On 5 September 2017 at 20:20, Christophe Lyon
wrote:
> On 5 September 2017 at 19:53, Kyrill Tkachov
> wrote:
>>
>> On 05/09/17 18:48, Bernd Edlinger wrote:
>>>
>>> On 09/05/17 17:02, Wilco Dijkstra wrote:
Bernd Edlinger wrote:
> Combine creates an invalid insn out of these two
This fixes a bogus check for a mode when the type matters. The
test can get fooled by vector ops with integral mode and thus we
later ICE trying to use wide-ints operating on vector constants.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
2017-09-06 Richard Biener
Jeff Law writes:
> On 08/22/2017 08:15 AM, Wilco Dijkstra wrote:
>> Jeff Law wrote:
>> On 07/26/2017 05:29 PM, Wilco Dijkstra wrote:
>>
But then the check size_align % MAX_SUPPORTED_STACK_ALIGNMENT != 0
seems wrong too given that round_push uses a different alignment to
align to.
Hi Thomas,
On 05/09/17 10:04, Thomas Preudhomme wrote:
Ping?
This is ok if a bootstrap and test run on arm-none-linux-gnueabihf shows
no problems.
Thanks,
Kyrill
Best regards,
Thomas
On 25/08/17 12:18, Thomas Preudhomme wrote:
Hi,
I've now also added a couple more changes:
* size to_
On 09/05/2017 09:46 PM, Mike Stump wrote:
I've included the dwarf people on the cc list. Seems like they may
have an opinion on the direction or the patch itself. I was fine
with the patch from the larger testsuite perspective.
Good idea, thank you! And thank you for your feedback. :-)
--
Pie
Hi all,
I've attached a new patch that addresses some of the issues raised with
my original patch.
On 08/23/2017 03:35 PM, Wilco Dijkstra wrote:
Richard Sandiford wrote:
Sorry for only noticing now, but the call to aarch64_legitimate_address_p
is asking whether the MEM itself is a legitimat
Richard Sandiford do you have any objections to the patch as it stands? It
doesn't appear as if anything is going to change in the mid-end anytime soon.
-Original Message-
From: Richard Sandiford [mailto:richard.sandif...@linaro.org]
Sent: Tuesday, August 22, 2017 9:11 AM
To: Richard Bie
Hi,
PR 82078 is another fallout from lazy setting of written flag in SRA.
The problem here is that we do not enqueue assignment links going out
of accesses of candidates that were disqualified before we start the
loop with sort_and_splice_var_accesses.
Given that the propagation is now a correctn
On Wed, 6 Sep 2017, Martin Jambor wrote:
> Hi,
>
> PR 82078 is another fallout from lazy setting of written flag in SRA.
> The problem here is that we do not enqueue assignment links going out
> of accesses of candidates that were disqualified before we start the
> loop with sort_and_splice_var_a
This patch fixes an omission in the handling of pragma Eliminate when applied
to a protected operation. The pragma was properly processed, but a call to
an eliminated protected operation was not flagged as an error, and the code
generator aborted on a call to an undefined operation.
Compiling:
On Tue, Sep 05, 2017 at 11:21:48PM -0600, Jeff Law wrote:
> --- a/gcc/tree-ssa-reassoc.c
> +++ b/gcc/tree-ssa-reassoc.c
> @@ -5763,14 +5763,15 @@ reassociate_bb (basic_block bb)
>"Width = %d was chosen for reassociation\n",
> width);
>
>
> - /* For bin
On 09/05/17 23:27, Wilco Dijkstra wrote:
> Bernd Edlinger wrote:
>> No, the split condition does not begin with "&& TARGET_32BIT...".
>> Therefore the split is enabled in TARGET_NEON after reload_completed.
>> And it is invoked from adddi3_neon for all alternatives without vfp
>> registers:
>
> Hm
AI12-0124 adds the notation Object'Image to the language, following the
semantics of GNAT-defined attribute 'Img. This patch fixes an omission in
the characterization of objects, which must include function calls and thus
attribute references for attributes that are functions, as well as predefined
This patch fixes a spurious error on the use of of a generic unit with formal
incomplete types, as a formal package in another generic unit, when the
actuals for the incomplete types are themselves formal incomplete types.
The treatment of incomplete subtypes that are created for such formals is n
This activates the reordering of components in record types with convention
Ada that was implemented some time ago in the compiler. The idea is to get
rid of blatant inefficiencies that the layout in textual order of the source
code can bring about, typically when the offset of components is not f
This patch fixes a bug in the resolution of set membership operations when
the expression and/or the alternatives on the right-hand side are overloaded.
If a given overloaded alternative is resolved to a unique type by intersection
with the types of previous alternatives, the type is used subsequen
Hi all,
This patch is split from part (1/2). It includes the patterns that have
been moved out of fold-const.c
It also removes an (almost entirely) redundant pattern:
(A / C1) +- (A / C2) -> A * (1 / C1 +- 1 / C2)
which was only used in special cases, either with combinations
of flags l
The inter-unit inlining done by the compiler requires a dedicated machinery
to deal with the public status of library-level entities, since it breaks
the private/plublic semantic barrier of the language.
This is a minor cleanup to this machinery, no functional changes.
Tested on x86_64-pc-linux-g
This patch fixes the handling of untagged discriminated derived types that
constrain some parent discriminants and rename others. The compiler failed
to handle a change of representation on the derived type, and generated
faulty code for the initialization procedure or such a derived type.
Executi
Hi all,
A minor improvement came to mind while updating other parts of this patch.
I've updated a testcase to make it more clear and a condition now uses a
call to is_division_by rather than manually checking those conditions.
Jackson
On 08/30/2017 05:32 PM, Jackson Woodruff wrote:
Hi all,
On 08/30/2017 01:46 PM, Richard Biener wrote:
On Wed, Aug 30, 2017 at 11:46 AM, Jackson Woodruff
wrote:
On 08/29/2017 01:13 PM, Richard Biener wrote:
On Tue, Aug 29, 2017 at 1:35 PM, Jackson Woodruff
wrote:
Hi all,
Apologies again to those CC'ed, who (again) received this twice.
Joseph:
This wasn't explicitly mentioned but the previous changes also replaced the
internal Table package used in the compiler by GNAT.Tables, resulting in a
large performance hit for the compiler because the memory management scheme
of the latter is very inefficient.
This restores the original implement
Primitive functions whose first formal is a controlling parameter, whose
other formals have defaults and whose result is an array type can lead to
ambiguities when the result of such a call is the prefix of an indexed
component. The interpretation that analyzes Obj.F (X, Y) into F (Obj)(X, Y)
is on
This patch fixes an omission in the handling of delayed aspects on derived
types. The type may inherit a representation aspect from its parent, but have
no explicit aspect specifications. At the point it is frozen, the parent is
frozen as well and its explicit aspects have been analyzed. The inheri
This patch ensures that if a pragma No_Return applies to a generic subprogram ,
all its instantiations are treated as No_Return subprograms as well.
Tested in ACATS 4.1D C651001.
Tested on x86_64-pc-linux-gnu, committed on trunk
2017-09-06 Ed Schonberg
* sem_ch12.adb (Analyze_Subprog
This patch fixes a compiler abort on a generic unit to which the aspect
No_Return applies.
Tested in ACATS 4.1D C651002.
Tested on x86_64-pc-linux-gnu, committed on trunk
2017-09-06 Ed Schonberg
* freeze.adb (Freeze_Entity): Do not generate a freeze
node for a generic unit, e
A typical error for new users of Ada is to call functions in a call
statement. Improve the error message for these users, to better indicate
what the error is in that case.
The following compilation raises the new message.
$ gcc -c main.adb
1. procedure Main is
2.function Lol
This patch implements legality rule in 6.5.1 (7/2): if a renaming as body
completes a nonreturning procedure declaration, the renamed procedure must
be nonreturning as well. Previously GNAT only produced a warning in such
cases.
Tested in ACATS test B651002.
Tested on x86_64-pc-linux-gnu, commit
The warning issued when accessing a string at a negative or null index
was misleading, suggesting to use S'First - 1 as correct index, which
it is obviously not. Add a detection for negative or null index when
accessing a standard string, so that an appropriate warning is issued.
Also add a corresp
This patch modifies the finalization machinery to ensure that the cursor of an
"of" iterator loop is properly finalized at the end of the loop. Previously it
was incorrectly assumed that such a cursor will never need finalization
ctions.
-- Source --
-- leak.adb
pragma
This patch fixes a bug in which if a derived type has a Default_Iterator
specified, and the parent type does not, then a "for ... of" loop causes
the compiler to crash. No small test case available.
Tested on x86_64-pc-linux-gnu, committed on trunk
2017-09-06 Bob Duff
* exp_ch5.adb (G
Michael Collison writes:
> Richard Sandiford do you have any objections to the patch as it stands?
> It doesn't appear as if anything is going to change in the mid-end
> anytime soon.
I think one of the suggestions was to do it in expand, taking advantage
of range info and TARGET_SHIFT_TRUNCATION
Hi all,
This patch merges loads and stores from D-registers that are of
different modes.
Code like this:
typedef int __attribute__((vector_size(8))) vec;
struct pair
{
vec v;
double d;
}
void
assign (struct pair *p, vec v)
{
p->v = v;
p->d
Refactor of all 'Image attributes for better error diagnostics and clarity.
Tested on x86_64-pc-linux-gnu, committed on trunk
2017-09-06 Justin Squirek
* exp_imgv.adb (Expand_Image_Attribute),
(Expand_Wide_Image_Attribute), (Expand_Wide_Wide_Image_Attribute):
Added cas
This patch renames lookup_fnfields_slot{,_nolazy} to
get_class_binding{,_direct}. It also removes a few now-unneeded checks
for CLASSTYPE_METHOD_VEC being non-null.
You may notice that the new names mention nothing about the kind of
member looked for. That's intentional. These functions wil
Hi Jakub
I compiled libjpeg-turbo and libdng_sdk with options "-g -O3 -Wall
-fsanitize-coverage=trace-pc,trace-cmp -fsanitize=address".
And run my fuzzer with pc and cmp feedbacks for hours. It works fine.
About __sanitizer_cov_trace_cmp{f,d} , yes, it isn't provided by llvm. But
once we trace i
When issuing a warning on a read of an uninitialized variable through
reading an attribute such as Loop_Entry, GNAT could crash. Now fixed.
GNAT issues a warning as expected on the following code:
$ gcc -c s.adb
1. package S is
2.
3.type Array_Range is range 1 .. 10;
This improves a little the algorithm used to compute the set of externally
visible entities in package bodies to make it less conservative in the
presence of local inlined subprograms. The typical effect is to eliminate
the out-of-line body if the subprogram is inlined at every call site:
package
The function Value of package GNAT.Calendar.Time_IO has been enhanced
to parse strings containing UTC date and time.
After this patch the following test works fine.
with Ada.Calendar; use Ada.Calendar;
with Ada.Text_IO; use Ada.Text_IO;
with GNAT.Calendar.Time_IO; use GNAT.Cale
This patch enahnces dimensionality checking to cover generic subprograms that
are intended to apply to types of different dimensions, such as an integration
function. Dimensionality checking is performed in each instance. and rely on
a special handling of conversion operations to prevent spurious d
This patch fixes the handling of overriding operations that have both an
explicit postcondition and an inherited classwide one.
Executing:
gnatmake -q -gnata post_class.adb
post_class
must yield:
raised SYSTEM.ASSERTIONS.ASSERT_FAILURE :
failed inherited postcondition from the_p
This patch modifies the finalization mechanism to recognize a heavily expanded
generalized indexing where the element type requires finalization actions.
-- Source --
-- types.ads
with Ada.Finalization; use Ada.Finalization;
package Types is
type Element is new Con
GNAT failed to issue an error on a Global/Depends aspect put on an object
declaration, which is only allowed for a task object. Instead it crashed.
Now fixed.
Tested on x86_64-pc-linux-gnu, committed on trunk
2017-09-06 Yannick Moy
* sem_prag.adb (Analyze_Depends_Global): Reinforce te
This makes the compiler generate an error message also in the case where one
of the specified components overlaps the parent field because its size has
been explicitly set by a size clause.
The compiler must issue an error on 32-bit platforms for the package:
1. package P is
2.
3.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk
sofar.
Richard.
2017-09-06 Richard Biener
PR tree-optimization/82108
* tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
for gap in the non-permutation SLP case.
* gcc.dg/vect/
On 05/09/17 18:40, Pedro Alves wrote:
On 09/05/2017 05:52 PM, Manuel López-Ibáñez wrote:
Yeah, ISTR it was close, though there were a couple things
that needed addressing still.
The wiki seems to miss a pointer to following iterations/review
of that patch (mailing list archives don't cross month
On 09/04/17 21:54, Bernd Edlinger wrote:
> Hi Kyrill,
>
> Thanks for your review!
>
>
> On 09/04/17 15:55, Kyrill Tkachov wrote:
>> Hi Bernd,
>>
>> On 18/01/17 15:36, Bernd Edlinger wrote:
>>> On 01/13/17 19:28, Bernd Edlinger wrote:
On 01/13/17 17:10, Bernd Edlinger wrote:
> On 01/13/1
On 06/09/17 13:44, Bernd Edlinger wrote:
> On 09/04/17 21:54, Bernd Edlinger wrote:
>> Hi Kyrill,
>>
>> Thanks for your review!
>>
>>
>> On 09/04/17 15:55, Kyrill Tkachov wrote:
>>> Hi Bernd,
>>>
>>> On 18/01/17 15:36, Bernd Edlinger wrote:
On 01/13/17 19:28, Bernd Edlinger wrote:
> On 01/
On Wed, 6 Sep 2017, Richard Biener wrote:
>
> This fixes a bogus check for a mode when the type matters. The
> test can get fooled by vector ops with integral mode and thus we
> later ICE trying to use wide-ints operating on vector constants.
>
> Bootstrapped and tested on x86_64-unknown-linux-
When fiddling around with vector lowering I found the following
adjusted testcase helpful testing proper vector lowering of
word_mode vector plus.
Tested on x86_64-unknown-linux-gnu, applied.
Richard.
2017-09-06 Richard Biener
* gcc.c-torture/execute/20050604-1.c: Adjust to be a bet
This patch corrects an issue where attributes applied to records were not
propagated to components within the records - causing incorrect code to be
generated by the backend. Additionally, this ticket fixes another issue with
pragma Volatile_Full_Access that allowed the attribute to be applied to a
This patch removes a spurious style warning on an operator declared in a
generic package when the package is used as a formal of a generic subprogram,
and the subprogream body includes a use clause on that package.
The following must compile quietly:
gcc -c -gnatyO generic_test.adb
---
with Gene
On 09/06/17 14:51, Richard Earnshaw (lists) wrote:
> On 06/09/17 13:44, Bernd Edlinger wrote:
>> On 09/04/17 21:54, Bernd Edlinger wrote:
>>> Hi Kyrill,
>>>
>>> Thanks for your review!
>>>
>>>
>>> On 09/04/17 15:55, Kyrill Tkachov wrote:
Hi Bernd,
On 18/01/17 15:36, Bernd Edlinger wr
The following replaces the weird PRE "DCE" algorithm by a simple
work-list based one seeded by inserted_exprs. This makes it possible
to get rid of the error-prone marking of stmts necessary and allows
re-ordering of elimination dead stmt removal and DCE again (I'm
in the process of developing a
Compared to prior versions of regular VxWorks (not AE/653), the VxWorks 7
header files expect the e500v2 family of CPUs to be designated in a slightly
different fashion.
With this on top of previously posted patches, a build for
e500v2-wrs-vxworks proceeds to completion.
Committing to mainline.
On 09/06/17 14:51, Richard Earnshaw (lists) wrote:
> On 06/09/17 13:44, Bernd Edlinger wrote:
>> On 09/04/17 21:54, Bernd Edlinger wrote:
>>> Hi Kyrill,
>>>
>>> Thanks for your review!
>>>
>>>
>>> On 09/04/17 15:55, Kyrill Tkachov wrote:
Hi Bernd,
On 18/01/17 15:36, Bernd Edlinger wr
Here's some cleanup of the SORTED_FIELDS vector initialization. Some
function renaming, to be more specific. The functionality change is a
minor bug in late enums. We only add them to the field vec, if there's
already a field vec. But of course, their addition could have cause the
class's T
This patch fixes an error in an assignmen statement to an entity of a mutable
type (variable or in-out parameter) when the righ-hand side of the assignment
is a conditioal expression, some of whose alternatives are aggregates. Prior
to this patch, not all components of the mutable object were prop
This preparatory patch fixes up a couple of places where a non-function
could start appearing in the METHOD_VEC. The warn_hidden change looks
bigger than necessary, because of indentation change. I noticed
check_classfn could check a template mismatch earlier, and avoid doing
some work.
app
This patch autogenerates arm-isa.h from new entries in arm-cpus.in.
This has the primary advantage that it makes the description file more
self-contained, but it also solves the 'array dimensioning' problem
that Tamar recently encountered. It adds two new constructs to
arm-cpus.in: features and f
This patch adds a bit more error checking to parsecpu.awk to ensure
that statements are not missing arguments or have excess arguments
beyond those permitted. It also slightly improves the handling of
errors so that we terminate properly if parsing fails and be as
helpful as we can while in the p
On Wed, Sep 06, 2017 at 07:47:29PM +0800, 吴潍浠(此彼) wrote:
> Hi Jakub
> I compiled libjpeg-turbo and libdng_sdk with options "-g -O3 -Wall
> -fsanitize-coverage=trace-pc,trace-cmp -fsanitize=address".
> And run my fuzzer with pc and cmp feedbacks for hours. It works fine.
> About __sanitizer_cov_tra
On Wed, Sep 06, 2017 at 04:37:18PM +0200, Jakub Jelinek wrote:
> Ok. Please make sure those entrypoints make it into the various example
> __sanitier_cov_trace* fuzzer implementations though, so that people using
> -fsanitize-coverage=trace-cmp in GCC will not need to hack stuff themselves.
> At l
Hi,
We have decided to apply the following patch to the embedded-7-branch as a
dependency patch to enable ARMv8-R support.
ChangeLog entry is as follows:
*** gcc/ChangeLog.arm ***
2017-09-04 Thomas Preud'homme
Backport from mainline
2017-07-04 Thomas Preud'homme
* conf
Hi,
We have decided to apply the following patch to the embedded-7-branch to enable
ARMv8-R support.
ChangeLog entry is as follows:
*** gcc/ChangeLog.arm ***
2017-09-04 Thomas Preud'homme
Backport from mainline
2017-07-06 Thomas Preud'homme
* config/arm/arm-cpus.in (arm
Hi,
We have decided to apply the following patch to the embedded-7-branch to enable
ARMv8-R support.
ChangeLog entry is as follows:
*** gcc/ChangeLog.arm ***
2017-09-04 Thomas Preud'homme
Backport from mainline
2017-07-14 Thomas Preud'homme
* config/arm/arm-isa.h (isa_
Hi,
We have decided to apply the following patch to the embedded-7-branch to enable
Arm Cortex-R52 support.
*** gcc/ChangeLog.arm ***
2017-09-04 Thomas Preud'homme
Backport from mainline
2017-07-14 Thomas Preud'homme
* config/arm/arm-cpus.in (cortex-r52): Add new entry.
GCC Maintainers:
The following patch adds support for the vec_xst_len_r() and
vec_xl_len_r() Powerr 9 builtins. The patch has been run on
powerpc64le-unknown-linux-gnu (Power 9 LE). No regressions were found
but it does seem to "fix" a couple of existing tests.
136a137
> FAIL: TestCgoCallbackGC
To match on vxworks*spe so it applies to VxWorks 7 as well.
Committing to mainline after verifying for e500v2-wrs-vxworks7
that we now include config/powerpcspe/vxworks.h instead of
config/rs6000/vxworks.h.
Olivier
2017-09-06 Olivier Hainque
* config.gcc (powerpc-wrs-vxworksspe): No
On 09/05/2017 03:16 PM, Jakub Jelinek wrote:
> Hi!
>
> If a DECL_THREAD_LOCAL_P decl has NULL DECL_INITIAL and
> -fzero-initialized-in-bss (the default), we ICE starting with
> r251602, which changed bss_initializer_p:
> + /* Do not put constants into the .bss section, they belong in a readonly
>
On 06/09/17 14:17, Bernd Edlinger wrote:
On 09/06/17 14:51, Richard Earnshaw (lists) wrote:
On 06/09/17 13:44, Bernd Edlinger wrote:
On 09/04/17 21:54, Bernd Edlinger wrote:
Hi Kyrill,
Thanks for your review!
On 09/04/17 15:55, Kyrill Tkachov wrote:
Hi Bernd,
On 18/01/17 15:36, Bernd Edl
This patch merges the lookup of function and non-function member lookup
into get_class_binding_direct. lookup_field_1 becomes an internal detail.
We grow a tri-valued argument to get_class_binding_direct:
<0 -- caller wants functions
=0 -- caller wants whatever is bound
>0 -- caller wants type
On Wed, Sep 06, 2017 at 09:29:25AM -0600, Jeff Law wrote:
> > Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> > trunk?
> >
> > 2017-09-05 Jakub Jelinek
> >
> > PR middle-end/82095
> > * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars
Richard,
The problem with this approach for Aarch64 is that TARGET_SHIFT_TRUNCATION_MASK
is based on SHIFT_COUNT_TRUNCATED which is normally 0 as it based on the
TARGET_SIMD flag.
-Original Message-
From: Richard Sandiford [mailto:richard.sandif...@linaro.org]
Sent: Wednesday, Septembe
Patch updated with all relevant comments and suggestions.
Bootstrapped and tested on arm-none-linux-gnueabihf, and aarch64-none-linux-gnu
and x86_64.
Ok for trunk?
2017-08-05 Kyrylo Tkachov
Michael Collison
* compare-elim.c: Include emit-rtl.h.
(can_merge_compar
Hi,
On Tue, Sep 05, 2017 at 11:27:25PM +0200, Jakub Jelinek wrote:
> On powerpc with sysv4 -fPIC we emit something like
> .LCL0:
> .long .LCTOC1-.LCF0
> before we start emitting the function, and in the prologue we emit
> .LCF0:
> and some code. This fails to assemble if the prologue is emi
GCC Maintainers:
The following patch adds support for a couple of requested builtins that
convert from float/double to int / long using the current rounding
mode.
The patch has been tested on powerpc64le-unknown-linux-gnu (Power 8 LE).
Please let me know if the following patch is acceptable. T
On Wed, Sep 06, 2017 at 11:10:07AM -0500, Segher Boessenkool wrote:
> >for (insn = get_insns (); insn; insn = NEXT_INSN (insn))
>
> {
>
> > if (INSN_P (insn))
> > @@ -25270,10 +25273,14 @@ uses_TOC (void)
> > sub = XEXP (sub, 0);
> > if (GET_CODE (sub) == UNSP
Add an alignment test to check that aligned alloca's really do get
correctly aligned. Some targets may not ensure SP is always a multiple
of STACK_BOUNDARY (particularly with outgoing arguments), which means
aligned alloca does not get correctly aligned. This can be fixed either
by aligning the o
On Wed, Sep 06, 2017 at 06:26:10PM +0200, Jakub Jelinek wrote:
> > Maybe this "switch to the other section" thing should be abstracted out?
> > Messing with in_cold_section_p is a bit dirty.
>
> But it reflects the reality, and is what final.c and varasm.c also do.
Yes, but those aren't target co
Michael Collison writes:
> Richard,
>
> The problem with this approach for Aarch64 is that
> TARGET_SHIFT_TRUNCATION_MASK is based on SHIFT_COUNT_TRUNCATED which is
> normally 0 as it based on the TARGET_SIMD flag.
Maybe I'm wrong, but that seems like a missed optimisation in itself.
Like you say
Richard Sandiford writes:
> Michael Collison writes:
>> Richard,
>>
>> The problem with this approach for Aarch64 is that
>> TARGET_SHIFT_TRUNCATION_MASK is based on SHIFT_COUNT_TRUNCATED which is
>> normally 0 as it based on the TARGET_SIMD flag.
>
> Maybe I'm wrong, but that seems like a missed
On Wed, Sep 06, 2017 at 01:48:38AM -0400, Michael Meissner wrote:
> Here is a respin of the patch to enable -mfloat128 on PowerPC Linux systems
> now
> that the libquadmath patch has been applied. I rebased the patches against
> the
> top of the trunk on Tuesday (subversion id 251609).
>
> I tw
Richard Sandiford writes:
> Richard Sandiford writes:
>> Michael Collison writes:
>>> Richard,
>>>
>>> The problem with this approach for Aarch64 is that
>>> TARGET_SHIFT_TRUNCATION_MASK is based on SHIFT_COUNT_TRUNCATED which is
>>> normally 0 as it based on the TARGET_SIMD flag.
>>
>> Maybe I'
Hi Paul,
thanks for your patch! It's really great to finally see PDTs come to
gfortran. You're a hero, man ;)
Also: Sorry about the silence. It's certainly not due to lack of
interest, but rather lack of time (day job and private life taking up
all of mine at the moment).
In my current situation
Thanks for your tireless efforts on this, Paul! I look forward to trying this
out after it hits the trunk.
Your phrase “last unimplemented F2003” feature bolsters my suspicion that it
might be ok to switch the features listed as “Partial” on the Fortran wiki to
“Yes." I suppose the differen
I was expecting that references to capture proxies would be resolved
in the reconstructed lambda by normal name lookup, but that doesn't
work in decltype, and processing the nested lambda really wants to
find the new capture proxy, not the captured variable.
Tested x86_64-pc-linux-gnu, applying to
Hi Carl,
On Wed, Sep 06, 2017 at 08:22:03AM -0700, Carl Love wrote:
> (define_insn "*stxvl"): add missing argument to the sldi instruction.
s/add/Add/ . This one-liner fix is approved right now, please commit
it as a separate patch.
> +(define_insn "addi_neg16"
> + [(set (match_operand:
Just a followup on this patch.
We did some run-time performance testing internally on this set of
change on sparc M8 machine with -mmisalign and -mno-misalign
based on the latest upstream gcc
for CPU2017 C/C++ SPEED run:
***without -O, -mmisalign slowdown the run-time performance about 4% on
av
This change broke bootstrap on AIX because sancov.c now references a
macro that is defined as a function on AIX. sancov.c needs to include
tm_p.h to pull in the target-dependent prototypes. The following
patch works for me. Is this okay?
* sancov.c: Include tm_p.h.
Index: sancov.c
The next main step in the SVE submission is to add support for
offsets and sizes that are a runtime invariant rather than a compile
time constant. This is an RFC about our approach for doing that.
It's an update of https://gcc.gnu.org/ml/gcc/2016-11/msg00031.html
(which covered more topics than th
On 09/06/2017 11:24 AM, Carl Love wrote:
> + "fctiw %1,%1; mfvsrd %0,%1; extsw %0,%0"
> + [(set_attr "type" "integer")
> + (set_attr "length" "4")])
Should be type "three" and length "12".
-Pat
On Wed, Sep 06, 2017 at 10:08:01PM +0200, David Edelsohn wrote:
> This change broke bootstrap on AIX because sancov.c now references a
> macro that is defined as a function on AIX. sancov.c needs to include
> tm_p.h to pull in the target-dependent prototypes. The following
> patch works for me.
Another old patch getting resurrected...
On 01/04/2017 06:50 AM, Richard Biener wrote:
> On Thu, Dec 22, 2016 at 7:26 AM, Jeff Law wrote:
>> This is the final patch in the kit to improve our DSE implementation.
>>
>> It's based on a observation by Richi. Namely that a read from bytes of
>> memo
On Wed, 2017-09-06 at 16:13 -0500, Pat Haugen wrote:
> On 09/06/2017 11:24 AM, Carl Love wrote:
> > + "fctiw %1,%1; mfvsrd %0,%1; extsw %0,%0"
> > + [(set_attr "type" "integer")
> > + (set_attr "length" "4")])
>
> Should be type "three" and length "12".
>
> -Pat
Pat:
Yes, that is wrong in m
This patch is OK with the spacing in the function prototype fixed as noted
to follow normal GNU standards.
--
Joseph S. Myers
jos...@codesourcery.com
On Thu, 17 Aug 2017, Martin Sebor wrote:
> +/* Check LAST_DECL and NODE of the same symbol for attributes that are
> + recorded in EXCL to be mutually exclusive with ATTRNAME, diagnose
> + them, and return true if any have been found. NODE can be a DECL
> + or a TYPE. */
> +
> +static bool
When we regenerate a lambda, the resulting op() doesn't have any
template information, so we can't delay instantiating default
arguments like we do for a normal template function. I believe this
is also the direction of the core working group for default arguments
in local extern function declarat
The analysis and original patch:
https://gcc.gnu.org/ml/gcc-patches/2017-09/msg00101.html
and the amended patch:
https://gcc.gnu.org/ml/gcc-patches/2017-09/msg00146.html
Thanks in advance.
--
Eric Botcazou
99 matches
Mail list logo
