Hi,
This is minor code refactorings in tree-vect-data-refs.c and tree-vect-loop.c.
Use LOOP_VINFO_DATAREFS and LOOP_VINFO_DDRS when possible and rename
several parameters to make code more consistent.
Bootstrapped and tested on aarch64-linux-gnu. OK?
Thanks,
Felix
gcc/
+2020-06-13 Felix
This patch adds a partial omp-tools.h from OpenMP project which
declares function prototypes and typedefs used in OMPD.
This patch also addressed all feedback.
2020-06-12 Tony Sim
libgomp/ChangeLog:
* Makefile.am: Added new header.
* Makefile.in: Regenerate.
*
This change by Clément Chigot corrects the value of arenaBaseOffset
for aix/ppc. The existing value was correct for aix/ppc64, but not
for 32-bit ppc. Bootstrapped and ran Go testsuite on
x86_64-pc-linux-gnu. Committed to master.
Ian
75452d68672ff7da6e5a167924b6aeb07b5b2ed1
diff --git
Hi!
On Thu, Jun 11, 2020 at 11:22:33PM -0500, will schmidt wrote:
> Fix codegen implementation for the builtin vec_pack_to_short_fp32.
> +;; Convert two vector F32 to packed vector f16.
> +(define_expand "convert_4f32_8f16"
> + [(set (match_operand:V8HI 0 "register_operand" "=v")
> +
On Fri, Jun 12, 2020 at 10:46:33AM +0200, Richard Biener wrote:
> On Fri, Jun 12, 2020 at 2:31 AM Segher Boessenkool
> wrote:
> > Why did this work before, and not anymore? That sounds like a missing
> > or broken test?
>
> It "worked" before because GIMPLE IL checking is not strict here
>
Hi,
when Fortran functions pass array descriptors they receive as a
parameter to another function, they actually rebuild it. Thanks to
work done mainly by Feng, IPA-CP can already handle the cases when
they pass directly the values loaded from the original descriptor.
Unfortunately, perhaps the
On Tue, May 19, 2020 at 6:35 AM CHIGOT, CLEMENT wrote:
>
> Description:
> On aix/ppc64, it's possible to allocate an array of 1<<50 int, because of
> the
> wider address space. Thus, increase it to 1<<59 like in Golang toolchain in
> order to trigger the panic.
>
> Changelog:
> 2020-05-19
On Jun 11, 2020, at 7:28 AM, Martin Jambor wrote:
>
> On Tue, Jun 09 2020, Mike Stump wrote:
>> I think I'd prefer the fix on the other side, if reasonable. I'd give
>> them some time to see about a fix there before selecting this patch.
>
> given Alexandre's email, are you OK with the patch?
Iain Sandoe wrote:
> Nathan Sidwell wrote:
>
>> On 6/8/20 5:17 AM, Iain Sandoe wrote:
>>> + r = gro_is_void_p ? integer_zero_node : rvalue (gro);
>>> + /* The return object is constructed, even if the gro is void. */
>>
>> Would error_mark_node work here? I presume we've already
On Fri, Jun 12, 2020 at 7:50 AM Przemyslaw Wirkus
wrote:
>
> Hi all,
>
> Pattern "(x | y) - y" can be optimized to simple "(x & ~y)" andn pattern.
Isn't it better to do this transformation on the gimple level and not
in a target specific form? Or at least do it in the RTL level in a
generic
On 6/12/20 11:49 AM, Richard Sandiford wrote:
Martin Liška writes:
On 6/12/20 3:22 PM, Richard Sandiford wrote:
Martin Liška writes:
diff --git a/gcc/tree-vect-patterns.c b/gcc/tree-vect-patterns.c
index 636ad59c001..eac372e6abc 100644
--- a/gcc/tree-vect-patterns.c
+++
On Fri, Jun 12, 2020 at 01:56:05PM -0400, y2s1982 wrote:
> This patch adds partial omp-tools.h continaing OMPD-specific functions and
> data types.
>
> 2020-06-12 Tony Sim
You'll need two spaces before < for the script to recognize it as ChangeLog
snippet start (though, it will recognize just
This patch adds partial omp-tools.h continaing OMPD-specific functions and data
types.
2020-06-12 Tony Sim
libgomp/ChangeLog:
* omp-tools.h: New file
---
libgomp/omp-tools.h | 278
1 file changed, 278 insertions(+)
create mode 100644
On Fri, Jun 12, 2020 at 7:30 AM 戎杰杰(无音) wrote:
> - The first version for vector extension and verified on rv64imafdcv linux
> target with qemu.
There is no RISC-V vector extension. There is only a draft proposal
to add one to the ISA. Our policy has always been to wait for draft
proposals to
Martin Liška writes:
> On 6/12/20 3:22 PM, Richard Sandiford wrote:
>> Martin Liška writes:
>>> diff --git a/gcc/tree-vect-patterns.c b/gcc/tree-vect-patterns.c
>>> index 636ad59c001..eac372e6abc 100644
>>> --- a/gcc/tree-vect-patterns.c
>>> +++ b/gcc/tree-vect-patterns.c
>>> @@ -5120,20
On June 12, 2020 5:29:40 PM GMT+02:00, "Martin Liška" wrote:
>On 6/12/20 3:02 PM, Richard Biener wrote:
>> For(gimple. *stmt : bb_vinfo->region_stmts())
>
>Hm, that will require region_stmts() returning one another object
>that will contain the begin/end methods.
>Is it what we want or do I miss
On Fri, 12 Jun 2020 at 17:27, Martin Liška wrote:
>
> On 6/12/20 4:19 PM, Marco Elver wrote:
> > On Fri, 12 Jun 2020, Jakub Jelinek wrote:
> >
> >> On Fri, Jun 12, 2020 at 04:07:57PM +0200, Marco Elver wrote:
> >>> gcc/ChangeLog:
> >>>
> >>> * params.opt: Add
On 6/12/20 3:02 PM, Richard Biener wrote:
For(gimple. *stmt : bb_vinfo->region_stmts())
Hm, that will require region_stmts() returning one another object
that will contain the begin/end methods.
Is it what we want or do I miss something?
Martin
On 6/12/20 3:22 PM, Richard Sandiford wrote:
Martin Liška writes:
diff --git a/gcc/tree-vect-patterns.c b/gcc/tree-vect-patterns.c
index 636ad59c001..eac372e6abc 100644
--- a/gcc/tree-vect-patterns.c
+++ b/gcc/tree-vect-patterns.c
@@ -5120,20 +5120,12 @@ vect_determine_precisions (vec_info
On 6/12/20 4:19 PM, Marco Elver wrote:
On Fri, 12 Jun 2020, Jakub Jelinek wrote:
On Fri, Jun 12, 2020 at 04:07:57PM +0200, Marco Elver wrote:
gcc/ChangeLog:
* params.opt: Add --param=tsan-instrument-func-entry-exit=.
* tsan.c (instrument_gimple): Make return value if func
Hi all,
Pattern "(x | y) - y" can be optimized to simple "(x & ~y)" andn pattern.
So, for the example code:
$ cat main.c
int
f_i(int x, int y)
{
return (x | y) - y;
}
long long
f_l(long long x, long long y)
{
return (x | y) - y;
}
typedef int v4si __attribute__ ((vector_size
On Fri, 2020-06-12 at 10:24 -0400, David Edelsohn wrote:
> Hi, Will
Hi,
>
> On Fri, Jun 12, 2020 at 12:22 AM will schmidt <
> will_schm...@vnet.ibm.com> wrote:
> >
> >
> > Hi,
> > Fix codegen implementation for the builtin
> > vec_pack_to_short_fp32.
> >
> > Regtests are underway
gcc/ChangeLog:
* config.gcc (extra_objs): Include vector pass.
* config/riscv/t-riscv : Likewise.
* config/riscv/riscv-pass.c : New.
* config/riscv/riscv.opt (-mdelete-vsetvl): Likewise.
---
gcc/config.gcc| 2 +-
gcc/config/riscv/riscv-pass.c |
gcc/ChangeLog:
* config/riscv/riscv-c.c (__riscv_vector): Define.
* config/riscv/riscv-opts.h (riscv_vlen): New.
* config/riscv/riscv-v.h (TARGET_VECTOR_VLEN): Define.
(UNITS_PER_VR_REG): Likewise.
(RISCV_VSEW_T): Likewise.
* config/riscv/riscv.opt
gcc/ChangeLog:
* config/riscv/riscv-v.h: New.
* config/riscv/riscv.h (FIRST_PSEUDO_REGISTER): Add vector register.
(FIRST_PSEUDO_REGISTER): Likewise.
(CALL_USED_REGISTERS): Likewise.
(reg_class): Likewise.
(REG_CLASS_NAMES): Likewise.
- The first version for vector extension and verified on rv64imafdcv linux
target with qemu.
gcc/ChangeLog:
* common/config/riscv/riscv-common.c: Parse vector extension.
* config/riscv/riscv.opt (TARGET_VECTOR): New.
---
gcc/common/config/riscv/riscv-common.c | 4
Hi, Will
On Fri, Jun 12, 2020 at 12:22 AM will schmidt wrote:
>
>
> Hi,
> Fix codegen implementation for the builtin vec_pack_to_short_fp32.
>
> Regtests are underway against powerpc64 (power7be,power8le,power9le).
> (this is a power9 builtin, so should be a noop for most of those).
> OK
On Fri, 12 Jun 2020, Jakub Jelinek wrote:
> On Fri, Jun 12, 2020 at 04:07:57PM +0200, Marco Elver wrote:
> > gcc/ChangeLog:
> >
> > * params.opt: Add --param=tsan-instrument-func-entry-exit=.
> > * tsan.c (instrument_gimple): Make return value if func entry
> > and exit should be
For real offloading compilers, the configure-time ENABLE_OFFLOAD
macro is set to true. If it is set, some additional code paths
are enabled – which can affect code generation with -fopenmp/-fopenacc.
As the PR shows, this can make differences for passing/failing a
testcase. – In principle, it
On Fri, Jun 12, 2020 at 04:07:57PM +0200, Marco Elver wrote:
> gcc/ChangeLog:
>
> * params.opt: Add --param=tsan-instrument-func-entry-exit=.
> * tsan.c (instrument_gimple): Make return value if func entry
> and exit should be instrumented dependent on param.
>
>
Adds param tsan-instrument-func-entry-exit, which controls if
__tsan_func_{entry,exit} calls should be emitted or not. The default
behaviour is to emit the calls.
This may be required by alternative race detection runtimes. One such
runtime is the Kernel Concurrency Sanitizer (KCSAN):
Installed as obvious.
ChangeLog:
* .gitignore: Add .clang-tidy.
---
.gitignore | 1 +
1 file changed, 1 insertion(+)
diff --git a/.gitignore b/.gitignore
index d9d3967a12c..d9eeaf2ed7a 100644
--- a/.gitignore
+++ b/.gitignore
@@ -36,6 +36,7 @@ TAGS.sub
.lvimrc
.clang-format
On June 12, 2020 2:55:49 PM GMT+02:00, "Martin Liška" wrote:
>On 6/12/20 1:43 PM, Richard Sandiford wrote:
>> Martin Liška writes:
>>> On 6/12/20 12:46 PM, Richard Sandiford wrote:
Martin Liška writes:
> Hello.
>
> I'm working one extension of SLP which will allow to vectorize
On 6/12/20 11:43 AM, Richard Biener wrote:
So ... how far are you with enforcing a split VEC_COND_EXPR?
Thus can we avoid the above completely (even as intermediate
state)?
Apparently, I'm quite close. Using the attached patch I see only 2 testsuite
failures:
FAIL: gcc.dg/tree-ssa/pr68714.c
Martin Liška writes:
> diff --git a/gcc/tree-vect-patterns.c b/gcc/tree-vect-patterns.c
> index 636ad59c001..eac372e6abc 100644
> --- a/gcc/tree-vect-patterns.c
> +++ b/gcc/tree-vect-patterns.c
> @@ -5120,20 +5120,12 @@ vect_determine_precisions (vec_info *vinfo)
>else
> {
>
Hi,
On Thu, 11 Jun 2020 at 22:37, Rainer Orth wrote:
>
> Hi Martin,
>
> > The compute_objsize() function started out as a thin wrapper around
> > compute_builtin_object_size(), but over time developed its own
> > features to compensate for the other function's limitations (such
> > as its
On Thu, Jun 11, 2020 at 8:24 PM Jeff Law wrote:
>
> On Sat, 2020-05-02 at 04:55 -0700, H.J. Lu wrote:
> > Currently patchable area is at the wrong place. It is placed immediately
> > after function label, before both .cfi_startproc and ENDBR. This patch
> > adds UNSPECV_PATCHABLE_AREA for
On 6/12/20 1:43 PM, Richard Sandiford wrote:
Martin Liška writes:
On 6/12/20 12:46 PM, Richard Sandiford wrote:
Martin Liška writes:
Hello.
I'm working one extension of SLP which will allow to vectorize multiple
BBs. This is a first step where I abstract _bb_vec_info::region_begin and
On Fri, 12 Jun 2020 at 12:29, Jakub Jelinek wrote:
>
> On Fri, Jun 12, 2020 at 09:21:59AM +0200, Marco Elver wrote:
> > Adds param tsan-instrument-func-entry-exit, which controls if
> > __tsan_func_{entry,exit} calls should be emitted or not. The default
> > behaviour is to emit the calls.
>
> If
Martin Liška writes:
> On 6/12/20 12:46 PM, Richard Sandiford wrote:
>> Martin Liška writes:
>>> Hello.
>>>
>>> I'm working one extension of SLP which will allow to vectorize multiple
>>> BBs. This is a first step where I abstract _bb_vec_info::region_begin and
>>> _bb_vec_info::region end by
On 6/12/20 12:46 PM, Richard Sandiford wrote:
Martin Liška writes:
Hello.
I'm working one extension of SLP which will allow to vectorize multiple
BBs. This is a first step where I abstract _bb_vec_info::region_begin and
_bb_vec_info::region end by providing an iterator.
Nice.
diff --git
This adds a reduced C testcase from libgomp.fortran/examples-4/target-1.f90
and libgomp.fortran/examples-4/target_data-1.f90 FAILs, showing a case
of SLP instance stmt sharing that affects the shared dataref alignment
info.
2020-06-12 Richard Biener
* gcc.dg/vect/bb-slp-44.c: New
Martin Liška writes:
> Hello.
>
> I'm working one extension of SLP which will allow to vectorize multiple
> BBs. This is a first step where I abstract _bb_vec_info::region_begin and
> _bb_vec_info::region end by providing an iterator.
Nice.
> diff --git a/gcc/tree-vectorizer.h
"Yangfei (Felix)" writes:
> +2020-06-12 Felix Yang
> +
> + PR tree-optimization/95570
> + * tree-vect-data-refs.c (vect_relevant_for_alignment_p): New function.
> + (vect_verify_datarefs_alignment): Call it to filter out data
> references
> + in the loop whose
On Fri, Jun 12, 2020 at 09:21:59AM +0200, Marco Elver wrote:
> Adds param tsan-instrument-func-entry-exit, which controls if
> __tsan_func_{entry,exit} calls should be emitted or not. The default
> behaviour is to emit the calls.
If you want that, I wonder if the spots you've chosen are the best
On 6/12/20 9:21 AM, Marco Elver wrote:
Adds param tsan-instrument-func-entry-exit, which controls if
__tsan_func_{entry,exit} calls should be emitted or not. The default
behaviour is to emit the calls.
I support the patch.
What about Jakub?
Martin
Hi!
On 2020-06-09T16:11:03+0200, Jakub Jelinek via Gcc-patches
wrote:
> On Tue, Jun 09, 2020 at 04:02:19PM +0200, Tobias Burnus wrote:
>> It turned out that this patch fails with LTO and partitions,
>> causing fails at runtime such as
>> libgomp: Duplicate node
>> via libgomp/splay-tree.c's
The following gets rid of the restriction in BB vectorization
where we needed to ensure underlying stmts vector types agree
for all uses in all SLP instances. That's done by recording
the vector type for SLP nodes in all cases and using that
when analyzing/code-generating by means of adjusting
The previous reorg missed a guard around the else clause access.
Committed as obvious.
2020-06-12 Richard Biener
PR tree-optimization/95633
* tree-vect-stmts.c (vectorizable_condition): Properly
guard the vec_else_clause access with EXTRACT_LAST_REDUCTION.
---
Adds param tsan-instrument-func-entry-exit, which controls if
__tsan_func_{entry,exit} calls should be emitted or not. The default
behaviour is to emit the calls.
This may be required by alternative race detection runtimes. One such
runtime is the Kernel Concurrency Sanitizer (KCSAN):
The tests for clear() and test_and_set() didn't cover all cases.
* testsuite/29_atomics/atomic_flag/clear/1.cc: Also test clear()
when the value is currently set.
* testsuite/29_atomics/atomic_flag/test_and_set/explicit.cc:
Actually check the return value.
On 11/05/20 09:11 +0100, Jonathan Wakely wrote:
On 08/05/20 17:05 +0200, Ulrich Drepper via Libstdc++ wrote:
This is not yet implemented. Here is a patch.
2020-05-08 Ulrich Drepper
* include/bits/atomic_base.h (atomic_flag): Implement test
memeber function.
*
On Thu, Jun 11, 2020 at 10:52 AM Martin Liška wrote:
>
> On 6/9/20 3:42 PM, Richard Biener wrote:
> > I think we need to fix that before merging.
>
> There's updated version of the patch that should handle the EH properly.
>
> Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
On Fri, Jun 12, 2020 at 2:31 AM Segher Boessenkool
wrote:
>
> On Tue, Jun 09, 2020 at 09:51:12AM -0600, Martin Sebor wrote:
> > >I think the backend declaration is wrong, the function only takes
> > >a void * argument and returns a long.
> >
> > Thanks. In his comment on the bug Segher (CC'd)
On Thu, Jun 11, 2020 at 9:04 PM Segher Boessenkool
wrote:
>
> Hi!
>
> On Thu, Jun 11, 2020 at 01:49:57PM +0200, Martin Liška wrote:
> > Since stmt_vec_info is not generated for debug info statements, these needs
> > to
> > be skipped in rs6000_density_test.
>
> > PR target/95627
> > *
AI12-0369 relaxes the rules for the Static_Barriers and Pure_Barriers
restrictions so that subcomponents of a protected object can be named
(subject to some restrictions) in a barrier expression. Note that one
effect of this is to undo a change that was made recently for AI12-0290.
Tested on
This mainly improves the code generated for the assignment of dynamic
record aggregates, which was always going through a temporary if the
target of the assignment was not a simple identifier (this is not the
case for dynamic array aggregates). This also extends a little the
set of acceptable
There are some cases where an explicit type conversion from an anonymous
access type to a named access type is legal but an implicit conversion
is not. Ada RM 8.6 requires that for such an implicit conversion (but
not for an explicit conversion), the operand of the conversion "shall
denote a view
Identifiers in the snames template file are ordered either logically or
alphabetically, but the Test_Case identifier was clearly out-of-order.
Likewise for pragmas. Cleanup before adding a new aspect; semantics in
unaffected.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Piotr
The current code was climbing the tree manually assuming that an array
attribute (e.g. 'First) would appear immediately, which isn't the case
for e.g. a selected component expression.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Arnaud Charlet
gcc/ada/
* sem_res.adb
This Ada 202x AI makes illegal some cases of out parameters whose type
has a Default_Value aspect.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Arnaud Charlet
gcc/ada/
* sem_res.adb (Resolve_Actuals): Check for AI12-0074.--- gcc/ada/sem_res.adb
+++
Such subprograms are not protected and have convention Intrinsic to
ensure that their 'Access isn't taken as per RM 6.3.1(10/2).
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Arnaud Charlet
gcc/ada/
* exp_ch6.adb (Expand_N_Subprogram_Declaration): Do nothing for
Windows Server 2019 do not need 500ms socket timeout correction. The
problem is that Windows API recommended function
IsWindowsVersionOrGreater is not comparing the build part which is only
difference between Windows 2016 and 2019 servers.
Tested on x86_64-pc-linux-gnu, committed on trunk
Use Standard'Maximum_Alignment * 2 unconditionally to accommodate
possible requirements by user code compiled with specific cpu/arch
options such as -mavx on x86 targets, which Gigi assumes the SS can
handle.
Such options influence what Standard'Maximum_Alignment returns and are
typically not
The required level of support for elementary streaming includes support
for 24 bits elements, which is added here.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Arnaud Charlet
gcc/ada/
* exp_strm.adb (Build_Elementary_Input_Call): Add support for 24
bits
We were detecting overlapping actuals in a call like "Proc (A, B, C)"
using two loops that iterated over all parameters, so we were examining
9 pairs:
A-A, A-B, A-C
B-A, B-B, B-C
C-A, C-B, C-C
Now the outer loop iterates over all parameters, while the inner loop
always starts from the next
This implements the AI under the -gnat2020 switch, which extends the
set of calls that can be made during the elaboration of library units
subject to the aspect/pragma Preelaborate, beyond the existing ones
to static functions. Calls to certain functions that are essentially
unchecked conversions
This patch fixes an infinite loop in the compiler on an instantiation
that includes formal packages and formal subprograms with aspect
specifications (but no pre- or postconditions), when compiling with
-gnatX.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Ed Schonberg
Replace a low-level membership test "Ekind_In (Id, E_Entry,
E_Entry_Family)" with a high-level "Is_Entry (Id)", which is more
readable. Semantics stays the same.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Piotr Trojanek
gcc/ada/
* contracts.adb, einfo.adb,
For subprograms we call Warn_On_Overlapping_Actuals in Resolve_Call;
however, for protected entries we didn't call it at all, because
Resolve_Call was executing Resolve_Entry_Call and returning early. As a
result, we didn't detect overlapping actuals in entry calls. Now we do.
Tested on
This patch fixes a crash in the compiler when handling an instantiation
with a formal subprogram that carries an aspect specification, when the
actual in the instantiation is an overloaded name.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Ed Schonberg
gcc/ada/
*
Routine Get_Discr_Value takes a name of a discriminant, not its entity,
so its parameter should be a Node_Id, not an Entity_Id. Also, fix typo
in its comment and refactor repeated calls in the body. Semantics is
unaffected.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Piotr
Detection of overlapping actuals was suppressed for internal units,
because we were getting spurious warnings on container units. Now this
suppression has no effect (most likely because ignore parameters of
generic types).
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Piotr
This patch adds checks to reject illegal overriding indicators that
appear on subprogram declarations in protected bodies, where such
a subprogram cannot override any operation. Fixes ACATS test B831006.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Ed Schonberg
gcc/ada/
This acats test shows that we are not checking 13.1(9.2/5) for
representation aspects, now fixed.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Arnaud Charlet
gcc/ada/
* aspects.ads (Is_Representation_Aspect): New array.
* sem_ch13.adb
This aligns the handling of Ada 2012 implicit dereferences in the
Is_OK_Variable_For_Out_Formal predicate wih that in Is_Variable,
and removes unreachable code as a result.
This also makes more use of the Get_Reference_Discriminant helper
function to retrieve the reference discriminant of a type.
Enable Put_Image for types that don't cause regressions (which is most
types).
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Bob Duff
gcc/ada/
* debug.adb: Remove usage of -gnatd_z.
* exp_attr.adb, exp_put_image.ads, exp_put_image.adb: Clean up
the
Hi Kwok!
On 2020-06-10T17:23:58+0100, Kwok Cheung Yeung wrote:
> On 10/06/2020 4:15 pm, Tobias Burnus wrote:
>> On 6/10/20 3:34 PM, Kwok Cheung Yeung wrote:
>>> This patch updates the previous entry on the website for devel/omp/gcc-9 to
>>> refer to the new branch instead. This removes
This implements the AI under the -gnat2020 switch, which extends the
Ada 2005 No_Return aspect/pragma to functions and generic functions.
As explained in the AI, functions subject to this aspect/pragma must
still have a return statement, but it must be of a specific form.
This also fixes the
This patch updates the Put_Image facilities in the Big_Numbers packages
to match the actual implementation of Put_Image.
Tested on x86_64-pc-linux-gnu, committed on trunk
2020-06-12 Bob Duff
gcc/ada/
* libgnat/a-nbnbin.adb, libgnat/a-nbnbin.ads,
libgnat/a-nbnbin__gmp.adb,
On Thu, Jun 11, 2020 at 5:24 PM Eric Botcazou wrote:
>
> > I am not very good at reasoning about reverse storage order stuff but
> > can it ever happen that this reverse is not the same as racc->reverse?
> >
> > In order for that to happen, we'd have to have an assignment (folded
> > memcpy?)
PING^2
On 6/3/20 11:14 AM, Martin Liška wrote:
On 6/1/20 1:40 PM, Martin Liška wrote:
Adding Honza as Uros recommended him for a review.
@Honza asked me personally to remind him this patch ;)
Martin
Thanks for the review.
On Fri, Jun 12, 2020 at 11:28 AM Jeff Law wrote:
>
> On Fri, 2020-06-05 at 13:46 +0800, Hongtao Liu via Gcc-patches wrote:
> > Hi:
> >
> > +/* Optimize vector MUL generation for V8QI, V16QI and V32QI
> > + under TARGET_AVX512BW. i.e. for v16qi a * b, it has
> > +
> > +
Hello.
I'm working one extension of SLP which will allow to vectorize multiple
BBs. This is a first step where I abstract _bb_vec_info::region_begin and
_bb_vec_info::region end by providing an iterator.
Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
Ready to be
On Thu, Jun 11, 2020 at 1:52 PM Eric Botcazou wrote:
>
> Hi,
>
> gimplify_modify_expr_rhs has an optimization whereby the assignment to an
> aggregate variable from a read-only object with a DECL_INITIAL is optimized
> into the direct assignment of the DECL_INITIAL, provided that no temporary is
Hi,
I reported a PR in gcc Bugzilla about the medium code model in aarch64. A
solution is proposed and some discussion has been posted.
The details of the discussion can be found here :
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95285
Wilco suggest me to make a PIC 48-bit code model by
"zhoukaipeng (A)" writes:
> Hi,
>
> This is a fix for pr95199.
> In vectorization, vinfo->ivexpr_map is supposed to catch those "IV base
> and/or step expressions". Just call cse_and_gimplify_to_preheader to handle
> gathering/scattering to avoid the extra variable.
>
> Bootstrap and tested on
"Roger Sayle" writes:
> diff --git a/gcc/simplify-rtx.c b/gcc/simplify-rtx.c
> index 28c2dc6..ccf5f6d 100644
> --- a/gcc/simplify-rtx.c
> +++ b/gcc/simplify-rtx.c
> @@ -3128,6 +3128,17 @@ simplify_binary_operation_1 (enum rtx_code code,
> machine_mode mode,
>
Hi Rasmus,
(switched from gcc to gcc-patches list)
> On 26 May 2020, at 16:52, Rasmus Villemoes wrote:
>
> Fix
>
> gcc-src/libgcc/config/gthr-vxworks.c:65:7: warning: implicit declaration of
> function 'taskDelay'; did you mean 'sysDelay'?
> [-Wimplicit-function-declaration]
> 65 |
In the meantime, here is a patch which fixes a regression I introduced
when fixing a regression with a memory leak. The important thing
here is to realize that we do not need to finalize (and deallocate)
multiple times for the same expression and the same component
in the same namespace. It
Hi,
This is a fix for pr95199.
In vectorization, vinfo->ivexpr_map is supposed to catch those "IV base and/or
step expressions". Just call cse_and_gimplify_to_preheader to handle
gathering/scattering to avoid the extra variable.
Bootstrap and tested on aarch64/x86_64 Linux platform. No new
91 matches
Mail list logo
