On 30 July 2016 at 02:57, Joseph Myers wrote:
> On Tue, 26 Jul 2016, Richard Sandiford wrote:
>
>> (which are really just extended tree codes). I suppose copysign is
>> a special case since we can always open code it, but in general we
>> shouldn't fall back to something that could generate a cal
If selective scheduling copies register restores it confuses dwarf2cfi.
2016-06-07 Segher Boessenkool
* sel-sched-ir.c (init_global_and_expr_for_insn): Don't copy
instructions with a REG_CFA_RESTORE note.
---
gcc/sel-sched-ir.c | 1 +
1 file changed, 1 insertion(+)
diff --git
Unfortunately even after the previous patch there are still a few cases
where regrename creates invalid code when used together with separate
shrink-wrapping. At noreturn exits regrename thinks it can use all
callee-saved registers, but that is not true. This patch disables
regrename for function
Doing cprop on frame-related instructions blows up spectacularly.
So don't.
2016-06-07 Segher Boessenkool
* regcprop.c (copyprop_hardreg_forward_1): Don't change
RTX_FRAME_RELATED_P instructions.
---
gcc/regcprop.c | 4
1 file changed, 4 insertions(+)
diff --git a/gcc/re
This is the main substance of this patch series.
Instead of doing all of the prologue and epilogue in one spot, it often
is better to do components of it at different places, so that they are
executed less frequently.
What exactly is a component is completely up to the target; this code
treats it
A restore is supposed to restore some certain register. Restoring it
into some other register will not work. Don't.
2016-06-07 Segher Boessenkool
* regrename.c (build_def_use): Invalidate chains that have a
REG_CFA_RESTORE on some instruction.
---
gcc/regrename.c | 6 ++
This implements the hooks for separate shrink-wrapping for rs6000.
It handles GPRs and LR. The GPRs get a component number corresponding
to their register number; LR gets component number 0.
This improves specint by 1.8%, specfp by 0.5%, separate benchmarks
much more. It improves the hot path in
cfgcleanup would try to join noreturn paths with different components
handled. This then fails in dwarf2cfi.
2016-06-07 Segher Boessenkool
* cfgcleanup.c (outgoing_edges_match): Don't join noreturn calls
if shrink_wrapped_separate.
---
gcc/cfgcleanup.c | 5 +
1 file chang
This patch adds a new command-line flag "-fshrink-wrap-separate", a status
flag "shrink_wrapped_separate", hooks for abstracting the target components,
and documentation for all those.
2016-06-07 Segher Boessenkool
* common.opt (-fshrink-wrap-separate): New flag.
* doc/invoke.t
Deleting restores (before a noreturn) that are dead confuses dwarf2cfi.
2016-06-07 Segher Boessenkool
* dce.c (delete_unmarked_insns): Don't delete instructions with
a REG_CFA_RESTORE note.
---
gcc/dce.c | 9 +
1 file changed, 9 insertions(+)
diff --git a/gcc/dce.c b/
This is the second version. Concern was renamed to component, and all
other comments were addressed (I hope). It still uses only two bitmaps
for the component placement, but now they are called needs_components
and has_components, which hopefully is easier to follow. The "can this
prologue be mo
On 31 July 2016 at 23:39, Joseph Myers wrote:
> On Sat, 30 Jul 2016, Manuel López-Ibáñez wrote:
>
>> > Building for different targets is fairly irrelevant here; the issue is
>> > building for different hosts, which is harder.
>>
>> What about my suggestion of forcing GCC to use the gnulib function
On Sat, 30 Jul 2016, Manuel López-Ibáñez wrote:
> > Building for different targets is fairly irrelevant here; the issue is
> > building for different hosts, which is harder.
>
> What about my suggestion of forcing GCC to use the gnulib functions by
> temporarily removing the system-wide functions
On 31/07/16 20:21 +0100, Jonathan Wakely wrote:
(I'm getting really annoyed with line numbers matching columns!)
testsuite/20_util/pair/cons/explicit_construct.cc is another one where
this happens. If you remove a blank line near the top of the file it
starts FAILing, because the error at expli
On 07/31/2016 10:28 AM, Jason Merrill wrote:
On Fri, Jul 29, 2016 at 7:22 PM, Martin Sebor wrote:
On 07/26/2016 12:53 PM, Jason Merrill wrote:
On 07/23/2016 01:18 PM, Martin Sebor wrote:
+ /* A pair of the first non-static non-empty data members following
+ either the flexible array me
This can use static_assert instead of a runtime assert().
* testsuite/20_util/conditional/requirements/typedefs.cc: Change to
compile-only test.
Tested x86_64-linux, committed to trunk.
commit 26ae1ac1207f9cdc6d533eaa3c0786a2a09f5dac
Author: Jonathan Wakely
Date: Sun Jul 31 20
On 28/07/16 22:06 +0100, Jonathan Wakely wrote:
The dg-options for FreeBSD and Dragonfly should have included
-std=gnu++11, or should have used dg-additional-options to add to,
rather than override, the common dg-options.
Use dg-additional-options in libstdc++ tests
* testsuite/17_intro/
On 28/07/16 22:06 +0100, Jonathan Wakely wrote:
The dg-prune-output directives in these tests are needed to prune some
additional errors that are only issued in C++98 mode, but the line
numbers had got out of date. I've replaced the hardcoded numbers with
strings that will stay valid.
Fix DR 4
On 31/07/16 19:46 +0100, Jonathan Wakely wrote:
diff --git
a/libstdc++-v3/testsuite/28_regex/basic_regex/ctors/basic/raw_string.cc
b/libstdc++-v3/testsuite/28_regex/basic_regex/ctors/basic/raw_string.cc
index 5625653..758f216 100644
--- a/libstdc++-v3/testsuite/28_regex/basic_regex/ctors/basic/
This test uses \n inside a raw string literal, which is not a newline
but just a backslash followed by 'n', and that's not a valid POSIX BRE
so the test fails when run with -std=c++11 rather than -std=gnu++11.
I've replaced \n with newlines, and also tested the "expected fail"
string, which works
On 31/07/16 18:10 +0100, Jonathan Wakely wrote:
PR libstdc++/72745
* include/std/array (get): Use positive message for static assertions.
* include/std/functional (_Safe_tuple_element_t): Fix indentation.
* include/std/tuple (tuple_element>): Add partial
sp
On 31 July 2016 at 22:01, Jan Hubicka wrote:
>> On Tue, 26 Jul 2016, Prathamesh Kulkarni wrote:
>>
>> > + warning_at (gimple_location (g), OPT_Wunused_value,
>> > + "Call from %s to %s has no effect",
>> > + e->caller->name (), e->callee->name ());
>>
>> Diagnostics s
PR libstdc++/72745
* include/std/array (get): Use positive message for static assertions.
* include/std/functional (_Safe_tuple_element_t): Fix indentation.
* include/std/tuple (tuple_element>): Add partial
specialization for invalid indices, with static ass
I'm moving { dg-error "..." "" { target *-*-* } NN } to line NN so it
can just be { dg-error "..." } with an implied line number. That makes
the tests less fragile if the test changes and the line moves.
* testsuite/20_util/duration/requirements/typedefs_neg3.cc: Move
dg-error to
> On Tue, 26 Jul 2016, Prathamesh Kulkarni wrote:
>
> > + warning_at (gimple_location (g), OPT_Wunused_value,
> > + "Call from %s to %s has no effect",
> > + e->caller->name (), e->callee->name ());
>
> Diagnostics should not start with capital letters. Function nam
On Fri, Jul 29, 2016 at 7:22 PM, Martin Sebor wrote:
On 07/26/2016 12:53 PM, Jason Merrill wrote:
On 07/23/2016 01:18 PM, Martin Sebor wrote:
+ /* A pair of the first non-static non-empty data members following
+ either the flexible array member, if found, or the zero-length
+ array
On Sun, Jul 31, 2016 at 02:28:51PM +0100, Jonathan Wakely wrote:
> On 28/07/16 10:20 +0300, Gleb Natapov wrote:
> > [resent with hopefully correct libstdc++ mailing list address this time]
> >
> > Here is my attempt to fix
> > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68297. The resulting patch
On 28/07/16 10:20 +0300, Gleb Natapov wrote:
[resent with hopefully correct libstdc++ mailing list address this time]
Here is my attempt to fix
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68297. The resulting patch
is a little bit long because I had to split and cxxabi.h
include files. The for
On 31/07/16 15:50 +0300, Ville Voutilainen wrote:
diff --git a/libstdc++-v3/include/bits/uses_allocator.h
b/libstdc++-v3/include/bits/uses_allocator.h
index b1ff58a..8fdeda7 100644
--- a/libstdc++-v3/include/bits/uses_allocator.h
+++ b/libstdc++-v3/include/bits/uses_allocator.h
@@ -108,8 +108,12
The library specializations for __int128 and __float128 are not
defined when __STRICT_ANSI__ is defined, so some tests fail when
compiled with -std=c++NN rather than -std=gnu++NN. This lets them
pass.
* testsuite/18_support/numeric_limits/40856.cc [__STRICT_ANSI__]: Do
not test 12
* testsuite/26_numerics/headers/cstdlib/54686.c: Rename to 54686.cc.
There doesn't seem to be any reason for this file to have a .c
extensions, so I've renamed it.
Committed to trunk.
commit 5c2f6316e6d2c8d1d0d4817e7e5850758dcf714c
Author: redi
Date: Sun Jul 31 13:03:46 2016 +
On 31 July 2016 at 14:53, Ville Voutilainen wrote:
>> Somebody didn't run the whole testsuite.
> I will send a patch shortly.
I'm finishing testing this patch on Linux-PPC64.
2016-07-31 Ville Voutilainen
Add missing variable traits, fix testsuite failures.
* include/bits/uses_allocat
On 31 July 2016 at 14:49, Jonathan Wakely wrote:
> A number of tests are now failing because their dg-error lines need
> adjusting:
>
> FAIL: 20_util/declval/requirements/1_neg.cc (test for errors, line 2259)
> FAIL: 20_util/declval/requirements/1_neg.cc (test for excess errors)
> FAIL: 20_util/d
On 29/07/16 22:47 +0300, Ville Voutilainen wrote:
Argh, I missed uses_allocator_v, is_placeholder_v and
is_bind_expression_v, aka the
ones in and . I'll send a follow-up patch in a
week or so unless
someone else fixes those in the meanwhile. :)
A number of tests are now failing because their d
On 07/31/2016 10:57 AM, Arnaud Charlet wrote:
Frankly at this stage, I do not think it makes sense to maintain an
Ada port for DJGPP and in particular maintain all these extra
special cases and #ifdefs.
I don't think this is a reasonable attitude to take with people who
are willing to do the wor
On 07/31/16 12:43, Bernd Edlinger wrote:
> I found out that the trouble starts one instruction earlier:
>
> (insn 19 40 20 2 (set (subreg:DI (reg:QI 93) 0)
> ) pr.c:8 50 {*movdi_aarch64}
> (expr_list:REG_DEAD (reg:DI 110 [ D.3037 ])
> (nil)))
>
oops, my e-mail missed one li
On 07/30/16 13:39, Segher Boessenkool wrote:
> On Sat, Jul 30, 2016 at 08:17:59AM +, Bernd Edlinger wrote:
>> Ok, I the incorrect REG_POINTER is done here:
>>
>> cse_main -> reg_scan -> reg_scan_mark_refs -> set_reg_attrs_from_value
>>
>> and here I see a bug, because if POINTERS_EXTEND_UNSIGNE
> > Frankly at this stage, I do not think it makes sense to maintain an
> > Ada port for DJGPP and in particular maintain all these extra
> > special cases and #ifdefs.
>
> I don't think this is a reasonable attitude to take with people who
> are willing to do the work to do it.
OK, let me reform
38 matches
Mail list logo
