On 08/23/2017 04:52 AM, Richard Sandiford wrote:
> This patch uses subreg_size_lowpart_offset in places that open-coded
> the calculation. The reload use (and the LRA one that was based on it)
> seemed to ignore the BYTES_BIG_ENDIAN != WORDS_BIG_ENDIAN case; it's not
> obvious whether that was del
Hello world,
after some thought, I think the PR can be fixed by something
far less invasive than my previous patch.
The new version of the patch simply issues an error for a
non-printable character (which should never be legal).
Anything else should be caught by other error reporting
routines.
On 08/20/2017 06:59 PM, Alan Modra wrote:
> On Sun, Aug 20, 2017 at 08:00:53AM -0500, Segher Boessenkool wrote:
>> Hi Alan,
>>
>> On Sat, Aug 19, 2017 at 12:19:35AM +0930, Alan Modra wrote:
>>> +Flushing registers to memory has performance implications and may be
>>> +an issue for time-sensitive co
On 08/21/2017 10:11 PM, Alan Modra wrote:
> On Mon, Aug 21, 2017 at 06:33:09PM +0100, Richard Sandiford wrote:
>> I think it's worth emphasising that tying operands doesn't change
>> whether an output needs an earlyclobber or not. E.g. for:
>
> Thanks for noticing this. It turns out that my Open
On 09/18/2017 05:38 AM, Richard Sandiford wrote:
> This patch stops combine from changing the mode of an existing register
> in-place if doing so would change the size of the underlying register
> allocation size, as given by REGMODE_NATURAL_SIZE. Without this,
> many tests fail in adjust_reg_mode
On Wed, 2017-10-11 at 15:51 -0400, Jason Merrill wrote:
> On Tue, Sep 26, 2017 at 3:27 PM, David Malcolm
> wrote:
> > * cp-tree.h (struct saved_scope): Add "location" field.
>
> saved_scope seems like the wrong place for this; it's only
> interesting
> at parse time, so having it saved du
Uros,
Attached is an updated patch. The main difference is in option name and
attribute name change. Other code is the same.
Igor
> -Original Message-
> From: Tsimbalist, Igor V
> Sent: Tuesday, September 19, 2017 5:06 PM
> To: Uros Bizjak ; gcc-patches@gcc.gnu.org
> Cc: Tsimbalist, Ig
Hi,
this patch makes profile probability use more precise representation
and takes care of places where 32bit artithmetic will possibly overflow.
This will make it possible to drop counts on edges next.
Bootstrapped/regtested x86_64-linux, comitted.
Honza
* profile-count.c (safe_scale_64
Attached is an updated patch according to your comments. New tests are
added to test ICF optimization in presence of nocf_check attribute.
Igor
> -Original Message-
> From: Tsimbalist, Igor V
> Sent: Tuesday, September 19, 2017 11:30 PM
> To: Uros Bizjak
> Cc: gcc-patches@gcc.gnu.org; T
The cp_expr class doesn't have a const-qualified operator * accessor.
This leaves one with a confusing error message if one ever tries to
dereference a constant cp_expr object:
const cp_expr thing = ...;
tree bob = *thing; // ERROR
What happens is the non-const operator * is not suitable, b
When the Go frontend imported a reference to an indirectly imported
type aliases, it was looking for the " = " before the optional package
name that appears for an indirect reference, but the exporter was
putting it after. This patch fixes that. The test case is
https://golang.org/cl/70290, and w
On Thu, Oct 12, 2017 at 10:40:22AM +0200, Uros Bizjak wrote:
> > So, if you aren't against it, I can extend the patch to handle the 4
> > other mask patterns; as for other modes, SImode is what is being handled
> > already, DImode is not a problem, because the FEs truncate the shift counts
> > to i
On Thu, Oct 12, 2017 at 08:21:46PM +0200, Thomas Koenig wrote:
>
> after some thought, I think the PR can be fixed by something
> far less invasive than my previous patch.
>
> The new version of the patch simply issues an error for a
> non-printable character (which should never be legal).
> Anyt
Hi!
The ia32intrin.h rotate intrinsics require the second argument to be
in between 1 and 31 (or 63), otherwise they invoke UB. But, we can do much
better while generating the same instruction when optimizing, so the
following patch uses the patterns we pattern recognize well and where
the second
Hi!
Marc in the PR mentioned that it is not really good that the recommended
rotate pattern is recognized only during forwprop1 and later, which is after
einline and that inlining or early opts could have changed stuff too much so
that we wouldn't recogize it anymore.
The following patch handles
Hi!
As mentioned in the PR, there are two bugs in these. One is that
the zero_extract destination is effectively another input operand (for the
remaining bits that are unchanged) and thus the constraint can't be =Q,
but has to be +Q.
And the other problem is that then LRA ICEs whenever it has 3 d
Enable building libgcc with CET options by default on Linux/x86 if
binutils supports CET v2.0.
It can be disabled with --disable-cet. It is an error to configure
GCC with --enable-cet if bintuiils doesn't support CET v2.0.
config/
* cet.m4: New file
gcc/
* config.gcc (extra_heade
Control-flow Enforcement Technology (CET), published by Intel, Introduces
the Shadow Stack feature, which ensures a return from a function is done
to exactly the same location from where the function was called. When EH
is present the control-flow transfer may skip some stack frames and the
shadow
Enable building libbacktrace with CET options.
libbacktrace/
* configure.ac: Add CET_FLAGS to EXTRA_FLAGS.
* aclocal.m4: Regenerate.
* Makefile.in: Likewise.
* configure: Likewise.
Igor
0009-Enable-building-libbacktrace-with-Intel-CET.PATCH
Description: 0009-En
Hi,
Update the logic in verify_gimple_comparision to allow a vector integer result
from a vector comparison, where it previously was limited to only allowing
compares with boolean results. This allows powerpc intrinsics such as this
one to build (after gimple folding):
vector bool int vec
Hi,
Add support for gimple folding of vec_cmp_{eq,ge,gt,le,ne} for
the integer data types.
This adds a handful of entries to the switch statement in builtin_function_type
for those builtins having unsigned arguments.
Three entries are added to vsx.md to enable vcmpne[bhw] instruction
Enable building libcilkrts with CET options.
libcilkrts/
* Makefile.am: Add AM_CXXFLAGS and XCXXFLAGS.
* configure.ac: Set CET_FLAGS, update XCFLAGS, XCXXFLAGS.
* Makefile.in: Regenerate.
* aclocal.m4: Likewise.
* configure: Likewise.
0010-Enable-building
Enable building libatomic with CET options.
libatomic/
* configure.ac: Set CET_FLAGS, update XCFLAGS.
* acinclude.m4: Add cet.m4 and enable.m4.
* configure: Regenerate.
* Makefile.in: Likewise.
* testsuite/Makefile.in: Likewise.
0011-Enable-building-libat
Enable building libgomp with CET options.
libgomp/
* configure.ac: Set CET_FLAGS, update XCFLAGS and FCFLAGS.
* acinclude.m4: Add cet.m4.
* configure: Regenerate.
* Makefile.in: Likewise.
* testsuite/Makefile.in: Likewise
0012-Enable-building-libgomp-with
Enable building libstdc++v3 with CET options.
libstdc++-v3/
* acinclude.m4: Add cet.m4.
* configure.ac: Set CET_FLAGS. Update EXTRA_CFLAGS.
* libsupc++/Makefile.am: Add EXTRA_CFLAGS.
* Makefile.in: Regenerate.
* configure: Likewise.
* doc/Makefile.in
Enable building libsanitizer with Intel CET options.
libsanitizer/
* acinclude.m4: Add enable.m4 and cet.m4.
* Makefile.in: Regenerate.
* asan/Makefile.am: Update AM_CXXFLAGS.
* asan/Makefile.in: Regenerate.
* configure: Likewise.
* configure.ac: Set
Enable building libvtv with Intel CET options.
libvtv/
* acinclude.m4: Add enable.m4 and cet.m4.
* libvtv/configure: Regenerate.
* libvtv/configure.ac: Set CET_FLAGS. Update XCFLAGS.
0015-Enable-building-libvtv-with-Intel-CET.PATCH
Description: 0015-Enable-buildi
Enable building libssp with Intel CET options.
libssp/
* Makefile.am: Update AM_CFLAGS.
* Makefile.in: Regenerate.
* configure: Likewise.
* aclocal.m4: Likewise.
* configure.ac: Set CET_FLAGS. Update XCFLAGS.
0016-Enable-building-libssp-with-Intel-CET.PAT
Enable building libquadmath with Intel CET options.
libquadmath/
* Makefile.am: Update AM_CFLAGS.
* Makefile.in: Regenerate:
* acinclude.m4: Add enable.m4 and cet.m4.
* configure: Regenerate.
* configure.ac: Set CET_FLAGS. Update XCFLAGS.
0017-Enable-buil
On Thu, 12 Oct 2017, Tsimbalist, Igor V wrote:
> Enable building libgcc with CET options by default on Linux/x86 if
> binutils supports CET v2.0.
> It can be disabled with --disable-cet. It is an error to configure
> GCC with --enable-cet if bintuiils doesn't support CET v2.0.
>
> config/
>
Enable building libmpx with Intel CET options.
libmpx/
* Makefile.in: Regenerate.
* acinclude.m4: Add enable.m4 and cet.m4.
* configure: Regenerate.
* configure.ac: Set CET_FLAGS. Update XCFLAGS.
* mpxrt/Makefile.am: Update libmpx_la_CFLAGS.
* mpxrt/
Hi,
this PR reports a couple of problems with the support of the DW_AT_endianity
attribute associated with the scalar_storage_order source attribute: it does
not persist through typedefs and it can contaminate native order DIEs.
The attached patch revamps it by associating native order DIEs and
On Thu, Oct 12, 2017 at 2:45 PM, David Malcolm wrote:
> - put the note on the string-literal, rather than the extern:
> note: 'extern "C"' linkage started here
> extern "C" {
> ^~~
Maybe a range spanning both tokens?
OK with or without that change.
Jason
Enable building libgfortran with Intel CET options.
libgfortran/
* acinclude.m4: Add enable.m4, cet.m4.
* configure: Regenerate.
* configure.ac: Set CET_FLAGS. Update AM_FCFLAGS, AM_CFLAGS,
CFLAGS.
0019-Enable-building-libgfortran-with-Intel-CET.PATCH
Description:
Enable building libobjc with Intel CET options.
libobjc/
* Makefile.in: Regenerate.
* aclocal.m4: Likeiwse.
* configure: Likewise.
* configure.ac: Set CET_FLAGS. Update XCFLAGS.
0020-Enable-building-libobjc-with-Intel-CET.PATCH
Description: 0020-Enable-building-l
Enable building libitm with Intel CET options.
libitm/
* Makefile.in: Regenerate.
* acinclude.m4: Add enable.m4 and cet.m4.
* config/x86/sjlj.S
(_ITM_beginTransaction): Save Shadow Stack pointer.
(GTM_longjmp): Restore Shadow Stack pointer.
* config/
On Fri, 2017-09-22 at 14:11 -0700, Andrew Pinski wrote:
> On Fri, Sep 22, 2017 at 11:39 AM, Jim Wilson
> wrote:
> >
> > On Fri, Sep 22, 2017 at 10:58 AM, Andrew Pinski
> > wrote:
> > >
> > > Two overall comments:
> > > * What about splitting register_offset into two different
> > > elements,
>
On 10/11/2017 12:13 AM, Martin Liška wrote:
> 2017-10-10 Martin Liska
>
> PR tree-optimization/82493
> * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
> (test_range_functions): New function.
> (sbitmap_c_tests): Likewise.
> * selftest-run-tests.c (self
On 08/23/2017 04:53 AM, Richard Sandiford wrote:
> This patch uses byte_lowpart_offset in places that open-coded the
> calculation.
>
> Tested on aarch64-linux-gnu and x86_64-linux-gnu, and by making sure
> that there were no differences in testsuite assembly output for one
> target per CPU. OK t
> -Original Message-
> From: Joseph Myers [mailto:jos...@codesourcery.com]
> Sent: Thursday, October 12, 2017 10:36 PM
> To: Tsimbalist, Igor V
> Cc: gcc-patches@gcc.gnu.org; Jeff Law ; i...@airs.com
> Subject: Re: [PATCH 07/22] Enable building libgcc with CET options.
>
> On Thu, 12 Oct
On Thu, 12 Oct 2017, Tsimbalist, Igor V wrote:
> * unwind.inc (_Unwind_RaiseException_Phase2): Use FRAMES_P_DECL,
> FRAMES_VAR_DECL_1, FRAMES_VAR_INC and FRAMES_P_UPDATE.
> (_Unwind_RaiseException): Use FRAMES_VAR_DECL, FRAMES_VAR_P and
> FRAMES_VAR.
> (_Unwind_ForcedU
"Tsimbalist, Igor V" writes:
> Enable building libbacktrace with CET options.
>
> libbacktrace/
> * configure.ac: Add CET_FLAGS to EXTRA_FLAGS.
> * aclocal.m4: Regenerate.
> * Makefile.in: Likewise.
> * configure: Likewise.
> +if test x$enable_cet = xyes; then
> + CET_FL
On Mon, Oct 09, 2017 at 03:46:17PM -0500, Pat Haugen wrote:
> Power9 has efficient unaligned load insns. The following patch fixes the
> cost to reflect that. There was no similar code for the unaligned_store
> case. Bootstrap/regtest on powerpc64le-linux with no new regressions.
> Ok for trunk?
From: Trevor Saunders
I had a go at updating Trevor's unique_ptr patch from July
( https://gcc.gnu.org/ml/gcc-patches/2017-07/msg02084.html )
One of the sticking points was what to call the namespace; there was
wariness about using "gtl" as the name.
Richi proposed (https://gcc.gnu.org/ml/gcc-p
Hi Will,
On Thu, Oct 12, 2017 at 03:03:12PM -0500, Will Schmidt wrote:
> * config/rs6000/rs6000.c: (rs6000_gimple_fold_builtin) Add support for
> folding of vector compares. (builtin_function_type) Add compare
> builtins to the list of functions having unsigned arguments.
>
On Thu, Oct 12, 2017 at 9:11 PM, Jakub Jelinek wrote:
> On Thu, Oct 12, 2017 at 10:40:22AM +0200, Uros Bizjak wrote:
>> > So, if you aren't against it, I can extend the patch to handle the 4
>> > other mask patterns; as for other modes, SImode is what is being handled
>> > already, DImode is not a
On Thu, Oct 12, 2017 at 9:49 PM, Jakub Jelinek wrote:
> Hi!
>
> As mentioned in the PR, there are two bugs in these. One is that
> the zero_extract destination is effectively another input operand (for the
> remaining bits that are unchanged) and thus the constraint can't be =Q,
> but has to be +
On Thu, Oct 12, 2017 at 9:39 PM, Jakub Jelinek wrote:
> Hi!
>
> The ia32intrin.h rotate intrinsics require the second argument to be
> in between 1 and 31 (or 63), otherwise they invoke UB. But, we can do much
> better while generating the same instruction when optimizing, so the
> following patc
On Thu, Oct 12, 2017 at 8:45 PM, Tsimbalist, Igor V
wrote:
> Uros,
>
> Attached is an updated patch. The main difference is in option name and
> attribute name change. Other code is the same.
Trivial changes (if they fall under "trivial" rule) don't need extra
approval, but OK nevertheless.
Tha
> Seems reasonable. As a result something like
> check_missing_nocf_check_attribute is going to just go away along with the
> code in *-typeck.c which called it, right? If so that seems like a nice
> cleanup.
Yes, you are right.
Updated patch is attached.
Igor
> -Original Message-
> F
On Thu, Oct 12, 2017 at 08:32:32AM +0200, Uros Bizjak wrote:
> On Wed, Oct 11, 2017 at 10:59 PM, Jakub Jelinek wrote:
> > As can be seen on the testcase below, the *3_mask
> > insn/splitter is able to optimize only the case when the and is
> > performed in SImode and then the result subreged into
On Thu, Oct 12, 2017 at 10:32 AM, Jakub Jelinek wrote:
> On Thu, Oct 12, 2017 at 08:32:32AM +0200, Uros Bizjak wrote:
>> On Wed, Oct 11, 2017 at 10:59 PM, Jakub Jelinek wrote:
>> > As can be seen on the testcase below, the *3_mask
>> > insn/splitter is able to optimize only the case when the and
On 10/11/2017 04:59 PM, Jason Merrill wrote:
> On Thu, Oct 5, 2017 at 12:53 PM, Martin Liška wrote:
>> On 10/05/2017 05:07 PM, Jason Merrill wrote:
>>> On Thu, Oct 5, 2017 at 6:31 AM, Martin Liška wrote:
As discussed 2 days ago on IRC with Jakub and Jonathan, C++ standard says
that havi
On Thu, Oct 12, 2017 at 10:40:42AM +0200, Martin Liška wrote:
> --- a/gcc/cp/constexpr.c
> +++ b/gcc/cp/constexpr.c
> @@ -1175,7 +1175,12 @@ cxx_eval_builtin_function_call (const constexpr_ctx
> *ctx, tree t, tree fun,
> {
> new_call = build_call_array_loc (EXPR_LOCATION (t), TREE_TY
The following avoids code-generation errors for modulo operations
resulting from our own constraints ending up as no-ops because
the type we code-generate in already imposes the modulo operation.
For the case in SPEC 2k6 triggering this we'd even know the
modulo constraint isn't necessary - we h
On Wed, Oct 11, 2017 at 3:43 PM, Richard Biener wrote:
>
> For PR82355 I introduced a fake dimension to ISL to allow CHRECs
> having an evolution in a loop that isn't fully part of the SESE
> region we are processing. That was easier than fending off those
> CHRECs (without simply giving up on SE
On 12/10/17 06:21, vladimir.mezent...@oracle.com wrote:
> From: Vladimir Mezentsev
>
> FMA (floating-point multiply-add) instructions are supported on aarch64.
> These instructions can produce different result if two operations executed
> separately.
> -ffp-contract=off doesn't allow the FMA ins
On Thu, 12 Oct 2017, Bin.Cheng wrote:
> On Wed, Oct 11, 2017 at 3:43 PM, Richard Biener wrote:
> >
> > For PR82355 I introduced a fake dimension to ISL to allow CHRECs
> > having an evolution in a loop that isn't fully part of the SESE
> > region we are processing. That was easier than fending o
On 10/11/2017 04:22 PM, Jakub Jelinek wrote:
> On Wed, Oct 11, 2017 at 03:36:40PM +0200, Martin Liška wrote:
>>> std::swap(addr1, addr2); ? I don't see it used in any of libsanitizer
>>> though, so not sure if the corresponding STL header is included.
>>
>> They don't use it anywhere and I had som
Hi.
There's one false positive I've noticed:
$ cat /tmp/ptr-cmp.c
int
__attribute__((noinline))
foo(char *p1, char *p2)
{
if (p2 != 0 && p1 > p2)
return 0;
return 1;
}
int main(int argc, char **argv)
{
return foo(argv[0], 0);
}
$ gcc /tmp/ptr-cmp.c -fsanitize=address,pointer-compare
On Thu, Oct 12, 2017 at 01:13:56PM +0200, Martin Liška wrote:
> + if (a1 == a2)
> +return;
> +
> + uptr shadow_offset1, shadow_offset2;
> + bool valid1, valid2;
> + {
> +ThreadRegistryLock l(&asanThreadRegistry());
> +
> +valid1 = GetStackVariableBeginning(a1, &shadow_offset1);
> +
Hello!
> The following patch fixes
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82353
>
> LRA did not update hard reg liveness on bb borders for hard regs which are
> part of insn patterns
> like CFLAGS reg. It was ok for inheritance in EBB which creates only moves
> and they usually
> have
On Thu, Oct 12, 2017 at 01:30:34PM +0200, Martin Liška wrote:
> There's one false positive I've noticed:
>
> $ cat /tmp/ptr-cmp.c
> int
> __attribute__((noinline))
> foo(char *p1, char *p2)
> {
> if (p2 != 0 && p1 > p2)
> return 0;
>
> return 1;
> }
Guess that is an argument for instrume
On Thu, Oct 12, 2017 at 12:13 PM, Richard Biener wrote:
> On Thu, 12 Oct 2017, Bin.Cheng wrote:
>
>> On Wed, Oct 11, 2017 at 3:43 PM, Richard Biener wrote:
>> >
>> > For PR82355 I introduced a fake dimension to ISL to allow CHRECs
>> > having an evolution in a loop that isn't fully part of the SE
On 10/11/2017 06:58 PM, Martin Sebor wrote:
> On 10/11/2017 11:26 AM, Joseph Myers wrote:
>> On Tue, 10 Oct 2017, Martin Sebor wrote:
>>
>>> The ideal solution for 1) would be a function pointer that can
>>> never be used to call a function (i.e., the void* equivalent
>>> for functions).[X]
>>
>> I
On 10/11/2017 03:57 AM, Martin Sebor wrote:
>
>
> Incidentally, void(*)(void) in C++ is a poor choice for this
> use case also because of the language's default function
> arguments. It's an easy mistake for a C++ programmer to make
> to assume that given, say:
>
> void foo (const char *s = "
On 10/11/2017 03:57 AM, Martin Sebor wrote:
>
>
> [X] This can be function that takes an argument of an incomplete
> type, such as:
>
> struct Incomplete;
> typedef void Uncallable (struct Incomplete);
>
> Any function can safely be converted to Uncallable* without
> the risk of being calle
On Wed, 11 Oct 2017, Jakub Jelinek wrote:
> Hi!
>
> This patch adds verification that vec_select in *.md files
> doesn't have any out of bounds indices in the selector.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, tested
> by building cc1 of aarch64, arm, powerpc64le and s390x cros
On Thu, Oct 12, 2017 at 4:40 AM, Martin Liška wrote:
> On 10/11/2017 04:59 PM, Jason Merrill wrote:
>> On Thu, Oct 5, 2017 at 12:53 PM, Martin Liška wrote:
>>> On 10/05/2017 05:07 PM, Jason Merrill wrote:
On Thu, Oct 5, 2017 at 6:31 AM, Martin Liška wrote:
> As discussed 2 days ago on I
On 06/10/17 13:44, Tamar Christina wrote:
> Hi All,
>
> this is a minor respin with changes echo'd from feedback from aarch64.
> I assume still OK for trunk.
>
> Regtested on arm-none-eabi, armeb-none-eabi,
> aarch64-none-elf and aarch64_be-none-elf with no issues found.
>
> Ok for trunk?
>
> g
On 06/10/17 13:45, Tamar Christina wrote:
> Hi All,
>
> this is a respin with the changes suggested. Note that this patch is no 8/8
> in the series.
>
> Regtested on arm-none-eabi, armeb-none-eabi,
> aarch64-none-elf and aarch64_be-none-elf with no issues found.
>
> Ok for trunk?
>
> gcc/tests
On Thu, Oct 5, 2017 at 3:17 PM, Bin Cheng wrote:
> Hi,
> This patch rewrites classification part of builtin partition so that nested
> builtin partitions are supported. With this extension, below loop nest:
> void
> foo (void)
> {
> for (unsigned i = 0; i < M; ++i)
> for (unsigned j = 0; j
On Thu, Oct 5, 2017 at 3:17 PM, Bin Cheng wrote:
> Hi,
> This patch merges adjacent memset builtin partitions if possible. It is
> a useful special case optimization transforming below code:
>
> #define M (256)
> #define N (512)
>
> struct st
> {
> int a[M][N];
> int c[M];
> int b[M][N];
>
On Wed, Oct 11, 2017 at 6:10 PM, Bin Cheng wrote:
> Hi,
> When reading the code I found it's could be confusing without comment.
> This patch adds comment explaining why we want merge PARALLEL type
> partitions in a SCC, even though the result partition can no longer
> be executed in parallel. It
> -Original Message-
> From: Richard Earnshaw (lists) [mailto:richard.earns...@arm.com]
> Sent: 12 October 2017 14:21
> To: Tamar Christina; James Greenhalgh
> Cc: gcc-patches@gcc.gnu.org; nd; Marcus Shawcroft
> Subject: Re: [PATCH][GCC][Testsuite][ARM][AArch64] Enable Dot Product
> for g
On Wed, Oct 11, 2017 at 02:16:38PM -0400, David Malcolm wrote:
> On Sat, 2017-08-05 at 01:39 -0400, Trevor Saunders wrote:
> > On Fri, Aug 04, 2017 at 08:55:50PM +0100, Jonathan Wakely wrote:
> > > On 01/08/17 23:09 -0400, Trevor Saunders wrote:
> > > > aiui C++03 is C++98 with a few additions to t
I made scheduling to fail when we end up with an empty domain but as
I forgot to actually check the return value of build_original_schedule
the fix was equivalent to just doing nothing to the schedule when
it has an empty domain. I verified that for the testcase it DCEs
the relevant stmt and that
.dg/tree-ssa/ldist-30.c: New test.
* gcc.dg/tree-ssa/ldist-31.c: New test.
>
> Ok with those changes.
>
> Thanks,
> Richard.
>
From 8271ce0851a60b38226e92558bca234774e5503e Mon Sep 17 00:00:00 2001
From: Bin Cheng
Date: Wed, 27 Sep 2017 13:00:59 +0100
Subject: [PATCH 6/7] loo
The type check seems premature (we're checking CHRECs already) and
we certainly can handle POINTER IVs just fine.
Bootstrap / regtest running on x86_64-unknown-linux-gnu.
SPEC CPU 2k6 sees ~100 more loop nest optimizations that way.
Ok?
[I'd rather have problematical testcases for those weird
On Oct 11, 2017 9:43 AM, "Richard Biener" wrote:
For PR82355 I introduced a fake dimension to ISL to allow CHRECs
having an evolution in a loop that isn't fully part of the SESE
region we are processing. That was easier than fending off those
CHRECs (without simply giving up on SESE regions wit
On Oct 12, 2017 9:29 AM, "Richard Biener" wrote:
The type check seems premature (we're checking CHRECs already) and
we certainly can handle POINTER IVs just fine.
Bootstrap / regtest running on x86_64-unknown-linux-gnu.
SPEC CPU 2k6 sees ~100 more loop nest optimizations that way.
Ok?
[I'd r
On Oct 12, 2017 9:08 AM, "Richard Biener" wrote:
I made scheduling to fail when we end up with an empty domain but as
I forgot to actually check the return value of build_original_schedule
the fix was equivalent to just doing nothing to the schedule when
it has an empty domain. I verified that
No functional changes.
2017-10-12 Uros Bizjak
* config/alpha/alpha.c (alpha_split_conditional_move):
Use std::swap instead of manually swapping.
(alpha_stdarg_optimize_hook): Ditto.
(alpha_canonicalize_comparison): Ditto.
Bootstrapped and regression tested on alphaev68-linux-g
On Oct 12, 2017 4:36 AM, "Richard Biener" wrote:
The following avoids code-generation errors for modulo operations
resulting from our own constraints ending up as no-ops because
the type we code-generate in already imposes the modulo operation.
For the case in SPEC 2k6 triggering this we'd even
On 10/12/2017 02:12 AM, Tsimbalist, Igor V wrote:
>> Seems reasonable. As a result something like
>> check_missing_nocf_check_attribute is going to just go away along with the
>> code in *-typeck.c which called it, right? If so that seems like a nice
>> cleanup.
> Yes, you are right.
>
> Update
On 10/12/2017 05:52 AM, Pedro Alves wrote:
On 10/11/2017 03:57 AM, Martin Sebor wrote:
[X] This can be function that takes an argument of an incomplete
type, such as:
struct Incomplete;
typedef void Uncallable (struct Incomplete);
Any function can safely be converted to Uncallable* witho
On Thu, 12 Oct 2017, Martin Sebor wrote:
> Yes. In light of this discussion I am thinking it might be
> worthwhile to bring up the issue of generic function pointers
> with WG14 for C2X.
I'm fine with the idea of having a standard solution that (unlike void (*)
(void)) cannot be called at all w
> -Original Message-
> From: Richard Earnshaw (lists) [mailto:richard.earns...@arm.com]
> Sent: 12 October 2017 13:58
> To: Tamar Christina; James Greenhalgh
> Cc: gcc-patches@gcc.gnu.org; nd; Marcus Shawcroft
> Subject: Re: [PATCH][GCC][AArch64] Dot Product SIMD patterns [Patch
> (5/8)]
Hi,
this patch fixes ix86_adjust_cost for zen support. In particular the original
code was accounting memory latencies incorrectly (3 for integer, 2 for FP unit)
while they are 4 for integer and 7 for FP on this CPU.
Using lower latencies makes scheduler overly pesimistic about CPU's ability
to e
Hi!
On Wed, Oct 11, 2017 at 06:41:05PM -0400, Vladimir Makarov wrote:
> > Tested on x86_64-linux -m32/-m64, and verified with cc1plus before your
> > change, ok for trunk?
BTW, I think it is quite fragile to scan for the reload messages, so I've
cooked up a runtime test that fails before your pat
On 10/12/2017 12:49 PM, Jakub Jelinek wrote:
Hi!
On Wed, Oct 11, 2017 at 06:41:05PM -0400, Vladimir Makarov wrote:
Tested on x86_64-linux -m32/-m64, and verified with cc1plus before your
change, ok for trunk?
BTW, I think it is quite fragile to scan for the reload messages, so I've
cooked up
On Thu, Oct 12, 2017 at 01:05:21PM -0400, Vladimir Makarov wrote:
>
>
> On 10/12/2017 12:49 PM, Jakub Jelinek wrote:
> > Hi!
> >
> > On Wed, Oct 11, 2017 at 06:41:05PM -0400, Vladimir Makarov wrote:
> > > > Tested on x86_64-linux -m32/-m64, and verified with cc1plus before your
> > > > change, o
On Wed, 2017-10-11 at 17:18 -0400, Jason Merrill wrote:
> On Tue, Sep 26, 2017 at 9:56 AM, David Malcolm
> wrote:
> > In r251026 (aka 3fe34694f0990d1d649711ede0326497f8a849dc,
> > "C/C++: show pertinent open token when missing a close token")
> > I copied part of cp_parser_error into cp_parser_req
While working with Martin L's bugfix/cleanup I came across a minor bug
in the DSE code.
Specifically it mis-handles references with negative offsets. Example
code can be found in gcc.dg/pr48335-4.c. Thankfully these are
relatively uncommon and we can just prune them from consideration
without
94 matches
Mail list logo
