Hi.
As Jakub correctly noted, I used a piggy backing to put the new function
to a file that is supposed to contain different functions. So that
I'm suggesting a new file. Moreover, I'm also adding dup2 fallback.
Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
Ready to be i
Hi!
The c_parser_omp_clause_dist_schedule function had a pasto in it, so instead
of checking for duplicate DIST_SCHEDULE clause it complained if there was a
schedule clause before it, which is valid on combined constructs as can be
seen on the first testcase.
Furthermore, I've discovered that Open
Hi!
Through code inspection I found that in C duplicate proc_bind wasn't
diagnosed (but in C++ it was). So, I've added following test coverage
for various other cases where duplication shall be diagnosed, and fixed
whatever that revealed (besides proc_bind also the target {enter,exit} data
if cla
On 8/8/19 4:17 PM, Jeff Law wrote:
> On 8/8/19 7:04 AM, Martin Liška wrote:
>> Hi.
>>
>> As requested by Richi, I'm suggesting to use new dump_printf
>> optimization info infrastructure.
>>
>> Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
>>
>> Ready to be installed?
>> Tha
normal_mv_distribution maintains the variance-covariance matrix param
in Cholesky-decomposed form. Existing param_type constructors, when
taking a full or lower-triangle varcov matrix, perform Cholesky
decomposition to convert it to the internal representation. This
internal representation is vis
I'm sending slightly updated version of the patch
where I allow -flto=auto in common_handle_option.
Martin
>From cc04dfc9dbf2ed91a021093d1d27b81848ea726b Mon Sep 17 00:00:00 2001
From: Martin Liska
Date: Mon, 5 Aug 2019 06:44:25 +0200
Subject: [PATCH] Add -flto=auto option value.
gcc/ChangeLog:
In LTO mode, if static library and dynamic library contains same
function and both libraries are passed as arguments, linker will link
the function in dynamic library no matter the sequence. This patch
will output LTO symbol node as UNDEF if BUILT_IN_NORMAL function FNDECL
is a math function, then
On Thu, Aug 8, 2019 at 4:17 PM Jeff Law wrote:
>
> On 8/8/19 7:04 AM, Martin Liška wrote:
> > Hi.
> >
> > As requested by Richi, I'm suggesting to use new dump_printf
> > optimization info infrastructure.
> >
> > Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
> >
> > Ready
On Fri, Aug 9, 2019 at 10:11 AM Martin Liška wrote:
>
> I'm sending slightly updated version of the patch
> where I allow -flto=auto in common_handle_option.
+One can also use @option{-flto=auto} to either use GNU make's
+job server mode to determine the number of parallel jobs, if available.
+Or
On Fri, Aug 9, 2019 at 9:50 AM Alexandre Oliva wrote:
> normal_mv_distribution maintains the variance-covariance matrix param
> in Cholesky-decomposed form. Existing param_type constructors, when
> taking a full or lower-triangle varcov matrix, perform Cholesky
> decomposition to convert it to t
On Fri, Aug 9, 2019 at 10:13 AM Xiong Hu Luo wrote:
>
> In LTO mode, if static library and dynamic library contains same
> function and both libraries are passed as arguments, linker will link
> the function in dynamic library no matter the sequence. This patch
> will output LTO symbol node as UN
In this PR we have two return paths from a function "map". The common
code sets to the value returned by one path, while the other
path does:
= map (&, ...);
We treated this call as tail recursion, losing the copy semantics
on the value returned by the recursive call.
We'd correctly reject
On Fri, 9 Aug 2019, Uros Bizjak wrote:
> On Mon, Aug 5, 2019 at 3:09 PM Uros Bizjak wrote:
>
> > > > > > > (define_mode_iterator MAXMIN_IMODE [SI "TARGET_SSE4_1"] [DI
> > > > > > > "TARGET_AVX512F"])
> > > > > > >
> > > > > > > and then we need to split DImode for 32bits, too.
> > > > > >
> > >
On Fri, Aug 9, 2019 at 10:33 AM Richard Sandiford
wrote:
>
> In this PR we have two return paths from a function "map". The common
> code sets to the value returned by one path, while the other
> path does:
>
> = map (&, ...);
>
> We treated this call as tail recursion, losing the copy seman
On Fri, Aug 09, 2019 at 11:25:30AM +0200, Richard Biener wrote:
> 0.65 │1e0: vpxor %xmm0,%xmm0,%xmm0
> 0.32 │ vpmaxs -0x10(%rsp),%xmm0,%xmm0
> 40.45 │ vmovd %xmm0,%eax
> 2.45 │ imul %r8d,%eax
Shouldn't we hoist the vpxor before the loop? Is it STV being done too lat
On Fri, Aug 09, 2019 at 11:28:32AM +0200, Richard Biener wrote:
> > Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
>
> OK.
Can't we have a CLOBBER also for the RESULT_DECL var in some cases and
on some paths and thus shouldn't we track the RESULT_DECL in
compute_live_vars/live_
On 09/08/19 10:20 +0200, Ulrich Drepper wrote:
On Fri, Aug 9, 2019 at 9:50 AM Alexandre Oliva wrote:
normal_mv_distribution maintains the variance-covariance matrix param
in Cholesky-decomposed form. Existing param_type constructors, when
taking a full or lower-triangle varcov matrix, perform
Hi Jakub,
I think this wrong code bug would be good to be fixed in 9.2.
Would you like me to go ahead, or should it wait for 9.3 ?
Thanks
Bernd.
On 8/7/19 3:32 PM, Vladimir Makarov wrote:
> On 8/5/19 4:37 PM, Bernd Edlinger wrote:
>> Hi!
>>
>>
>> PR 91109 is a wrong-code bug, where LRA is usin
On Fri, Aug 09, 2019 at 10:31:57AM +, Bernd Edlinger wrote:
> I think this wrong code bug would be good to be fixed in 9.2.
>
> Would you like me to go ahead, or should it wait for 9.3 ?
Wait for 9.2.1 reopening, even if we'd roll another RC, I'd be afraid that
for RA changes, especially ones
On Fri, 9 Aug 2019, Richard Biener wrote:
> On Fri, 9 Aug 2019, Uros Bizjak wrote:
>
> > On Mon, Aug 5, 2019 at 3:09 PM Uros Bizjak wrote:
> >
> > > > > > > > (define_mode_iterator MAXMIN_IMODE [SI "TARGET_SSE4_1"] [DI
> > > > > > > > "TARGET_AVX512F"])
> > > > > > > >
> > > > > > > > and then
On Fri, 9 Aug 2019, Jakub Jelinek wrote:
> On Fri, Aug 09, 2019 at 11:25:30AM +0200, Richard Biener wrote:
> > 0.65 │1e0: vpxor %xmm0,%xmm0,%xmm0
> > 0.32 │ vpmaxs -0x10(%rsp),%xmm0,%xmm0
> > 40.45 │ vmovd %xmm0,%eax
> > 2.45 │ imul %r8d,%eax
>
> Shouldn't we hoist
On Thu, Aug 8, 2019 at 12:04 PM Martin Liška wrote:
>
> On 8/7/19 1:57 PM, Richard Biener wrote:
> > On Tue, Aug 6, 2019 at 5:43 PM Martin Liska wrote:
> >
> > This warrants a comment like
> >
> > /* This function originally abstracts from simply comparing
> > get_deref_alias_set
> > so th
On Thu, Aug 8, 2019 at 2:24 PM Michael Matz wrote:
>
> Hi,
>
> On Thu, 8 Aug 2019, Martin Liška wrote:
>
> > > So docs have
> > >
> > > @defmac JUMP_ALIGN (@var{label})
> > > The alignment (log base 2) to put in front of @var{label}, which is
> > > a common destination of jumps and has no fallthru
On Tue, Aug 6, 2019 at 5:44 PM Martin Liska wrote:
>
>
> gcc/ChangeLog:
Hum. I don't like the "abstraction" - how is it going to help you to not
duplicate all the code? What's wrong with doing this all in ICF?
Richard.
> 2019-07-24 Martin Liska
>
> * fold-const.c (operand_equal_p):
On Wed, Aug 7, 2019 at 6:33 PM Steve Ellcey wrote:
>
> While trying to use the -flto and -fwhole-program flags I ran into problems
> understanding what they do. I would like to update the documentation but I
> still don't understand these flags enough to be able to describe their
> behaviour. He
Hi,
On 08/08/19 16:51, Jason Merrill wrote:
On 8/6/19 8:28 AM, Paolo Carlini wrote:
apparently this is now easy to do, likely because a while ago I made
sure that we consistently have meaningful locations for TYPE_DECLs too.
(I went through grokdeclarator and confirmed that when the third
ar
Hi,
On Fri, 9 Aug 2019, Richard Biener wrote:
> Of course I'm still afraid that the other code exists for a reason
> (tuning/hack/whatever...).
>
> Note that with the patch we're now applying LOOP_ALIGN to L2 here:
> if (a)
> foo = bar;
> L2:
> blah;
>
> because there's a jump-around an
On 8/9/19 10:19 AM, Richard Biener wrote:
> OK with that. I still think that making -flto use a jobserver if detected
> (but _not_ use the number of CPU cores by default) makes
> sense as an independent change.
In order to address that, I'm suggesting following patch that I've been
testing.
Mart
Hi.
The patch prevents crashes caused by fact that do_per_function_toporder
uses get_uid () to register all dead cgraph_nodes. That does not work
now as cgraph_nodes are directly released via ggc_free and so that one
will see a garbage here. Second steps is to register all cgraph hooks
and correct
On 8/9/19 2:13 PM, Michael Matz wrote:
> Hi,
>
> On Fri, 9 Aug 2019, Richard Biener wrote:
>
>> Of course I'm still afraid that the other code exists for a reason
>> (tuning/hack/whatever...).
>>
>> Note that with the patch we're now applying LOOP_ALIGN to L2 here:
>> if (a)
>> foo = bar;
>
On Fri, 9 Aug 2019, Richard Biener wrote:
> On Fri, 9 Aug 2019, Richard Biener wrote:
>
> > On Fri, 9 Aug 2019, Uros Bizjak wrote:
> >
> > > On Mon, Aug 5, 2019 at 3:09 PM Uros Bizjak wrote:
> > >
> > > > > > > > > (define_mode_iterator MAXMIN_IMODE [SI "TARGET_SSE4_1"] [DI
> > > > > > > > >
On 8/9/19 2:38 PM, Martin Liška wrote:
> On 8/9/19 10:19 AM, Richard Biener wrote:
>> OK with that. I still think that making -flto use a jobserver if detected
>> (but _not_ use the number of CPU cores by default) makes
>> sense as an independent change.
>
> In order to address that, I'm suggesti
On 8/9/19 10:13 AM, Richard Biener wrote:
> On Thu, Aug 8, 2019 at 4:17 PM Jeff Law wrote:
>>
>> On 8/8/19 7:04 AM, Martin Liška wrote:
>>> Hi.
>>>
>>> As requested by Richi, I'm suggesting to use new dump_printf
>>> optimization info infrastructure.
>>>
>>> Patch can bootstrap on x86_64-linux-gnu
On Fri, Aug 9, 2019 at 3:00 PM Richard Biener wrote:
> > > > > > > > > > (define_mode_iterator MAXMIN_IMODE [SI "TARGET_SSE4_1"] [DI
> > > > > > > > > > "TARGET_AVX512F"])
> > > > > > > > > >
> > > > > > > > > > and then we need to split DImode for 32bits, too.
> > > > > > > > >
> > > > > > > >
Hi.
The patch is about prevention of LTO section name clashing.
Now we have a situation where body of 2 functions is streamed
into the same ELF section. Then we'll end up with smashed data.
Patch can bootstrap on x86_64-linux-gnu and survives regression tests.
Ready to be installed?
Thanks,
Mart
On 8/9/19 6:41 AM, Martin Liška wrote:
Hi.
The patch prevents crashes caused by fact that do_per_function_toporder
uses get_uid () to register all dead cgraph_nodes. That does not work
now as cgraph_nodes are directly released via ggc_free and so that one
will see a garbage here. Second steps is
Jakub Jelinek writes:
> On Fri, Aug 09, 2019 at 11:28:32AM +0200, Richard Biener wrote:
>> > Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
>>
>> OK.
>
> Can't we have a CLOBBER also for the RESULT_DECL var in some cases and
> on some paths and thus shouldn't we track the RESUL
On Fri, Aug 09, 2019 at 04:19:38PM +0100, Richard Sandiford wrote:
> > Can't we have a CLOBBER also for the RESULT_DECL var in some cases and
> > on some paths and thus shouldn't we track the RESULT_DECL in
> > compute_live_vars/live_vars_at_stmt
> > in addition to the local vars (sure, tree-ssa-li
The addsi3_compare_op[12] patterns currently only have constraints to
pick the 32-bit variants of the instructions. Although the assembler
may sometimes opportunistically match a 16-bit t2 instruction, there's
no real control over that within the compiler. Consequently we might
emit a 32-bit
PR target/91386 is a situation where a peephole2 pattern substitution
is discarded late because the selected instructions contain
frame-related notes that we cannot redistribute (because the pattern
has more than one insn in the output). Unfortunately, the original
insns were being modified durin
GCC 9 optimizes a subset of expression of the form
(0 == strcmp(a, b)) based on the length and/or size of
the arguments but it doesn't take advantage of all
the opportunities there. For example in the following,
although it folds the first test to false it doesn't fold
the second one:
char a[4
On Fri, Aug 09, 2019 at 10:17:12AM -0600, Martin Sebor wrote:
> --- a/gcc/gengtype-state.c
> +++ b/gcc/gengtype-state.c
> @@ -79,6 +79,14 @@ enum state_token_en
>STOK_NAME /* hash-consed name or identifier. */
> };
>
> +/* Suppress warning: ISO C forbids zero-size array
On 7/1/19 7:47 PM, Martin Sebor wrote:
>
> Jeff, I looked into your question/suggestion for me last week
> when we spoke, to introduce some sort of a recursion limit for
> get_range_strlen_dynamic. It's easily doable but before we go
> down that path I did some testing to see how bad it can get a
On 8/9/19 10:22 AM, Jakub Jelinek wrote:
On Fri, Aug 09, 2019 at 10:17:12AM -0600, Martin Sebor wrote:
--- a/gcc/gengtype-state.c
+++ b/gcc/gengtype-state.c
@@ -79,6 +79,14 @@ enum state_token_en
STOK_NAME /* hash-consed name or identifier. */
};
+/* Suppress warnin
On Fri, Aug 09, 2019 at 10:51:09AM -0600, Martin Sebor wrote:
> That said, we should change this code one way or the other.
> There is even less of a guarantee that other compilers support
> writing past the end of arrays that have non-zero size than
> that they recognize the documented zero-length
On 7/10/19 5:54 PM, Martin Sebor wrote:
>> So if I'm reading things correctly, it appears gimple-ssa-sprintf.c is
>> no longer a distinct pass. Instead it co-exists with the strlen pass.
>> Right?
>
> Yes. strlen just calls into sprintf to handle the calls.
OK. Just wanted to make sure I unders
Hello,
The attached patch is a proposal to plug a hole in create_trace_edges
(dwarf2cfi.c), which doesn't handle casesi dispatch insns.
The visible misbehavior we observed is a failure in a cross configuration
of a recent acats test for Ada, a very simplified sketch of which is provided
below.
T
This fixes two minor problems with the new testcases. The first is
that almost all other tests, including all vec* tests, for powerpc use
names with dashes, not underscores. The more important one is the the
vec-rotate-1.c and vec-rotate-3.c tests need the -maltivec flag.
Committing to trunk.
On Fri, Aug 9, 2019 at 12:15 AM Martin Liška wrote:
>
> As Jakub correctly noted, I used a piggy backing to put the new function
> to a file that is supposed to contain different functions. So that
> I'm suggesting a new file. Moreover, I'm also adding dup2 fallback.
>
> Patch can bootstrap on x86
On Fri, Aug 09, 2019 at 11:05:42AM -0700, Ian Lance Taylor wrote:
> > * Makefile.in: Add filedescriptor.c.
> > * filedescriptor.c: New file.
> > * lrealpath.c (is_valid_fd): Remove.
>
>
> I don't understand the dup2 fallback. It looks backward: if dup2(fd,
> fd) will retu
On Fri, Aug 9, 2019 at 11:13 AM Jakub Jelinek wrote:
>
> On Fri, Aug 09, 2019 at 11:05:42AM -0700, Ian Lance Taylor wrote:
> > > * Makefile.in: Add filedescriptor.c.
> > > * filedescriptor.c: New file.
> > > * lrealpath.c (is_valid_fd): Remove.
> >
> >
> > I don't understan
On 8/9/19 11:07 AM, Olivier Hainque wrote:
> Hello,
>
> The attached patch is a proposal to plug a hole in create_trace_edges
> (dwarf2cfi.c), which doesn't handle casesi dispatch insns.
>
> The visible misbehavior we observed is a failure in a cross configuration
> of a recent acats test for Ada
On Fri, 28 Jun 2019, Tejas Joshi wrote:
> +CASE_CFN_ROUNDEVEN:
> +CASE_CFN_ROUNDEVEN_FN:
> + if (!REAL_VALUE_ISNAN (*arg) || !flag_errno_math)
Checking flag_errno_math here does not make sense. roundeven never sets
errno (at least, TS 18661-1 makes it implementation-defined whether
On 8/7/19 8:15 AM, Vladimir Makarov wrote:
> On 8/7/19 7:36 AM, senthilkumar.selva...@microchip.com wrote:
>> Hi,
>>
>> gcc/testsuite/c-c++-common/pr60101.c fails with an ICE for the
>> avr target, because of a gcc_assert firing at reload1.c:4233
>>
>> The assert (in the patch below) looks
Attached is a very rough and only superficially barely tested
prototype of the __builtin_warning intrinsic we talked about
the other day. The built-in is declared like so:
int __builtin_warning (int loc,
const char *option,
const char *txt, ...
On 8/5/19 4:47 AM, Dragan Mladjenovic wrote:
> From: "Dragan Mladjenovic"
>
> gcc/ChangeLog:
>
> 2019-08-05 Dragan Mladjenovic
>
> * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
> TARGET_SOFT_FLOAT.
> * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
>
On 8/5/19 4:49 AM, Dragan Mladjenovic wrote:
> From: "Dragan Mladjenovic"
>
> libgcc/ChangeLog:
>
> 2019-08-05 Dragan Mladjenovic
>
> * config/mips/gnustack.h: Check for TARGET_LIBC_GNUSTACK also.
>
> gcc/ChangeLog:
>
> 2019-08-05 Dragan Mladjenovic
>
> * config.in: Regener
On 7/23/19 10:16 AM, Martin Jambor wrote:
> This are all modifications to the testsuite required to get to the
> state described in the cover letter of the entire IPA-SRA
> patch-series. Please note that ipa/ipa-sra-2.c and ipa/ipa-sra-6.c
> should actually be svn rm-ed instead as they try to invo
On Sat, 29 Jun 2019, Segher Boessenkool wrote:
> So I'd say that yes, void * and char * are interchangeable as arguments
> to variable argument functions as well.
They are explicitly interchangeable as arguments to variable argument
functions using va_arg; the definition of va_arg specifies that
On Mon, 5 Aug 2019, Dragan Mladjenovic wrote:
> diff --git a/gcc/configure.ac b/gcc/configure.ac
> index c620dd2..ab080c8 100644
> --- a/gcc/configure.ac
> +++ b/gcc/configure.ac
> @@ -6143,6 +6143,18 @@ if test x$gcc_cv_libc_provides_hwcap_in_tcb = xyes;
> then
> [Define if your target
On 7/23/19 3:57 PM, H.J. Lu wrote:
[ Snip ]
> Here is the updated patch to improve register allocator and RTL
> expressions independently.
>
> Any comments?
>
> Thanks.
>
>
> -- H.J.
>
>
> 0001-i386-Separate-costs-of-pseudo-registers-from-hard-re.patch
>
> From 79834daf252cecfc3ee51acd864641
On 7/27/19 3:22 AM, Uros Bizjak wrote:
> On Wed, Jul 24, 2019 at 5:03 PM Jeff Law wrote:
>
>>> Clearly this approach will run into register allocation issues
>>> but it looks cleaner than writing yet another STV-like pass
>>> (STV itself is quite awkwardly structured so I refrain from
>>> touchin
On 8/9/19 10:58 AM, Jakub Jelinek wrote:
On Fri, Aug 09, 2019 at 10:51:09AM -0600, Martin Sebor wrote:
That said, we should change this code one way or the other.
There is even less of a guarantee that other compilers support
writing past the end of arrays that have non-zero size than
that they
On Fri, 9 Aug 2019, Jeff Law wrote:
> > 2019-08-05 Dragan Mladjenovic
> >
> > * config.in: Regenerated.
> > * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
> > for TARGET_LIBC_GNUSTACK.
> > * configure: Regenerated.
> > * configure.ac: Define TARGET_LIBC_GNUST
On 6/24/19 3:35 PM, Uecker, Martin wrote:
>
>
> Hi,
>
> here is a new version of this patch. It makes "-fno-trampolines"
> work for C which then makes it possible to use nested functions
> without executable stack. The only change in this version is in
> the documentation.
>
> Maybe it could be
On Fri, Aug 9, 2019 at 3:01 PM Jeff Law wrote:
>
> On 7/23/19 3:57 PM, H.J. Lu wrote:
> [ Snip ]
> > Here is the updated patch to improve register allocator and RTL
> > expressions independently.
> >
> > Any comments?
> >
> > Thanks.
> >
> >
> > -- H.J.
> >
> >
> > 0001-i386-Separate-costs-of-pseu
On 8/8/19 7:05 PM, Jeff Law wrote:
On 7/31/19 6:36 PM, Martin Sebor wrote:
More extensive testing of the last week's strlen patch for
PR91183 on various non-mainstream targets and with better tests
has exposed a few gaps and a couple of bugs. The attached patch
addresses all in one change. I c
The attach patch uses a temporary to possibly point at a
pointer that should not be pointed at something sometimes.
Regression tested on x86_64-*-freebsd. OK to commit?
2019-08-09 Steven G. Kargl
PR fortran/88072
* misc.c (gfc_typename): Do not point to something that ought no
Hi,
This patch renames libcall_type to d_libcall_type, fixing PR d/90893.
Bootstrapped and regression tested on x86_64-linux-gnu.
Committed to trunk as r274249.
--
Iain
---
gcc/d/ChangeLog:
PR d/90893
* runtime.cc (enum libcall_type): Rename to...
(enum d_libcall_type)
69 matches
Mail list logo