Hi,
On ThunderX, load (and store) pair that does a pair of two word
(32bits) load/stores is slower in some cases than doing two
load/stores. For some internal benchmarks, it provides a 2-5%
improvement.
This patch disables the forming of the load/store pairs for SImode if
we are tuning for Thun
On Fri, Aug 05, 2016 at 07:56:02AM +0200, Daniel Krügler wrote:
> 2016-08-05 6:49 GMT+02:00 Gleb Natapov :
> > Instead of throwing an exception allocate its memory and initialize it
> > explicitly. Makes std::make_exception_ptr more efficient since no stack
> > unwinding is needed.
>
> [..]
>
> >
On 08/04/2016 06:52 PM, Nathan Sidwell wrote:
> On 08/04/16 12:10, Martin Liška wrote:
>> On 08/04/2016 05:13 PM, Nathan Sidwell wrote:
>>> On 08/04/16 10:42, Martin Liška wrote:
>>>
I decided to use a new enum, hope it's better?
>>>
>>> that's fine. But you know, if you set the enum values a
On 08/05/2016 09:30 AM, Martin Liška wrote:
> Hi.
>
> Sorry for the mistake with the enum, that was silly ;)
> New patch version also handles the unnecessary braces.
>
> Martin
I attached a wrong patch, sending the new one.
Martin
>From 6a811bf138aa29e6b821de21de61f77afec5a0d6 Mon Sep 17 00:00:
There is no need to avoid threading to a loop header, the threading
code can cope with this just fine. Noticed when working on PR72772.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
2016-08-05 Richard Biener
* tree-ssa-threadupdate.c (thread_block_
On Thu, Aug 4, 2016 at 5:19 PM, Aldy Hernandez wrote:
> On 08/04/2016 08:57 AM, Richard Biener wrote:
>>
>> On Thu, Aug 4, 2016 at 1:30 PM, Aldy Hernandez wrote:
>>>
>>> Howdy!
>>>
>>> As part of my -Walloca-larger-than=life work, I've been running said pass
>>> over gcc, binutils, and gdb, and t
On Fri, 5 Aug 2016, Prathamesh Kulkarni wrote:
> On 4 August 2016 at 12:39, Richard Biener wrote:
> > On Thu, 4 Aug 2016, Prathamesh Kulkarni wrote:
> >
> >> Hi,
> >> The attached patch fixes pr22051-2.c which regressed due to
> >> r238754. Matthew, could you please confirm if this patch fixes th
On Fri, 5 Aug 2016, Prathamesh Kulkarni wrote:
> Ah, the mail failed to be delivered to gcc-patches, sorry for the double-post.
> On 5 August 2016 at 01:26, Prathamesh Kulkarni
> wrote:
> > Hi,
> > Is the attached patch OK ?
> > Since we want to extend based on signop, I removed ORing with wi::ma
On Fri, 5 Aug 2016, Prathamesh Kulkarni wrote:
> Hi,
> This patch makes streamer_read_wi and streamer_write_wi non-static,
> and exports them from lto-streamer.h. I suppose this hunk could be committed
> independently of the ipa-bitwise-cp patch ?
> Bootstrap+test in progress on x86_64-unknown-lin
On Thu, Aug 4, 2016 at 6:48 PM, Ian Lance Taylor wrote:
> On Thu, Aug 4, 2016 at 1:11 AM, Uros Bizjak wrote:
>> On Thu, Aug 4, 2016 at 12:53 AM, Ian Lance Taylor wrote:
>>> On Thu, Jul 28, 2016 at 4:24 AM, Uros Bizjak wrote:
A new testsuite failure is introduced:
FAIL: text/
If edges are swapped so are PHI args, the following adjusts the pattern
for this case.
Committed.
Richard.
2016-08-05 Richard Biener
* gcc.dg/tree-ssa/ivopt_5.c: Make robust against edge swapping.
Index: gcc/testsuite/gcc.dg/tree-ssa/ivopt_5.c
==
On 08/04/2016 07:03 PM, Nathan Sidwell wrote:
> On 08/04/16 12:43, Nathan Sidwell wrote:
>
>> How about:
>> gcov_t expected;
>> atomic_load (&counter[0], val, ...);
>> gcov_t delta = val == value ? 1 : -1;
>> atomic_add (&counter[1], delta); <-- or atomic_add_fetch
>> if (delta < 0) {
>> /* c
This fixes a bug in the _Callable SFINAE constraint of std::function,
where it was using an rvalue _Func in the result_of expression, when
the target is actually invoked as an lvalue.
I'm also making std::function use std::__invoke() directly, instead of
going via __callable_functor() and/or mem_
Hi Richard and Prathamesh,
On 05/08/16 18:23, Richard Biener wrote:
On Fri, 5 Aug 2016, Prathamesh Kulkarni wrote:
Hi,
This patch makes streamer_read_wi and streamer_write_wi non-static,
and exports them from lto-streamer.h. I suppose this hunk could be committed
independently of the ipa-bitwi
On 27/07/16 18:06 -0400, Jason Merrill wrote:
On Wed, Jul 27, 2016 at 8:05 AM, Jonathan Wakely wrote:
Consider:
template T declval();
int& r1 = declval();
int&& r2 = declval();
int& r3 = declval();
This produces three quite different errors:
refs.cc:3:25: error: invalid initialization of
Irfan,
Thanks for the case sharing. It is a persuasive reason not to error out
-mno-SPB.
Nathan's patch changes default behaviour that -mSPB will be implied when
-mno-PDITR is provided. So with this patch your project need to explicitly
specify -mno-SPB to make it work as before. IMHO default
On 04/08/16 22:16, Bernd Edlinger wrote:
> Hi,
>
> this patch introduces a new target hook that allows the target's
> INITIAL_ELIMINATION_OFFSET function to use cached values instead of
> re-computing the frame layout every time.
>
> I have updated the documentation a bit and hope it is clearer
On Tue, Jul 26, 2016 at 2:55 PM, James Greenhalgh
wrote:
>
> OK? As this is an ABI break, I'm not proposing for it to go back to GCC 6,
> though it will apply cleanly there if the maintainers support that.
>
What do you mean by "ABI break"? AFAICS, with this patch, it conforms to
AAPCS.
The sub
On 04/08/16 16:56, Richard Earnshaw (lists) wrote:
> On 04/08/16 12:06, Wilco Dijkstra wrote:
>> Improve stack adjustment by reusing a temporary move immediate
>> from the epilog if the register is still valid in the epilog. This generates
>> smaller code for leaf functions:
>>
>> mov
On Fri, Aug 05, 2016 at 11:00:39AM +0100, Yao Qi wrote:
> On Tue, Jul 26, 2016 at 2:55 PM, James Greenhalgh
> wrote:
> >
> > OK? As this is an ABI break, I'm not proposing for it to go back to GCC 6,
> > though it will apply cleanly there if the maintainers support that.
> >
>
> What do you mean
On 2016.08.03 at 15:47 +0200, Richard Biener wrote:
> On Wed, Aug 3, 2016 at 6:00 AM, Patrick Palka wrote:
> > VRP currently has functionality to eliminate case labels that lie
> > completely outside of the switch operand's value range. This patch
> > complements this functionality by teaching VR
This avoids regressing gcc.dg/tree-ssa/pr21417.c with the fix for
PR72772 where after it a forwarder block no longer is present.
It's easy to simply create it when FSM threading faces the situation
that the edge ending the path enters a loop.
I also fixed the costs for obviously related anon SSA
This fixes PR72772 by avoing placing a degenerate PHI in each
forwarder block loop init creates when creating simple preheaders.
The solution is to simply split the single loop entry edge which
is also way cheaper than using the forwarder block creation path.
You've seen a load of fallout fixes a
The following patch avoids overaligned types created from IPA parameter
replacement. It is said to help mipsel which still suffers from the
backend-looks-at-type-alignment-for-parameter-passing-ABI bug.
Bootstrap and regtest pending on x86_64-unknown-linux-gnu.
mips testing appreciated.
Richar
On 05/08/16 01:37 +0100, Jonathan Wakely wrote:
The core of the new guidelines for writing tests is this:
[begin]
Test cases that use features of a particular C++ standard should
specify the minimum required standard as an effective target:
// { dg-do run { target c++11 } }
or
// { dg-req
> Hi Honza,
>
> On 04/08/16 23:05, Jan Hubicka wrote:
> >>I didn't look at the propagation part but eventually the IPA-CP
> >>lattice gets quite big. Also the alignment lattice is very
> >>similar to the bits lattice so why not merge those two? But
> >
> >This was always the original idea to rep
On 08/05/16 00:56, Irfan Ahmad wrote:
1. In the absence of lazy binding (that is almost always the case in embedded
systems), GOT is practically read-only - it needs to be modified only during
'practically' as in 'almost', or is it really read only?
linking by the dynamic linker, after that
Hello!
Attached patch was inspired by assembly from PR 72805 testcase.
Currently, the compiler generates:
test:
vpternlogd $0xFF, %zmm0, %zmm0, %zmm0
vpxord %zmm1, %zmm1, %zmm1
vpcmpd $1, %zmm1, %zmm0, %k1
kmovw %k1, %eax
movzwl %ax, %eax
Hello.
This corrects a record field name where comment does not correspond to it's
name.
make check -k RUNTESTFLAGS="gcov.exp" works fine.
Ready to install?
Martin
>From 1733e26c619e65ef95604f04f026be0006e9c4e0 Mon Sep 17 00:00:00 2001
From: marxin
Date: Fri, 5 Aug 2016 14:18:47 +0200
Subject:
I've committed this patch to fix a failed assertion. unify's (pt.c) default
case asserts EXPR_P(parm), but TRAIT_EXPR is tcc_exceptional, so fails. However
from unify's POV it's no more exceptional than, say, SIZEOF_EXPR, so should be
allowed at that point.
nathan
2016-08-05 Nathan Sidwell
On 08/05/16 03:32, Martin Liška wrote:
On 08/05/2016 09:30 AM, Martin Liška wrote:
Hi.
Sorry for the mistake with the enum, that was silly ;)
New patch version also handles the unnecessary braces.
Martin
I attached a wrong patch, sending the new one.
This one looks good, thanks!
nathan
Hi,
generally speaking, the ipa-cp.c and ipa-cp.[hc] bits look reasonable,
but I have a few comments:
On Thu, Aug 04, 2016 at 12:06:18PM +0530, Prathamesh Kulkarni wrote:
> Hi,
> This is a prototype patch for propagating known/unknown bits
> inter-procedurally.
> for integral types which propaga
On 08/05/16 04:55, Martin Liška wrote:
Thank you for very intensive brainstorming ;) Well I still believe that
following code
is not thread safe, let's image following situation:
yeah, you're right.
we could do better by using compare_exchange storing value, and detect the race
I mentione
On 08/05/16 08:22, Martin Liška wrote:
Hello.
This corrects a record field name where comment does not correspond to it's
name.
make check -k RUNTESTFLAGS="gcov.exp" works fine.
Ready to install?
heh, 'line_next' is perfectly good name for the next function. No, wait, it's
stupid!
your p
On 08/05/2016 02:38 PM, Nathan Sidwell wrote:
> On 08/05/16 04:55, Martin Liška wrote:
>
>> Thank you for very intensive brainstorming ;) Well I still believe that
>> following code
>> is not thread safe, let's image following situation:
>>
>
> yeah, you're right.
>
>>> we could do better by us
On 08/05/16 11:29, Richard Earnshaw (lists) wrote:
> On 04/08/16 22:16, Bernd Edlinger wrote:
>> Hi,
>>
>> this patch introduces a new target hook that allows the target's
>> INITIAL_ELIMINATION_OFFSET function to use cached values instead of
>> re-computing the frame layout every time.
>>
>> I hav
Sometimes the case labels of a switch have a different type than the
switch operand, e.g. in the test case below, the case labels have type t
and the switch operand has type int. Because the verifier expects that
case labels of a switch each have the exact same type, it's important to
avoid changin
On 08/05/16 08:48, Martin Liška wrote:
Ok, after all the experimenting and inventing "almost" thread-safe code, I
incline to not to include __gcov_one_value_profiler_body_atomic
counter. The final solution is cumbersome and probably does not worth doing.
Moreover, even having a thread-safe imp
Richard,
Here is updated patch which implements your proposal - I pass loop
instead of stmt to determine either REF is defined inside LOOP nest or
not. I checked that for pr70729-nest.cc the reference this->S_n for
statements which are out of innermost loop are not considered as
independent as y
On 08/04/16 22:27, Jeff Law wrote:
> On 07/21/2016 10:29 AM, Bernd Edlinger wrote:
But I think we have a use case where "X" means really more possible
registers (i.e. includes ss2, mmx etc.) than "g" (only general
registers). Otherwise, in the test cases of pr59155 we would not
Hi Joost,
I am not familiar with fortran code.
Maybe Thomas can do something in his new patch?
Regards,
Renlin
On 28/07/16 12:34, VandeVondele Joost wrote:
Thanks.. I wonder if you could add the testcase in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71961#c11
to the testsuite, as it catch
On Fri, 5 Aug 2016, Richard Biener wrote:
>
> This avoids regressing gcc.dg/tree-ssa/pr21417.c with the fix for
> PR72772 where after it a forwarder block no longer is present.
> It's easy to simply create it when FSM threading faces the situation
> that the edge ending the path enters a loop.
>
On Fri, 5 Aug 2016, Richard Biener wrote:
> On Fri, 5 Aug 2016, Richard Biener wrote:
>
> >
> > This avoids regressing gcc.dg/tree-ssa/pr21417.c with the fix for
> > PR72772 where after it a forwarder block no longer is present.
> > It's easy to simply create it when FSM threading faces the situ
On Fri, Aug 5, 2016 at 3:08 PM, Patrick Palka wrote:
> Sometimes the case labels of a switch have a different type than the
> switch operand, e.g. in the test case below, the case labels have type t
> and the switch operand has type int. Because the verifier expects that
> case labels of a switch
On 08/05/2016 03:14 PM, Nathan Sidwell wrote:
> On 08/05/16 08:48, Martin Liška wrote:
>
>> Ok, after all the experimenting and inventing "almost" thread-safe code, I
>> incline to not to include __gcov_one_value_profiler_body_atomic
>> counter. The final solution is cumbersome and probably does
On Fri, Aug 5, 2016 at 3:28 PM, Yuri Rumyantsev wrote:
> Richard,
>
> Here is updated patch which implements your proposal - I pass loop
> instead of stmt to determine either REF is defined inside LOOP nest or
> not. I checked that for pr70729-nest.cc the reference this->S_n for
> statements whic
* include/std/chrono (floor, ceil, round, abs): New for C++17.
* testsuite/20_util/duration_cast/rounding.cc: New test.
* testsuite/20_util/time_point_cast/rounding.cc: New test.
* doc/xml/manual/status_cxx2017.xml: Update status table.
* doc/html/manual/sta
On 04/08/16 13:33 +0100, Jonathan Wakely wrote:
On 03/08/16 20:11 +0100, Jonathan Wakely wrote:
Another feature we already support, so just define the macro.
* include/bits/shared_ptr_base.h (__cpp_lib_enable_shared_from_this):
Define feature-test macro.
* testsuite/20_u
On 05/08/16 13:49, Bernd Edlinger wrote:
> On 08/05/16 11:29, Richard Earnshaw (lists) wrote:
>> On 04/08/16 22:16, Bernd Edlinger wrote:
>>> Hi,
>>>
>>> this patch introduces a new target hook that allows the target's
>>> INITIAL_ELIMINATION_OFFSET function to use cached values instead of
>>> re-c
On Fri, Aug 05, 2016 at 11:15:24AM +0100, James Greenhalgh wrote:
> On Fri, Aug 05, 2016 at 11:00:39AM +0100, Yao Qi wrote:
> > On Tue, Jul 26, 2016 at 2:55 PM, James Greenhalgh
> > wrote:
> > >
> > > OK? As this is an ABI break, I'm not proposing for it to go back to GCC 6,
> > > though it will a
On 05/08/16 15:17, James Greenhalgh wrote:
> On Fri, Aug 05, 2016 at 11:15:24AM +0100, James Greenhalgh wrote:
>> On Fri, Aug 05, 2016 at 11:00:39AM +0100, Yao Qi wrote:
>>> On Tue, Jul 26, 2016 at 2:55 PM, James Greenhalgh
>>> wrote:
OK? As this is an ABI break, I'm not proposing for it
On 01/08/16 15:31, Kyrill Tkachov wrote:
On 11/07/16 18:55, Bernd Schmidt wrote:
On 07/11/2016 04:52 PM, Kyrill Tkachov wrote:
Based on that, I think that code block is a useful optimisation, we just
need
to take care with immediates.
What do you think?
Yeah, I think the patch is ok.
Th
I think only thing you're missing is that we probably don't want to be
using alloca in new code anyway. And if we're going through the trouble
of fixing old code, we ought to just remove alloca. So building up this
kind of class is probably taking folks in the wrong direction.
After spending
On 26/07/16 14:55, James Greenhalgh wrote:
>
> Hi,
>
> It looks like we've not been handling structures of 16-bit floating-point
> data correctly for AArch64. For some reason we end up passing them
> packed in to integer registers. That is to say, on trunk and GCC 6, for:
>
> struct x {
>
I was surprised by the always_inline trick. I suppose it makes
sense but I wouldn't have expected to be able to rely on it. Out
of curiosity I tested it with other compilers. It turns out that
it only works with some like GCC and IBM XLC++, but not others
like Clang or Sun CC. In recursive c
On Thu, Aug 4, 2016 at 1:48 PM, Richard Biener
wrote:
> On Thu, Aug 4, 2016 at 10:40 AM, Bin.Cheng wrote:
>> On Wed, Aug 3, 2016 at 11:17 PM, Jeff Law wrote:
>>> On 08/03/2016 10:35 AM, Bin Cheng wrote:
Hi,
When I introduced parameter STOP for expand_simple_operations, I also
On 08/05/2016 03:37 PM, Aldy Hernandez wrote:
> My question is, how wed are we to an alloca based approach? Can we
> replace the non-critical uses by std::string (diagnostics,
> BLAH_check_failed(), building a string to manipulate the output of
> env("SOME_VAR"), etc?).
My suggestion too.
( I h
On 08/05/16 16:06, Richard Earnshaw (lists) wrote:
>
> Thanks, that's pretty much what I expected would be the case.
>
> Could I suggest:
>
> This target hook is called once each time the frame layout needs to be
> recalculated. The calculations can be cached by the target and can then
> be used b
On 08/04/2016 04:49 PM, Thomas Schwinge wrote:
Global Reviewers are welcome to review OpenACC/OpenMP/offloading patches.
But that doesn't help if that's then not happening in reality. (With the
exception of Bernd, who then did review such patches for a while, but
also seems to have stopped with
On Fri, Aug 5, 2016 at 12:32 AM, Martin Liška wrote:
> On 08/05/2016 09:30 AM, Martin Liška wrote:
>> Hi.
>>
>> Sorry for the mistake with the enum, that was silly ;)
>> New patch version also handles the unnecessary braces.
>>
>> Martin
>
> I attached a wrong patch, sending the new one.
This br
On Wed, Aug 3, 2016 at 8:36 AM, Jonathan Wakely wrote:
> On 03/08/16 15:02 +0300, Gleb Natapov wrote:
>>
>> On Wed, Aug 03, 2016 at 12:47:30PM +0100, Jonathan Wakely wrote:
>>> Huh. If I'm reading the ABI spec correctly, we should terminate if the
>>> copy constructor throws.
>>>
>> I'll make it t
On Fri, Aug 5, 2016 at 8:24 AM, Nathan Sidwell wrote:
> I've committed this patch to fix a failed assertion. unify's (pt.c) default
> case asserts EXPR_P(parm), but TRAIT_EXPR is tcc_exceptional, so fails.
> However from unify's POV it's no more exceptional than, say, SIZEOF_EXPR, so
> should be
OK.
On Fri, Aug 5, 2016 at 5:08 AM, Jonathan Wakely wrote:
> On 27/07/16 18:06 -0400, Jason Merrill wrote:
>>
>> On Wed, Jul 27, 2016 at 8:05 AM, Jonathan Wakely
>> wrote:
>>>
>>> Consider:
>>>
>>> template T declval();
>>>
>>> int& r1 = declval();
>>> int&& r2 = declval();
>>> int& r3 = declval
On 08/05/2016 08:37 AM, Aldy Hernandez wrote:
After looking at all this code, I realize no size will fit all if we
want something clean. So, I take my approach back. I think we're best
served by a couple different approaches, depending on usage.
Yup.
My question is, how wed are we to an al
Andrew Pinski writes:
> On Mon, Jul 27, 2015 at 3:36 AM, James Greenhalgh
> wrote:
>> On Mon, Jul 27, 2015 at 10:52:58AM +0100, pins...@gmail.com wrote:
>>> > On Jul 27, 2015, at 2:26 AM, Jiong Wang wrote:
>>> >
>>> > Andrew Pinski writes:
>>> >
>>> >>> On Fri, Jul 24, 2015 at 2:07 AM, Jiong Wa
Ping.
On 28/07/2016 17:36, Adhemerval Zanella wrote:
> From: Adhemerval Zanella
>
> This patch adds the split-stack support on aarch64 (PR #67877). As for
> other ports this patch should be used along with glibc and gold support.
>
> The support is done similar to other architectures: a __priv
On Fri, 5 Aug 2016, Andrew Pinski wrote:
>> I attached a wrong patch, sending the new one.
> This broke cross aarch64-elf-gnu building.
It also broke native i586-unknown-freebsd10.3 (which features
clang as the system compiler).
/scratch/tmp/gerald/gcc-HEAD/gcc/gcov.c:468:25: error: use of undec
On Sat, Jul 23, 2016 at 10:30 AM, Alexandre Oliva wrote:
> We used to emit, in debug information, the values bound to pointer to
> member function template parameters only when they were NULL or
> virtual member functions, because those can be represented with
> DW_AT_const_value.
>
> In order to
On August 5, 2016 6:23:27 PM GMT+02:00, Jeff Law wrote:
>On 08/05/2016 08:37 AM, Aldy Hernandez wrote:
>>
>> After looking at all this code, I realize no size will fit all if we
>> want something clean. So, I take my approach back. I think we're
>best
>> served by a couple different approaches,
On August 5, 2016 4:42:48 PM GMT+02:00, Aldy Hernandez wrote:
>
>> I was surprised by the always_inline trick. I suppose it makes
>> sense but I wouldn't have expected to be able to rely on it. Out
>> of curiosity I tested it with other compilers. It turns out that
>> it only works with some li
On Fri, 2016-08-05 at 19:55 +0200, Richard Biener wrote:
>
> Please don't use std::string. For string building you can use
> obstacks.
>
Just out of curiosity ... why? I remember there was some discussion
about it, what was the conclusion? Is that now a general rule or does
it depend on the
On Fri, 8 Jul 2016, Sameera Deshpande wrote:
> I have created new branch unified-autovect based on ToT.
>
> Please find attached the patch adding information about new branch
> "unified-autovect" in the documentation. Is it ok to commit?
Yes, please!
(Once a branch exists, you don't need to ask
On Aug 5, 2016, Jason Merrill wrote:
> On Sat, Jul 23, 2016 at 10:30 AM, Alexandre Oliva wrote:
>> With this patch, when we can't emit a DW_AT_const_value, we emit each
>> "member" of the pointer to member function "record" as a
>> DW_OP_stack_value DW_OP_piece, as long as the referenced member
On 08/05/16 12:06, Jason Merrill wrote:
On Fri, Aug 5, 2016 at 8:24 AM, Nathan Sidwell wrote:
I've committed this patch to fix a failed assertion. unify's (pt.c) default
case asserts EXPR_P(parm), but TRAIT_EXPR is tcc_exceptional, so fails.
However from unify's POV it's no more exceptional th
On August 5, 2016 8:15:54 PM GMT+02:00, Oleg Endo wrote:
>On Fri, 2016-08-05 at 19:55 +0200, Richard Biener wrote:
>
>>
>> Please don't use std::string. For string building you can use
>> obstacks.
>>
>
>Just out of curiosity ... why? I remember there was some discussion
>about it, what was th
PR 72812 points out that Go can generate non-ASCII characters in
assembly code. This is a consequence of the fact that Go permits
identifiers to contain non-ASCII Unicode code points. The GNU
assembler doesn't seem to mind, but the Solaris assembler does.
This patch changes the GCC Go interface
On 08/04/2016 08:32 AM, Renlin Li wrote:
Hi all,
In the case of PR64971
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64971),
the compiler ICE when compiling gcc.c-torture/compile/pr37433.c with
ilp32 abi.
As we know, in aarch64 ilp32, the ptr_mode is SImode while Pmode is
still DImode. It mean
On Wed, Aug 03, 2016 at 12:02:54AM +0930, Alan Modra wrote:
> This is a hack I tried to avoid having to poke at lra code for
> pr71680..
>
> The idea of adding force_reg here was that it removes subregs from
> fix_trunc, emitting the subreg mode conversion in a separate set insn.
> That avoids the
On Fri, Aug 05, 2016 at 08:27:39AM -0700, Andrew Pinski wrote:
> On Fri, Aug 5, 2016 at 12:32 AM, Martin Liška wrote:
> > On 08/05/2016 09:30 AM, Martin Liška wrote:
> >> Hi.
> >>
> >> Sorry for the mistake with the enum, that was silly ;)
> >> New patch version also handles the unnecessary braces
On 08/05/2016 01:45 AM, Richard Biener wrote:
There is no need to avoid threading to a loop header, the threading
code can cope with this just fine. Noticed when working on PR72772.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
2016-08-05 Richard Biener
On 08/05/2016 12:18 AM, Andrew Pinski wrote:
This patch disables the forming of the load/store pairs for SImode if
we are tuning for ThunderX. I used the tuning flags route so it can
be overridden if needed later on or if someone else wants to use the
same method for their core.
+ if (mode ==
On 07/28/2016 09:46 PM, Patrick Palka wrote:
This patch improves the forward jump threader's ability to thread
GIMPLE_SWITCHes by making the VRP simplification callback attempt to
determine which case label will be taken.
For example, if the index operand of a switch has a value range ~[5,6]
alo
On Fri, 5 Aug 2016, Ian Lance Taylor wrote:
> PR 72812 points out that Go can generate non-ASCII characters in
> assembly code. This is a consequence of the fact that Go permits
> identifiers to contain non-ASCII Unicode code points. The GNU
> assembler doesn't seem to mind, but the Solaris asse
On 08/05/2016 07:36 AM, Richard Biener wrote:
@@ -560,6 +562,14 @@ fsm_find_control_statement_thread_paths
edge taken_edge = profitable_jump_thread_path (path, bbi, name,
arg);
if (taken_edge)
{
+ /* If the taken edge is a loop entry avoid mashing two
+
On 08/02/2016 03:15 PM, Segher Boessenkool wrote:
> On Tue, Aug 02, 2016 at 03:03:42PM -0500, Pat Haugen wrote:
>> On 07/29/2016 10:47 AM, Kelvin Nilsen wrote:
>>> + "xsxexpdp %0,%x1"
>>> + [(set_attr "type" "fp")])
>>
>> Type should be 'integer'.
>
> It has VSX regs as input, integer is worse t
On Fri, Aug 5, 2016 at 2:13 PM, Joseph Myers wrote:
> On Fri, 5 Aug 2016, Ian Lance Taylor wrote:
>
>> PR 72812 points out that Go can generate non-ASCII characters in
>> assembly code. This is a consequence of the fact that Go permits
>> identifiers to contain non-ASCII Unicode code points. The
On 06/06/2016 04:19 AM, Jan Hubicka wrote:
Hi,
while looking into profile mismatches introduced by the backward threading pass
I noticed that the heuristics seems quite simplistics. First it should be
profile sensitive and disallow duplication when optimizing cold paths. Second
it should use est
This is a patch to fix some testsuite failures reported for arm:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69847
The patch was bootstrapped and tested on x86-64 and ppc64.
Committed as rev. 239180.
Index: ChangeLog
===
---
On Aug 4, 2016, Jason Merrill wrote:
> On Mon, Aug 1, 2016 at 10:57 PM, Alexandre Oliva wrote:
>> * langhooks.h (struct lang_hooks_for_decls): Add
>> function_decl_defaulted_inclass_p and
>> function_decl_defaulted_outofclass_p.
> Maybe one hook that returns unsigned rather than two bool hooks
On 08/04/2016 10:27 AM, Kelvin Nilsen wrote:
> + "xsxexpdp %0,%x1"
> + [(set_attr "type" "vecsimple")])
> + "xsxsigdp %0,%x1"
> + [(set_attr "type" "vecsimple")])
'integer' for both, reasoning in response on initial thread for this patch.
> + "xvxexp %x0,%x1"
> + [(set_attr "type" "fpsimpl
On Aug 5, 2016, Jason Merrill wrote:
>> With this patch, when we can't emit a DW_AT_const_value, we emit each
>> "member" of the pointer to member function "record" as a
>> DW_OP_stack_value DW_OP_piece, as long as the referenced member
>> function is output in the same translation unit, otherwi
On 08/03/2016 11:33 PM, Michael Meissner wrote:
> {
> - if (BYTES_BIG_ENDIAN)
> -return "xxpermdi %x0,%x1,%x2,0";
> + if (which_alternative == 0)
> +return (BYTES_BIG_ENDIAN
> + ? "xxpermdi %x0,%x1,%x2,0"
> + : "xxpermdi %x0,%x2,%x1,0");
> +
> + else if (which_alternativ
On Fri, Aug 05, 2016 at 11:35:11AM +0930, Alan Modra wrote:
> diff --git a/gcc/config/rs6000/rs6000.c b/gcc/config/rs6000/rs6000.c
> index 779ba1f..c59d07a 100644
> --- a/gcc/config/rs6000/rs6000.c
> +++ b/gcc/config/rs6000/rs6000.c
> @@ -7717,8 +7717,34 @@ mem_operand_gpr (rtx op, machine_mode mod
On Fri, Aug 05, 2016 at 02:20:40PM +0930, Alan Modra wrote:
> Here are the reload costs for the various alternatives of
> movsf_hardfloat:
> "=!r, !r, m, f, ww, ww, !r, f, wb, m, wY, wu, Z,?wn, ?r,*c*l, !r, *h"
> "r, m, r, f, ww, j, j, m, wY, f, wb, Z, wu, r, wn, r, *h, 0"
> 617
To avoid a circular dependency between and I've
moved std::__invoke to its own header.
In order to use std::__invoke in std::apply it needs to be constexpr,
which is easy, but the standard says that std::invoke isn't constexpr.
So we have to intentionally disallow std::invoke in constant
expres
Go packages use build tags (see the section on Build Constraints at
https://golang.org/pkg/go/build/) to select which files to build on
specific systems.
Previously the libgo Makefile explicitly listed the set of files to
compile for each package. For packages that use build tags, this
required a
On Thu, 2016-08-04 at 14:22 -0600, Jeff Law wrote:
> On 08/04/2016 01:24 PM, David Malcolm wrote:
>
> > > Do you realize that this isn't used for ~700 lines after this
> > > point?
> > > Is
> > > there any sensible way to factor some code here to avoid the
> > > coding
> > > disconnect. I realize
On Fri, Aug 05, 2016 at 04:27:36PM -0500, Pat Haugen wrote:
> On 08/02/2016 03:15 PM, Segher Boessenkool wrote:
> > On Tue, Aug 02, 2016 at 03:03:42PM -0500, Pat Haugen wrote:
> >> On 07/29/2016 10:47 AM, Kelvin Nilsen wrote:
> >>> + "xsxexpdp %0,%x1"
> >>> + [(set_attr "type" "fp")])
> >>
> >> T
On Fri, Aug 05, 2016 at 06:01:47PM -0500, Segher Boessenkool wrote:
> On Fri, Aug 05, 2016 at 02:20:40PM +0930, Alan Modra wrote:
> > Here are the reload costs for the various alternatives of
> > movsf_hardfloat:
> > "=!r, !r, m, f, ww, ww, !r, f, wb, m, wY, wu, Z,?wn, ?r,*c*l, !r, *h"
> > "
On Sat, Aug 06, 2016 at 10:53:33AM +0930, Alan Modra wrote:
> On Fri, Aug 05, 2016 at 06:01:47PM -0500, Segher Boessenkool wrote:
> > On Fri, Aug 05, 2016 at 02:20:40PM +0930, Alan Modra wrote:
> > > Here are the reload costs for the various alternatives of
> > > movsf_hardfloat:
> > > "=!r, !r, m
1 - 100 of 104 matches
Mail list logo
