On Wed, Jan 03, 2018 at 05:00:41PM -0700, Martin Sebor wrote:
> This is an example where having a solution for bug 78014 would
> be helpful. I.e., a -Wformat checker to help enforce the use
No, because %zu isn't portable enough for all the hosts we support.
What we could do is define SIZE_T_PRIN
Hi!
Another spot where we need to punt on BLKmode, otherwise we try to create
a subreg with BLKmode, or ZERO/SIGN_EXTEND with BLKmode.
Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
Eventually, we should try to handle both hw and generic vectors in the debug
info better, bu
gcc/
config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*)
configuration.
config/epiphany/rtems.h: New file.
libgcc/
config.host (epiphany-*-elf*): Add (epiphany-*-rtems*)
configuration.
---
gcc/config.gcc | 16
gcc/config/epip
Hi Nathan,
> This fixes a tree dumping ICE involving static thunk fns. Copying the
> thunked-to fn's context suffices.
the new test FAILs on Solaris/SPARC:
FAIL: g++.dg/ipa/pr83667.C -std=gnu++11 scan-ipa-dump inline "summary for
void c::*.LTHUNK0"
FAIL: g++.dg/ipa/pr83667.C -std=gnu++1
On Thu, Jan 04, 2018 at 09:55:10AM +0100, Rainer Orth wrote:
> Hi Nathan,
>
> > This fixes a tree dumping ICE involving static thunk fns. Copying the
> > thunked-to fn's context suffices.
>
> the new test FAILs on Solaris/SPARC:
>
> FAIL: g++.dg/ipa/pr83667.C -std=gnu++11 scan-ipa-dump inline
Bootstrapping on Solaris/SPARC with /bin/as fails in make compare. The
difference is always like this:
* With -gtoggle (stage2), there's no .text section in the assembler
output.
* Without -gtoggle (stage3), the assembler output gets
--- table.s.gtoggle 2017-08-28 14:07:23.811292680 +
I'd missed a TARGET_VIS2 condition when replacing vec_perm_const_optab
with TARGET_VECTORIZE_VEC_PERM_CONST in r256093.
Tested by Rainer on sparc-sun-solaris* (thanks) and installed as obvious.
Sorry the dumb mistake.
Richard
2018-01-04 Richard Sandiford
gcc/
PR target/83675
tree-ssa-forwprop.c was asserting that a VEC_PERM_EXPR fold on three
VECTOR_CSTs would always succeed, but it's possible for it to fail
with variable-length vectors.
Tested on aarch64-linux-gnu, x86_64-linux-gnu and powerpc64le-linux-gnu.
Also tested by comparing the before and after assembly outp
Hello!
According to [1], invalid canonization from ASHIFT to MULT was removed
a while ago. Some instruction patterns (like the one below) needs to
be updated from MULT to ASHIFT, and alpha was left behind.
(define_insn "*sadd"
[(set (match_operand:I48MODE 0 "register_operand" "=r,r")
(plus:
The first BIT_FIELD_REF folding pattern assumed without checking that
operands satisfy tree_fits_uhwi_p. The second pattern does check this:
/* On constants we can use native encode/interpret to constant
fold (nearly) all BIT_FIELD_REFs. */
if (CONSTANT_CLASS_P (arg0)
Following on from:
* tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
for VR_RANGE only; don't allow VR_ANTI_RANGE.
(extract_range_from_binary_expr_1): Don't call
extract_range_from_multiplicative_op_1 if !range_int_cst_p.
there was a later call to extra
An update of the patch originally posted here, deep in the original
77 machine_mode chain:
https://gcc.gnu.org/ml/gcc-patches/2017-07/msg01407.html
Tested on aarch64-linux-gnu. OK to install?
Richard
2018-01-04 Richard Sandiford
gcc/
* doc/rtl.texi: Document machine_mode wrapper
This makes the searchers use the same optimization for std::byte as
for narrow character types, and also uses it when the predicate is
std::equal_to rather than the default std::equal_to.
I wanted to call the new trait __is_byte_size, for pun value, but it
isn't true for everything with size 1, b
On January 4, 2018 9:36:33 AM GMT+01:00, Jakub Jelinek wrote:
>Hi!
>
>Another spot where we need to punt on BLKmode, otherwise we try to
>create
>a subreg with BLKmode, or ZERO/SIGN_EXTEND with BLKmode.
>
>Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
OK.
Richard.
>Even
Ping**2
Richard Sandiford writes:
> Ping
>
> Richard Sandiford writes:
>> This patch makes the AArch64 vec_perm_const code use the new
>> vec_perm_indices routines, instead of checking each element individually.
>> This means that they extend naturally to variable-length vectors.
>>
>> Also, aar
Useing exists(file_status) inside exists(const path&, error_code&)
means we call status_known twice, because exists(file_status) calls it
internally. By inlining the exists(file_status) logic into the outer
function we avoid the redundant call.
* include/bits/fs_ops.h (exists(const path&,
Recently, Google Project Zero disclosed several classes of attack
against speculative execution. One of these, known as variant-1
(CVE-2017-5753), allows explicit bounds checks to be bypassed under
speculation, providing an arbitrary read gadget. Further details can
be found on the GPZ blog [1] an
This patch implements support for __builtin_load_no_speculate on
AArch64. On this architecture we inhibit speclation by emitting a
combination of CSEL and a hint instruction that ensures the CSEL is
full resolved when the operands to the CSEL may involve a speculative
load.
* config/aarc
This patch adds generic support for the new builtin
__builtin_load_no_speculate. It provides the overloading of the
different access sizes and a default fall-back expansion for targets
that do not support a mechanism for inhibiting speculation.
* builtin_types.def (BT_FN_I1_CONST_VPTR_CO
This patch implements despeculation on ARM. We only support it when
generating ARM or Thumb2 code (we need conditional execution); and we
only support it for sizes up to DImode. For unsupported cases we
fall back to the generic code generation sequence so that a suitable
failure warning is emitt
ping (note also Jeff's reply
https://gcc.gnu.org/ml/gcc-patches/2017-11/msg01916.html)
From: Wilco Dijkstra
Sent: 15 November 2017 15:36
To: Richard Biener
Cc: GCC Patches; nd
Subject: Re: [PATCH] Simplify floating point comparisons
Richard Biener wrote:
> On Tue, Oct 17, 2017 at 6:28 PM, W
Hi,
https://gcc.gnu.org/PR83677 reports that generation of xxpermr is always
wrong. It effectively inverts the order of the two input registers from
what they should be. This patch addresses that and provides a test case
modified from the original report.
Bootstrapped and tested on powerpc64le-
This patch implements support for __builtin_load_no_speculate on
AArch64. On this architecture we inhibit speclation by emitting a
combination of CSEL and a hint instruction that ensures the CSEL is
full resolved when the operands to the CSEL may involve a speculative
load.
* config/aarc
This is a back-port of the __builtin-load_no_speculate builtin to the GCC-7
branch.
Recently, Google Project Zero disclosed several classes of attack
against speculative execution. One of these, known as variant-1
(CVE-2017-5753), allows explicit bounds checks to be bypassed under
speculation, pr
This patch adds generic support for the new builtin
__builtin_load_no_speculate. It provides the overloading of the
different access sizes and a default fall-back expansion for targets
that do not support a mechanism for inhibiting speculation.
So that users can know that this version of GCC sup
This patch implements despeculation on ARM. We only support it when
generating ARM or Thumb2 code (we need conditional execution); and we
only support it for sizes up to DImode. For unsupported cases we
fall back to the generic code generation sequence so that a suitable
failure warning is emitt
This patch fixes an inverted condition in r256093. I checked for other
instances of the same thing and there didn't seem to be any.
Tested on arm-linux-gnueabihf and applied as obvious.
Richard
2018-01-04 Richard Sandiford
gcc/
PR target/83680
* config/arm/arm.c (arm_vector
The following patch fixes an issue Sebastian ran into trying to build Ada
for a new 64-bit PowerPC RTEMS target using the ELFv2 ABI. As I discussed
in the bug, the rs6000 backend has certain assumptions that 64-bit targets
expect that we have FP hardware in the cases where we are using explicit
FP
Hi Michael,
Not a review of the full patch, just a comment about the patterns:
Michael Collison writes:
> +(define_expand "aarch64_fmll_lane_lowv2sf"
> + [(set (match_operand:V2SF 0 "register_operand" "")
> + (unspec:V2SF [(match_operand:V2SF 1 "register_operand" "")
> +
On 01/04/2018 03:55 AM, Rainer Orth wrote:
Hi Nathan,
This fixes a tree dumping ICE involving static thunk fns. Copying the
thunked-to fn's context suffices.
The dump has
IPA function summary for void c::*.LLTHUNK0(...)/1
thanks.
does this patch work for you?
nathan
--
Nathan Sidwel
Hi
The bug reported a particular test di-longlong64-sync-1.c failing when
run on arm-linux-gnueabi with options -mthumb -march=armv5t -O[g,1,2,3]
and -mthumb -march=armv6 -O[g,1,2,3].
According to what I could see, the crash was caused because of the
explicit VOIDmode argument that was sent
Hi Nathan,
> On 01/04/2018 03:55 AM, Rainer Orth wrote:
>> Hi Nathan,
>>
>>> This fixes a tree dumping ICE involving static thunk fns. Copying the
>>> thunked-to fn's context suffices.
>>
>
>> The dump has
>>
>> IPA function summary for void c::*.LLTHUNK0(...)/1
>
> thanks.
>
> does this patch wo
On 01/04/2018 01:30 AM, Jakub Jelinek wrote:
On Wed, Jan 03, 2018 at 05:00:41PM -0700, Martin Sebor wrote:
This is an example where having a solution for bug 78014 would
be helpful. I.e., a -Wformat checker to help enforce the use
No, because %zu isn't portable enough for all the hosts we sup
On Thu, Jan 04, 2018 at 08:50:30AM -0700, Martin Sebor wrote:
> On 01/04/2018 01:30 AM, Jakub Jelinek wrote:
> > On Wed, Jan 03, 2018 at 05:00:41PM -0700, Martin Sebor wrote:
> > > This is an example where having a solution for bug 78014 would
> > > be helpful. I.e., a -Wformat checker to help enf
Hi Sudi,
On 04/01/18 15:35, Sudakshina Das wrote:
Hi
The bug reported a particular test di-longlong64-sync-1.c failing when
run on arm-linux-gnueabi with options -mthumb -march=armv5t -O[g,1,2,3]
and -mthumb -march=armv6 -O[g,1,2,3].
According to what I could see, the crash was caused because
On 01/03/2018 02:49 PM, Michele Pezzutti wrote:
Hi.
On 01/02/2018 05:43 PM, Michele Pezzutti wrote:
On 01/02/2018 02:28 AM, Ed Smith-Rowland wrote:
I like the patch.
I have a similar one in the tr29124 branch.
Anyway, I got held up and I think it's good to have new folks
looking into this
On 01/03/2018 02:23 PM, Jakub Jelinek wrote:
> On Mon, Oct 23, 2017 at 06:26:12PM +0100, Richard Sandiford wrote:
>> 2017-10-23 Richard Sandiford
>> Alan Hayward
>> David Sherwood
> ...
>
>> --- /dev/null2017-10-21 08:51:42.385141415 +0100
>> +++ gcc/testsuite/gcc.
The SVE support for the new CONST_VECTOR encoding needs to be able
to extract the first N bits of the vector and duplicate it. This patch
adds a simplify_subreg rule for that.
The code is covered by the gcc.target/aarch64/sve_slp_*.c tests.
Tested on aarch64-linux-gnu, x86_64-linux-gnu and power
On 01/04/2018 01:51 AM, Sebastian Huber wrote:
> gcc/
> config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*)
> configuration.
> config/epiphany/rtems.h: New file.
>
> libgcc/
> config.host (epiphany-*-elf*): Add (epiphany-*-rtems*)
> configuration.
Seems to me like t
On 01/04/2018 02:15 AM, Rainer Orth wrote:
> Bootstrapping on Solaris/SPARC with /bin/as fails in make compare. The
> difference is always like this:
>
> * With -gtoggle (stage2), there's no .text section in the assembler
> output.
>
> * Without -gtoggle (stage3), the assembler output gets
>
On 01/03/2018 01:54 PM, Jakub Jelinek wrote:
> Hi!
>
> My recent dwarf2out.c, dbxout.c and rs6000.c uses of crtl->has_bb_partition
> all assume that if it is true then the function actually has 2 partitions
> and NOTE_INSN_SWITCH_SECTIONS is present, but as the following testcase
> shows, that isn
On 01/04/2018 03:02 AM, Richard Sandiford wrote:
> tree-ssa-forwprop.c was asserting that a VEC_PERM_EXPR fold on three
> VECTOR_CSTs would always succeed, but it's possible for it to fail
> with variable-length vectors.
>
> Tested on aarch64-linux-gnu, x86_64-linux-gnu and powerpc64le-linux-gnu.
This patch improves register allocation of fma by preferring to update the
accumulator register. This is done by adding fma insns with operand 1 as the
accumulator. The register allocator considers copy preferences only in operand
order, so if the first operand is dead, it has the highest chance
On 01/04/2018 03:22 AM, Richard Sandiford wrote:
> An update of the patch originally posted here, deep in the original
> 77 machine_mode chain:
>
> https://gcc.gnu.org/ml/gcc-patches/2017-07/msg01407.html
>
> Tested on aarch64-linux-gnu. OK to install?
>
> Richard
>
>
> 2018-01-04 Richard
On 01/04/2018 03:06 AM, Richard Sandiford wrote:
> The first BIT_FIELD_REF folding pattern assumed without checking that
> operands satisfy tree_fits_uhwi_p. The second pattern does check this:
>
> /* On constants we can use native encode/interpret to constant
> fold (nearly) all B
Martin Sebor writes:
> In addition to assuming that built-in functions are called with
> the correct number of arguments (bug 83603), the restrict pass
> also assumes that they are called with arguments of the expected
> types. When a built-in is declared with no prototype and called
> with argum
The aarch64_legitimate_constant_p tests for HIGH and CONST seem
to be the wrong way round: (high (const ...)) is valid rtl that
could be passed in, but (const (high ...)) isn't. As it stands,
we disallow anchor+offset but allow (high anchor+offset).
TBH I can't remember whether this caused a test
On Thu, 4 Jan 2018, Richard Earnshaw wrote:
> 1 - generic modifications to GCC providing the builtin function for all
> architectures and expanding to an implementation that gives the
> logical behaviour of the builtin only. A warning is generated if
> this expansion path is used that
On 01/03/2018 02:29 PM, Martin Sebor wrote:
> Prathamesh's fix restores the optimization for the test case
> reported in the bug (thanks!) but it isn't sufficient to bring
> GCC 8 completely up to par with 7. Prior GCC versions are able
> to compute the string length in the test case below but GCC
On 01/03/2018 06:57 AM, Sudakshina Das wrote:
> Hi
>
> This patch add support for the missing transformation of (x | y) == x ->
> (y & ~x) == 0.
> The transformation for (x & y) == x case already exists in
> simplify-rtx.c since 2014 as of r218503 and this patch only adds a
> couple of extra patte
On 01/04/2018 03:11 AM, Richard Sandiford wrote:
> Following on from:
>
> * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
> for VR_RANGE only; don't allow VR_ANTI_RANGE.
> (extract_range_from_binary_expr_1): Don't call
> extract_range_from_multiplicati
Hey Martin,
thanks for implementing this!
One quick note, given that the intermediate file contains multiple
"file" sections, maybe it would make more sense to put the "version:" at
the very beginning of the file (maybe outputting it from
https://github.com/gcc-mirror/gcc/blob/7ea14e0563d16460368
Hi!
I'd like to ping 2 patches:
http://gcc.gnu.org/ml/gcc-patches/2017-12/msg01558.html
- PR83604 VBMI2 fixes
http://gcc.gnu.org/ml/gcc-patches/2017-12/msg01559.html
- PR83604 GFNI & misc fixes
Thanks
Jakub
On 01/03/2018 12:08 AM, Prathamesh Kulkarni wrote:
> On 3 January 2018 at 12:33, Prathamesh Kulkarni
> wrote:
>> On 2 January 2018 at 17:49, Jakub Jelinek wrote:
>>> On Tue, Jan 02, 2018 at 05:39:17PM +0530, Prathamesh Kulkarni wrote:
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr
On 01/02/2018 06:19 AM, Richard Biener wrote:
> On Wed, Dec 20, 2017 at 2:56 PM, Wilco Dijkstra
> wrote:
>> This patch fixes PR83491 - if an SSA expression contains 2 identical float
>> constants, the division reciprocal optimization will ICE. Fix this by
>> explicitly
>> checking for SSA_NAME
On 01/04/2018 10:52 AM, Richard Sandiford wrote:
Martin Sebor writes:
In addition to assuming that built-in functions are called with
the correct number of arguments (bug 83603), the restrict pass
also assumes that they are called with arguments of the expected
types. When a built-in is declar
I've committed the following patch that fixes a regression
for simplification of cshift when the array is zero sized.
Commits were made to both the 6-branch and 7-branch. The
regression does not occur on trunk.
2018-01-04 Steven G. Kargl
PR Fortran/83679
* simplify.c (gfc_simp
Hello world,
here is a patch for simplifying eoshift. Along the way, I discovered
two ICEs caused by wrong arguments to eoshift, which are now also
caught by new checks and a corresponding test case.
This already passed all of the *eoshift* tests and is
regression-testing as I write this. OK if
Here is another update to the t99 scheduling file (thunderx2-t99.md).
There were some instruction types (mostly asimd) that did not have
schedules and other types that had duplicate schedules. With this
patch there should be one schedule for every type and no duplicates.
I did some SPEC2017 runs
On 01/04/2018 06:16 PM, Ed Smith-Rowland wrote:
On 01/03/2018 02:49 PM, Michele Pezzutti wrote:
Hi.
On 01/02/2018 05:43 PM, Michele Pezzutti wrote:
On 01/02/2018 02:28 AM, Ed Smith-Rowland wrote:
I like the patch.
I have a similar one in the tr29124 branch.
Anyway, I got held up and I t
I wrote:
here is a patch for simplifying eoshift. Along the way, I discovered
two ICEs caused by wrong arguments to eoshift, which are now also
caught by new checks and a corresponding test case.
This already passed all of the *eoshift* tests and is
regression-testing as I write this. OK if it
And the test cases
diff --git
a/libstdc++-v3/testsuite/tr1/5_numerical_facilities/special_functions/09_cyl_bessel_j/check_value.cc
b/libstdc++-v3/testsuite/tr1/5_numerical_facilities/special_functions/09_cyl_bessel_j/check_value.cc
index 26f4dd3..e340b78 100644
---
a/libstdc++-v3/testsuite/tr1
On Wed, Jan 03, 2018 at 03:14:10PM +0100, Martin Liška wrote:
> 2018-01-03 Martin Liska
>
> PR ipa/82352
> * g++.dg/ipa/pr82352.C: New test.
> ---
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/ipa/pr82352.C
> @@ -0,0 +1,93 @@
> +// PR ipa/82352
> +// { dg-do compile }
> +// { dg-opti
Hi!
While the testcase could be perhaps handled with some extra effort (the
issue there is just CSE of an early possibly throwing trapping addition,
so maybe_cleanup_or_replace_eh_stmt + gimple_purge_dead_eh_edges +
TODO_cleanup_cfg might do the job, but I'm afraid the pass wouldn't know
what to d
Hi!
This ICEs because split2 splits a HImode rotate by 8 into an instruction
that isn't really valid (bswaphi_lowpart insn, which has preferred_for_size
enabled just the first alternative that uses +Q, while the splitter creates
insn with %r9) and the second alternative with +r should be disabled.
(just a log of my thoughts while reading the patch)
On Thu, 4 Jan 2018, Wilco Dijkstra wrote:
Richard Biener wrote:
On Tue, Oct 17, 2017 at 6:28 PM, Wilco Dijkstra wrote:
+(if (flag_unsafe_math_optimizations)
+ /* Simplify (C / x op 0.0) to x op 0.0 for C > 0. */
+ (for op (lt le gt ge)
LRA has code to try to prevent cycling, by avoiding reloads that
look too similar to the instruction being reloaded. E.g. if we
have a R<-C move for some constant C, reloading the source with
another R<-C move is unlikely to be a good idea.
However, this safeguard unnecessarily triggered in tests
Hi Jakub,
Okay by me. Thanks for fixing this! Sorry I missed this bug on the list.
Feel free to CC me on any SLSR bugs.
-- Bill
> On Jan 4, 2018, at 3:23 PM, Jakub Jelinek wrote:
>
> Hi!
>
> While the testcase could be perhaps handled with some extra effort (the
> issue there is just CSE o
PR lto/83121 reports an ICE deep inside the linemap code when -Wodr
reports on a type mismatch.
The root cause is that the warning can access the DECL_SOURCE_LOCATION
of a streamed-in decl before the lto_location_cache has been applied.
lto_location_cache::input_location stores RESERVED_LOCATION_
On Thu, Jan 4, 2018 at 10:27 PM, Jakub Jelinek wrote:
> Hi!
>
> This ICEs because split2 splits a HImode rotate by 8 into an instruction
> that isn't really valid (bswaphi_lowpart insn, which has preferred_for_size
> enabled just the first alternative that uses +Q, while the splitter creates
> in
This fixes filesystem::remove and filesystem::remove_all for the cases
where the file to be removed doesn't exist. The standard says it
should not be an error, but we were reporting errors.
PR libstdc++/83626
* src/filesystem/ops.cc (remove(const path&, error_code&))): Remove
This is a patch for PR target/83466, an ILP32 aarch64 bug. I verified
that it did not cause any regressions in the GCC testsuite and Andreas
verified that it fixed the original failure that he reported. The
basic idea is to do the addition in ptr_mode (SImode in ILP32) instead
of Pmode (DImode in
This patch is the beginning step to switching the PowerPC long double support
from IBM extended double to IEEE 128-bit floating point on PowerPC servers. It
will be necessary to have this patch or a similar patch to allow the GLIBC team
to begin their modifications in GLIBC 2.28, so that by the ti
What if the file to be removed is externally removed between the
symlink_status and the ::remove call? This is probably QoI because
filesystem race, but it seems reasonable to double check errno if
::remove fails and not fail if the failure is due to the file not
existing.
PR 82922 asks to enable -Wstrict-prototypes. The attached
patch handles the errors in an x86_64 bootstrap. With it,
GCC bootstraps successfully with --enable-languages=all,jit,
but there are many FAILs in the test suite but I think those
could be handled by a script so unless there are objection
On Thu, Jan 4, 2018 at 9:46 AM, Wilco Dijkstra wrote:
> This patch improves register allocation of fma by preferring to update the
> accumulator register. This is done by adding fma insns with operand 1 as the
> accumulator. The register allocator considers copy preferences only in
> operand
>
gcc/
* config/nios2/nios2.h (nios2_section_threshold): Guard by not
USED_FOR_TARGET.
---
gcc/config/nios2/nios2.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/gcc/config/nios2/nios2.h b/gcc/config/nios2/nios2.h
index 858d4a41cd7..f62b2ce3176 100644
--- a/gcc/
gcc/
PR target/83681
* config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
by not USED_FOR_TARGET.
(make_pass_resolve_sw_modes): Likewise.
---
gcc/config/epiphany/epiphany.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/gcc/config/ep
78 matches
Mail list logo
