On Tue, Aug 22, 2017 at 8:26 AM, Andreas Schwab wrote:
> On Aug 21 2017, Daniel Santos wrote:
>
>> This is a problem that occured because of this code in
>> ix86_option_override_internal:
>>
>> if (!opts_set->x_ix86_abi)
>> opts->x_ix86_abi = DEFAULT_ABI;
>
> Why is that a problem? Note op
On Tue, Aug 22, 2017 at 4:10 AM, Daniel Santos wrote:
> This is a problem that occured because of this code in
> ix86_option_override_internal:
>
> if (!opts_set->x_ix86_abi)
> opts->x_ix86_abi = DEFAULT_ABI;
>
> I tested this along with my other patches. OK for trunk?
>
> * config
On Aug 21 2017, Daniel Santos wrote:
> This is a problem that occured because of this code in
> ix86_option_override_internal:
>
> if (!opts_set->x_ix86_abi)
> opts->x_ix86_abi = DEFAULT_ABI;
Why is that a problem? Note opts_set vs opts.
Andreas.
--
Andreas Schwab, SUSE Labs, sch...@su
On Tue, Aug 22, 2017 at 4:00 AM, Daniel Santos wrote:
> We currently error when -mx32 and -mabi=sysv and we encounter a function
> with attribute ms_abi, but we are not erroring on -mx32 and -mabi=ms
> (either explicitly or when it is the default on Windows). In fact, it
> generates code that run
On Tue, Aug 22, 2017 at 01:41:21PM +0930, Alan Modra wrote:
> + "#n=%1 ap=%8=%12 lda=%13 x=%7=%10 y=%0=%2 alpha=%9 o16=%11\n"
> + "#a0=%3 a1=%4 a2=%5 a3=%6"
> + :
> + "+m" (*(double (*)[n]) y),
> + "+r" (n), // 1
Another small revision. That needs to be "+&r" (n), in c
On Mon, Aug 21, 2017 at 06:33:09PM +0100, Richard Sandiford wrote:
> I think it's worth emphasising that tying operands doesn't change
> whether an output needs an earlyclobber or not. E.g. for:
Thanks for noticing this. It turns out that my OpenBLAS example
actually ought to have an early-clobb
This is a problem that occured because of this code in
ix86_option_override_internal:
if (!opts_set->x_ix86_abi)
opts->x_ix86_abi = DEFAULT_ABI;
I tested this along with my other patches. OK for trunk?
* config/i386/i386-opts.h (enum calling_abi): Modify so that no legal
v
We currently error when -mx32 and -mabi=sysv and we encounter a function
with attribute ms_abi, but we are not erroring on -mx32 and -mabi=ms
(either explicitly or when it is the default on Windows). In fact, it
generates code that runs, but is of an undfined ABI.
I'm also changing pr64409.c beca
It took me a while to figure out how to do this so I figured that it should be
in the docs. OK for trunk?
* doc/install.texi: Add more details on selecting multiple tests.
Thanks,
Daniel
Signed-off-by: Daniel Santos
---
gcc/doc/install.texi | 10 ++
1 file changed, 10 insertio
Ping. Original patch here:
https://gcc.gnu.org/ml/gcc-patches/2017-08/msg00766.html
On 08/09/2017 10:14 AM, Jeff Law wrote:
On 08/06/2017 05:08 PM, Martin Sebor wrote:
Well, simply because the way as implemented isn't a must-alias query
but maybe one that's good enough for warnings (reduces false positives
but surely doesn't eliminate them).
I'm very interested in reducing
On 08/21/2017 01:51 AM, Richard Biener wrote:
On Sat, 19 Aug 2017, Andreas Schwab wrote:
On Aug 17 2017, Richard Biener wrote:
I was notifed I broke proper handling of undefined overflow in
multiplicative ops handling. The following resurrects previous
behavior (and adds a testcase).
Boots
Following feedback on std::list patch this one had the same problem of
unused code being deleted. So here is a new version.
Ok to commit ?
François
On 28/07/2017 18:45, François Dumont wrote:
Hi
There was a little issue in this patch, here is the correct version.
François
On 23/07/201
On 18/08/2017 22:04, Jonathan Wakely wrote:
On 28/07/17 18:42 +0200, François Dumont wrote:
Hi
Completing execution of the testsuite revealed a bug. So here is
the correct version of this patch.
François
On 21/07/2017 19:14, François Dumont wrote:
Hi
Here is a proposal for 2 optimiz
2017-08-21 16:35 GMT+04:00 Georg-Johann Lay :
> "address" attribute only must be specified with VARs,
> yet the compiler dived into attribute analysis for
> non-VARs, resulting in ICE.
>
> This patch also adds OPT_Wattributes as warning filter.
>
> Ok to apply?
Approved.
Please apply.
>
> Johann
> According to Eric:
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53037#c32
>
> "if (mode_align >= TYPE_ALIGN (type))" was intentional. I am
> not familiar with STRICT_ALIGNMENT target and there is no
> testcase to show why it is needed.
Yes, the mode promotion triggers the alignment promot
Some more dead code in class lookup. It's not been well formed to allow
class fields to be the type-name for a very long time now.
I booted with a gcc-unreachable inside the elided loop condition, before
ripping the loop out completely.
Again, applied to trunk to get wider exposure.
nathan
On August 21, 2017 7:46:09 PM GMT+02:00, Richard Sandiford
wrote:
>Richard Biener writes:
>> On Tue, Aug 8, 2017 at 10:20 PM, Richard Kenner
>> wrote:
Correct. It is truncated for integer shift, but not simd shift
instructions. We generate a pattern in the split that only
>generates
>
Richard Biener writes:
> On Tue, Aug 8, 2017 at 10:20 PM, Richard Kenner
> wrote:
>>> Correct. It is truncated for integer shift, but not simd shift
>>> instructions. We generate a pattern in the split that only generates
>>> the integer shift instructions.
>>
>> That's unfortunate, because it wo
Thanks for doing this.
Alan Modra writes:
> This is a revised version of
> https://gcc.gnu.org/ml/gcc-patches/2017-03/msg01562.html limited to
> showing just the scratch register aspect, as a followup to
> https://gcc.gnu.org/ml/gcc-patches/2017-08/msg01174.html
>
> * doc/extend.texi (Exte
This small patch adds asserts that we never get to look for a field in a
non-class. I'm moving this code around on the name-lookup branch, but
as Jakub found & fixed a previous suprise of this ilk, I thought I'd put
this straight on trunk to get wider exposure of the invariant.
I also noticed
I missed a c++-03 requirement about friends. Fixed thusly.
Also fixed the recent debug9 failure for stabs debugging. That test
appears to be a dwarf test, but as there's no dg-requires-dwarf the
simplest fix was turning off block partitioning optimization (enabled on
x86 automatically at O2
On 2017-08-21 23:37 +0800, Xi Ruoyao wrote:
> On 2017-04-25 09:30 -0600, Jeff Law wrote:
> > On 04/14/2017 06:44 PM, Xi Ruoyao wrote:
> > > On 2017-04-14 15:00 +0800, Xi Ruoyao wrote:
> > > > On 2017-04-13 09:05 +0200, Richard Biener wrote:
> > > >
> > > > > Did you verify LTO bootstrap still work
On 2017-04-25 09:30 -0600, Jeff Law wrote:
> On 04/14/2017 06:44 PM, Xi Ruoyao wrote:
> > On 2017-04-14 15:00 +0800, Xi Ruoyao wrote:
> > > On 2017-04-13 09:05 +0200, Richard Biener wrote:
> > >
> > > > Did you verify LTO bootstrap still works with the patch?
> > >
> > > I've just done a LTO boot
On Mon, Aug 21, 2017 at 12:59 AM, Richard Biener
wrote:
> On Sat, Aug 19, 2017 at 10:18 PM, H.J. Lu wrote:
>> Don't override alignment specified by user with the same value to
>> preserve TYPE_USER_ALIGN. This fixes PR 53037 tests on Sparc.
>>
>> Does it look right?
>
> Doesn't match do_type_ali
Hello!
Attached patch adds btr/bts/btc patterns with variable count operands,
so combine is able to synthesize these instructions from shift+logic
operations.
2017-08-21 Uros Bizjak
PR target/46091
* config/i386/i386.md (*btsq_imm): Rename from *btsq.
(*btrq_imm): Rename from *btr
We made the interator range access functions constexpr for C++17, but
didn't test it, and so didn't notice that std::__iterator_category
isn't constexpr. This fixes it, and improve tests a bit.
PR libstdc++/81912
* include/bits/stl_iterator_base_types.h (__iterator_category): Add
This was agreed upon and now that early LTO debug landed here it is.
Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
In case the gcc_unreachable () gets through I'll instead remove the
if as noted in my followup TODO.
Richard.
2017-08-21 Richard Biener
* common.opt (
These two patches are OK.
--
Joseph S. Myers
jos...@codesourcery.com
This patch adds a partial_subreg_p predicate to go alongside
paradoxical_subreg_p.
The first two changes to cse_insn preserve the current behaviour,
but the condition seems strange. Shouldn't we be able to continue
to cse if the inner modes of the two subregs have the same size?
The patch also p
This patch fixes pr 81899. It was an underlying bug exposed by my
removal of LANG_TYPE_SPECIFIC from BOUND_TEMPLATE_TEMPLATE_PARM.
B_T_T_P happened to not set CLASSTYPE_USE_TEMPLATE, even though, by
definition, they are template instantiations. However they also
happened to not be DECL_HIDDE
This patch makes more use of the existing paradoxical_subreg_p
predicate and also adds a version that operates on outer and
inner modes.
Some of the affected tests were based on GET_MODE_SIZE rather than
GET_MODE_PRECISION and so the patch could change the result for modes
that have the same size
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
2017-08-21 Richard Biener
PR tree-optimization/81900
* tree-ssa-pre.c (compute_antic_aux): Properly compute changed
for blocks with abnormal predecessors.
(compute_antic): Do not set visite
On Mon, Aug 14, 2017 at 11:19 AM, Bin Cheng wrote:
> Hi,
> This simple patch adds new field in struct dependence_info. The new field
> indicates if non-dependence information is only valid for fixed memory access
> length of this reference. There is a concern that this costs an additional
> byte
ping ?
https://gcc.gnu.org/ml/gcc-patches/2017-07/msg01063.html
Christophe
On 18 July 2017 at 14:50, Christophe Lyon wrote:
> Hello,
>
> I've received a complaint that GCC for AArch64 would generate
> vectorized code relying on unaligned memory accesses even when using
> -mstrict-align. This is
On Mon, Aug 14, 2017 at 11:19 AM, Bin Cheng wrote:
> Hi,
> This simple patch adds new interface returning adjacent vertices for a vertex
> in graph.
> Bootstrap and test in series. Is it OK?
The comment of the function doesn't match its implementation. Why did
you choose
to use the dfs helpers
On 08/18/2017 04:28 AM, Yvan Roux wrote:
https://gcc.gnu.org/ml/gcc-patches/2017-07/msg00730.html
Thanks,
Yvan
gcc/testsuite
2017-07-13 Yvan Roux
PR c++/80287
* g++.dg/pr80287.C: New test.
ok
--
Nathan Sidwell
On Mon, Aug 21, 2017 at 1:14 PM, Richard Sandiford
wrote:
> This patch simplifies the alignment calculations in pad_below.
> The first arm of the "if" was:
>
> - taking GET_MODE_BITSIZE (always equal to GET_MODE_SIZE * BITS_PER_UNIT),
> - rounding up to the next multiple of PARM_BOUNDARY
> - conve
On Mon, Aug 21, 2017 at 1:12 PM, Richard Sandiford
wrote:
> Later patches will add support for frame sizes that are a run-time
> invariant but not a compile-time constant. This then raised the
> question of whether the frame size argument to the function_prologue/
> epilogue hooks should be updat
On Mon, Aug 21, 2017 at 11:58 AM, Richard Sandiford
wrote:
> Richard Biener writes:
>> On Fri, Aug 18, 2017 at 1:04 PM, Richard Sandiford
>> wrote:
>>> Richard Biener writes:
On Fri, Aug 18, 2017 at 10:10 AM, Richard Sandiford
wrote:
> There are several places that test whether:
Hello,
this is a ping for that patch:
https://gcc.gnu.org/ml/gcc-patches/2017-07/msg01958.html
Thanks.
On Mon, Aug 21, 2017 at 3:59 AM, Szabolcs Nagy wrote:
> On 17/08/17 15:56, H.J. Lu wrote:
>> On Thu, Aug 17, 2017 at 6:52 AM, Joseph Myers
>> wrote:
>>> On Sat, 8 Jul 2017, H.J. Lu wrote:
>>>
+@item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++
only)}
+@opindex W
Hi Jerry,
That looks good to me - OK for trunk and for backporting.
Thanks for the patch.
Paul
PS Did you have time to think about that rather more difficult bug,
involving a mix of DT descriptors and intrinsic descriptors for the
declared type?
On 21 August 2017 at 03:23, Jerry DeLisle wrote
"address" attribute only must be specified with VARs,
yet the compiler dived into attribute analysis for
non-VARs, resulting in ICE.
This patch also adds OPT_Wattributes as warning filter.
Ok to apply?
Johann
gcc/
PR target/81910
* config/avr/avr.c (avr_handle_addr_attribute):
Since alignment of int is 2 bytes for m68k, expect either 2 or 4
alignments in warning.
OK for trunk?
H.J.
---
* g++.dg/pr53037-1.C: Expect either 2 or 4 alignments in
warning.
* gcc.dg/pr53037-1.c: Likewise.
---
gcc/testsuite/g++.dg/pr53037-1.C | 6 +++---
gcc/testsuite/
When warn_if_not_aligned_p is true, a warning will be issued on function
declaration later. There is no need to warn function alignment when
warn_if_not_aligned_p is true.
OK for trunk?
H.J.
--
* c-attribs.c (common_handle_aligned_attribute): Don't warn
function alignment if warn
This patch simplifies the alignment calculations in pad_below.
The first arm of the "if" was:
- taking GET_MODE_BITSIZE (always equal to GET_MODE_SIZE * BITS_PER_UNIT),
- rounding up to the next multiple of PARM_BOUNDARY
- converting bits to bytes and
- subtracting the GET_MODE_SIZE
so was in eff
Later patches will add support for frame sizes that are a run-time
invariant but not a compile-time constant. This then raised the
question of whether the frame size argument to the function_prologue/
epilogue hooks should be updated accordingly.
It turned out that only two targets actually used
On Mon, Aug 21, 2017 at 11:54 AM, Richard Sandiford
wrote:
> ...so that it's possible to use TYPE_MODE in tree.h.
>
> Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
Ok.
Richard.
> Richard
>
>
> 2017-08-21 Richard Sandiford
>
> gcc/
> * stor-layout.h (vector_type_mo
On 17/08/17 15:56, H.J. Lu wrote:
> On Thu, Aug 17, 2017 at 6:52 AM, Joseph Myers wrote:
>> On Sat, 8 Jul 2017, H.J. Lu wrote:
>>
>>> +@item -Wpacked-not-aligned @r{(C, C++, Objective-C and Objective-C++ only)}
>>> +@opindex Wpacked-not-aligned
>>> +@opindex Wno-packed-not-aligned
>>> +Warn if a s
On Fri, 12 May 2017, Richard Biener wrote:
>
> This is a heads-up that I am in the process of implementing the last
> of Jasons review comments on the dwarf2out parts of early LTO debug
> support. I hope to post final patches early next week after thoroughly
> re-testing everything.
>
> Note th
Richard Biener writes:
> On Fri, Aug 18, 2017 at 1:04 PM, Richard Sandiford
> wrote:
>> Richard Biener writes:
>>> On Fri, Aug 18, 2017 at 10:10 AM, Richard Sandiford
>>> wrote:
There are several places that test whether:
TYPE_PRECISION (t) == GET_MODE_PRECISION (TYPE_MODE (t
...so that it's possible to use TYPE_MODE in tree.h.
Tested on aarch64-linux-gnu and x86_64-linux-gnu. OK to install?
Richard
2017-08-21 Richard Sandiford
gcc/
* stor-layout.h (vector_type_mode): Move to...
* tree.h (vector_type_mode): ...here.
* stor-layout.c (vect
The read-md.c iterator callbacks previously used a void * to record the
position at which the iterator value should be installed. This doesn't
scale easily to the SUBREG_BYTE representation used by a later patch,
so this patch replaces the void * with both an rtx and an operand
number. The operan
The CONST_WIDE_INT case in const_rtx_hash_1 started the hash
with the precision of the mode, but the mode is always VOIDmode.
Tested on aarch64-linux-gnu and x86_64-linux-gnu. Applied as obvious.
Richard
2017-08-21 Richard Sandiford
Alan Hayward
David Sherwood
g
On Sat, Aug 19, 2017 at 10:18 PM, H.J. Lu wrote:
> Don't override alignment specified by user with the same value to
> preserve TYPE_USER_ALIGN. This fixes PR 53037 tests on Sparc.
>
> Does it look right?
Doesn't match do_type_align so it introduces inconsistencies. The documentation
for TYPE_U
On Sat, 19 Aug 2017, Andreas Schwab wrote:
> On Aug 17 2017, Richard Biener wrote:
>
> > I was notifed I broke proper handling of undefined overflow in
> > multiplicative ops handling. The following resurrects previous
> > behavior (and adds a testcase).
> >
> > Bootstrapped and tested on x86_6
On Fri, Aug 18, 2017 at 8:36 PM, Bill Schmidt
wrote:
> Hi,
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81488 reports a problem with
> SLSR where
> too many memory resources are required to complete SLSR processing of
> conditional
> candidates. The code in question contains large trees of
This PR shows an issue in the must-alias oracle (stmt_kills_ref_p) which
when determining must-alias by finding a common base in two reference
trees doesn't handle trailing arrays correctly. The fix is to do
array-at-struct-end detection on-the-fly by recording the innermost
ARRAY_REF we drop whe
This avoids redundant ref finding.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
2017-08-21 Richard Biener
* tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
(mem_ref_in_stmt): Remove.
(determine_max_movement): Use ref index to
60 matches
Mail list logo
