> Hi!
>
> On Thu, Jun 01, 2017 at 01:35:56PM +0200, Jan Hubicka wrote:
> > + /* FIXME: shrink wrapping violates this sanity check. */
> > + gcc_checking_assert ((num >= 0
> > + && (num <= REG_BR_PROB_BASE
> > + || den <= REG_BR_PROB_BASE)
This patch by Than McIntosh fixes the types that the Go frontend
passes to the middle-end for interface method tables. The Go frontend
was using Go function types for values that are actually C function
pointers. The value is a pointer either way, but clearly the type
should be correct. Bootstra
Hi Nathan,
> On Jun 5, 2017, at 12:58 , Nathan Sidwell wrote:
>
> On 06/02/2017 11:58 AM, Olivier Hainque wrote:
>> Hello,
>
>> 2017-06-02 Olivier Hainque
>> ada/
>> * vx_crtbegin_auto.c: New file.
>> * vx_crtbegin.c: New file.
>> * vx_crtbegin.inc: New file.
>> * vx
On Tue, 6 Jun 2017, Kyrill Tkachov wrote:
Another vec_merge simplification that's missing is transforming:
(vec_merge (vec_duplicate x) (vec_concat (y) (z)) (const_int N))
into
(vec_concat x z) if N == 1 (0b01) or
(vec_concat y x) if N == 2 (0b10)
Do we have a canonicalization somewhere that g
On 06/05/2017 07:53 PM, Martin Sebor wrote:
On 06/05/2017 01:13 PM, Martin Sebor wrote:
On 06/05/2017 10:07 AM, Martin Sebor wrote:
Maybe I should use a different approach and instead of trying
to see if a function is deleted use trivially_xible to see if
it's usable. That will mean changing t
GCC Maintainers:
The support for the vec_mulo and vec_mule that was recently submitted
has a couple of bugs. Specifically, they were implemented with
int/unsigned int args and return int/unsigned int. The return types
should have been long long/unsigned long long. Additionally it was
noted that
On Tue, Jun 6, 2017 at 10:34 AM, Martin Sebor wrote:
> On 06/06/2017 10:59 AM, H.J. Lu wrote:
>>
>> On Tue, Jun 6, 2017 at 9:10 AM, Martin Sebor wrote:
>>>
>>> On 06/06/2017 10:07 AM, Martin Sebor wrote:
On 06/05/2017 11:45 AM, H.J. Lu wrote:
>
>
> On Mon, Jun 5, 2017 a
Spelling correction blew up for me on the modules branch because
suggest_alternatives_for was using the raw namespace accessor. That led
me to reconsider the behaviour. This patch implements qualified name
lookup (which it used to do), but ignoring using directives (which is
new). That way
Tony,
This patch generally looks good to me -- it clearly is an incremental
improvement. One of the libiberty maintainers, such as Ian, needs to
approve the patch.
https://gcc.gnu.org/ml/gcc-patches/2017-05/msg01181.html
+ if (strcmp (name, ".text") == 0)
+textptr = scnptr;
The ab
Tony,
This patch seems like a reasonable start for the minimal libbacktrace
framework needed by GCC Go. The libbacktrace maintainer, Ian, needs
to approve the patch.
https://gcc.gnu.org/ml/gcc-patches/2017-05/msg01186.html
I'll defer to Ian on the style of the fileline.c changes.
Thanks, David
David Edelsohn writes:
> This patch generally looks good to me -- it clearly is an incremental
> improvement. One of the libiberty maintainers, such as Ian, needs to
> approve the patch.
As AIX maintainer, I think you have the authority to approve patches
like this, which only affect your OS.
On 06/06/2017 04:57 PM, H.J. Lu wrote:
On Tue, Jun 6, 2017 at 10:34 AM, Martin Sebor wrote:
On 06/06/2017 10:59 AM, H.J. Lu wrote:
On Tue, Jun 6, 2017 at 9:10 AM, Martin Sebor wrote:
On 06/06/2017 10:07 AM, Martin Sebor wrote:
On 06/05/2017 11:45 AM, H.J. Lu wrote:
On Mon, Jun 5, 201
On May 18, 2017, Alexandre Oliva wrote:
> When an insn used by combine has multiple SETs, only the non-REG_UNUSED
> set is used: others will end up dropped on the floor. We have to take
> note of the dropped REG_UNUSED SETs, clearing their cached values, so
> that, even if the REGs remain used (
On 06/06/2017 12:55 PM, Richard Sandiford wrote:
> This patch tries to clarify some of the restrictions on define_insn
> conditions, and also on the use of "#".
>
> OK to install?
>
> Richard
>
>
> 2017-06-06 Richard Sandiford
>
> gcc/
> * doc/md.texi: Clarify the restrictions on a de
On 06/06/2017 04:31 AM, Marek Polacek wrote:
> Another case of extract_muldiv running off the rails. Here it did a wrong
> distribution; turning
>
> ((A * x) - (B * x)) * -6
>
> into
>
> (A' * x) - (B' * x)
>
> incurred an overflow in the subtraction. The fix is essentially the same
> as
On 05/23/2017 05:28 AM, Nathan Sidwell wrote:
> On 05/23/2017 06:48 AM, Aldy Hernandez wrote:
>
>> The class can live outside of his work, as can be demonstrated by the
>> attached patch. With it, I was able to rewrite the post-VRP range
>> information to use this class and get rid of VR_ANTI_RAN
On 05/23/2017 08:34 AM, Jakub Jelinek wrote:
> On Tue, May 23, 2017 at 10:29:58AM -0400, David Malcolm wrote:
>>> Do we really want methods starting with capital letters?
>>> I understand why you can't use union, but I don't think we use
>>> CamelCase
>>> anywhere.
>>
>> FWIW in the JIT, I have a c
Note I'd be _much_ more sympathetic to simply canonicalizing all of
bzero and bcopy
to memset / memmove and be done with all the above complexity.
Attached is an updated patch along these lines. Please let me
know if it matches your expectations.
FWIW, although I don't feel too strongly about
On Tue, Jun 06, 2017 at 09:38:04AM +0200, Richard Biener wrote:
> On Sat, 3 Jun 2017, Richard Biener wrote:
>
> > On June 3, 2017 1:38:14 AM GMT+02:00, Michael Meissner
> > wrote:
> > >On Fri, Jun 02, 2017 at 03:22:27PM +0200, Richard Biener wrote:
> > >>
> > >> This implements vectorization of
Hi
the following patch makes cfgbuild to preserve profile when loops are
introduced at RTL level (not very well, but at least do not throw it all
away) and also avoids re-computing probabilities when there are no
changes to CFG.
Bootstrapped/regtested x86_64-linux. Comitted.
Honza
Index: cfgbuil
> Patch 6 breaks no-vfa-vect-57.c on powerpc.
Which CPU model (power6/7/8?) and which compile options (-maltivec/
-mpower8-vector?) have been used for running and compiling the test? As
discussed in
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80925
this has an influence on the cost function and
Hi,
I would like to ping this patch:
https://gcc.gnu.org/ml/gcc-patches/2017-05/msg01374.html
Thanks,
-Maxim
On Sat, 3 Jun 2017, Richard Biener wrote:
> On June 3, 2017 1:38:14 AM GMT+02:00, Michael Meissner
> wrote:
> >On Fri, Jun 02, 2017 at 03:22:27PM +0200, Richard Biener wrote:
> >>
> >> This implements vectorization of SLP inductions (in the not outer
> >loop
> >> vectorization case for now).
>
> On Thu, Jun 1, 2017 at 4:35 AM, Jan Hubicka wrote:
> > Index: profile.c
> > ===
> > --- profile.c (revision 248684)
> > +++ profile.c (working copy)
> > @@ -67,6 +67,10 @@ along with GCC; see the file COPYING3.
> >
> > #include
Hi,
this patch makes the count initialization loop to also visit EXIT_BLOCK.
Profiledbootstrapped/regtested x86_64-linux, comitted.
Honza
* profile.c (compute_branch_probabilities): Also initialize
EXIT_BLOCK profile.
Index: profile.c
=
Hi,
this patch makes return block profile to be initialized after splitting.
profiledbootstrapped/regtested x86_64-linux, comitted.
Honza
* ipa-split.c (split_function): Initialize return bb profile.
Index: ipa-split.c
===
--
Hi,
gimple_find_sub_bbs misses any profile updating code. While this is legacy
function it seems it is in tree for quite a while. This patch adds similar
logic to RTL's variant.
Profiledbootstraped/regtested x86_64-linux, comitted.
Honza
* predict.c (tree_estimate_probability_bb): Add L
Hi,
shrink-wrap misses profile update in handle_simple_exit.
Profiledbootstrapped/regtested x86_64-linux, comitted.
Honza
* shrink-wrap.c (handle_simple_exit): Update profile.
(try_shrink_wrapping): Upate profile.
Index: shrink-wrap.c
=
Hi all,
I'm trying to improve some of the RTL-level handling of vector lane operations
on aarch64 and that
involves dealing with a lot of vec_merge operations. One simplification that I
noticed missing
from simplify-rtx are combinations of vec_merge with vec_duplicate.
In this particular case:
101 - 129 of 129 matches
Mail list logo
