> Ok with those changes.
>Jakub
Thanks. Committed revision 204934.
-Y
On Sat, Nov 16, 2013 at 7:59 AM, John David Anglin wrote:
> As far as I can tell, libsanitizer works on hppa-linux. So, the change
> could be added to the llvm tree.
> However, I'm unlikely to test anything in the tree unless someone tells me
> there's something to test.
Submitted the hppa patch
So, the validation is ok with this patch, I'm just not able to say if
the original performance issue is still fixed with it. Could you
check it Kyrylo ?
Yvan
2013-11-17 Yvan Roux
* config/arm/arm.md (store_minmaxsi): Use only when
optimize_function_for_size_p.
On 15 Novembe
Ping.
On 7 November 2013 15:56, Yvan Roux wrote:
> Hi,
>
> this patch fixed an LRA cycling due to secondary reload (Thumb mode).
> Notice that this patch is a prerequisite to turn on LRA by default on
> ARM. Bootstrapped on a9 and a15 without any regression in the
> testsuite as LRA is off by de
Hi all,
here is a straightforward patch to teach 'get_expr_storage_size' about
type-bound procedures (which are handled internally as
procedure-pointer components of the corresponding vtab). In that sense
the patch should handle both TBPs as well as PPCs.
Regtested on x86_64-unknown-linux-gnu. Ok
> Yes, it passed all tests with -m32, -mx32, -m64 on Linux/x86-64.
> Installed on trunk. Is this OK to backport to 4.8 branch after
> a few days?
Actually there are two issues with your change:
Using time_t for tv_nsec looks actually wrong, the definition on (my)
linux is:
struct timespec {
> Actually there are two issues with your change:
>
> Using time_t for tv_nsec looks actually wrong, the definition on (my)
> linux is:
>
> struct timespec {
> __kernel_time_t tv_sec; /* seconds */
> longtv_nsec;/* nanoseconds */
> };
>
Jeff Law writes:
> On 11/16/13 05:53, Richard Sandiford wrote:
>> After the patch that went in yesterday, all calls to host_integerp and
>> tree_low_cst pass a constant "pos" argument. This series replaces each
>> function with two separate ones:
> [ ... ]
> So I've almost entirely ignored the wh
Alan Modra writes:
> diff -urp gcc3/libffi/testsuite/libffi.call/cls_double_va.c
> gcc4/libffi/testsuite/libffi.call/cls_double_va.c
> --- gcc3/libffi/testsuite/libffi.call/cls_double_va.c 2013-11-15
> 23:03:07.193964372 +1030
> +++ gcc4/libffi/testsuite/libffi.call/cls_double_va.c 2013-11-15
Hi,
This patch adds infrastructure to allow us to model the correct
types for the intrinsics in arm_neon.h. Thus far we have relied
on ugly casts between data types, and treated every intrinsic
as taking 'signed' vectors.
So we have intrinsics in arm_neon.h which look like:
uint32x4_t
vaddw_hig
On 18/11/13 08:37, Yvan Roux wrote:
So, the validation is ok with this patch, I'm just not able to say if
the original performance issue is still fixed with it. Could you
check it Kyrylo ?
Hi Yvan,
I'll run the benchmark today to confirm the performance, but from compiling some
code sequences
> Hi Yvan,
> I'll run the benchmark today to confirm the performance, but from compiling
> some code sequences that exhibited the bad behaviour in the past, I see that
> this patch still fixes the issues. store_minmaxsi is not generated when
> optimising for speed.
Ok Cool, Thanks Kyrill
Cheers,
Ping, as wasting 8 bytes of RAM isn't ignorable on embedded system.
OK to trunk stage 1?
> -Original Message-
> From: Tom Tromey [mailto:tro...@redhat.com]
> Sent: Thursday, October 10, 2013 21:32
> To: Jakub Jelinek
> Cc: Joey Ye; p...@bothner.com; a...@redhat.com; H.J. Lu; gcc-patches;
Ping
2013/11/8 Ilya Enkovich :
> Hi,
>
> Here is an updated patch version with no langhook.
>
> Regarding TLS objects issue - I do not think compiler should compensate the
> absence of instrumentation in libraries. Compiler should be responsible for
> initialization of Bounds Tables for .tdata
Ping
2013/11/7 Ilya Enkovich :
> 2013/11/7 Jeff Law :
>> On 10/31/13 03:15, Ilya Enkovich wrote:
>>>
>>> Hi,
>>>
>>> Here is a patch which adds support for bound constant to be used as
>>> DECL_INITIAL for constant static bounds generated by compiler.
>>>
>>> Thanks,
>>> Ilya
>>> --
>>>
>>> gcc/
>
Thanks for the changes.
> @@ -8162,7 +8162,7 @@ fold_builtin_logarithm (location_t loc,
> /* Prepare to do logN(exp10(exponent) -> exponent*logN(10). */
> {
> REAL_VALUE_TYPE dconst10;
> - real_from_integer (&dconst10, VOIDmode, 10, SIGNED);
> +
Hi,
Here is a patch to adopt some builtins expanding to Pointer Bounds Checker.
Patch mostly disables inlining of instrumented calls to string function. Also
adds support for _NOBND_NOCHK versions of string functions (this version does
not check and copy bounds; therefore may be inlined as re
This fixes an ICE on an obscure case with the -gnatc switch.
Tested on x86_64-suse-linux, applied on the mainline.
2013-11-18 Eric Botcazou
* gcc-interface/decl.c (gnat_to_gnu_entity) : Deal with an
error mark as renamed object in type annotating mode.
--
Eric BotcazouInde
Hi,
Here is a patch to expand bounded arguments for calls, input bounds for bounded
params and returned bounds.
Thanks,
Ilya
--
2013-11-15 Ilya Enkovich
* calls.c: Include tree-chkp.h, rtl-chkp.h.
(arg_data): Add fields for bounds information
and to store value pushed
This fixes an old issue in the compiler, whereby it unexpectedly generates a
read of a volatile variable with scalar type passed as Out parameter to a
subprogram; the subtlety being that the side-effects of the parameter viewed
as a name still need to be evaluated prior to the call.
Tested on x
I'm looking into using LTO to pass information between compilers for
different targets, for the OpenACC implementation. This area of the
compiler seems somewhat obfuscated by overengineering, and I'd like to
simplify it first to make it easier to work with.
lto_gimple_out and lto_finish_out aren't
Ping^2
Thanks,
bin
On Tue, Nov 12, 2013 at 3:08 PM, bin.cheng wrote:
> Ping in this one.
> Hi Bernd, could you please help us on this one?
> Sorry for the inconvenience.
>
> Thanks,
> bin
>
>> -Original Message-
>> From: Bin.Cheng [mailto:amker.ch...@gmail.com]
>> Sent: Monday, November
> diff --git a/gcc/ChangeLog b/gcc/ChangeLog
> index 5cb07b7..754f882 100644
> --- a/gcc/ChangeLog
> +++ b/gcc/ChangeLog
> @@ -1,3 +1,13 @@
> +2013-11-17 Martin Liska
> + Jan Hubicka
> +
> + * cgraphunit.c (node_cmp): New function.
> + (expand_all_functions): Function ordering
Tested on x86_64-suse-linux, applied on the mainline.
2013-11-18 Eric Botcazou
* gcc-interface/trans.c (TARGET_ABI_OPEN_VMS): Delete as redundant.
--
Eric BotcazouIndex: gcc-interface/trans.c
===
--- gcc-interface/tran
Hi,
Here is a patch to add instumented code support in inliner. Main changes are:
- Replace BUILT_IN_CHKP_ARG_BND calls in inlined code with actually passed
bounds
- Replace BUILT_IN_CHKP_BNDRET in caller with returned bounds
- Add bounds copy for assignments generated by inliner
Thanks,
Ping!
Regards
Senthil
On Mon, Nov 04, 2013 at 06:45:19PM +0530, Senthil Kumar Selvaraj wrote:
> The AVR backend does not generate position independent code, yet it
> happily accepts -fpic, -fPIC, -fpie and -fPIE. The generated code
> doesn't change at all. Also, it accepts the -shared option to g
Hi,
Here is a patch to disable splitting when bounds transfer is required for
splitted function.
Thanks,
Ilya
--
2013-11-13 Ilya Enkovich
* ipa-split.c: Include tree-chkp.h.
(consider_split): Do not split when splitted
part needs bounds transfer.
diff --git a/gcc/ip
Hi,
Here is a patch to disable propagation of bounded values.
Thanks,
Ilya
--
2013-11-13 Ilya Enkovich
* ipa-prop.c: Include tree-chkp.h.
(ipa_compute_jump_functions_for_edge): Do not propagate bounded args.
diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
index eb464e4..81e1237
Hi,
On Fri, 15 Nov 2013 13:30:51, Richard Biener wrote:
>> That looks like always pretending it is a bit field.
>> But it is not a bit field, and bitregion_start=bitregion_end=0
>> means it is an ordinary value.
>
> I don't think it is supposed to mean that. It's supposed to mean
> "the access is
Hi,
Here is a patch to disable tail recursion transformation when bounds are passed
by call. The reason is BUILT_IN_CHKP_ARG_BND which should always get default
SSA_NAME of PARM_DECL as an argument.
Thanks,
Ilya
--
2013-11-15 Ilya Enkovich
* tree-tailcall.c: Include tree-chkp.h.
Hi,
Here is a patch remove bounded params from candidates for reduction.
Thanks,
Ilya
--
2013-11-15 Ilya Enkovich
* tree-sra.c: Include tree-chkp.h.
(find_param_candidates): Bounded param is not
a candidate.
diff --git a/gcc/tree-sra.c b/gcc/tree-sra.c
index d97af13.
On Sat, Nov 16, 2013 at 1:55 PM, Richard Sandiford
wrote:
> Add tree_fits_shwi_p and tree_fits_uhwi_p. The implementations are taken
> directly from host_integerp.
tree_ is a bit generic - you only ever return true for INTEGER_CSTs,
so please use int_fits_[su]hwi_p please (mimicing int_fits_type
On Sat, Nov 16, 2013 at 2:03 PM, Richard Sandiford
wrote:
> Add tree_to_shwi and tree_to_uhwi. Initially tree_to_uhwi returns a
> HOST_WIDE_INT, so that it's a direct replacement for tree_low_cst.
> Patch 10 makes it return unsigned HOST_WIDE_INT instead.
Possibly same naming issue (though we al
On Sat, Nov 16, 2013 at 2:05 PM, Richard Sandiford
wrote:
> Like patch 2, but using sed to replace "tree_low_cst (x, 0)" with
> "tree_to_shwi (x)".
Ok with whatever the naming bikeshedding arrives at.
Richard.
> Thanks,
> Richard
>
>
> gcc/c-family/
> * c-common.c, c-format.c, c-omp.c,
On Sat, Nov 16, 2013 at 2:06 PM, Richard Sandiford
wrote:
> Like the previous patch, but for tree_low_cst (x, 1) -> tree_to_uhwi_p (x).
Same ok.
Thanks,
Richard.
> Thanks,
> Richard
>
>
> gcc/ada/
> * gcc-interface/decl.c, gcc-interface/utils.c, gcc-interface/utils2.c:
> Replace
On Sat, Nov 16, 2013 at 2:12 PM, Richard Sandiford
wrote:
> Remove the old functions, which are now unused.
Ok.
Thanks,
Richard.
> Thanks,
> Richard
>
>
> gcc/
> * tree.h (host_integerp, tree_low_cst): Delete.
> * tree.c (host_integerp, tree_low_cst): Delete.
>
> Index: gcc/tree
Hi,
Here is a patch to support BUILT_IN_CHKP_BNDRET and BUILT_IN_CHKP_BIND_BOUNDS
in BUILT_IN_STACK_SAVE result uses.
Thanks,
Ilya
--
2013-11-15 Ilya Enkovich
* tree-ssa-ccp.c (insert_clobber_before_stack_restore): Handle
BUILT_IN_CHKP_BNDRET and BUILT_IN_CHKP_BIND_BOUNDS cal
On Sat, Nov 16, 2013 at 2:10 PM, Richard Sandiford
wrote:
> Handle tree_low_cst references that weren't caught by the sed.
Same ok.
Thanks,
Richard.
> Thanks,
> Richard
>
>
> gcc/ada/
> * gcc-interface/cuintp.c (UI_From_gnu): Use tree_to_shwi rather than
> tree_low_cst.
>
> gcc/
On Sat, Nov 16, 2013 at 2:25 PM, Richard Sandiford
wrote:
> This is probably the only non-obvious part of the series. I went through
> all callers to tree_to_uhwi to see whether they were used in a context
> where signedness mattered. If so, I tried to adjust the casting to match.
>
> This mostl
Hi,
Here is a patch to replace BUILT_IN_CHKP_BNDRET with assignment when optimizing
strchr.
Thanks,
Ilya
--
2013-11-13 Ilya Enkovich
* tree-ssa-strlen.c: Include tree-chkp.h.
(handle_builtin_strchr): Remove retbnd call
when strchr call is removed.
diff --git a/gcc/t
On Mon, Nov 18, 2013 at 11:50:45AM +0100, Richard Biener wrote:
> On Sat, Nov 16, 2013 at 1:55 PM, Richard Sandiford
> wrote:
> > Add tree_fits_shwi_p and tree_fits_uhwi_p. The implementations are taken
> > directly from host_integerp.
>
> tree_ is a bit generic - you only ever return true for I
Richard Biener writes:
> On Sat, Nov 16, 2013 at 2:03 PM, Richard Sandiford
> wrote:
>> Add tree_to_shwi and tree_to_uhwi. Initially tree_to_uhwi returns a
>> HOST_WIDE_INT, so that it's a direct replacement for tree_low_cst.
>> Patch 10 makes it return unsigned HOST_WIDE_INT instead.
>
> Possib
> I actually think tree_fits is better, int_fits* looks too ambiguous to me,
> what kind of integer (tree, int, HWI, ...) you actually mean.
Seconded, all the more so that int_fits_type_p requires an INTEGER_CST as
argument, unlike tree_fits_[su]hwi_p which accept any tree as argument.
--
Eric
Mike,
On 25/10/13 00:37, Mike Stump wrote:
On Oct 24, 2013, at 2:26 AM, Vidya Praveen wrote:
On Mon, Oct 21, 2013 at 06:40:28PM +0100, Mike Stump wrote:
On Oct 21, 2013, at 3:28 AM, Vidya Praveen wrote:
Tests gcc.dg/20050922-1.c and gcc.dg/20050922-2.c includes stdlib.h. This can
be a issue
On Mon, Nov 18, 2013 at 10:08 AM, Richard Sandiford
wrote:
> Jeff Law writes:
>> On 11/16/13 05:53, Richard Sandiford wrote:
>>> After the patch that went in yesterday, all calls to host_integerp and
>>> tree_low_cst pass a constant "pos" argument. This series replaces each
>>> function with two
On Mon, Nov 18, 2013 at 11:10 AM, Bernd Schmidt wrote:
> I'm looking into using LTO to pass information between compilers for
> different targets, for the OpenACC implementation. This area of the
> compiler seems somewhat obfuscated by overengineering, and I'd like to
> simplify it first to make i
On Mon, Nov 18, 2013 at 5:36 AM, Jeff Law wrote:
> On 11/17/13 04:28, Steven Bosscher wrote:
>>
>>
>> TRAP_CONDITION (PATTERN (i3)) == const1_rtx
>>
>> But shouldn't the check be on const_true_rtx? Or does combine put a
>> const1_rtx there?
>
> I took const1_rtx from control_flow_insn_p. That's u
Hi,
Here is a patch to avoid ICE in debug info generator for instrumented code.
Thanks,
Ilya
--
2013-11-15 Ilya Enkovich
* dbxout.c (dbxout_type): Ignore POINTER_BOUNDS_TYPE.
* dwarf2out.c (gen_subprogram_die): Ignore bound args.
(gen_type_die_with_usage): Skip pointer
On 11/04/2013 04:31 AM, bin.cheng wrote:
> 2013-11-01 Bin Cheng
>
> * tree-ssa-loop-ivopts.c (enum ainc_type): New.
> (address_cost_data): New field.
> (get_address_cost): Compute auto-increment rtx cost in ainc_costs.
> Use ainc_costs for auto-increment rtx patterns.
>
Hi,
This modified test case exposes a bug in the already approved part of the
strict-volatile-bitfields patch:
#include
typedef struct {
char pad;
int arr[0];
} __attribute__((packed)) str;
str *
foo (int* src)
{
str *s = malloc (sizeof (str) + sizeof (int));
s->arr[0] = 0x12345678;
On Fri, Nov 15, 2013 at 08:41:38PM +0100, Jakub Jelinek wrote:
> Hi!
>
> This patch adds support for -fsanitize=leak and -static-liblsan options.
> If combined with -fsanitize=address, it does nothing, otherwise it links
> in liblsan, a new shared+static library (on x86_64-linux only so far,
> the
Richard Biener writes:
> On Mon, Nov 18, 2013 at 10:08 AM, Richard Sandiford
> wrote:
>> Jeff Law writes:
>>> On 11/16/13 05:53, Richard Sandiford wrote:
After the patch that went in yesterday, all calls to host_integerp and
tree_low_cst pass a constant "pos" argument. This series rep
Hi,
Here is a patch to add size relocation and instruction to obtain object's size
in i386 target.
Thanks,
Ilya
--
2013-11-15 Ilya Enkovich
* config/i386/i386.md (UNSPEC_SIZEOF): New.
(move_size_reloc_): New.
* config/i386/predicates.md (size_relocation): New.
> On Mon, Nov 18, 2013 at 11:10 AM, Bernd Schmidt
> wrote:
> > I'm looking into using LTO to pass information between compilers for
> > different targets, for the OpenACC implementation. This area of the
> > compiler seems somewhat obfuscated by overengineering, and I'd like to
> > simplify it fi
Hi!
info->dlpi_addr == 0 is true for executables (but those have also
info->dlpi_name set to ""), but not just for those, it is true also
for successfully prelinked shared libraries or any other libraries that
have been linked at non-zero offset and the dynamic linker managed to mmap
them where th
On 11/18/2013 02:03 PM, Jan Hubicka wrote:
>> On Mon, Nov 18, 2013 at 11:10 AM, Bernd Schmidt
>> wrote:
>>> The following patch simplifies this by getting rid of
>>> all_lto_gen_passes. We could simplify a little further if we decided we
>>> really don't need two different timevars for different
On Fri, 15 Nov 2013, Richard Biener wrote:
> On Fri, 15 Nov 2013, Jakub Jelinek wrote:
>
> > On Fri, Nov 15, 2013 at 02:56:51PM +0100, Richard Biener wrote:
> > > Now that there is (finally :() a wrong-code testcase for the
> > > PR54570 issue we can no longer ignore it (bah). So the following
>
On Mon, Nov 18, 2013 at 2:19 PM, Bernd Schmidt wrote:
> On 11/18/2013 02:03 PM, Jan Hubicka wrote:
>>> On Mon, Nov 18, 2013 at 11:10 AM, Bernd Schmidt
>>> wrote:
The following patch simplifies this by getting rid of
all_lto_gen_passes. We could simplify a little further if we decided w
Hi!
This patch adds libbacktrace symbolizer to libsanitizer, with which we
can avoid spawning and running external an external program (especially when
it is not even provided by gcc itself).
I've kept the possibility to override the symbolizer by magic symbols
(InternalSymbolizer), and as I have
On Wed, Nov 13, 2013 at 12:13:48AM +0100, Marek Polacek wrote:
> --- gcc/config/bootstrap-ubsan.mk.mp 2013-11-12 13:46:13.345182065 +0100
> +++ gcc/config/bootstrap-ubsan.mk 2013-11-12 13:46:49.812314016 +0100
> @@ -2,6 +2,6 @@
>
> STAGE2_CFLAGS += -fsanitize=undefined
> STAGE3_CFLAGS += -
On Mon, Nov 18, 2013 at 1:02 AM, Arnaud Charlet wrote:
>> Actually there are two issues with your change:
>>
>> Using time_t for tv_nsec looks actually wrong, the definition on (my)
>> linux is:
>>
>> struct timespec {
>> __kernel_time_t tv_sec; /* seconds */
>> lon
This is something I've been meaning to change for years, see
http://gcc.gnu.org/ml/libstdc++/2007-10/msg00180.html for a prior
post. Currently __shared_ptr specializations still
dispatch through __gnu_cxx::__exchange_and_add_dispatch for ref-count
updates. The attached patch means the _S_single po
On Mon, Nov 18, 2013 at 02:51:41PM +0100, Jakub Jelinek wrote:
> On Wed, Nov 13, 2013 at 12:13:48AM +0100, Marek Polacek wrote:
> > --- gcc/config/bootstrap-ubsan.mk.mp2013-11-12 13:46:13.345182065
> > +0100
> > +++ gcc/config/bootstrap-ubsan.mk 2013-11-12 13:46:49.812314016 +0100
> > @@
> >> struct timespec {
> >> __kernel_time_t tv_sec; /* seconds */
> >> longtv_nsec;/* nanoseconds */
> >> };
>
> This is wrong for x32 where tv_nsec is long long, not
> long. There are a couple places where long should be
> long long for
On Fri, Nov 15, 2013 at 12:34:14PM -0800, Ian Lance Taylor wrote:
> On Fri, Nov 15, 2013 at 11:41 AM, Jakub Jelinek wrote:
> >
> > This patch adds support for -fsanitize=leak and -static-liblsan options.
> > If combined with -fsanitize=address, it does nothing, otherwise it links
> > in liblsan, a
On Mon, Nov 18, 2013 at 01:15:02PM +0100, Marek Polacek wrote:
> Moreover, it fixes the bootstrap-ubsan failure when
> building fixincl - I guess the gnu-user.h hunk was missing. Thanks.
Sorry, turned out this is bogus. It really will need some
configure/Makefile fix.
Marek
Hi,
Here is a patch introducing i386 target versions of Pointer Bounds Checker
builtins.
Thanks,
Ilya
--
2013-11-15 Ilya Enkovich
* config/i386/i386-builtin-types.def (BND): New.
(ULONG): New.
(BND_FTYPE_PCVOID_ULONG): New.
(VOID_FTYPE_BND_PCVOID): New.
On Mon, Nov 18, 2013 at 5:57 AM, Arnaud Charlet wrote:
>> >> struct timespec {
>> >> __kernel_time_t tv_sec; /* seconds */
>> >> longtv_nsec;/* nanoseconds */
>> >> };
>>
>> This is wrong for x32 where tv_nsec is long long, not
>> long.
Hi,
this patch adds support to anti-ranges to determine_block_size.
This makes it possible to bound size of block from bellow that is
useful to avoid need for small size code path and it also allows
us to inline the following:
void *a;
void *b;
t(int c)
{
if (c<10)
memcpy (a,b,c);
}
Because
> Ada was using long for time_t and
>
> type timeval is array (1 .. 2) of C.long
>
> It assumes that the type of tv_nsec is the same as tv_sec.
Yes, and that was indeed wrong/dangerous.
> > --- s-osinte-solaris-posix.ads (revision 298928)
> > +++ s-osinte-solaris-posix.ads (working copy)
> >
On Sun, Nov 17, 2013 at 3:38 PM, Jan Hubicka wrote:
> Hi,
> this is version I comitted. It also adds a testcase and enables the support
> in i386 backend.
>
> Honza
>
> * doc/md.texi (setmem, movstr): Update documentation.
> * builtins.c (determine_block_size): New function.
>
On Mon, Nov 18, 2013 at 6:17 AM, Arnaud Charlet wrote:
>> >
>> > type timespec is record
>> >tv_sec : time_t;
>> > - tv_nsec : long;
>> > + tv_nsec : time_t;
>> > end record;
>> > pragma Convention (C, timespec);
>> >
>> > But I'd rather settle on which type is appr
> >> __syscall_slong_t is a Linux specific type. We
> >> can add tv_nsec_t, which should be the same
> >> as time_t for all the current targets.
> >
> > Introducing tv_nsec_t looks reasonable to me.
> >
>
> Can you make the change?
>
> Thanks.
Not right now, I have lots of other things to do.
On Fri, Nov 15, 2013 at 11:10:18AM +0100, Jakub Jelinek wrote:
> 2013-11-15 Jakub Jelinek
>
> * cfgexpand.c (struct stack_vars_data): Add asan_base and asan_alignb
> fields.
> (expand_stack_vars): For -fsanitize=address, use (and set initially)
> data->asan_base as base
On Mon, Nov 18, 2013 at 02:51:41PM +0100, Jakub Jelinek wrote:
> On Wed, Nov 13, 2013 at 12:13:48AM +0100, Marek Polacek wrote:
> > --- gcc/config/bootstrap-ubsan.mk.mp2013-11-12 13:46:13.345182065
> > +0100
> > +++ gcc/config/bootstrap-ubsan.mk 2013-11-12 13:46:49.812314016 +0100
> > @@
On 16/11/13 01:20, Aldy Hernandez wrote:
On 11/15/13 15:05, Aldy Hernandez wrote:
Why all this, and not keep everything but
GF_OMP_FOR_KIND_SIMD and GF_OMP_FOR_KIND_DISTRIBUTE
as they were, and just use:
GF_OMP_FOR_KIND_DISTIRBUTE = 1 << 0,
GF_OMP_FOR_KIND_SIMD = 2 << 0,
GF_
I would agree that the example is just for the case cost model makes
correct estimation But how can we assure ourself that it won't have any
mistakes in the future?
I believe it'll be Ok to introduce an extra flag as Jakub proposed for the
dedicated simd-forced vectorization to use unlimited cost
On Sat, Nov 16, 2013 at 12:33 AM, Jan Hubicka wrote:
>> When testing with -freorder-blocks-and-partition enabled, I hit a
>> verification failure in an LTO profiledbootstrap. Edge forwarding
>> performed when we went into cfg layout mode after bb reordering
>> (during compgotos) created a situatio
On Mon, 18 Nov 2013, Jan Hubicka wrote:
> Hi,
> this patch adds support to anti-ranges to determine_block_size.
> This makes it possible to bound size of block from bellow that is
> useful to avoid need for small size code path and it also allows
> us to inline the following:
> void *a;
> void *b;
On Mon, 18 Nov 2013, Sergey Ostanevich wrote:
> I would agree that the example is just for the case cost model makes
> correct estimation But how can we assure ourself that it won't have any
> mistakes in the future?
We call it bugs and not mistakes and we have bugzilla for it.
Richard.
> I bel
This fixes some minor ugliness I noticed while looking at this code. It
renames a poorly named global variable (shadowed by some function args
in the same file), replaces a cast of an integer value to a pointer, and
removes an undocumented and unused return value.
Bootstrapped and tested on x86_64
Hi,
tested x86_64-linux, committed to mainline.
Paolo.
/
2013-11-18 Paolo Carlini
PR c++/53473
* g++.dg/cpp0x/constexpr-noexcept7.C: New.
Index: g++.dg/cpp0x/constexpr-noexcept7.C
===
--- g++
On Mon, Nov 18, 2013 at 3:53 PM, Teresa Johnson wrote:
> From a bb layout perspective it seems like it would be beneficial to
> do compgotos before layout. Was the current position just to try to
> reduce compile time by keeping the block unified as long as possible?
It was more a hack that got ou
Hi,
Here is a patch with i386 version of Pointer Bounds Checker hooks.
Thanks,
Ilya
--
2013-11-15 Ilya Enkovich
* config/i386/i386.c (ix86_builtin_mpx_function): New.
(ix86_load_bounds): New.
(ix86_store_bounds): New.
(ix86_fn_abi_va_list_bounds_size): New.
The testcase c-c++-common/cilk-plus/PS/body.c needs an effective target
check for fopenmp before adding -fopenmp to its options, otherwise it'll
fail on bare-metal targets like arm-none-eabi.
This patch adds that.
Ok to commit?
Thanks,
Kyrill
2013-11-18 Kyrylo Tkachov
* c-c++-common
(resending in plain text)
On Mon, Nov 18, 2013 at 7:49 PM, Alexey Samsonov wrote:
Hi Jakub,
Unfortunately, recently there were a few enhancements to
sanitizer_symbolizer_posix_libcdep.cc (and friends),
in LLVM trunk, and now it looks different from gcc version (apparently, the
changes were
On 18/11/13 09:14, Kyrill Tkachov wrote:
On 18/11/13 08:37, Yvan Roux wrote:
So, the validation is ok with this patch, I'm just not able to say if
the original performance issue is still fixed with it. Could you
check it Kyrylo ?
Hi Yvan,
I'll run the benchmark today to confirm the performance
On Mon, Nov 18, 2013 at 07:49:03PM +0400, Alexey Samsonov wrote:
> > As for sanitizer, the reason I haven't implemented SendCommand method for
> > the libbacktrace symbolizer is that the library doesn't provide the info
> > as text, but as individual values passed to the callback, so printing
> > t
On Mon, 2013-11-18 at 10:05 +0100, Andreas Krebbel wrote:
> With the patch the htm-nofloat-2 testcase fails. Due to the
> "returns_twice" flag on tbegin the optimizers fail to fold the
> compares of the condition code and the s390_optimize_nonescaping_tx
> routine in turn fails to optimize the sim
On Mon, Nov 18, 2013 at 02:52:34PM +0100, Jakub Jelinek wrote:
> Oh, one more thing, please update gcc/doc/, the -fsanitize= description is
> far from up to date there.
Ok, the following (incremental) hopefully improves the docs. Joseph, would
you mind having a look at this? Thanks,
2013-11-18
Hi,
there was many changes in this area. The following are ones I can think of.
Please fell free to suggest more changes.
We probably should mention Teresa's splitting work once it is complete
and new micro-architectures targetd by x86 backend.
Honza
Index: changes.html
=
On Mon, Nov 18, 2013 at 07:50:33PM +0400, Alexey Samsonov wrote:
> On Mon, Nov 18, 2013 at 7:49 PM, Alexey Samsonov wrote:
> Unfortunately, recently there were a few enhancements to
> sanitizer_symbolizer_posix_libcdep.cc (and friends),
> in LLVM trunk, and now it looks different from gcc versi
On Mon, Nov 18, 2013 at 5:11 AM, Jakub Jelinek wrote:
>
> info->dlpi_addr == 0 is true for executables (but those have also
> info->dlpi_name set to ""), but not just for those, it is true also
> for successfully prelinked shared libraries or any other libraries that
> have been linked at non-zero
This incremental patch brings the implementation of reference binding
instrumentation, that is, it detects cases like
int *p = NULL;
int &r = *p;
or
auto &&rr = *p;
and similarly. It does so by adding a COMPOUND_EXPR to the decl.
As standard says, "A reference shall be initialized to refer to a
On 11/08/13 02:02, Ilya Enkovich wrote:
Hi,
Here is an updated patch version with no langhook.
Regarding TLS objects issue - I do not think compiler should compensate the
absence of instrumentation in libraries. Compiler should be responsible for
initialization of Bounds Tables for .tdata se
On Mon, Nov 18, 2013 at 08:29:05AM -0800, Ian Lance Taylor wrote:
> On Mon, Nov 18, 2013 at 5:11 AM, Jakub Jelinek wrote:
> >
> > info->dlpi_addr == 0 is true for executables (but those have also
> > info->dlpi_name set to ""), but not just for those, it is true also
> > for successfully prelinked
On Mon, 18 Nov 2013, Marek Polacek wrote:
> +@item @option{-fsanitize=shift}
> +
> +This option enables checking that the result of a shift operation is
> +not undefined. Note that what exactly is considered undefined differs
> +slightly between C and C++, as well as between ANSI C and C99, etc.
On Mon, Nov 18, 2013 at 5:39 AM, Jakub Jelinek wrote:
>
> I've kept the possibility to override the symbolizer by magic symbols
> (InternalSymbolizer), and as I have no idea how the llvm buildsystem etc.
> works and what is the possibility there to add libbacktrace, this just
> requires users to c
On Mon, Nov 18, 2013 at 8:46 AM, Jakub Jelinek wrote:
>
> 2013-11-18 Jakub Jelinek
>
> * elf.c (SHN_UNDEF): Define.
> (elf_initialize_syminfo): Add base_address argument. Ignore symbols
> with st_shndx == SHN_UNDEF. Add base_address to address fields.
> (elf_ad
On Mon, Nov 18, 2013 at 6:19 AM, H.J. Lu wrote:
> On Sun, Nov 17, 2013 at 3:38 PM, Jan Hubicka wrote:
>> Hi,
>> this is version I comitted. It also adds a testcase and enables the support
>> in i386 backend.
>>
>> Honza
>>
>> * doc/md.texi (setmem, movstr): Update documentation.
>>
1 - 100 of 204 matches
Mail list logo
