On LWG reflector it was pointed out that returning copy of F, may also
impact the result
of overload resolution, as without forwarding pointer parameters can be
initialized from 0:
#include
struct S {
void operator()(long, long); // 1
void operator()(int, void*); // 2
};
void foo() {
For any minimum value of a signed type, its negation (with wraparound) results
in the same value, behaving like zero. Representing the unordered result with
this minimum value, along with 0 for equal, 1 for greater, and -1 for less
in partial_ordering, allows its value to be reversed using unary ne
Bootstrapped and regtested on x86_64 and s390. Ok for mainline?
-- >8 --
In case an asm operand is an error node, constraints etc. are still
validated. Furthermore, all other operands are gimplified, although an
error is returned in the end anyway. For hard register constraints an
operand is r
On Wed, Aug 27, 2025 at 6:32 AM H.J. Lu wrote:
>
> Source operands of 2 *tls_dynamic_gnu2_call_64_di patterns in
>
> (insn 10 9 11 3 (set (reg:DI 100)
> (unspec:DI [
> (symbol_ref:DI ("caml_state") [flags 0x10] 0x7fe10e1d9e40 caml_state>)
> ] UNSPEC_TLSDESC))
The problem here is after r16-101, the 2 functions containing alloca/VLA
start to be cloned and then we un-VLA happens in using_vararray so this
is no longer testing what it should be testing.
The obvious fix is to mark using_vararray and using_alloca as noclone too.
Pushed as obvious after a qui
Since kind == vec_perm may not be a real vec_perm, just a broadcast or
simple load in BB vectorizer.
Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}.
Ready push to trunk.
gcc/ChangeLog:
* config/i386/i386.cc (ix86_vector_costs::finish_cost):
Restrict tune avx256_avoid_ve
On Wed, Aug 27, 2025 at 11:59 AM H.J. Lu wrote:
>
> On Tue, Aug 26, 2025 at 8:50 PM Hongtao Liu wrote:
> >
> > On Wed, Aug 27, 2025 at 6:30 AM H.J. Lu wrote:
> > >
> > > For a basic block with only a debug marker:
> > >
> > > (note 3 0 2 2 [bb 2] NOTE_INSN_BASIC_BLOCK)
> > > (note 2 3 5 2 NOTE_I
On Tue, Aug 26, 2025 at 8:50 PM Hongtao Liu wrote:
>
> On Wed, Aug 27, 2025 at 6:30 AM H.J. Lu wrote:
> >
> > For a basic block with only a debug marker:
> >
> > (note 3 0 2 2 [bb 2] NOTE_INSN_BASIC_BLOCK)
> > (note 2 3 5 2 NOTE_INSN_FUNCTION_BEG)
> > (debug_insn 5 2 16 2 (debug_marker) "x.c":6:3
On Wed, Aug 27, 2025 at 6:30 AM H.J. Lu wrote:
>
> For a basic block with only a debug marker:
>
> (note 3 0 2 2 [bb 2] NOTE_INSN_BASIC_BLOCK)
> (note 2 3 5 2 NOTE_INSN_FUNCTION_BEG)
> (debug_insn 5 2 16 2 (debug_marker) "x.c":6:3 -1 (nil))
>
> emit the TLS call after debug marker.
>
> gcc/
>
>
On Linux/x86_64,
001cd39749f94ece8276b63f91eb864babb81a5d is the first bad commit
commit 001cd39749f94ece8276b63f91eb864babb81a5d
Author: Andi Kleen
Date: Sun Aug 3 17:35:39 2025 -0700
Use x86 GFNI for vectorized constant byte shifts/rotates
caused
FAIL: gcc.target/i386/shift-gf2p8affine
On Linux/x86_64,
c77d79785d32b30502c8bcd6aa53b63d715bb124 is the first bad commit
commit c77d79785d32b30502c8bcd6aa53b63d715bb124
Author: David Faust
Date: Wed Aug 6 09:24:40 2025 -0700
ctf: avoid overflow for array num elements [PR121411]
caused
FAIL: gcc.dg/debug/ctf/ctf-array-7.c (tes
On Thu, Aug 14, 2025, 12:17 PM Gopi Kumar Bulusu wrote:
>
> namaskaaram
>
> This patch has been reviewed by Neal Frager - his comments are present in
> bugzilla PR 118280
>
> If needed I can submit an updated patch with the patch comment updated as
> follows
>
> Reviewed-by: Neal Frager
>
>
> dh
More testsuite hygiene. Some of the thead tests are expecting to find
xtheadvdot in the extension set, but it's not defined as a valid
extension anywhere. I'm just removing xtheadvdot. Someone more
familiar with these cores can add it back properly if they're so inclined.
Second, there's a
As Andrew Pinski noted in PR60674:
devirt-23.C started to pass with r16-101-g132d01d96ea9d6.
So just need to update the testcase removing the xfail and close this
bug as fixed.
The reason why this was not fixed until r16-101-g132d01d96ea9d6 is
because the call is from main which is kno
On Tue, Aug 26, 2025 at 11:54:34AM -0700, Jerry D wrote:
> The attached patch provides an error with -std=f95 or higher.
>
> Regression tested on x86_64.
>
> OK for trunk?
>
Yes. Thanks for the patch.
--
Steve
This patch tries to add support for a variant of SAT_TRUNC where
negative numbers are clipped to 0 instead of NARROW_TYPE_MAX_VALUE.
This form is seen in x264, aka
UT clip (T a)
{
return a & (UT)(-1) ? (-a) >> 31 : a;
}
Where sizeof(UT) < sizeof(T)
I'm unable to get the SAT_TRUNC pattern to ap
This patch contains testcases for PR120378 after the change made to
support the vnclipu variant of the SAT_TRUNC pattern.
PR target/120378
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/autovec/pr120378-1.c: New test.
* gcc.target/riscv/rvv/autovec/pr120378-2.c: New test
Respin of
https://gcc.gnu.org/pipermail/gcc-patches/2025-August/691426.html but
separates the middle end vectorizer changes from the RISC-V target
specific changes.
The patch changes the riscv loop codegen from
vsetvli a5,a2,e32,m1,ta,mu
vle32.v v1,0(a1)
sllia4,a5,2
"H.J. Lu" writes:
> Move pr121656.c to gcc.dg/torture and replace weak attribute with noipa
> attribute. Verified by reverting
>
> 56ca14c4c4f Fix invalid right shift count with recent ifcvt changes
>
> to trigger
>
> FAIL: gcc.dg/torture/pr121656.c -O1 execution test
> FAIL: gcc.dg/torture/p
Move pr121656.c to gcc.dg/torture and replace weak attribute with noipa
attribute. Verified by reverting
56ca14c4c4f Fix invalid right shift count with recent ifcvt changes
to trigger
FAIL: gcc.dg/torture/pr121656.c -O1 execution test
FAIL: gcc.dg/torture/pr121656.c -O2 execution test
FAI
Shreya and I were working through some testsuite failures and noticed
that many of the current failures on the pioneer were just silly. We
have tests that expect to see full architecture strings in their
expected output when the bulk (some might say all) of the architecture
string is irrelev
Source operands of 2 *tls_dynamic_gnu2_call_64_di patterns in
(insn 10 9 11 3 (set (reg:DI 100)
(unspec:DI [
(symbol_ref:DI ("caml_state") [flags 0x10] )
] UNSPEC_TLSDESC)) "x.c":7:16 1674 {*tls_dynamic_gnu2_lea_64_di}
(nil))
(insn 11 10 12 3 (parallel [
For a basic block with only a debug marker:
(note 3 0 2 2 [bb 2] NOTE_INSN_BASIC_BLOCK)
(note 2 3 5 2 NOTE_INSN_FUNCTION_BEG)
(debug_insn 5 2 16 2 (debug_marker) "x.c":6:3 -1 (nil))
emit the TLS call after debug marker.
gcc/
PR target/121668
* config/i386/i386-features.cc (ix86_
Ping patch:
| Date: Thu, 31 Jul 2025 02:25:19 -0400
| Subject: [PATCH, V8] Fix PR 118541, do not generate unordered fp compares
| Message-ID:
https://gcc.gnu.org/pipermail/gcc-patches/2025-July/691248.html
--
Michael Meissner, IBM
PO Box 98, Ayer, Massachusetts, USA, 01432
email: meiss...@linu
Ping #3 to add -mcpu=future support to the PowerPC posted on July 23rd,
2025:
| Date: Wed, 23 Jul 2025 15:17:41 -0400
| Subject: [PATCH, V4] Add -mcpu=future to the PowerPC
| Message-ID:
https://gcc.gnu.org/pipermail/gcc-patches/2025-July/690398.html
I am back now, and I would like to see if we
Jakub Jelinek wrote:
It of course makes a lot of sense. declare simd is used mostly on the host
in the wild, outside of target regions.
On "normal" Linux for instance in /usr/include/bits/math-vector.h,
albeit most of the time -fopenmp is not active such that the "simd"
attribute is used inste
On Tue Aug 26, 2025 at 9:53 PM CEST, Joseph Myers wrote:
> There has been a significant discussion on the reflector in the past few
> days concerning jumping backwards across a defer block within a scope.
> How do you handle this case, what is the rationale for the decisions made
> about that c
Similar to the existing ptr32plus, add a proc returning whether the
target uses 64-bit or larger pointers. It can be useful immediately in
the CTF testsuite, and likely elsewhere.
Tested on x86_64-pc-linux-gnu and with cross for arm-eabi cortex-m3
in the CTF testsuite.
gcc/testsuite/
*
There has been a significant discussion on the reflector in the past few
days concerning jumping backwards across a defer block within a scope.
How do you handle this case, what is the rationale for the decisions made
about that case, and are there appropriate testcases for it included?
int v
On Tue, 26 Aug 2025, Tomasz Kaminski wrote:
>
>
> On Tue, Aug 26, 2025 at 5:50 PM Patrick Palka wrote:
> On Tue, 26 Aug 2025, Jonathan Wakely wrote:
>
> > On 20/08/25 10:13 +0200, Tomasz Kamiński wrote:
> > > This patch refactors the implementation of bind_front and bind_back
On 8/26/25 11:54, H.J. Lu wrote:
> On Mon, Aug 25, 2025 at 12:11 PM David Faust wrote:
>>
>>
>>
>> On 8/25/25 10:33, Indu Bhagat wrote:
>>> On 8/25/25 8:08 AM, David Faust wrote:
On 8/22/25 23:49, Indu Bhagat wrote:
> On 8/6/25 9:26 AM, David Faust wrote:
>> CTF array enc
The test fails to compile on 32-bit targets because the arrays are too
large. Restrict to targets where the array index type is 64-bits.
Also note the relevant PR in the test comment.
Tested on x86_64-pc-linux-gnu host and arm-eabi cross target.
Pushed as obvious.
PR debug/121411
gcc/te
On 2025-08-26 15:27, Richard Earnshaw wrote:
On 10/08/2025 12:51, Torbjörn SVENSSON wrote:
Ok for trunk and releases/gcc-15?
For releases/gcc-15, I would also want to cherry-pick r16-562-
g20c25919132 that converts
the test to use function body instead of three scan-assembler.
Changes sinc
On Mon, Aug 25, 2025 at 12:11 PM David Faust wrote:
>
>
>
> On 8/25/25 10:33, Indu Bhagat wrote:
> > On 8/25/25 8:08 AM, David Faust wrote:
> >>
> >>
> >> On 8/22/25 23:49, Indu Bhagat wrote:
> >>> On 8/6/25 9:26 AM, David Faust wrote:
> CTF array encoding uses uint32 for number of elements.
The attached patch provides an error with -std=f95 or higher.
Regression tested on x86_64.
OK for trunk?
Regards,
Jerry
Author: Jerry DeLisle
Date: Tue Aug 26 11:36:25 2025 -0700
Fortran: H edit descriptor error with -std=f95
PR fortran/114611
gcc/fortran/ChangeLog:
On 8/26/25 08:47, Tobias Burnus wrote:
Hi Sandra,
Sandra Loosemore wrote:
On 8/26/25 05:26, Jakub Jelinek wrote:
I wouldn't emit sorry for targets where we already implement it
"correctly",
i.e. at least all targets which have NULL
targetm.simd_clone.compute_vecsize_and_simdlen
There the argu
On Tue, Aug 26, 2025 at 12:46:51PM -0600, Sandra Loosemore wrote:
> > There is also no means to obtain
> > targetm.simd_clone.compute_vecsize_and_simdlen
> > on the non-offload compiler for the offload compiler (but see below).
>
> Right, this is what I was getting at. It doesn't make sense fo
Hi, Kees,
> On Aug 26, 2025, at 13:25, Kees Cook wrote:
>
> The __attribute__((__copy__)) functionality was crashing when copying
> sanitizer-related attributes because these attributes violated the standard
> GCC attribute infrastructure by storing INTEGER_CST values directly instead
> of wrapp
Changes in v4:
* Move global symbols to correct list.
* Fix up commit annotations: Changelog, title.
Changes in v3:
* Add to global symbols list.
Changes in v2:
* tai_clock and gps_clock remain "is_steady = false".
* Note adjustments to the exported global symbols list are
not yet included
The __attribute__((__copy__)) functionality was crashing when copying
sanitizer-related attributes because these attributes violated the standard
GCC attribute infrastructure by storing INTEGER_CST values directly instead
of wrapping them in TREE_LIST like all other attributes.
Wrap sanitizer attr
Change in v3:
* Add two new symbols to config/abi/pre/gnu.ver
Changes in v2:
* tai_clock and gps_clock remain "is_steady = false".
* Note adjustments to the exported global symbols list are
not yet included in this patch.
This moves std::tai_clock::now() and std::tai_clock::now()
definitions
On Tue, Aug 26, 2025 at 03:54:13PM +0300, Claudiu Zissulescu-Ianculescu wrote:
> I think the tests are missing.
Argh. I will send a v3...
--
Kees Cook
On Tue, 26 Aug 2025 at 17:21, Nathan Myers wrote:
>
> Change in v3:
> * Add two new symbols to config/abi/pre/gnu.ver
>
> Changes in v2:
> * tai_clock and gps_clock remain "is_steady = false".
> * Note adjustments to the exported global symbols list are
> not yet included in this patch.
>
> T
This patch refactors the implementation of bind_front and bind_back to avoid
using std::tuple for argument storage. Instead, bound arguments are now:
* stored directly if there is only one,
* within a dedicated _Bound_arg_storage otherwise.
_Bound_arg_storage is less expensive to instantiate and a
On Tue, Aug 26, 2025 at 5:50 PM Patrick Palka wrote:
> On Tue, 26 Aug 2025, Jonathan Wakely wrote:
>
> > On 20/08/25 10:13 +0200, Tomasz Kamiński wrote:
> > > This patch refactors the implementation of bind_front and bind_back to
> > > avoid using std::tuple for argument storage. Instead, bound a
> -Original Message-
> From: soum...@nvidia.com
> Sent: Monday, August 25, 2025 12:00 PM
> To: gcc-patches@gcc.gnu.org
> Cc: ktkac...@nvidia.com; rdsandif...@googlemail.com; pins...@gmail.com
> Subject: [v3 PATCH 0/6] aarch64: Support for user-defined aarch64 tuning
>
> From: Soumya AR
>
On Tue, 26 Aug 2025, Jonathan Wakely wrote:
> On 20/08/25 10:13 +0200, Tomasz Kamiński wrote:
> > This patch refactors the implementation of bind_front and bind_back to
> > avoid using std::tuple for argument storage. Instead, bound arguments are
> > now:
> > * stored directly if there is only one
On Wed, 20 Aug 2025 at 11:08, Tomasz Kamiński wrote:
>
> This patch adjusts the implementation of bind_front(f) and bind_back(f)
> (with zero bound arguments), so it returns an auto(f) (a copy of the
> functor), rather than a specialization of Bind_front/Bind_back. This change
> is mostly unobserv
On Tue, 26 Aug 2025 at 16:29, Tomasz Kamiński wrote:
>
> This patch introduces a new function, _M_fill_append, which is invoked when
> copies of the same value are appended to the end of a vector. Unlike
> _M_fill_insert(end(), n, v), _M_fill_append never permute elements in place,
> so it does no
On Tue, 19 Aug 2025 at 16:12, Tomasz Kamiński wrote:
>
> Previously, an empty functor (EmptyIdFunc) stored inside a
> std::move_only_function being first member of a Composite class could have the
> same address as the base of the EmptyIdFunc type (see included test cases),
> resulting in two obje
On Tue, Aug 26, 2025 at 4:30 PM Patrick Palka wrote:
> On Tue, 26 Aug 2025, Tomasz Kaminski wrote:
>
> >
> >
> > On Tue, Aug 26, 2025 at 3:54 PM Patrick Palka wrote:
> > On Wed, 20 Aug 2025, Tomasz Kamiński wrote:
> >
> > > This patch refactors the implementation of bind_front and
>
This patch introduces a new function, _M_fill_append, which is invoked when
copies of the same value are appended to the end of a vector. Unlike
_M_fill_insert(end(), n, v), _M_fill_append never permute elements in place,
so it does not require:
* vector element type to be assignable;
* a copy of t
Hi All,
The following patch has been bootstrapped and regtested on powerpc64le-linux.
Changes from V4:
* Refined AMO tests with proper variable names.
* Added tests for missed checks.
* Reworded comments.
Changes from V3:
Replaced named operands with positional operands in inline assembly for be
On Wed, 20 Aug 2025, Tomasz Kamiński wrote:
> This patch adjusts the implementation of bind_front(f) and bind_back(f)
> (with zero bound arguments), so it returns an auto(f) (a copy of the
> functor), rather than a specialization of Bind_front/Bind_back. This change
> is mostly unobservable, as st
On 20/08/25 10:13 +0200, Tomasz Kamiński wrote:
This patch refactors the implementation of bind_front and bind_back to
avoid using std::tuple for argument storage. Instead, bound arguments are
now:
* stored directly if there is only one,
* within a dedicated _Bound_arg_storage otehrwise.
"otehr
On Tue, Aug 26, 2025 at 04:09:11PM +0200, Anna (navi) Figueiredo Gomes wrote:
> --- a/gcc/c-family/c-common.h
> +++ b/gcc/c-family/c-common.h
> @@ -97,11 +97,11 @@ enum rid
>RID_IMAGINARY,
>
>/* C */
> - RID_INT, RID_CHAR, RID_FLOAT,RID_DOUBLE, RID_VOID,
> - RID_ENUM,RID_S
> Why do you reformat everything?
> The enumerators aren't sorted alphabetically, just add the new one
> after RID_BITINT.
i thought it was, but now i see it's not really, it can go in the end.
>
>> @@ -434,20 +434,21 @@ extern machine_mode c_default_pointer_mode;
>> #define D_CXXONLY 0x0002
Hi Sandra,
Sandra Loosemore wrote:
On 8/26/25 05:26, Jakub Jelinek wrote:
I wouldn't emit sorry for targets where we already implement it
"correctly",
i.e. at least all targets which have NULL
targetm.simd_clone.compute_vecsize_and_simdlen
There the argument transformation is identity.
That
On Tue, 19 Aug 2025, Tomasz Kamiński wrote:
> Previously, an empty functor (EmptyIdFunc) stored inside a
> std::move_only_function being first member of a Composite class could have the
> same address as the base of the EmptyIdFunc type (see included test cases),
> resulting in two objects of the
On Tue, 26 Aug 2025, Tomasz Kaminski wrote:
>
>
> On Tue, Aug 26, 2025 at 3:54 PM Patrick Palka wrote:
> On Wed, 20 Aug 2025, Tomasz Kamiński wrote:
>
> > This patch refactors the implementation of bind_front and bind_back to
> > avoid using std::tuple for argument storage. I
represent the start and end of scope which a local jump is not allowed
to cross as a jump barrier, moving this logic out into a function allows
for defer statements to use the same codepath
gcc/c/ChangeLog:
* c-decl.cc (struct c_jump_barrier): New structure.
(struct c_spot_binding
On Tue, Aug 26, 2025 at 3:54 PM Patrick Palka wrote:
> On Wed, 20 Aug 2025, Tomasz Kamiński wrote:
>
> > This patch refactors the implementation of bind_front and bind_back to
> > avoid using std::tuple for argument storage. Instead, bound arguments are
> > now:
> > * stored directly if there is
On 8/26/25 05:26, Jakub Jelinek wrote:
On Mon, Aug 25, 2025 at 08:24:18PM -0600, Sandra Loosemore wrote:
On 8/25/25 00:13, Jakub Jelinek wrote:
Yes, GCC doesn't have it implemented fully, but that doesn't mean it should
be ripped off, the implementation should be simply finished.
Well, I have
based on the technical specification n3589[1], and set behind the
command line flag '-fdefer-ts'
deferred statements execute at the end of the scope they're added on,
similar to the cleanup attribute, and they have much of the same
constraints as statement expressions, with added limitation that n
for future use with defer statements, as they do not have a declaration
associated, and are bound to the scope with a BIND_EXPR
gcc/c/ChangeLog:
* c-typeck.cc (push_cleanup): Adjust.
Signed-off-by: Anna (navi) Figueiredo Gomes
---
gcc/c/c-typeck.cc | 10 --
1 file changed, 8 in
* remove the need of -std=c2y for -fdefer-ts to take effect
---
* added a test jumping into a defer block from above.
* reworded the lookup_label_for_goto error message to mention the error
is a jump into the defer block, instead of out.
* moved the 'defer' keyword behind a -fdefer-ts flag, and a
On 26/08/2025 13:39, Jennifer Schmitz wrote:
> The function gimple_folder::fold_pfalse that was introduced in
> g5289540ed58e42ae66255e31f22afe4ca0a6e15e
> inappropriately folds
> 1) svbrka/b with _m if the inactive operand is pfalse
> 2) svpmov_lane with _m if the non-governing predicate operand i
On Wed, 20 Aug 2025, Tomasz Kamiński wrote:
> This patch refactors the implementation of bind_front and bind_back to
> avoid using std::tuple for argument storage. Instead, bound arguments are
> now:
> * stored directly if there is only one,
> * within a dedicated _Bound_arg_storage otehrwise.
I
Got it, thanks!
Patrick Palka 于2025年8月23日 周六02:59写道：
> On Fri, 22 Aug 2025, Yrong wrote:
>
> > friendly ping~
> >
> > Yrong 于2025年8月19日周二 23:58写道：
> > Thanks for the review! But I don't have a commit access, could you
> help me to commit this patch? Many thanks!
>
> The patch will be pushe
We can fold svbrka/b to pfalse, if the governing predicate and the operand
are both ptrue.
We implemented this optimization during gimple folding in
svbrk_unary_impl::fold.
The patch was bootstrapped and tested on aarch64-linux-gnu, no regression.
OK for trunk?
Signed-off-by: Jennifer Schmitz
g
The function gimple_folder::fold_pfalse that was introduced in
g5289540ed58e42ae66255e31f22afe4ca0a6e15e
inappropriately folds
1) svbrka/b with _m if the inactive operand is pfalse
2) svpmov_lane with _m if the non-governing predicate operand is pfalse.
To fix this, we added extra checks in the fol
On Tue, Aug 26, 2025 at 3:23 PM Jakub Jelinek wrote:
> On Mon, Jul 14, 2025 at 12:11:18PM +0200, Tomasz Kaminski wrote:
> > We have a preference to use [[likely]] attribute when possible.
>
> Done.
>
> > > + for (; __first != __last; ++__first, ++__result)
> > > + {
> > > +
On 10/08/2025 12:51, Torbjörn SVENSSON wrote:
Ok for trunk and releases/gcc-15?
For releases/gcc-15, I would also want to cherry-pick r16-562-g20c25919132 that
converts
the test to use function body instead of three scan-assembler.
Changes since v1:
- Removed the acceptance of LDR as it's only
Hi All,
The following patch has been bootstrapped and regtested on powerpc64le-linux.
Previously, vec_slo/vec_sll always default to V4SI, inserting unwanted
VIEW_CONVERT_EXPR int casts. This caused widening of char/short vectors,
constants
exceeding vspltisb/xxspltib range.
For example:
v
On Mon, 25 Aug 2025 at 13:42, Tomasz Kamiński wrote:
>
> This patch introduces a new function, _M_fill_append, which is invoked when
> copies of the same value are appended to the end of a vector. Unlike
> _M_fill_insert(end(), n, v), _M_fill_append never permute elements in place,
> so it does no
On Tue, 26 Aug 2025, Zhou Zhao wrote:
>
> 在 2025/8/26 下午6:52, Richard Biener 写道:
> > On Tue, 26 Aug 2025, Zhou Zhao wrote:
> >
> >> 在 2025/8/26 下午3:37, Richard Biener 写道:
> >>> On Thu, 21 Aug 2025, Zhou Zhao wrote:
> >>>
> This patch is a respond of the patch posted at
> https://gcc.gnu
On Thu, 21 Aug 2025 at 17:10, Tomasz Kamiński wrote:
>
> The move constructors for stop_source and stop_token are equivalent to
> copying and clearing the raw pointer, as they are wrappers for a
> counted-shared state.
>
> For jthread, the move constructor performs a member-wise move of stop_sourc
On 8/26/25 6:39 AM, Alfie Richards wrote:
The 08/19/2025 14:06, Jason Merrill wrote:
On 8/19/25 4:49 AM, Alfie Richards wrote:
Adds the target_version and target_clones attributes to diagnostic messages
for target_version semantics.
This is because the target_version/target_clones attributes a
On 8/26/25 6:33 AM, Alfie Richards wrote:
The 08/22/2025 12:22, Jason Merrill wrote:
On 8/19/25 4:48 AM, Alfie Richards wrote:
This patch changes the semantics of target_version and target_clones attributes
to match the behavior described in the Arm C Language extension.
The changes to behavio
On Tue, Aug 26, 2025 at 5:31 PM Jan Hubicka wrote:
>
> > > > > > In general we should have a look at register pressure, I
> > > > > > suppose issue_rate / m_num_reductions ensures we're never
> > > > > > getting close to this in practice.
> > > > >
> > > > > Bootstrapped and regtested on x86_64-pc
Hi Kees,
I think the tests are missing.
Best wishes,
Claudiu
On 8/26/25 9:17 AM, Kees Cook wrote:
> The __attribute__((__copy__)) functionality was crashing when copying
> sanitizer-related attributes because these attributes violated the standard
> GCC attribute infrastructure by storing INTEGE
On 8/26/25 6:06 AM, Richard Earnshaw wrote:
The compiler is getting too smart!
We try, we try :-)
But this test is really intended
to test that we generate BICS instead of BIC+CMP, so make the test use
something that we can't subsequently fold away into a bit minipulation
of a store-flag va
On 8/26/25 4:02 AM, Paul-Antoine Arras wrote:
This pattern enables the combine pass (or late-combine, depending on the case)
to merge a vec_duplicate into a div RTL instruction. The vec_duplicate is the
dividend operand.
Before this patch, we have two instructions, e.g.:
vfmv.v.f v2,
On 8/20/25 12:53 PM, Paul-Antoine Arras wrote:
This pattern enables the combine pass (or late-combine, depending on the case)
to merge a vec_duplicate into a mult RTL instruction.
Before this patch, we have two instructions, e.g.:
vfmv.v.f v2,fa0
vfmul.vv v1,v1,v2
After, we
Recent changes from Kito have an unused parameter. On the assumption
that he's going to likely want it as part of the API, I've simply
removed the parameter's name until such time as Kito needs it.
This should restore bootstrapping to the RISC-V port. Committing now
rather than waiting for t
The compiler is getting too smart! But this test is really intended
to test that we generate BICS instead of BIC+CMP, so make the test use
something that we can't subsequently fold away into a bit minipulation
of a store-flag value.
I've also added a couple of extra tests, so we now cover both th
在 2025/8/26 下午6:52, Richard Biener 写道:
On Tue, 26 Aug 2025, Zhou Zhao wrote:
在 2025/8/26 下午3:37, Richard Biener 写道:
On Thu, 21 Aug 2025, Zhou Zhao wrote:
This patch is a respond of the patch posted at
https://gcc.gnu.org/pipermail/gcc-patches/2025-January/673051.html
as some suggestion by
Dear Jakub, dear Sandra, hello world,
Jakub Jelinek wrote:
Starting with OpenMP 5.2, the spec explicitly says differently for C++:
"The function variant is determined by base language standard name lookup
rules ([basic.lookup]) of variant-name using the argument types at the call
site..."
I th
On Mon, Aug 25, 2025 at 08:24:18PM -0600, Sandra Loosemore wrote:
> On 8/25/25 00:13, Jakub Jelinek wrote:
> > Yes, GCC doesn't have it implemented fully, but that doesn't mean it should
> > be ripped off, the implementation should be simply finished.
>
> Well, I have no idea how to make this work
The following addresses a bogus swapping of SLP operands of a
reduction operation which gets STMT_VINFO_REDUC_IDX out of sync
with the SLP operand order. In fact the most obvious mistake is
that we simply swap operands even on the first stmt even when
there's no difference in the comparison operat
On Tue, 26 Aug 2025, Zhou Zhao wrote:
>
> 在 2025/8/26 下午3:37, Richard Biener 写道:
> > On Thu, 21 Aug 2025, Zhou Zhao wrote:
> >
> >> This patch is a respond of the patch posted at
> >> https://gcc.gnu.org/pipermail/gcc-patches/2025-January/673051.html
> >> as some suggestion by Richard Biener, I h
The 08/19/2025 14:06, Jason Merrill wrote:
> On 8/19/25 4:49 AM, Alfie Richards wrote:
> > Adds the target_version and target_clones attributes to diagnostic messages
> > for target_version semantics.
> >
> > This is because the target_version/target_clones attributes affect the
> > identity
> >
The 08/22/2025 12:22, Jason Merrill wrote:
> On 8/19/25 4:48 AM, Alfie Richards wrote:
> > This patch changes the semantics of target_version and target_clones
> > attributes
> > to match the behavior described in the Arm C Language extension.
> >
> > The changes to behavior are:
> >
> > - The s
This pattern enables the combine pass (or late-combine, depending on the case)
to merge a vec_duplicate into a div RTL instruction. The vec_duplicate is the
dividend operand.
Before this patch, we have two instructions, e.g.:
vfmv.v.f v2,fa0
vfdiv.vv v1,v2,v1
After, we get only on
Hi Yuao,
Yuao Ma wrote:
I addressed some review comments and included additional feedback.
Thanks for the update. I know that the patch is incomplete, but as I
started with bootstrapping, I start with the following. I bet the issues
are known to you:
First, your patch does not bootstrap* h
在 2025/8/26 下午3:37, Richard Biener 写道:
On Thu, 21 Aug 2025, Zhou Zhao wrote:
This patch is a respond of the patch posted at
https://gcc.gnu.org/pipermail/gcc-patches/2025-January/673051.html
as some suggestion by Richard Biener, I have adopted these suggestions
and regenerated the patch.
In
> > > > > In general we should have a look at register pressure, I
> > > > > suppose issue_rate / m_num_reductions ensures we're never
> > > > > getting close to this in practice.
> > > >
> > > > Bootstrapped and regtested on x86_64-pc-linu-gnu{-m32,}.
> > >
> > > This looks reasonable from my side
This patch allows profiles input in '--with-arch'. With this change,
profile names such as `rva22*`, `rvb23*`, or `rvi*` can be passed
directly to `--with-arch` without triggering configure errors. This
makes it easier to configure and build a toolchain targeting RISC-V
profiles, since the input
> -Original Message-
> From: Richard Biener
> Sent: Tuesday, August 26, 2025 8:59 AM
> To: gcc-patches@gcc.gnu.org
> Cc: Tamar Christina
> Subject: [PATCH][aarch64] Compute vect_reduc_type off SLP node instead of
> stmt-info
>
> The following changes the vect_reduc_type API to work on th
The following makes sure to read from the lanes_ifn member only
when necessary (and thus it was set).
Bootstrap and regtest running on x86_64-unknown-linux-gnu.
* tree-vect-stmts.cc (vectorizable_store): Access lanes_ifn
only when VMAT_LOAD_STORE_LANES.
(vectorizable_load)
1 - 100 of 103 matches
Mail list logo
