Hi,
IRA initialization shows high in profiles even when building lto objects. This
patch simply
delays RTL backend initialization until we really decide to output a function.
In some cases
this avoids the initialization completely (like in the case of LTO but also
user target attributes)
and t
On 16 June 2014 15:26, Kyrill Tkachov wrote:
> Hi all,
>
> I noticed that a few saturating math intrinsics in arm_neon.h for aarch64
> have the wrong types, i.e. not what's mandated by the ACLE spec.
>
> This patch fixes that by adjusting the types of the builtin functions that
> those intrinsics
On Thu, Jun 19, 2014 at 07:19:31PM +0200, Jakub Jelinek wrote:
> > + case IFN_UBSAN_BOUNDS:
> > + ubsan_expand_bounds_btn (&gsi);
> > + break;
> > default:
>
> Why *_btn instead of *_ifn ?
Remnant from when I was using __builtin.ubsan instead of the internal
On Fri, Jun 20, 2014 at 10:43:04AM +0200, Marek Polacek wrote:
> +
> +/* Internal function code. */
> +ENUM_BITFIELD(internal_fn) ifn : 5;
Any reason for the " : 5" here? I mean, the union also contains
unsigned int, so it doesn't hurt if you use full 32 bits for it there,
and it should
Hi all,
we have backported a set of AArch64 costs model related revisions in
the linaro/gcc-4_9-branch at r211843. The backported revisions are:
210493 : [AArch64 costs 1/18] Refactor aarch64_address_costs.
210494 : [AArch64 costs 2/18] Add cost tables for Cortex-A57
210495 : [AArch64 costs 3/18
On Fri, Jun 20, 2014 at 10:57:47AM +0200, Jakub Jelinek wrote:
> On Fri, Jun 20, 2014 at 10:43:04AM +0200, Marek Polacek wrote:
> > +
> > +/* Internal function code. */
> > +ENUM_BITFIELD(internal_fn) ifn : 5;
>
> Any reason for the " : 5" here? I mean, the union also contains
> unsigned
On Fri, Jun 20, 2014 at 11:34:26AM +0200, Marek Polacek wrote:
> On Fri, Jun 20, 2014 at 10:57:47AM +0200, Jakub Jelinek wrote:
> > On Fri, Jun 20, 2014 at 10:43:04AM +0200, Marek Polacek wrote:
> > > +
> > > +/* Internal function code. */
> > > +ENUM_BITFIELD(internal_fn) ifn : 5;
> >
>
[I'm at last back to this...]
> > With [1, -x + INF] as the resulting range? But it can be bogus if x is
> > itself equal to +INF (unlike the input range [x + 1, +INF] which is
> > always correct)
>
> Hmm, indeed.
>
> > so this doesn't look valid to me. I don't see how we can get away
> > witho
On Fri, Jun 20, 2014 at 11:39:23AM +0200, Jakub Jelinek wrote:
> On Fri, Jun 20, 2014 at 11:34:26AM +0200, Marek Polacek wrote:
> > On Fri, Jun 20, 2014 at 10:57:47AM +0200, Jakub Jelinek wrote:
> > > On Fri, Jun 20, 2014 at 10:43:04AM +0200, Marek Polacek wrote:
> > > > +
> > > > +/* Internal
Hi,
This patch fixes https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61530.
Tested on x86_64-unknown-linux-gnu, no regressions.
Ok to commit?
-Maxim
gcc/ChangeLog:
2014-06-20 Yury Gribov
Max Ostapenko
* asan.c (build_check_stmt): Add condition.
gcc/testsuite/ChangeLog:
2014-06-20
On Fri, Jun 20, 2014 at 02:04:25PM +0400, Maxim Ostapenko wrote:
>
> This patch fixes https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61530.
>
> Tested on x86_64-unknown-linux-gnu, no regressions.
>
> Ok to commit?
>
> -Maxim
> gcc/ChangeLog:
>
> 2014-06-20 Yury Gribov
> Max Ostap
On 06/20/2014 02:07 PM, Jakub Jelinek wrote:
On Fri, Jun 20, 2014 at 02:04:25PM +0400, Maxim Ostapenko wrote:
This patch fixes https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61530.
Tested on x86_64-unknown-linux-gnu, no regressions.
Ok to commit?
-Maxim
gcc/ChangeLog:
2014-06-20 Yury Gribov
> From: Richard Biener [mailto:richard.guent...@gmail.com]
> Sent: Tuesday, June 10, 2014 5:05 PM
>
> Backports are welcome - please post a patch.
>
Sorry for the delay. Here you are:
diff --git a/gcc/testsuite/gcc.c-torture/execute/pr61375.c
b/gcc/testsuite/gcc.c-torture/execute/pr61375.c
new
This patch fixes invalid sanitization of trailing byte in
__builtin_strlen ( https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61547).
Tested on x86_64-unknown-linux-gnu, no regressions.
Ok to commit?
-Maxim
gcc/ChangeLog:
2014-06-20 Yury Gribov
Max Ostapenko
PR sanitizer/61547
* asan
From: Trevor Saunders
Hi,
this patch allows you to define the type the hash table stores as elements
instead of the type elements point at by having your hash descriptor define the
type store_values_directly. It turns out trying to implement both cases with
the same code is really confusing, so
From: Trevor Saunders
Hi,
This patch adds a hash_map class so we can consolidate the boiler plate around
using hash_table as a map, it also allows us to get rid of pointer_map which I
do in this patch by converting its users to hash_map.
bootstrapped + regtested without regression on x86_64-unk
On Fri, Jun 20, 2014 at 08:41:22AM +0200, Jan Hubicka wrote:
> Hi,
> this patch moves init and fini priorities to symbol table instead of trees.
> They are already in on-side hashtables, but the hashtables are now maintaned
> by symbol table. This is needed for correctness with LTO.
>
> Currently
On Fri, Jun 20, 2014 at 02:49:12PM +0400, Maxim Ostapenko wrote:
> This patch fixes invalid sanitization of trailing byte in __builtin_strlen (
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61547).
>
> Tested on x86_64-unknown-linux-gnu, no regressions.
> +
> +#include
> +
> +char a[2] = "0";
>
On Fri, Jun 20, 2014 at 2:57 PM, Jakub Jelinek wrote:
> On Fri, Jun 20, 2014 at 02:49:12PM +0400, Maxim Ostapenko wrote:
>> This patch fixes invalid sanitization of trailing byte in __builtin_strlen (
>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61547).
>>
>> Tested on x86_64-unknown-linux-gnu,
Hi,
On Thu, Jun 19, 2014 at 06:18:47PM +0200, Bernd Edlinger wrote:
> Hi,
>
> from a recent discussion on g...@gcc.gnu.org I have learned that the default
> of
> --param allow-store-data-races is still 1, and it is causing problems.
> Therefore I would like to suggest to change the default of th
While adding some new type modifiers I did find a typo in my original
patch, so I decided to add some testcases to make sure no regressions
were introduced. This is the same patch as the original, but with the
typo in modified_type_die fixed that could accidentally add an extra
layer of type modifi
On 19 June 2014 13:27, Alan Lawrence wrote:
> This updates the .md files to generate V1DFmode patterns instead of DFmode
> for create and reinterpret, and the corresponding __builtins.
>
> The various other float64x1_t intrinsics can then be rewritten, generally
> I've tried to use gcc vector exte
On 19 June 2014 13:30, Alan Lawrence wrote:
> Similarly, this makes int64x1_t a proper vector type, updating arm_neon.h
> with many explicit vector construction/destruction operations (also
> including some range checking using __builtin_aarch64_im_lane_boundsi).
>
> Change the vabs_s64 intrinsic
> What about bootstrap though?
Bootstrap in progress.
>> +__SIZE_TYPE__ strlen (const char *p) {
>> + /* Simulate error */
>> + if (p == a)
>> +return 1;
> Why this? Can't you instead just use
__attribute__((no_sanitize_address, noinline))
> on it instead?
Done.
Ok to commit if boots
On Fri, Jun 20, 2014 at 04:55:43PM +0400, Maxim Ostapenko wrote:
> > What about bootstrap though?
>
> Bootstrap in progress.
>
> >> +__SIZE_TYPE__ strlen (const char *p) {
> >> + /* Simulate error */
> >> + if (p == a)
> >> +return 1;
>
> > Why this? Can't you instead just use __attribute
On 06/19/2014 01:29 PM, Marc Glisse wrote:
That looks weird to me. The const&& version is a better match than the
const&, so we should pick that one in overload resolution, but if we
remove the const&& version, the other one seems valid to me
Hmm, you're right, I was confused.
Jason
On 06/20/2014 04:59 PM, Jakub Jelinek wrote:
On Fri, Jun 20, 2014 at 04:55:43PM +0400, Maxim Ostapenko wrote:
What about bootstrap though?
Bootstrap in progress.
+__SIZE_TYPE__ strlen (const char *p) {
+ /* Simulate error */
+ if (p == a)
+return 1;
Why this? Can't you instead just u
On 19 June 2014 14:32, Yufeng Zhang wrote:
> Hi,
>
> This patch updates a number of aapcs64 tests to make them big-endian
> friendly. Changes are mainly:
>
> * checking the W regs instead of X regs for integral arguments less than 8
> bytes
> * correcting the corresponding stack location checks i
> "Samuel" == Samuel Bronson writes:
Samuel> [Am I really supposed to CC this to gcc@ like binutils/MAINTAINERS
Samuel> says I should?]
I think just for files that are intended to be put in both trees and
shared.
Samuel> Individual users will still have to:
Samuel> 1. Install git-merge-cha
On 20/06/14 09:41, Marcus Shawcroft wrote:
On 16 June 2014 15:26, Kyrill Tkachov wrote:
Hi all,
I noticed that a few saturating math intrinsics in arm_neon.h for aarch64
have the wrong types, i.e. not what's mandated by the ACLE spec.
This patch fixes that by adjusting the types of the built
Hi all,
Now that Alan fixed the float64x1_t machinery, this patch implements
some low-hanging intrinsics
in arm_neon.h.
Tested aarch64-none-elf and bootstrapped on aarch64-linux.
Ok for trunk?
Thanks,
Kyrill
2014-06-20 Kyrylo Tkachov
* config/aarch64/arm_neon.h (vfma_f64): New intri
Hi,
we have merged the gcc-4_8-branch into linaro/gcc-4_8-branch up to
revision 210799 as r211850.
Thanks
Yvan
I don't have any comments on this patch.
--
Joseph S. Myers
jos...@codesourcery.com
> "Trevor" == Trevor Saunders writes:
Trevor> I'm curious, what is the reason you choose not to write this in C++11 or
Trevor> later? Its distributed with gcc, so the only case where you aren't
Trevor> building with the in tree compiler and libraries is when your cross
Trevor> compiling gcc,
On Thu, Jun 19, 2014 at 07:47:54PM +0200, Jakub Jelinek wrote:
> On Thu, Jun 19, 2014 at 04:56:53PM +0200, Marek Polacek wrote:
> > + /* Don't instrument this FMA-like array in non-strict
>
> Also, please don't use FMA to mean flexible member array, it is
> flexible array member, but more import
I tested variant to use additional the split pass for memory combining
for indirect jumps. By this we don't need to add a second peephole2
pass. Other advantage of this version is that even for -O1 we do
combining.
ChangeLog
2014-06-20 Kai Tietz
PR target/39284
* config/i386/i386.md
Hi,
I have noticed there is a minor flaw with the include path when cloog is
installed in-tree.
That is, the cloog-include directory is added twice, first with absolute path,
and then
again with relative path, but with one "../" to little, so this is useless when
compiling
sources in the gcc d
On 06/20/2014 08:56 AM, Kai Tietz wrote:
> +(define_split
> + [(set (match_operand:W 0 "register_operand")
> +(match_operand:W 1 "memory_operand"))
> + (set (pc) (match_dup 0))]
> + "!TARGET_X32 && peep2_reg_dead_p (2, operands[0])"
> + [(set (pc) (match_dup 1))])
> +
Huh? You can't
When implementing -fsanitize=bounds I noticed a whole slew of
errors about accessing u.fld[] field in rtx_def. Turned out this
is indeed a bug, the array should have a size of 8; u.hwint[] array
had similar issue. Thus fixed, plus I added some verification code
to genpreds.c (can't do it in genge
2014-06-20 19:23 GMT+02:00 Richard Henderson :
> On 06/20/2014 08:56 AM, Kai Tietz wrote:
>> +(define_split
>> + [(set (match_operand:W 0 "register_operand")
>> +(match_operand:W 1 "memory_operand"))
>> + (set (pc) (match_dup 0))]
>> + "!TARGET_X32 && peep2_reg_dead_p (2, operands[0])"
So revert to use a second peephole2 pass before final split before sched2 pass.
Ok for apply
ChangeLog
2014-06-20 Kai Tietz
PR target/39284
* passes.def (peephole2): Add second peephole2 pass before
split before sched2 pass.
* config/i386/i386.md (peehole2): To combine
ind
On 06/20/2014 10:52 AM, Kai Tietz wrote:
> 2014-06-20 Kai Tietz
>
> PR target/39284
> * passes.def (peephole2): Add second peephole2 pass before
> split before sched2 pass.
> * config/i386/i386.md (peehole2): To combine
> indirect jump with memory.
> (split2): Likewise.
2014-06-20 19:55 GMT+02:00 Richard Henderson :
> On 06/20/2014 10:52 AM, Kai Tietz wrote:
>> 2014-06-20 Kai Tietz
>>
>> PR target/39284
>> * passes.def (peephole2): Add second peephole2 pass before
>> split before sched2 pass.
>> * config/i386/i386.md (peehole2): To combine
>>
On 06/20/14 12:07, Kai Tietz wrote:
2014-06-20 19:55 GMT+02:00 Richard Henderson :
On 06/20/2014 10:52 AM, Kai Tietz wrote:
2014-06-20 Kai Tietz
PR target/39284
* passes.def (peephole2): Add second peephole2 pass before
split before sched2 pass.
* config/i386/i386.md (pe
On 06/20/2014 03:11 PM, Jason Merrill wrote:
On 06/19/2014 01:29 PM, Marc Glisse wrote:
That looks weird to me. The const&& version is a better match than the
const&, so we should pick that one in overload resolution, but if we
remove the const&& version, the other one seems valid to me
Hmm, y
Now that we're calling build_this in build_over_call, it needs to happen
on the template path as well.
Tested x86_64-pc-linux-gnu, applying to trunk.
commit daf445b34181c222baa792e7310fc4af2d26ec3c
Author: Jason Merrill
Date: Thu Jun 19 14:29:51 2014 +0200
PR c++/61556
* call.c (bu
On 19-06-14 20:41, Richard Henderson wrote:
On 06/19/2014 11:25 AM, Tom de Vries wrote:
On 19-06-14 05:53, Richard Henderson wrote:
On 06/01/2014 03:00 AM, Tom de Vries wrote:
+aarch64_emit_call_insn (rtx pat)
+{
+ rtx insn = emit_call_insn (pat);
+
+ rtx *fusage = &CALL_INSN_FUNCTION_USAGE
014-06-20 Maciej W. Rozycki
gcc/
* config/rs6000/rs6000.md: Append `DONE' to preparation
statements of `bswap' pattern splitters.
Okay.
Thanks, David
On Fri, Jun 20, 2014 at 07:36:41PM +0200, Marek Polacek wrote:
> 2014-06-20 Marek Polacek
>
> * genpreds.c (verify_rtx_codes): New function.
> (main): Call it.
> * rtl.h (RTX_FLD_WIDTH, RTX_HWINT_WIDTH): Define.
> (struct rtx_def): Use them.
Note, e.g. Coverity also com
On Fri, 20 Jun 2014, David Edelsohn wrote:
> gcc/
> * config/rs6000/rs6000.md: Append `DONE' to preparation
> statements of `bswap' pattern splitters.
>
> Okay.
Committed, thanks for your review.
Maciej
On 06/20/14 13:01, Jakub Jelinek wrote:
On Fri, Jun 20, 2014 at 07:36:41PM +0200, Marek Polacek wrote:
2014-06-20 Marek Polacek
* genpreds.c (verify_rtx_codes): New function.
(main): Call it.
* rtl.h (RTX_FLD_WIDTH, RTX_HWINT_WIDTH): Define.
(struct rtx_def):
There aren't too many users of the cmpelim pass, and previously they were all
small embedded targets without an FPU.
I'm a bit surprised that Ramana decided to enable this pass for aarch64, as
that target is not so limited as the block comment for the pass describes.
Honestly, whatever is being de
Hi
I would like to finally propose this patch before the one on
_Rb_tree, as a separate one.
I have adopted the same evolution on the tracker_allocator with
even a perfect forwarding constructor to allow its usage on top of the
uneq_allocator which take a personality parameter. Doin
Hello,
as reported in PR 61503, there seems to be a typo in the shrd pattern. I
think it is quite unlikely to cause any problem, because the pattern is 1
instruction too long for combine to recognize it (by the way, if someone
has suggestions for PR 55583...). But it is still better to fix it.
On Fri, Jun 20, 2014 at 09:01:14PM +0200, Jakub Jelinek wrote:
> On Fri, Jun 20, 2014 at 07:36:41PM +0200, Marek Polacek wrote:
> > 2014-06-20 Marek Polacek
> >
> > * genpreds.c (verify_rtx_codes): New function.
> > (main): Call it.
> > * rtl.h (RTX_FLD_WIDTH, RTX_HWINT_WIDTH): Defi
On Fri, Jun 20, 2014 at 01:55:41PM -0600, Jeff Law wrote:
> >like spot. Most RTLs are allocated through rtx_alloc and the size
> >is determined from RTX_HDR_SIZE (i.e. offsetof) and/or RTX_CODE_SIZE,
> >so your rtl.h change IMHO shouldn't affect anything but make the
> >expmed.c init_expmed_rtl st
On Fri, Jun 20, 2014 at 01:14:52PM +0800, Hale Wang wrote:
> 2014-06-20 Hale Wang
>
> * gcc.target/arm/lto/: New folder to verify the LTO option.
> * gcc.target/arm/lto/pr61123-enum-size_0.c: New test case.
> * gcc.target/arm/lto/pr61123-enum-size_1.c: Likewise.
>
On Fri, Jun 20, 2014 at 11:10:18PM +0200, Jakub Jelinek wrote:
> On Fri, Jun 20, 2014 at 01:55:41PM -0600, Jeff Law wrote:
> > >like spot. Most RTLs are allocated through rtx_alloc and the size
> > >is determined from RTX_HDR_SIZE (i.e. offsetof) and/or RTX_CODE_SIZE,
> > >so your rtl.h change IMH
Hi!
As mentioned in the PR, some? KVM versions disable some CPU flags,
supposedly so that it can be migrated to any other x86-64 hw.
Thus, it announces only sse2 and lm, but already not sse3, ssse3
nor 3dnow, and (unfortunately) identifies itself as GenuineIntel
family 6 model 13. There is no 64-
2014-06-20 20:14 GMT+02:00 Jeff Law :
> On 06/20/14 12:07, Kai Tietz wrote:
>>
>> 2014-06-20 19:55 GMT+02:00 Richard Henderson :
>>>
>>> On 06/20/2014 10:52 AM, Kai Tietz wrote:
2014-06-20 Kai Tietz
PR target/39284
* passes.def (peephole2): Add second peephole2
On Tue, Jun 17, 2014 at 11:29:01PM +0100, Ramana Radhakrishnan wrote:
> On Sun, May 18, 2014 at 10:23 PM, Aurelien Jarno wrote:
> > On ARM soft-float, the float to double conversion doesn't convert a sNaN
> > to qNaN as the IEEE Std 754 standard mandates:
> >
> > "Under default exception handling,
On Fri, Jun 20, 2014 at 2:42 PM, Jakub Jelinek wrote:
> Hi!
>
> As mentioned in the PR, some? KVM versions disable some CPU flags,
> supposedly so that it can be migrated to any other x86-64 hw.
> Thus, it announces only sse2 and lm, but already not sse3, ssse3
> nor 3dnow, and (unfortunately) ide
On Sat, 21 Jun 2014, Aurelien Jarno wrote:
> Index: gcc/testsuite/gcc.dg/pr59833.c
> ===
> --- gcc/testsuite/gcc.dg/pr59833.c(revision 0)
> +++ gcc/testsuite/gcc.dg/pr59833.c(working copy)
> @@ -0,0 +1,15 @@
> +/* PR target/59
The following patch is based on and extends the guality test from my
earlier "Pass type modifiers as flags arguments. Add guality type test."
patch.
gcc/ChangeLog
PR debug/59051
* dwarf2out.h (enum dw_mod_flag): Add dw_mod_restrict.
* dwarf2out.c (dw_mod_decl_flags): Handl
Patch Updated.
Sri
On Mon, Jun 9, 2014 at 3:55 PM, Sriraman Tallam wrote:
> Ping.
>
> On Mon, May 19, 2014 at 11:11 AM, Sriraman Tallam wrote:
>> Ping.
>>
>> On Thu, May 15, 2014 at 11:34 AM, Sriraman Tallam
>> wrote:
>>> Optimize access to globals with -fpie, x86_64 only:
>>>
>>> Currently,
Hello,
after some lengthly investigation it turned out that aliases on AIX doesn't
behave in the way we expect. In particular creating a static alias of a global
symbol has no effect. This is somewhat special behaviour of AIX's .set
pseudo-op I think I can get this fixed by simply emitting alterna
66 matches
Mail list logo
