On 27/10/16 19:41 +0100, Jonathan Wakely wrote:
On 27/10/16 20:33 +0200, Uros Bizjak wrote:
Attached patch improves sendfile syscall compatibility with (older)
Solaris 12, where non-null third argument is required. It also paves
the way for compatibility with Solaris 10/11, where otherwise
addit
On Thu, Oct 27, 2016 at 8:43 PM, Jonathan Wakely wrote:
>> Is there a good reason to call it "always_zero_offset" rather than
>> something that fits on one line, like "offset"?
>>
>> OK for trunk anyway, thanks.
>
>
>
> (I actually already sent a similar patch, see
> https://gcc.gnu.org/ml/gcc-pa
Hi!
In simd_clone_adjust we weren't taking into account the possibility of no
edges to the exit block. If inbranch, we still want to create the loop over
the simd lanes, though of course don't have any edges from the body to the
increment bb - e.g. if one calls the function with all zeros mask, i
Hi!
This PR has been reported as something related to OpenMP, but in the end
I think it is unrelated, the bug I see is in the select type parsing.
The problem is that if select type is the very first stmt in the TU,
we parse it and before actually accepting that ST_SELECT_TYPE, we perform
various
Hi!
REE breaks the following testcase on KNL, where we only accept
QImode/HImode in k0-k7 registers (SImode/DImode is only supported
with -mavx512bw). We have
(set (reg:HI k0) (mem:HI ...))
...
(set (reg:DI rax) (zero_extend:DI (reg:HI k0)))
and we optimize it into:
(set (reg:DI rax) (zero_extend
Hi!
The following testcase ICEs on x86_64-linux with -O1, the problem is
that we expand assignment from COMPONENT_REF of MEM_REF into a V4SImode
SSA_NAME. The MEM_REF has non-addressable DCmode var inside of it, and
type of a struct containing a single V4SImode element.
The bug seems to be that
On Thu, Oct 27, 2016 at 09:07:49PM +0200, Jakub Jelinek wrote:
>
> This PR has been reported as something related to OpenMP, but in the end
> I think it is unrelated, the bug I see is in the select type parsing.
>
> The problem is that if select type is the very first stmt in the TU,
> we parse i
On Thu, Oct 27, 2016 at 12:45:25PM -0700, Steve Kargl wrote:
> Thanks for the detailed analysis. The patch looks ok to me.
> I would prefer functional and cosmetic changes to be committed
> separately, but in this case the cosmetic changes are small.
>
> > + {
> > + std::swap (ns, gfc_curr
On Wed, Oct 26, 2016 at 06:51:54PM -0400, Michael Meissner wrote:
> (zero_extendsi2): Reorder pattern, so RLDICL comes before
> the FPR and VSX loads, but before MTVSRWZ. Remove ??, ! from the
> constraints. Add MFVSRWZ and XXEXTRACTUW instructions to support
> small integ
On Thu, 27 Oct 2016, Bernd Edlinger wrote:
> Hi,
>
> by code reading I became aware that libgcc can call count_leading_zeros
> in certain cases which can give undefined results. This happens on
> signed int128 -> float or double conversions, when the int128 is in the range
> INT64_MAX+1 to UINT6
This patch is designed to improve code generation for "and" instructions with
certain immediate operands.
For the following test case:
int f2(int x)
{
x &= 0x0ff8;
x &= 0xff001fff;
return x;
}
the trunk aarch64 compiler generates:
mov w1, 8184
movkw1, 0xf00, lsl 16
and
The posix_memalign declaration in gcc/i386/config/pmm_malloc.h is
decorated with 'throw ()', which occasionally causes declaration
conflict errors (some header files, not part of GCC, that declare
posix_memalign, do not have the throw decoration). An example of this
can be seen at https://github.c
On Thu, Oct 27, 2016 at 02:55:51PM -0500, Segher Boessenkool wrote:
> On Wed, Oct 26, 2016 at 06:51:54PM -0400, Michael Meissner wrote:
> > (zero_extendsi2): Reorder pattern, so RLDICL comes before
> > the FPR and VSX loads, but before MTVSRWZ. Remove ??, ! from the
> > constraints. A
The attached patch enhances the compile-time detection of buffer
overflow in functions like __builtin___memcpy_chk to consider
non-constant lengths known to be in a certain range and warn
when the lower bound of the range doesn't fit in the destination
object.
The patch does the same thing for th
Hi,
{
tree val = ipa_get_jf_constant (jfunc);
if (TREE_CODE (val) == INTEGER_CST)
{
+ value_range vr;
if (TREE_OVERFLOW_P (val))
val = drop_tree_overflow (val);
- jfunc->vr_known = true;
- jfunc->m_vr.type = VR_RANGE;
-
Hi,
On 28/10/16 01:58, Jan Hubicka wrote:
gcc/testsuite/ChangeLog:
2016-10-25 Kugan Vivekanandarajah
* gcc.dg/ipa/vrp7.c: New test.
gcc/ChangeLog:
2016-10-25 Kugan Vivekanandarajah
* ipa-cp.c (ipa_get_jf_pass_through_result): Skip unary expressions.
(propagate
Hi,
This patch uses replaces TREE_CODE(x) == VAR_DECL by VAR_P(x) in arm.c.
Bootstrap+tested on arm-linux-gnueabihf.
OK to commit ?
Thanks,
Prathamesh
2016-10-28 Prathamesh Kulkarni
* config/arm/arm.c (arm_const_not_ok_for_debug_p): Use VAR_P.
diff --git a/gcc/config/arm/arm.c b/gcc/c
101 - 117 of 117 matches
Mail list logo
