Hi Richard
> >> > ChangeLog:
> >> > 2021-01-08 Qian jianhua
> >> >
> >> > gcc/
> >> > * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
> >> > * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
> >> > (a64fx_regmove_cost, a64fx_vector_cost): New.
> >> > (a64fx_tunings): Us
On Mon, 11 Jan 2021, Jeff Law wrote:
On 1/9/21 5:43 PM, Maciej W. Rozycki wrote:
On Mon, 21 Dec 2020, Jakub Jelinek wrote:
This patch adds the ~(X - Y) -> ~X + Y simplification requested
in the PR (plus also ~(X + C) -> ~X + (-C) for constants C that can
be safely negated.
This regress
Hi!
We already had x != 0 && y != 0 to (x | y) != 0 and
x != -1 && y != -1 to (x & y) != -1 and
x < 32U && y < 32U to (x | y) < 32U, this patch adds signed
x < 0 && y < 0 to (x | y) < 0. In that case, the low/high seem to be
always the same and just in_p indices whether it is >= 0 or < 0,
also, a
Hi!
As reported by Matthias, --enable-link-serialization=1 can currently start
two concurrent links first (e.g. gnat1 and cc1).
The problem is that make var = value values seem to work differently between
dependencies and actual rules (where it was tested).
As the language make fragments can be in
Hi!
The following patch adds patterns (in the end I went with define_insn rather
than combiner define_split + define_insn_and_split I initially hoped or
define_insn_and_split) to represent (so far 128-bit only) permutations
like { 0 16 1 17 2 18 3 19 4 20 5 21 6 22 7 23 } where the second
operand
Hi!
As the testcase shows, my latest changes caused ICE on that testcase.
The problem is that arith_overflow_check_p now can change the use_stmt
argument (has a reference), so that if it succeeds (returns non-zero),
it points it to the GIMPLE_COND or EQ/NE or COND_EXPR assignment from the
TRUNC_DI
On Tue, 12 Jan 2021, Jakub Jelinek wrote:
> Hi!
>
> We already had x != 0 && y != 0 to (x | y) != 0 and
> x != -1 && y != -1 to (x & y) != -1 and
> x < 32U && y < 32U to (x | y) < 32U, this patch adds signed
> x < 0 && y < 0 to (x | y) < 0. In that case, the low/high seem to be
> always the same
On Tue, 12 Jan 2021, Jakub Jelinek wrote:
> Hi!
>
> As reported by Matthias, --enable-link-serialization=1 can currently start
> two concurrent links first (e.g. gnat1 and cc1).
> The problem is that make var = value values seem to work differently between
> dependencies and actual rules (where i
On Tue, 12 Jan 2021, Jakub Jelinek wrote:
> Hi!
>
> As the testcase shows, my latest changes caused ICE on that testcase.
> The problem is that arith_overflow_check_p now can change the use_stmt
> argument (has a reference), so that if it succeeds (returns non-zero),
> it points it to the GIMPLE_
On Tue, Jan 12, 2021 at 10:33 AM Jakub Jelinek wrote:
>
> Hi!
>
> The following patch adds patterns (in the end I went with define_insn rather
> than combiner define_split + define_insn_and_split I initially hoped or
> define_insn_and_split) to represent (so far 128-bit only) permutations
> like {
On Tue, Jan 12, 2021 at 11:42:44AM +0100, Uros Bizjak wrote:
> > The following patch adds patterns (in the end I went with define_insn rather
> > than combiner define_split + define_insn_and_split I initially hoped or
> > define_insn_and_split) to represent (so far 128-bit only) permutations
> > li
This is a repost of:
https://gcc.gnu.org/pipermail/gcc-patches/2020-February/539763.html
which was initially posted during stage 4. (And yeah, I only just
missed stage 4 again.)
IMO it would be better to fix the bug directly (as the patch tries
to do) instead of wait for a more thorough redes
On 2021/1/11 6:28 PM, Jakub Jelinek wrote:
On Wed, Dec 16, 2020 at 11:06:10PM +0800, Chung-Lin Tang wrote:
we have some other sollve_vv tests for OpenMP 5.0, which tests the occurrence of
'delete' map kind on the target directive.
That is strange.
In OpenMP 5.0, I certainly see the target d
The patch is about an extensive debugging facility that I see
beneficial for GCOV issue.
Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
I'm going to push it.
Thanks,
Martin
gcc/ChangeLog:
* gcov.c (source_info::debug): New.
(print_usage): Add --debug (-D
On Tue, Jan 12, 2021 at 11:47:48AM +0100, Jakub Jelinek via Gcc-patches wrote:
> > OTOH, perhaps some of the new testcases can be handled in x86
> > target_fold_builtin? In the long term, maybe target_fold_shuffle can
> > be introduced to map __builtin_shufle to various target builtins, so
> > the
Similarly to 7f967bd2a7ba156ede3fbb147e66dea5fb7137a6, we need to
compare string with strcmp.
Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
I'm going to push it as it's the same as what I did in
7f967bd2a7ba156ede3fbb147e66dea5fb7137a6.
Martin
gcc/ChangeLog:
P
On Tue, Jan 12, 2021 at 11:42:44AM +0100, Uros Bizjak via Gcc-patches wrote:
> You can use post-reload define_insn_and_split here. This way,
> gen_lowpart on all arguments, including output, can be used. So,
> instead of generating an insn template, the patterns you introduced
> should split to "re
On Tue, Jan 12, 2021 at 2:40 PM Jakub Jelinek wrote:
>
> On Tue, Jan 12, 2021 at 11:42:44AM +0100, Uros Bizjak via Gcc-patches wrote:
> > You can use post-reload define_insn_and_split here. This way,
> > gen_lowpart on all arguments, including output, can be used. So,
> > instead of generating an
This fixes the check that disqualifies BB vectorization because of
required unrolling to match up with the later exact_div we do. To
not disable the ability to split groups that do not match up
exactly with a choosen vector type this also introduces a soft-fail
mechanism to vect_build_slp_tree_1 w
Iain Sandoe writes:
> Hi,
>
> The armv8_arm manual [C6.2.226, ROR (immediate)] uses a # in front
> of the immediate rotation quantity.
>
> Although, it seems, GAS is able to infer the # (or is leninent about
> its absence) assemblers based on the LLVM back end expect it and error out.
>
> tested o
On 1/9/21 3:33 PM, Alexandre Oliva wrote:
On Jan 7, 2021, Richard Biener wrote:
On Wed, Jan 6, 2021 at 12:34 PM Alexandre Oliva wrote:
On Jan 4, 2021, Richard Biener wrote:
Thus, please remove uses of BREAK_FROM_IMM_USE_STMT
together with this patch.
And RETURN_FROM_IMM_USE_STMT, I sup
Hello.
As seen in the PR, we need to fix also virtual PHIs, otherwise
TODO_cfg will skip edges for a missing PHI argument.
Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
Ready to be installed?
Thanks,
Martin
gcc/ChangeLog:
PR tree-optimization/98455
* g
Hi Iain,
> Having a look, actually I can just use the presence of TEST_OUTPUT to be
> a gate for whether to prune all output or not.
that's quite nice indeed.
> Maybe this can be improved later to extract the contents of TEST_OUTPUT,
> but for now, it has caught a few hidden bugs in the tests th
Hi Richard,
Richard Sandiford wrote:
Iain Sandoe writes:
The armv8_arm manual [C6.2.226, ROR (immediate)] uses a # in front
of the immediate rotation quantity.
Although, it seems, GAS is able to infer the # (or is leninent about
its absence) assemblers based on the LLVM back end expect it
) On Tue, Jan 12, 2021 at 3:50 PM Martin Liška wrote:
>
> Hello.
>
> As seen in the PR, we need to fix also virtual PHIs, otherwise
> TODO_cfg will skip edges for a missing PHI argument.
>
> Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
>
> Ready to be installed?
OK - doe
Iain Sandoe writes:
> Hi Richard,
>
> Richard Sandiford wrote:
>
>> Iain Sandoe writes:
>
>>> The armv8_arm manual [C6.2.226, ROR (immediate)] uses a # in front
>>> of the immediate rotation quantity.
>>>
>>> Although, it seems, GAS is able to infer the # (or is leninent about
>>> its absence) a
Hi everyone,
I've reworked the patch to merged dragonfly and AIX
models into the new one named "ieee_1003.1-2008".
It seems okay on the AIX part but if someone can test
on Dragonfly and Freebsd I would be glad. Configure
needs to be regenerated, first.
For now, I've used #ifdef inside the code
On 12/01/21 15:14 +, CHIGOT, CLEMENT wrote:
Hi everyone,Â
I've reworked the patch to merged dragonfly and AIX
models into the new one named "ieee_1003.1-2008".Â
It seems okay on the AIX part but if someone can test
on Dragonfly and Freebsd I would be glad. Configure
needs to be regenerate
> And I don't know for the Changelog/Patch, how renamed
> files must be handled ? I've retrieved my git commit with
> "git format-patch --no-renames" and thus the Changelog
> is made of "Removed" and "New File". Is it okay or is there
> anything special to add ? I didn't find the answer quickly
> >Hi everyone,Â
> >
> >I've reworked the patch to merged dragonfly and AIX
> >models into the new one named "ieee_1003.1-2008".Â
> >It seems okay on the AIX part but if someone can test
> >on Dragonfly and Freebsd I would be glad. Configure
> >needs to be regenerated, first.
>
> Presumably it
On Tue, Jan 12, 2021 at 10:25 AM Jonathan Wakely wrote:
>
> On 12/01/21 15:14 +, CHIGOT, CLEMENT wrote:
> >Hi everyone,
> >
> >I've reworked the patch to merged dragonfly and AIX
> >models into the new one named "ieee_1003.1-2008".
> >It seems okay on the AIX part but if someone can test
> >on
Hi Jonathan,
> On 12/01/21 15:14 +, CHIGOT, CLEMENT wrote:
>>Hi everyone,Â
>>
>>I've reworked the patch to merged dragonfly and AIX
>>models into the new one named "ieee_1003.1-2008".Â
>>It seems okay on the AIX part but if someone can test
>>on Dragonfly and Freebsd I would be glad. Config
Hi Clement,
> I've reworked the patch to merged dragonfly and AIX
> models into the new one named "ieee_1003.1-2008".
> It seems okay on the AIX part but if someone can test
> on Dragonfly and Freebsd I would be glad. Configure
> needs to be regenerated, first.
>
> For now, I've used #ifdef insi
On Tue, Jan 12, 2021 at 3:55 AM Martin Liška wrote:
>
> The patch is about an extensive debugging facility that I see
> beneficial for GCOV issue.
>
> Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
> I'm going to push it.
>
> Thanks,
> Martin
>
> gcc/ChangeLog:
>
>
The following patch fixes
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97969
The patch was successfully bootstrapped on x86-64.
[PR97969] LRA: Transform pattern `plus (plus (hard reg, const), pseudo)` after elimination
LRA can loop infinitely on targets without `reg + imm` insns. Register e
Hi all,
just pushed the following as obvious.
Regards
Andrea
>From 1aff68d54c33ae10fbbb52adb50527baf7b2f627 Mon Sep 17 00:00:00 2001
From: Andrea Corallo
Date: Tue, 12 Jan 2021 17:52:52 +0100
Subject: [PATCH] Fix typo in function-abi.h
gcc/Changelog
2021-01-12 Andrea Corallo
On 1/12/21 5:27 PM, H.J. Lu wrote:
This breaks build on 32-bit hosts:
Sorry for that.
Should be fixed with:
commit 248feb2fa2c0dfc8950566010b30351879c9d057
Author: Martin Liska
Date: Tue Jan 12 18:16:05 2021 +0100
gcov: fix printf format for 32-bit hosts
gcc/ChangeLog:
On 12/01/21 10:44 -0500, David Edelsohn wrote:
On Tue, Jan 12, 2021 at 10:25 AM Jonathan Wakely wrote:
On 12/01/21 15:14 +, CHIGOT, CLEMENT wrote:
>Hi everyone,
>
>I've reworked the patch to merged dragonfly and AIX
>models into the new one named "ieee_1003.1-2008".
>It seems okay on the A
Hi Jonathan,
>> On 12/01/21 15:14 +, CHIGOT, CLEMENT wrote:
>>>Hi everyone,Â
>>>
>>>I've reworked the patch to merged dragonfly and AIX
>>>models into the new one named "ieee_1003.1-2008".Â
>>>It seems okay on the AIX part but if someone can test
>>>on Dragonfly and Freebsd I would be glad.Â
On Tue, Jan 12, 2021 at 12:41 PM Rainer Orth
wrote:
>
> Hi Jonathan,
>
> >> On 12/01/21 15:14 +, CHIGOT, CLEMENT wrote:
> >>>Hi everyone,Â
> >>>
> >>>I've reworked the patch to merged dragonfly and AIX
> >>>models into the new one named "ieee_1003.1-2008".Â
> >>>It seems okay on the AIX part b
We indent with tabs, not spaces. This fixes it.
Committed.
2021-01-12 Segher Boessenkool
* MAINTAINERS: Fix spacing.
---
MAINTAINERS | 8
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index e88808f..6ec346c 100644
--- a/MAINTAINERS
Hi!
On Tue, Jan 12, 2021 at 03:44:22PM +0800, Qian Jianhua wrote:
> ChangeLog:
>
> 2021-01-12 Qian Jianhua
>
> * MAINTAINERS (Write After Approval): Add myself
Welcome!
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -451,6 +451,7 @@ Daniel Jacobowitz
> Andreas
Hi everyone,
In a previous email thread (subject: Possible bug re:
ISO_Fortran_binding.h) it was determined that there was a bug in
ISO_Fortran_binding.c. When attempting to use the ISO Fortran binding
from C to create an array of type uint8_t or signed char, a crash
would result unless the elemen
This patch fixes a regression on sh4 introduced by the rtl-ssa stuff.
The port had a pattern:
(define_insn "movsf_ie"
[(set (match_operand:SF 0 "general_movdst_operand"
"=f,r,f,f,fy, f,m, r, r,m,f,y,y,rf,r,y,<,y,y")
(match_operand:SF 1 "general_movsrc_oper
Hi David,
>> >> Presumably it could also be tested on GNU/Linux and Solaris, since
>> >> they implement the POSIX 2008 APIs needed.
>> >
>> > I'll give the patch a whirl on Solaris. However, we will need to
>> > distinguish between 11.3 (which is XPG6 only) and 11.4 (which support
>> > XPG7).
>>
A recent improvement to MEM_REF formatting introduced a few invalid
assumptions that didn't get exposed during testing and ended up
causing ICEs in downstream packages and some distress to their
maintainers. I have committed the attached patch to remove two
of these assumptions and unblock the pa
Hi,
Just check in to see whether you have any comments and suggestions on this:
FYI, I have been continue with Approach D implementation since last week:
D. Adding calls to .DEFFERED_INIT during gimplification, expand the
.DEFFERED_INIT during expand to
real initialization. Adjusting uninitia
On Sat, Jan 09, 2021 at 07:44:31PM +0100, Matthias Klose wrote:
> These warnings, including the suggested fixes are seen on power*-linux builds.
>
> warning: misspelled term 'builtin function' in format; use 'bult-in function'
> instead [-Wformat-diag]
This one is wrong. Almost all the functions
This libgo patch by Paul Murphy ensures that openat always uses the
variadic libc wrapper. On powerpc64le, not using the wrapper caused a
failure in TestUnshareUidGidMapping due to stack corruption which
resulted in a bogus execve syscall. This fixes GCC PR 98610.
Bootstrapped and ran Go testsuit
Hi!
On Sun, Jan 10, 2021 at 02:18:24PM -0700, Martin Sebor wrote:
> Symbols/identifiers should be formatted using the appropriate
> directives or quoted in %< %>.
We do not have a way to mark up the mathematical symbols [], but we do
require those to express ranges (which is a bad idea /an sich/,
Dear all,
when playing around with the issues exposed by PR93340, particularly visible
in the tree dump, I tried to find ways to simplify substrings in those cases
where they are eligible as designator, which is required e.g. in DATA
statements.
Given my limited understanding, I finally arrived
This patch removes a vestigial use of dk_no_check from
cp_parser_late_parsing_for_member, which ideally should have been
removed as part of the PR41437 patch that improved access checking
inside templates. This allows us to correctly reject f1 and f2 below in
the testcase access34.C below (whereas
On 1/12/21 12:35 PM, Richard Sandiford wrote:
> This patch fixes a regression on sh4 introduced by the rtl-ssa stuff.
> The port had a pattern:
>
> (define_insn "movsf_ie"
> [(set (match_operand:SF 0 "general_movdst_operand"
> "=f,r,f,f,fy, f,m, r, r,m,f,y,y,rf,r,
On Tue, 12 Jan 2021, Segher Boessenkool wrote:
> On Sat, Jan 09, 2021 at 07:44:31PM +0100, Matthias Klose wrote:
> > These warnings, including the suggested fixes are seen on power*-linux
> > builds.
> >
> > warning: misspelled term 'builtin function' in format; use 'bult-in
> > function'
> > i
This patch fixes a typo in the subdf3 pattern that meant it had a
non-standard name and thus the compiler would emit a libcall rather than
the proper hardware instruction for DFmode subtraction.
Tested with standalone AMD GCN target. I will commit shortly.
Julian
2021-01-13 Julian Brown
gcc/
GCN has a reciprocal-approximation instruction but no
hardware divide. This patch adjusts the open-coded reciprocal
approximation/Newton-Raphson refinement steps to use fused multiply-add
instructions as is necessary to obtain a properly-rounded result, and
adds further refinement steps to correctl
This patch fixes a corner case in the AMD GCN md-reorg pass when the
EXEC register is live on entry to a BB, and could be clobbered by code
inserted by the pass before a use in (e.g.) a different BB.
I don't have a standalone test case demonstrating this failure mode,
but I did observe it with an
This patch removes code to fix the v0 register in
gcn_conditional_register_usage that was missed out of the previous patch
removing the need for that:
https://gcc.gnu.org/pipermail/gcc-patches/2019-November/534284.html
Tested with standalone AMD GCN target. I will commit shortly.
Julian
2021-
Hi!
On Wed, Jan 13, 2021 at 12:49:42AM +, Joseph Myers wrote:
> On Tue, 12 Jan 2021, Segher Boessenkool wrote:
> > On Sat, Jan 09, 2021 at 07:44:31PM +0100, Matthias Klose wrote:
> > > These warnings, including the suggested fixes are seen on power*-linux
> > > builds.
> > >
> > > warning: m
Add contains member function to basic_string_view and basic_string.
The new method is enabled for -std=gnu++20, gnu++2b and c++2b. This allows
users to access the method as a GNU extension to C++20. The conditional
test may be reduced to "__cplusplus > 202011L" once GCC has a c++2b switch.
libs
Since certain members of a class are a complete-class context
[class.mem.general]p7, we delay their parsing untile the whole class has
been parsed. For instance, NSDMIs and noexcept-specifiers. The order
in which we perform this delayed parsing matters; we were first parsing
NSDMIs and only they
Another ICE with delayed noexcept parsing, but a bit gnarlier.
A function definition marked with __attribute__((used)) ought to be
emitted even when it is not referenced in a TU. For a member function
template marked with __attribute__((used)) this means that it will
be instantiated: in instantia
Hi everyone,
Sorry, my previous email erroneously referred to unsigned chars / uint8_t, when
I in fact meant signed chars / int8_t. The actual patch works, but the test
case files have ‘unit’ in the file names. I’ll provide a corrected patch
tomorrow to fix the file names.
Harris
Hello Alexandre,
On 19/10/2020 13:52, Alexandre Oliva wrote:
On Oct 19, 2020, Andreas Schwab wrote:
-nostdinc a-nallfl.ads -o a-nallfl.o
a-nallfl.ads:48:13: warning: intrinsic binding type mismatch on return value
a-nallfl.ads:48:13: warning: intrinsic binding type mismatch on argument 1
a-na
> On Wed, Jan 13, 2021 at 02:45:19, Segher Boessenkool wrote:
>
> Hi!
>
> On Tue, Jan 12, 2021 at 03:44:22PM +0800, Qian Jianhua wrote:
> > ChangeLog:
> >
> > 2021-01-12 Qian Jianhua
> >
> > * MAINTAINERS (Write After Approval): Add myself
>
> Welcome!
>
> > --- a/MAINTAINERS
> > +++ b/M
Hi!
The following patch implements what I've talked about, i.e. to no longer
force operands of vec_perm_const into registers in the generic code, but let
each of the (currently 8) targets force it into registers individually,
giving the targets better control on if it does that and when and allowi
Hi!
In the following testcase we only optimize f2 and f7 to btrl, although we
should optimize that way all of the functions. The problem is the type
demotion/narrowing (which is performed solely during the generic folding and
not later), without it we see the AND performed in SImode and match it
Hi!
These simplifications are only simplifications if the (~D ^ C) or (D ^ C)
expressions fold into constants, but in that case they decrease number of
operations by 1.
For the first transformation I guess I should add reverse simplification
when C and D are arbitrary (constants would fold before
On Wed, 13 Jan 2021, Jakub Jelinek wrote:
> Hi!
>
> The following patch implements what I've talked about, i.e. to no longer
> force operands of vec_perm_const into registers in the generic code, but let
> each of the (currently 8) targets force it into registers individually,
> giving the target
On Wed, Jan 13, 2021 at 08:26:49AM +0100, Richard Biener wrote:
> On Wed, 13 Jan 2021, Jakub Jelinek wrote:
>
> > Hi!
> >
> > The following patch implements what I've talked about, i.e. to no longer
> > force operands of vec_perm_const into registers in the generic code, but let
> > each of the (
On Wed, 13 Jan 2021, Jakub Jelinek wrote:
> Hi!
>
> These simplifications are only simplifications if the (~D ^ C) or (D ^ C)
> expressions fold into constants, but in that case they decrease number of
> operations by 1.
>
> For the first transformation I guess I should add reverse simplificatio
On Tue, 12 Jan 2021, Qing Zhao wrote:
> Hi,
>
> Just check in to see whether you have any comments and suggestions on this:
>
> FYI, I have been continue with Approach D implementation since last week:
>
> D. Adding calls to .DEFFERED_INIT during gimplification, expand the
> .DEFFERED_INIT d
72 matches
Mail list logo
