c still depends on the public unwind.h API.
This makes it easier to build libgo separately. Bootstrapped and ran
Go testsuite on x86_64-p
This patch by Than McIntosh refactors the lists of libgo package to
live in separate files, and reworks generation of runtime.inc to live
in a separate shell file. Bootstrapped and ran Go testsuite on
x86_64-pc-linux-gnu. Committed to mainline.
Ian
Index: libgo/Makefile.am
==
The file libgo/go/runtime/stack.go was removed from the upstream
gofrontend repository back in April, 2017. Somehow it was not removed
from GCC. I've now committed a change to remove it. Bootstrapped and
ran Go testsuite on x86_64-pc-linux-gnu.
Ian
Index: libgo/go/runtime/stack.go
=
This improves the code for a switch statement on targets that sign-extend
function arguments, such as RISC-V. Given a simple testcase
extern void asdf(int);
void foo(int x) {
switch (x) {
case 0: asdf(10); break;
case 1: asdf(11); break;
case 2: asdf(12); break;
case 3: asdf(13); break;
In the RISC-V backend, there is code that does a zero-extend by default for
QImode compare arguments. This is because zero-extend is a single and insn,
whereas sign-extend is two shifts. However, if we have two values that are
already sign extended, or a sign extended value and a CONST_INT, then
This patch by Tobias Klauser adds libgo support for NIOS II.
Bootstrapped and ran Go testsuite on x86_64-pc-linux-gnu. Committed
to mainline.
Ian
Index: gcc/go/gofrontend/MERGE
===
--- gcc/go/gofrontend/MERGE (revision 259805)
++
This libgo patch by Than McIntosh adds explicit references to various
types and constants defined by the header files included by sysinfo.c
(used to drive the generation of gen-sysinfo.go as part of the libgo
build via the GCC "-fdump-go-spec" option).
The intent is to enable clients to gather the
Hello
On 05/01/2018 06:32 AM, Magnus Granberg wrote:
> New patch
> libgcc/ChangeLog:
>
> 2018-05-01 Magnus Granberg
>
> * config/i386/resms64.h: Add .note.GNU-stack section
> * config/i386/resms64f.h: Likewise.
> * config/i386/resms64fx.h: Likewise.
> * config/i386/resms
torsdag 3 maj 2018 kl. 01:07:51 CEST skrev Daniel Santos:
> Hello
>
> On 05/01/2018 06:32 AM, Magnus Granberg wrote:
> > New patch
> > libgcc/ChangeLog:
> >
> > 2018-05-01 Magnus Granberg
> >
> > * config/i386/resms64.h: Add .note.GNU-stack section
> > * config/i386/resms64f.h: Likew
On 05/02/2018 06:17 PM, Magnus Granberg wrote:
> torsdag 3 maj 2018 kl. 01:07:51 CEST skrev Daniel Santos:
>> Hello
>>
>> On 05/01/2018 06:32 AM, Magnus Granberg wrote:
>>> New patch
>>> libgcc/ChangeLog:
>>>
>>> 2018-05-01 Magnus Granberg
>>>
>>> * config/i386/resms64.h: Add .note.GNU-sta
In libgo, we earlier disabled tests of the go tool that required
package build IDs, because we didn't have them. But now we do, so
enable the tests. Bootstrapped and ran Go testsuite on
x86_64-pc-linux-gnu. Committed to mainline.
Ian
Index: gcc/go/gofrontend/MERGE
==
64 matches
Mail list logo
