Hi,
we have agreed on this list recently that the versionable flag should
be moved from the inline summary to the local part of struct
cgraph_node because it has nothing to do with inlining and is computed
by ipa-prop and used only by ipa-cp. This patch does exactly that.
Bootstrapped and tested
Hi,
the patch below improves the comparisons of BINFOs in IPA-CP. The
problem is that we can read different BINFOs for the same type (or a
base type component) from the LTO summaries because BINFOs coming from
different compilation units are not unified. Because we now have the
BINFO_VTABLE avai
On Sat, Sep 3, 2011 at 11:11 PM, Uros Bizjak wrote:
> On Sat, Sep 3, 2011 at 5:42 PM, Michael Matz wrote:
>
>>> > I've decided to not use four new bits from target_flags, and instead
>>> > created a new mask (recip_mask). Four bits would have fit in target
>>> > bits right now, but in the futur
> So what's your opinion on the "bug" that triggered the patch in question?\
> Namely extract_muldiv_1 folding
>
> (((10240 - (sizetype) first) + 1) * 8) /[cl] 8
>
> to
>
> ((sizetype) first * 0x0fff8 + 81928) /[cl] 8
>
> to
>
> ((sizetype) first * 2305843009213693951 + 10241)
I t
> And as you are here now, can you shed some light on the weird
> decision to make sizetype TYPE_UNSIGNED but sign-extended
> at the same time? ;)
Basically, to make division work properly given that the type is unsigned.
> So what's your opinion on the "bug" that triggered the patch in question?
> Namely extract_muldiv_1 folding
>
> (((10240 - (sizetype) first) + 1) * 8) /[cl] 8
>
> to
>
> ((sizetype) first * 0x0fff8 + 81928) /[cl] 8
>
> to
>
> ((sizetype) first * 2305843009213693951 + 10241)
>
>
On Sat, Sep 3, 2011 at 5:42 PM, Michael Matz wrote:
>> > I've decided to not use four new bits from target_flags, and instead
>> > created a new mask (recip_mask). Four bits would have fit in target
>> > bits right now, but in the future we might want to add more
>> > specialization, like modes
Hello,
this patch is obviously related to PR c++/2316 ("g++ fails to overload on
language linkage") but seems fairly independent. Currently, the demangler
recognizes 'Y' for extern "C" functions and ignores it. The patch makes it
print extern "C" after the function type:
_Z1aIFYviEEvPT_
void
On Sat, Sep 3, 2011 at 10:37 PM, Richard Guenther
wrote:
> On Sat, Sep 3, 2011 at 9:47 PM, Richard Kenner
> wrote:
>> Let me jump in on this a little bit, since much of the code in this area
>> was originally written by me.
>>
>>> Are all sizetype (sub-)expressions always of value in that range?
On Sat, Sep 3, 2011 at 9:47 PM, Richard Kenner
wrote:
> Let me jump in on this a little bit, since much of the code in this area
> was originally written by me.
>
>> Are all sizetype (sub-)expressions always of value in that range?
>> What do we do about the fact that sizetype is unsigned, so -x a
Let me jump in on this a little bit, since much of the code in this area
was originally written by me.
> Are all sizetype (sub-)expressions always of value in that range?
> What do we do about the fact that sizetype is unsigned, so -x always
> overflows for x != 0? Thus, do we need to disable all
> tree-ssa-forwprop.c I presume? That one misses calling fold_stmt when
> it propagates things.
Yep.
>
> > :
> > iftmp.1_9 = 0;
> >
> > :
> > # iftmp.1_2 = PHI <1(4), 0(5)>
> > D.2739_10 = __builtin_constant_p (iftmp.1_2);
>
> But I wonder how it manages to do something to the above ... (are
On Sat, Sep 3, 2011 at 6:41 PM, Jan Hubicka wrote:
>> Jan Hubicka wrote:
>>
>> > (edge_execution_predicate): Rewrite as...
>> > (set_cond_stmt_execution_predicate): ... this function; handle
>> > __builtin_constant_p.
>>
>> This causes ICEs when building recent Linux kernels with the
>
On Sat, Sep 3, 2011 at 5:31 PM, Michael Matz wrote:
> Hi,
>
> On Sat, 3 Sep 2011, Richard Guenther wrote:
>
>> >> As I feared the call to get_var_ann in set_is_used right now really
>> >> is still needed, privatizing it hence isn't that straight forward.
>> >
>> > After pondering I concluded that
On Sat, Sep 3, 2011 at 5:08 PM, Eric Botcazou wrote:
>> Well, for real-world code I believe that. But see all the recent testcases
>> for corner-cases of our signed-overflow stuff, they all require
>> hand-crafted testcases involving INT_MIN, no inlining and even -ftrapv.
>> What I meant to say i
Hi,
.init_array is enabled only if linker can take input .ctors sections with
input .init_array sections to generate the single output .init_array
section in the right order so that GCC is compatible with existing .o
files. Not all linkers support it even if they support .init_array
section. Sin
> Jan Hubicka wrote:
>
> > (edge_execution_predicate): Rewrite as...
> > (set_cond_stmt_execution_predicate): ... this function; handle
> > __builtin_constant_p.
>
> This causes ICEs when building recent Linux kernels with the
> CONFIG_TRACE_BRANCH_PROFLING option. This reduced test
This change defines a "return" insn pattern on pa to improve return
optimization when no epilogue is needed. Tested on hppa2.0w-hp-hpux11.11
and hppa-unknown-linux-gnu. Committed to trunk.
Defining a "return" pattern fixes PR middle-end/50232 on PA.
Dave
--
J. David Anglin
Hi,
On Sat, 3 Sep 2011, Uros Bizjak wrote:
> > I've decided to not use four new bits from target_flags, and instead
> > created a new mask (recip_mask). Four bits would have fit in target
> > bits right now, but in the future we might want to add more
> > specialization, like modes for which
Hi,
On Sat, 3 Sep 2011, Richard Guenther wrote:
> >> As I feared the call to get_var_ann in set_is_used right now really
> >> is still needed, privatizing it hence isn't that straight forward.
> >
> > After pondering I concluded that it's not necessary to call
> > set_is_used for variables with
> Well, for real-world code I believe that. But see all the recent testcases
> for corner-cases of our signed-overflow stuff, they all require
> hand-crafted testcases involving INT_MIN, no inlining and even -ftrapv.
> What I meant to say is, given Ada can construct arbitrary layouted types it
>
> Then your original patch is ok with removing the size == 0 check I added.
Bootstrapped/regtested on x86_64-suse-linux and applied, thanks.
--
Eric Botcazou
... Sent again, with correct Cc and subject line ...
Hello!
> Here is a patch which adds few more splits for AGU stalls avoidance on
> Atom. It also fixes cost model and detects AGU stalls more
> efficiently.
>
> Bootstrapped and checked on x86_64-linux.
>
> 2011-09-02 Enkovich Ilya
>
>
Hello!
> Here is a patch which adds few more splits for AGU stalls avoidance on
> Atom. It also fixes cost model and detects AGU stalls more
> efficiently.
>
> Bootstrapped and checked on x86_64-linux.
>
> 2011-09-02 Enkovich Ilya
>
> * config/i386/i386-protos.h (ix86_lea_outperforms): Ne
Recently, I came across PR 44646, where I wrote a bit more than a year
ago: "I have an embryonic patch for it". Well, I found it again, did
some polishing - and here it is.
This patch implements the parsing/diagnostic for "DO[,] CONCURRENT
for-all-header", e.g.
do concurrent (i = 1:5)
A
On Wed, Aug 31, 2011 at 6:16 PM, Michael Matz wrote:
> I'd like to have tighter control over the individual situations that
> -mrecip handles, and I think the user might appreciate this too. Hence
> I've introduced four new target options -mrecip-div, -mrecip-sqrt,
> -mrecip-vec-div and -mrecip-
On Sat, Sep 3, 2011 at 11:24 AM, Eric Botcazou wrote:
>> Well, even when sign-extended there is a constant you can't negate
>> without overflow. I would start digging for a testcase with
>> such case - but as said, testcases involving TYPE_IS_SIZETYPE are
>> very hard to generate for me.
>
> We r
On Sat, Sep 3, 2011 at 11:09 AM, Eric Botcazou wrote:
>> I'm not sure we want to create a the replacement decl with DECL_SIZE zero
>> though, so I suppose instead of making sure align is BITS_PER_UNIT
>> shouldn't we make sure that size is at least 1? I fear we might run into
>> other odd problem
> Well, even when sign-extended there is a constant you can't negate
> without overflow. I would start digging for a testcase with
> such case - but as said, testcases involving TYPE_IS_SIZETYPE are
> very hard to generate for me.
We run thousands of Ada tests every night on many platforms and ne
> I'm not sure we want to create a the replacement decl with DECL_SIZE zero
> though, so I suppose instead of making sure align is BITS_PER_UNIT
> shouldn't we make sure that size is at least 1? I fear we might run into
> other odd problems with zero-size autos.
The cfgexpand.c code already does
On Sat, Sep 3, 2011 at 3:51 AM, Michael Matz wrote:
> Hi,
>
> On Fri, 2 Sep 2011, Michael Matz wrote:
>
>> > > Ok. Time to make get_var_ann private?
>>
>> As I feared the call to get_var_ann in set_is_used right now really is
>> still needed, privatizing it hence isn't that straight forward.
>
>
On Fri, Sep 2, 2011 at 8:40 PM, Michael Matz wrote:
> Hi,
>
> On Fri, 2 Sep 2011, Michael Matz wrote:
>
>> Hi,
>>
>> On Fri, 2 Sep 2011, Richard Guenther wrote:
>>
>> > > Currently regstrapping on x86_64-linux (without Ada). Okay for trunk?
>> >
>> > Ok. Time to make get_var_ann private?
>>
>> Y
On Fri, Sep 2, 2011 at 7:42 PM, Ulrich Weigand wrote:
> Jan Hubicka wrote:
>
>> (edge_execution_predicate): Rewrite as...
>> (set_cond_stmt_execution_predicate): ... this function; handle
>> __builtin_constant_p.
>
> This causes ICEs when building recent Linux kernels with the
>
On Fri, Sep 2, 2011 at 6:55 PM, Eric Botcazou wrote:
>> And this is the real fix. Richard, do you want me to apply (part of it)?
A varaint of the patch is ok,
> In fact I'd even propose to revert the fold_builtin_alloca_for_var part of
> your patch and apply mine entirely, as eliminating alloca
34 matches
Mail list logo
