On Thu, Sep 4, 2025 at 5:50 AM Kyrylo Tkachov wrote:
>
> Hi all,
>
> Unlike Advanced SIMD, SVE has instruction to perform smin, smax, umin, umax
> on 64-bit elements. Thus, we can use them with the fixed-width V2DImode
> expander. Most of the machinery is already there on the define_insn side,
>
On Fri, Sep 5, 2025 at 1:50 AM Andrew Pinski
wrote:
>
> SRA likes to create VCE(a) when it comes to bool. This confuses
> a few different passes including jump threading and uninitialization
> warning. This removes the VCE in one case where it will help.
> Values outside of 0/1 with the VCE will p
Duly swung! Pushed as r16-3589.
Thanks
Paul
On Thu, 4 Sept 2025 at 17:21, Jerry D wrote:
> On 9/4/25 6:22 AM, Paul Richard Thomas wrote:
> > Hi All,
> >
> > Although PR87362 is marked as fixed, the error becomes rather more
> explicit with
> > this patch, which I actually developed for PR1024
Lower the partial limbs of extended _BitInts like the full limbs for most
operations, so that explicit extensions can be inserted only where they
are really needed.
gcc/ChangeLog:
* gimple-lower-bitint.cc (struct bitint_large_huge): Remove
the abi_load_p parameter of limb_access.
On Thu, Sep 4, 2025 at 7:17 PM Jakub Jelinek wrote:
>
> On Thu, Sep 04, 2025 at 05:59:25PM +0100, Iain Sandoe wrote:
> > PR c++/119060
> >
> > gcc/ChangeLog:
> >
> > * builtins.cc (expand_builtin): Handle BUILT_IN_OBSERVABLE_CHKPT.
> > * builtins.def (BUILT_IN_OBSERVABLE_CHKPT):
Previously the spec
https://github.com/riscv-non-isa/riscv-toolchain-conventions/pull/70
has changed the save/restore routines to save/restore the registers which
are really used for ILP32E/LP64 rather than always save/restore all
of ra/s0/s1.
I also found here that lacks the implementation for lp
Bootstrapped and regtested on x86_64-pc-linux-gnu{-m32,}.
Ready push to trunk.
gcc/ChangeLog:
* config/i386/i386-expand.cc (expand_vec_perm_vpermil): Extend
to handle V8SImode.
(avx_vpermilp_parallel): Extend to handle vector integer modes
with same vector size and
On Fri, Sep 5, 2025 at 9:21 AM Kito Cheng wrote:
>
> On Thu, Sep 4, 2025 at 11:50 PM Robin Dapp wrote:
> >
> > > The layout will be different between VLEN=128 and VLEN=256 (and also
> > > any larger VLEN)
> > >
> > > Give a practical example:
> > > vec1 allocated into v8, and v9, the reg layout w
Implement RISC-V-specific KCFI backend.
- Function preamble generation using .word directives for type ID storage
at offset from function entry point (no alignment NOPs needed due to
fix 4-byte instruction size).
- Scratch register allocation using t1/t2 (x6/x7) following RISC-V
procedure c
Michal Jireš writes:
> On 9/4/25 8:53 PM, Sam James wrote:
>> Sam James writes:
>>
>>> Andi Kleen writes:
>>>
Sam James writes:
> Michal Jires writes:
>
>> I did handle node->iterate_referring, but forgot cnode->callers.
>>
>> Only change are contents of the new
On Thu, Sep 4, 2025 at 11:50 PM Robin Dapp wrote:
>
> > The layout will be different between VLEN=128 and VLEN=256 (and also
> > any larger VLEN)
> >
> > Give a practical example:
> > vec1 allocated into v8, and v9, the reg layout will be:
> >
> > VLEN = 128
> > v8 = [0, 1, 2, 3]
> > v9 = [4, 5, 6
On Thu, Sep 04, 2025 at 05:50:45PM -0700, Andrew Pinski wrote:
> On Thu, Sep 4, 2025 at 5:27 PM Kees Cook wrote:
> > +
> > + /* Unknown builtin type - this should never happen in a well-formed C
> > program. */
> > + debug_tree (type);
> > + internal_error ("mangle: Unknown builtin type in fu
To support the KCFI type-id which needs to convert unique function
prototypes into unique 32-bit values, add a subset of the Itanium C++
mangling ABI for C typeinfo of function prototypes, but then do
hashing, which is needed by KCFI to get a 32-bit hash value for a
given function prototype. Option
On Thu, Sep 4, 2025 at 5:27 PM Kees Cook wrote:
>
> To support the KCFI type-id which needs to convert unique function
> prototypes into unique 32-bit values, add a subset of the Itanium C++
> mangling ABI for C typeinfo of function prototypes, but then do
> hashing, which is needed by KCFI to get
在 2025/9/4 下午7:48, Lulu Cheng 写道:
在 2025/8/22 下午4:14, Xi Ruoyao 写道:
In a CAS operation, even if expected != *memory we still need to do an
atomic load of *memory into output. But I made a mistake in the initial
implementation, causing the output to contain junk in this situation.
Like a nor
Implements the Linux Kernel Control Flow Integrity ABI, which provides a
function prototype based forward edge control flow integrity protection
by instrumenting every indirect call to check for a hash value before
the target function address. If the hash at the call site and the hash
at the target
SRA likes to create VCE(a) when it comes to bool. This confuses
a few different passes including jump threading and uninitialization
warning. This removes the VCE in one case where it will help.
Values outside of 0/1 with the VCE will produce undefined code so
removing the VCE is always fine as we
Implement ARM 32-bit KCFI backend supporting ARMv7+:
- Function preamble generation using .word directives for type ID storage
at -4 byte offset from function entry point (no prefix NOPs needed due to
4-byte instruction alignment).
- Use movw/movt instructions for 32-bit immediate loading.
-
Implement x86_64-specific KCFI backend:
- Implies -mindirect-branch-register since KCFI needs call target in
a register for typeid hash loading.
- Function preamble generation with type IDs positioned at -(4+prefix_nops)
offset from function entry point.
- Function-aligned KCFI preambles usi
Hi!
Here is v2, which is substantially changed compared to the RFC[1].
This series implements[2][3] the Linux Kernel Control Flow Integrity
ABI, which provides a function prototype based forward edge control flow
integrity protection by instrumenting every indirect call to check for
a hash value
Implement AArch64-specific KCFI backend.
- Function preamble generation using .word directives for type ID storage
at offset from function entry point (no default alignment NOPs needed
due to fixed 4-byte instruction size).
- Trap debugging through ESR (Exception Syndrome Register) encoding
Hi Honza,
> On 5 Sep 2025, at 1:30 am, Jan Hubicka wrote:
>
> External email: Use caution opening links or attachments
>
>
> Hi,
> with -fpartial-profling we ICE building perlbench and gcc from spec2k17 since
> afdo_annotate_cfg applies knowlede about zero profiles too early. This patch
> mo
On Thu, Sep 04, 2025 at 10:53:48PM +0200, Harald Anlauf wrote:
> Hi Paul!
>
> Am 04.09.25 um 20:46 schrieb Paul Richard Thomas:
> > Hi All,
> >
> > PDT components with default initializers must have type parameter and
> > length expressions that reduce to compile time integer constants. The chunk
> On Sep 4, 2025, at 16:15, Jakub Jelinek wrote:
>
> On Thu, Sep 04, 2025 at 07:47:17PM +, Qing Zhao wrote:
>>> On Sep 4, 2025, at 11:01, Jakub Jelinek wrote:
>>>
>>> On Thu, Sep 04, 2025 at 02:45:16PM +0200, Jakub Jelinek via Gcc wrote:
On Wed, Sep 03, 2025 at 03:38:53PM +0200, Jaku
Hi Paul!
Am 04.09.25 um 20:46 schrieb Paul Richard Thomas:
Hi All,
PDT components with default initializers must have type parameter and
length expressions that reduce to compile time integer constants. The chunk
in expr.cc verifies that this is the case for array bounds and character
lengths.
Hi Torbjorn,
Sorry for the delay
On Mon, 1 Sept 2025 at 17:14, Torbjorn SVENSSON
wrote:
>
>
>
> On 2025-09-01 16:59, Christophe Lyon wrote:
> > On Wed, 27 Aug 2025 at 13:25, Torbjorn SVENSSON
> > wrote:
> >>
> >>
> >>
> >> On 2025-08-18 19:24, Christophe Lyon wrote:
> >>> Like we do in othe
On Mon, 11 Aug 2025, mmalcom...@nvidia.com wrote:
> + /* Ensure that this doesn't get optimised out of the COMPOUND_EXPR we
> + * define below. It appears on first glance that the fact the
> + * initialisation argument is a function call would mean this
> + * automatically
On 9/4/25 8:53 PM, Sam James wrote:
Sam James writes:
Andi Kleen writes:
Sam James writes:
Michal Jires writes:
I did handle node->iterate_referring, but forgot cnode->callers.
Only change are contents of the newly separated
mark_symbol_referenced_from_asm
Thanks, I'll try the n
Sam James writes:
> Andi Kleen writes:
>
>> Sam James writes:
>>
>>> Michal Jires writes:
>>>
I did handle node->iterate_referring, but forgot cnode->callers.
Only change are contents of the newly separated
mark_symbol_referenced_from_asm
>>>
>>> Thanks, I'll try the new pa
On 9/1/25 3:53 PM, Aleksandar Rakic wrote:
Hi Jeff,
[PATCH v2 02/12] Fix unsafe comparison against stack_pointer_rtx:
https://sourceware.org/pipermail/gcc-patches/2025-March/677829.html
Is this still an issue? I thought this was fixed a while back in the
renamer. I don't necessarily thing
Hi All,
PDT components with default initializers must have type parameter and
length expressions that reduce to compile time integer constants. The chunk
in expr.cc verifies that this is the case for array bounds and character
lengths.
This error checking results in pdt_26.f03 segfaulting because
This patch refactors ranges::_Partial to be implemented using _Bind_back_t.
This allows it to benefit from the changes in r16-3398-g250dd5b5604fbc,
specifically making the closure trivially copyable. Since _Bind_back_t
already provides an optimized implementation for a single bound argument,
specia
Sorry, this patch seems to be causing a regression in the testsuite.
I'll come up with a fixed version.
On 04/09/2025 13:14, Paul-Antoine Arras wrote:
This pattern enables the combine pass (or late-combine, depending on the case)
to merge a vec_duplicate into a plus RTL instruction.
Before thi
On Thu, 4 Sept 2025, 18:02 Iain Sandoe, wrote:
> The facility (with the original shorter name) has been in use on the
> contracts development branch for almost a year, and has been tested in
> isolation on x86_64-darwin and powerpc64le-linux.
> OK for trunk?
> thanks
> Iain
>
> --- 8< ---
>
> Thi
On Wed, 3 Sep 2025, Alejandro Colomar wrote:
> Hi Joseph,
>
> On Wed, Sep 03, 2025 at 03:44:28PM +, Joseph Myers wrote:
> > On Wed, 3 Sep 2025, Alejandro Colomar wrote:
> >
> > > Hi Joseph,
> > >
> > > I'd like to ping about this thread.
> >
> > As far as I know, nothing has been resolved
Sam James writes:
> Michal Jires writes:
>
>> I did handle node->iterate_referring, but forgot cnode->callers.
>>
>> Only change are contents of the newly separated
>> mark_symbol_referenced_from_asm
>
> Thanks, I'll try the new patch now.
>
> With the workaround I mentioned earlier, I managed t
On 03/09/2025 13:33, Kyrylo Tkachov wrote:
Hi Karl,
On 2 Sep 2025, at 16:16, Karl Meakin wrote:
gcc/ChangeLog:
* config/aarch64/aarch64-sme.md (@aarch64_sme_write_zt): New
insn.
(aarch64_sme_lut_zt): Likewise.
* config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): New type format
On Thu, 4 Sept 2025 at 09:57, Richard Biener wrote:
>
> On Thu, 4 Sep 2025, Richard Biener wrote:
>
> > On Thu, Sep 4, 2025 at 10:27 AM Jonathan Wakely wrote:
> > >
> > > On Thu, 4 Sept 2025 at 08:26, Richard Biener wrote:
> > > >
> > > > On Wed, 3 Sep 2025, Jakub Jelinek wrote:
> > > >
> > > >
On Wed, Sep 03, 2025 at 09:24:22PM -0700, Kees Cook wrote:
> > If the hacker knows these, it should be quite easy for them to come up with
> > a
> > matched typeid, is it?
>
> The hashes aren't considered secret -- they need to be known/match between
> compilation units, and even across languag
On Wed, Sep 3, 2025 at 6:03 PM Jonathan Wakely wrote:
> On Wed, 3 Sept 2025 at 16:30, Luc Grosheintz
> wrote:
> >
> > In r16-2328-g29d53f6213e0a1 we fixed a bug related to user-defined
> > objects that can convert to an integers only via an rvalue reference.
> > The same commit also implemented
Jan Hubicka writes:
> With -O2 we automatically enable several loop optimizations with
> -fprofile-use.
> The rationale is that those optimizations at -O3 only mainly since they may
> hurt performance or not pay back in code size when used blindly on all loops.
> Profile feedback gives us data o
bits/binders.h is already mapped in libstdc++-v3/doc/doxygen/stdheader.cc.
libstdc++-v3/ChangeLog:
* include/Makefile.am: Add bits/binders.h
* include/Makefile.in: Add bits/binders.h
* include/std/functional (std::_Indexed_bound_arg, std::_Binder)
(std::__make_boun
v2: Moves change from std::indirect to std::__indirect to first patch,
so second patch is plain move, without sneaky changes.
Patch 3 is not modified.
I haven't implemented Nathan suggestions to use enum instead of bool,
in my opinion this add miniscule compile time for no user benefit.
And we alr
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
On Tue, Aug 26, 2025 at 06:22:31PM +, Qing Zhao wrote:
> 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
On Thu, 4 Sept 2025 at 10:04, Tomasz Kamiński wrote:
>
> v2: Moves change from std::indirect to std::__indirect to first patch,
Ah, invoke ... I was confused what this had to do with std::indirect! :-)
> so second patch is plain move, without sneaky changes.
> Patch 3 is not modified.
>
> I have
On Tue, 19 Aug 2025 at 16:24, Patrick Palka wrote:
>
> On Wed, 16 Jul 2025, Tomasz Kaminski wrote:
>
> >
> >
> > On Tue, Jul 15, 2025 at 6:13 PM Patrick Palka wrote:
> > Tested on x86_64-pc-linux-gnu, does this look OK for trunk only
> > (since it impacts ABI)?
> >
> > Changes i
On Thu, 4 Sept 2025 at 09:40, Richard Biener wrote:
>
> On Thu, Sep 4, 2025 at 10:27 AM Jonathan Wakely wrote:
> >
> > On Thu, 4 Sept 2025 at 08:26, Richard Biener wrote:
> > >
> > > On Wed, 3 Sep 2025, Jakub Jelinek wrote:
> > >
> > > > On Wed, Sep 03, 2025 at 06:41:08PM +0200, Richard Biener w
On 18/08/25 13:55, dhr...@nvidia.com wrote:
External email: Use caution opening links or attachments
From: Dhruv Chawla
This patch is a respin of the RFC originally posted at
https://gcc.gnu.org/pipermail/gcc-patches/2025-June/686835.html.
The patch reads the file names from the GCOV file an
On Thu, Sep 04, 2025 at 05:59:25PM +0100, Iain Sandoe wrote:
> PR c++/119060
>
> gcc/ChangeLog:
>
> * builtins.cc (expand_builtin): Handle BUILT_IN_OBSERVABLE_CHKPT.
> * builtins.def (BUILT_IN_OBSERVABLE_CHKPT): New.
>
> gcc/c-family/ChangeLog:
>
> * c-common.cc: Add __b
On Thu, Sep 4, 2025 at 12:28 AM Jeff Law wrote:
>
>
>
> On 9/3/25 1:27 PM, Sam James wrote:
> > Back in GCC 8, with r8-5241-g8697bf9f46f361 (-gstatement-frontiers),
> > r8-6658-g58006663903200, r8-6657-gbd2b9f1e2d67ec
> > (-gvariable-location-views),
> > some advanced GNU extensions to DWARF were
The _Bind_front and _Bind_back class templates are now merged into a single
_Binder implementation that accepts _Back as a template parameter. This makes
the bind_back implementation available in C++20 mode, allowing it to be used
for range adaptor closures.
With zero bound arguments, bind_back an
Richard Biener writes:
> On Wed, 3 Sep 2025, Richard Sandiford wrote:
>
>> Tamar Christina writes:
>> > We also don't ever force unrolling for predicated SVE because for
>> > predicated SVE we have to balance predicate throughput limitations
>> > of any given CPU. Having the user unroll factor b
Jakub Jelinek writes:
> On Thu, Sep 04, 2025 at 09:58:14AM +0200, Richard Biener wrote:
>> I'm in favor of disabling but I also fear the code will bitrot quickly if so?
>> We might also want to have a set of testcases for the -fcompare-debug
>> issues that are fixed by the change of defaults?
>>
Sam James writes:
> Michal Jires writes:
>
>> I did handle node->iterate_referring, but forgot cnode->callers.
>>
>> Only change are contents of the newly separated
>> mark_symbol_referenced_from_asm
>
> Thanks, I'll try the new patch now.
>
> With the workaround I mentioned earlier, I managed t
Hello,
This is the second version of the patch proposed for master aiming to fix
PR119702. I request the review of this patch.
The following sequence of assembly in powerpc64le
vspltisw 0,1
vsld 2,2,0
is replaced by this
vaddudm 2,2,2
whenever there is a vector left shi
This is a partial implementation of P2781R9. It adds std::cw and
std::constant_wrapper, but doesn't modify __integral_constant_like for
span/mdspan.
libstdc++-v3/ChangeLog:
* include/bits/version.def (constant_wrapper): Add.
* include/bits/version.h: Regenerate.
* include/
The following removes back-and-forth of state in
vect_create_epilog_for_reduction and code that's pointless, in
particular around double reduction handling which isn't that
special as it seems.
Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
Richard.
* tree-vect-loop.cc (v
The facility (with the original shorter name) has been in use on the
contracts development branch for almost a year, and has been tested in
isolation on x86_64-darwin and powerpc64le-linux.
OK for trunk?
thanks
Iain
--- 8< ---
This implements the library parts of P1494 as amended by P3641. For G
This patch (with the original shorter name) has been in use on the contracts
development branch for almost a year, and has been tested in isolation on
x86_64-darwin and powerpc64le-linux.
OK for trunk?
thanks
Iain
--- 8< ---
The paper provides a mechanism that serves to demarc epochs within the c
This implements the "observable_checkpoint()" functionality that serves
as an epoch delimiter in the source code, where it stipulates that any
code before this (and the checkpoint) are considered observable.
When code that would exhibit undefined behaviour if reached is seen then
that can result i
This is a slightly amended patch that fixes modes and instruction type
attribute. Here is the relevant snippet:
diff --git gcc/config/riscv/autovec-opt.md gcc/config/riscv/autovec-opt.md
index d4335dc04ba..67f4d9ce3a8 100644
--- gcc/config/riscv/autovec-opt.md
+++ gcc/config/riscv/autovec-opt.md
I did handle node->iterate_referring, but forgot cnode->callers.
Only change are contents of the newly separated
mark_symbol_referenced_from_asm
---
This new pass heuristically detects symbols referenced by toplevel
assembly to prevent their optimization.
Heuristics is done by comparing
On Thu, Sep 4, 2025 at 2:23 PM Luc Grosheintz
wrote:
> A usecase for P2781R9 is more ergonomic creation of span and mdspan with
> mixed static and dynamic extents, e.g.:
>
> span(ptr, cw<3>)
> extents(cw<3>, 5, cw<7>)
> mdspan(ptr, cw<3>, 5, cw<7>)
>
> should be deduced as:
> span
Dear Richard,No, I don’t have access to gcc git. Anyway, I updated the patch description and attached here the new patch.Best regards,Matteo
tree-optimization-121595.patch
Description: Binary data
On Sep 2, 2025, at 1:57 PM, Richard Biener wrote:On Fri, Aug 29, 2025 at 2:20 PM Matteo Nicoli wrote
> Toplevel assembly is now streamed as partitioned instead of into the
> first partition.
>
> gcc/ChangeLog:
>
> * lto-cgraph.cc (output_symtab): Remove asm_nodes_out.
> * lto-streamer-out.cc (lto_output_toplevel_asms): Use
> partitioning.
> (create_order_remap): Remove as
Add an expander for isnan using integer arithmetic. Since isnan is
just a compare, enable it only with -fsignaling-nans to avoid
generating spurious exceptions. This fixes part of PR66462.
int isnan1 (float x) { return __builtin_isnan (x); }
Before:
fcmps0, s0
csetw0, v
> 1to1 partitioning now also partitions toplevel assembly.
> Other partitionings keep the old behavior of putting all
> toplevel assembly into single partition.
>
> gcc/ChangeLog:
>
> * lto-cgraph.cc (compute_ltrans_boundary): Add asm_node.
>
> gcc/lto/ChangeLog:
>
> * lto-partition
On 9/4/25 1:58 AM, Richard Biener wrote:
I'm in favor of disabling but I also fear the code will bitrot quickly if so?
Quite likely. But given consumer support doesn't seem to be on the way
for any reasonable time horizon, that may not be that big of a problem.
We might also want to hav
On Wed, Sep 3, 2025 at 11:20 PM Patrick Palka wrote:
> On Wed, 3 Sep 2025, Tomasz Kamiński wrote:
>
> > The _Bind_front and _Bind_back class templates are now merged into a
> single
> > _Binder implementation that accepts _Back as a template parameter. This
> makes
> > the bind_back implementatio
> This patch replaces symtab_node with toplevel_node in lto_symtab_encoder
> and modifies all places where lto_symtab_encoder is used to handle
> (ignore) asm_node.
>
> gcc/ChangeLog:
>
> * ipa-icf.cc (sem_item_optimizer::write_summary): Use
> toplevel_node.
> (sem_item_optimize
On Wed, Sep 3, 2025 at 11:37 PM Patrick Palka wrote:
>
> On Wed, 3 Sep 2025, Tomasz Kamiński wrote:
>
> > This patch refactors ranges::_Partial to be implemented using
> _Bind_back_t.
> > This allows it to benefit from the changes in r16-3398-g250dd5b5604fbc,
> > specifically making the closure t
> Minor simplification as preparation for next patch.
>
> gcc/lto/ChangeLog:
>
> * lto-partition.cc (lto_balanced_map): Simplify.
OK,
balanced partitining needs either cleanups or obsoletting by the cached
algorithm.
It was really meant as quick and fast implementation of something that
wor
> asm_node and symbol_node will now inherit from toplevel_node.
> This is now useful for lto partitioning, in future it should be also
> useful for toplevel extended assembly.
>
> gcc/ChangeLog:
>
> * cgraph.h (enum symtab_type): Replace with toplevel_type.
> (enum toplevel_type): New
On Wed, 3 Sept 2025 at 19:31, Jonathan Wakely wrote:
>
> On Wed, 3 Sept 2025 at 19:26, Jonathan Wakely wrote:
> >
> > On Wed, 3 Sept 2025 at 19:09, Jonathan Wakely wrote:
> > >
> > > On Tue, 19 Aug 2025 at 16:17, Patrick Palka wrote:
> > > >
> > > > LGTM! Perhaps we want to backport this, not sur
> We use lto_file_data in 1to1 partitioning, so we need to not zero it
> out. Nothing depends on lto_file_data being NULL.
>
> gcc/ChangeLog:
>
> * cgraph.cc (cgraph_node::release_body): Keep lto_file_data.
> (cgraph_node::remove): likewise.
> * lto-section-in.cc (lto_free_funct
The following changes how we detect double reductions, in particular
not setting vect_double_reduction_def on the outer PHIs when the inner
loop doesn't satisfy double reduction constraints. It also simplifies
the setup a bit by not having to detect wheter we process an inner
loop of a double redu
On Thu, Sep 4, 2025 at 2:49 PM Jonathan Wakely wrote:
> On Thu, 4 Sept 2025 at 10:04, Tomasz Kamiński wrote:
> >
> > v2: Moves change from std::indirect to std::__indirect to first patch,
>
> Ah, invoke ... I was confused what this had to do with std::indirect! :-)
>
I have no idea why indirect
From: Pan Li
This patch would like to combine the vec_duplicate + vmadd.vv to the
vmadd.vx. From example as below code. The related pattern will depend
on the cost of vec_duplicate from GR2VR. Then the late-combine will
take action if the cost of GR2VR is zero, and reject the combination
if t
When producing output, the libstdc++ format implementation only uses _Sink_iter
specializations. Since users cannot construct basic_format_context, this is the
only iterator type actually used. The __format_padded helper relies on this
property to efficiently pad sequences from tuples and ranges.
Insufficient validation of the operands in vec_set__internal
means that the optimizers can transform the exanded code into
something that is invalid. We then emit code based on the incorrect
RTL assuming that it is still valid. A valid pattern can only have a
single bit set in the immediate opera
Part of what is needed to fix the BZ, tested on x86_64-darwin and
powerpc64le-linux, pushed to trunk, thanks,
Iain
--- 8< ---
Currently, for Darwin unwind and EH frames are emitted without use
of .cfi_xxx instructions; the emitted frames also contain the
string 'ascii'. For the purpose of this t
Hi all,
Unlike Advanced SIMD, SVE has instruction to perform smin, smax, umin, umax
on 64-bit elements. Thus, we can use them with the fixed-width V2DImode
expander. Most of the machinery is already there on the define_insn side,
supporting V2DImode operands of the SVE pattern. We just need to
> Sorting by node order in lto partitioning is incorrectly reversed.
> For default balanced partitioning this caused all noreorder symbols
> to be partitioned into a single partition where they were sorted again,
> but correctly.
Yep, up to now the order was not really that important, but processi
This patch adds runtime thread count detection to auto-parallelization.
-ftree-parallelize-loops=0 option generates parallelized loops without
specifying a fixed thread count, deferring this decision to program execution
time where it is controlled by the OMP_NUM_THREADS environment variable.
The
Hi,
with -fpartial-profling we ICE building perlbench and gcc from spec2k17 since
afdo_annotate_cfg applies knowlede about zero profiles too early. This patch
moves it after the early exit when profile is 0 everywhere and also fixes
formatting issue in the next block.
Bootstrapped/regtesed x86_64
On 9/4/25 6:22 AM, Paul Richard Thomas wrote:
Hi All,
Although PR87362 is marked as fixed, the error becomes rather more explicit with
this patch, which I actually developed for PR102457.
Regtests on FC42/x86_64 - OK for mainline
Paul
Yes, OK
Swing Away.
Jerry
On Fri, 01 Aug 2025 at 19:01 -0400, Patrick Palka wrote:
Three small review comments below:
On Fri, 1 Aug 2025, Nathan Myers wrote:
Changes in v8:
* Adjust template indentation to match rest of file
* Change std Mandates conditions from "requires" to static_asserts.
* Make _Bind_fn_t defini
On Thu, Sep 4, 2025 at 9:05 AM Robin Dapp wrote:
>
> > Given we use a poly_int64 for bound_epilog elsewhere now the best thing to
> > do would be to have a poly_int64 for bound_prolog as well. For the scaling
> > we'd use estimated_poly_value (align_in_elems) then (I guess for alignment
> > the d
Michal Jires writes:
> I did handle node->iterate_referring, but forgot cnode->callers.
>
> Only change are contents of the newly separated
> mark_symbol_referenced_from_asm
Thanks, I'll try the new patch now.
With the workaround I mentioned earlier, I managed to build but got this
when booting
The layout will be different between VLEN=128 and VLEN=256 (and also
any larger VLEN)
Give a practical example:
vec1 allocated into v8, and v9, the reg layout will be:
VLEN = 128
v8 = [0, 1, 2, 3]
v9 = [4, 5, 6, 7]
VLEN=256
v8 = [0, 1, 2, 3, 4, 5, 6, 7]
v9 = [?, ?, ?, ?, ?, ?, ?, ?]
Then you c
Document missing isinf optab.
gcc:
PR middle-end/101852
* doc/md.texi: Document isinf optab.
---
diff --git a/gcc/doc/md.texi b/gcc/doc/md.texi
index
973c0dd302964966a91fa8dbab85930d6dbeec9e..a9c3354891551101d25ba2e6656711dbd9c5dd09
100644
--- a/gcc/doc/md.texi
+++ b/gcc/doc/m
On Thu, Sep 04, 2025 at 02:45:16PM +0200, Jakub Jelinek via Gcc wrote:
> On Wed, Sep 03, 2025 at 03:38:53PM +0200, Jakub Jelinek via Gcc wrote:
> > But there is one thing the paper doesn't care about, which looks like a show
> > stopper to me, in particular the stuff -Wtrivial-auto-var-init warning
Here is an updated patch that fixes scan dumps in the testsuite:
diff --git
gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/floating-point-add-2.c
gcc/testsuite/gcc.target/riscv/rvv/autovec/vls/floating-point-add-2.c
index 042dd0d5ccc..00b9222e765 100644
--- gcc/testsuite/gcc.target/riscv/rvv/a
> On Sep 4, 2025, at 00:24, Kees Cook wrote:
>
>> At the same time, A wrapper type is created for the original function type,
>> whose typename is
>> “__kcfi_wrapper_type_id”.
>>
>> I am confused:
>>
>> 1. Why the additional wrapper type is needed? why the original function
>> type + “kcf
On Thu, Sep 4, 2025 at 2:20 PM Luc Grosheintz
wrote:
> This is a partial implementation of P2781R9. It adds std::cw and
> std::constant_wrapper, but doesn't modify __integral_constant_like for
> span/mdspan.
>
> libstdc++-v3/ChangeLog:
>
> * include/bits/version.def (constant_wrapper): Ad
On 9/4/25 2:20 PM, Luc Grosheintz wrote:
This is a partial implementation of P2781R9. It adds std::cw and
std::constant_wrapper, but doesn't modify __integral_constant_like for
span/mdspan.
libstdc++-v3/ChangeLog:
* include/bits/version.def (constant_wrapper): Add.
* include/
Hi All,
Although PR87362 is marked as fixed, the error becomes rather more explicit
with this patch, which I actually developed for PR102457.
Regtests on FC42/x86_64 - OK for mainline
Paul
Change.Logs
Description: Binary data
diff --git a/gcc/fortran/decl.cc b/gcc/fortran/decl.cc
index 1e91b57
> It's not really a big deal but at least a bit surprising. If it's just that
> then pre-approved.
Thanks Robin, I will rename it and commit if no surprise from test.
Pan
-Original Message-
From: Robin Dapp
Sent: Thursday, September 4, 2025 8:34 PM
To: Li, Pan2 ; Robin Dapp ;
gcc-pa
So before we had vmacc_vx and now madd can be included. Is this somehow
different to pred_mul_plus (without vx)? "mul then plus" sounds like there
is some operand order that differs from the regular order but the
multiplication is always first IIRC? The difference is just which operand is
bei
1 - 100 of 123 matches
Mail list logo
