Hi!
On Thu, 13 Nov 2014 17:44:40 -0800, Cesar Philippidis
wrote:
> On 11/13/2014 08:43 AM, Jakub Jelinek wrote:
> > Can you please avoid the TODOs in the source? If it is not the right
> > thing, either do something better, or file a PR to schedule such work for
> > the future.
Should we use t
On Thu, Nov 13, 2014 at 05:54:46PM -0700, Jeff Law wrote:
> On 11/12/14 15:33, Jakub Jelinek wrote:
> >This patch makes the -fdiagnostics-color= default configurable, and
> >changes the default (if no configure option is specified for it)
> >to --with-diagnostics-color=auto. The previous behavior
On 11/13/2014 10:52 PM, Jakub Jelinek wrote:
> On Thu, Nov 13, 2014 at 05:44:40PM -0800, Cesar Philippidis wrote:
>> Thanks. I couldn't figure out how to assign the bugs in the PR. Maybe my
>> account doesn't have permission to do so. Regardless, I'll work on them.
>
> Use your @gcc.gnu.org accoun
On Tue, Nov 11, 2014 at 11:46:55AM +0100, Richard Biener wrote:
> > BTW: There are lots of places where std::swap can be used, a nice
> > search-and-replace task for someone to start with gcc development. ;)
>
> Agreed ;) Note that we have to be careful to avoid pulling all of libstdc++
> into al
On 11/14/2014 10:02 AM, Jakub Jelinek wrote:
On Fri, Nov 14, 2014 at 09:46:14AM +0300, Yury Gribov wrote:
Hi all,
This patch fixes https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63802 by only
limiting minimal type alignment with BIGGEST_ALIGNMENT for types with no
__attribute__((aligned)).
Boots
On 11/06/14 05:32, Ilya Enkovich wrote:
Hi,
This patch adds instrumentation optimization to use string function with no
checks and/or bounds copy.
Thanks,
Ilya
--
gcc/
2014-11-06 Ilya Enkovich
* tree-chkp-opt.c (chkp_get_nobnd_fndecl): New.
(chkp_get_nochk_fndecl): New.
On Fri, Nov 14, 2014 at 09:46:14AM +0300, Yury Gribov wrote:
> Hi all,
>
> This patch fixes https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63802 by only
> limiting minimal type alignment with BIGGEST_ALIGNMENT for types with no
> __attribute__((aligned)).
>
> Bootstrapped and regtested on x64. Ok
On 11/06/14 05:24, Ilya Enkovich wrote:
Hi,
This patch adds support of instrumented builtin calls in expand. Calls are
mostly expanded as calls. But some of them reuse existing string function
calls expand functions (memcpy expand was slightly refactored for that).
This is the last enabling
On Thu, Nov 13, 2014 at 05:44:40PM -0800, Cesar Philippidis wrote:
> Thanks. I couldn't figure out how to assign the bugs in the PR. Maybe my
> account doesn't have permission to do so. Regardless, I'll work on them.
Use your @gcc.gnu.org account instead, then you have far more permissions
in bugz
On 11/06/14 05:10, Ilya Enkovich wrote:
Hi,
This patch enables instrumentation of chosen builtin calls.
Thanks,
Ilya
--
2014-11-06 Ilya Enkovich
* ipa-chkp.c (chkp_versioning): Clone builtin functions.
(chkp_instrument_normal_builtin): New.
(chkp_add_bounds_to_call_s
Hi all,
This patch fixes https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63802 by
only limiting minimal type alignment with BIGGEST_ALIGNMENT for types
with no __attribute__((aligned)).
Bootstrapped and regtested on x64. Ok for trunk?
-Y
>From 7e5d09453dcff22f591162e1b5c5a115b17b0014 Mon Sep 1
On 11/06/14 04:48, Ilya Enkovich wrote:
--
2014-11-06 Ilya Enkovich
* tree-core.h (built_in_class): Add builtin codes to be used
by Pointer Bounds Checker for instrumented builtin functions.
* tree-streamer-in.c: Include ipa-chkp.h.
(streamer_get_builtin_tree):
On 11/08/14 17:35, Andrew Waterman wrote:
Thank you for approving the patch. I neglected to mention that I do
not have write access. Would you or someone else be so kind as to
commit this?
Done. Thanks for your patience.
Jeff
On 10/20/14 22:06, Maxim Kuvyrkov wrote:
Hi,
Ramana, this change requires benchmarking, which I can't easily do
at
the moment. I would appreciate any benchmarking results that you can
share. In particular, the value of PARAM_SCHED_AUTOPREF_QUEUE_DEPTH
needs to be tuned/confirmed for Cortex-A15.
For i[34567]86-*-linux* and x86_64-*-linux* targets, config.gcc assumes
modern glibc and set default_gnu_indirect_function to yes. That is wron
for i[34567]86-*-linux-android* and and x86_64-*-linux-android* targets.
This patch fixes it. Tested on Linux/x86-64, Linux/ia32 and cross-teste
for i686
On Thu, Nov 13, 2014 at 8:59 PM, Jeff Law wrote:
> On 11/13/14 21:46, H.J. Lu wrote:
>
>>
>> Hi Jeff,
>>
>> I believe the patch is correct. Not all glibcs support IFUNC. It doesn't
>> mean
>> we should disable IFUNC for all glibcs. By default, we do a configure
>> time
>> check for IFUNC. We DO
On 11/13/14 07:09, Kyrill Tkachov wrote:
I've updated the documentation for the hook.
The testcase I was looking at involves fusing the AArch64 adrp+add
instructions and depends on the backend implementation of the matching
code, under review currently at
https://gcc.gnu.org/ml/gcc-patches/2014
On 11/13/14 21:46, H.J. Lu wrote:
Hi Jeff,
I believe the patch is correct. Not all glibcs support IFUNC. It doesn't mean
we should disable IFUNC for all glibcs. By default, we do a configure time
check for IFUNC. We DO want to use configure time check for IFUNC for
Android NDK.
I
I'm sorry,
On 11/13/14 13:06, mliska wrote:
gcc/ChangeLog:
2014-11-13 Martin Liska
* fibonacci_heap.h: New file.
* ipa-inline.c (update_edge_key): New heap API is used.
(update_caller_keys): Likewise.
(update_callee_keys): Likewise.
(lookup_recursive_calls): Like
On Thu, Nov 13, 2014 at 8:46 PM, H.J. Lu wrote:
> On Thu, Nov 13, 2014 at 9:36 AM, Jeff Law wrote:
>> On 11/12/14 03:02, Alexander Ivchenko wrote:
>>>
>>> Hi,
>>>
>>> Bionic - Android libc - supports indirect functions right now, but
>>> they are disabled in gcc;
>>>
>>> We cannot do the configur
On Thu, Nov 13, 2014 at 9:36 AM, Jeff Law wrote:
> On 11/12/14 03:02, Alexander Ivchenko wrote:
>>
>> Hi,
>>
>> Bionic - Android libc - supports indirect functions right now, but
>> they are disabled in gcc;
>>
>> We cannot do the configure-time check for that, because there is only
>> one version
On Fri, Nov 14, 2014 at 12:40:23AM +0100, Jan Hubicka wrote:
> Hello,
> in general I like this addition - I was not aware that sreal has no support
> for negative values.
> This would be serious maintainance burden if sreals was used for profile
> updating - it is very
> easy to get negative temp
-bool
-linux_has_ifunc_p (void)
-{
- return OPTION_BIONIC ? false : HAVE_GNU_INDIRECT_FUNCTION;
-}
That is what prevent us from doing so. We need to remove OPTION_BIONIC
from that hook and that is what the patch does.
2014-11-14 8:32 GMT+04:00 H.J. Lu :
> On Thu, Nov 13, 2014 at 8:27 PM, Andrew
On 11/13/14 13:10, mliska wrote:
gcc/ChangeLog:
2014-11-13 Martin Liska
* ipa-inline.c (edge_badness): long is replaced by sreal
as fibonacci_heap template type.
(update_edge_key): Likewise.
(inline_small_functions): Likewise.
* sreal.h (inline sreal o
On 11/13/14 03:17, mliska wrote:
gcc/ChangeLog:
2014-11-13 Martin Liska
* var-tracking.c (vt_find_locations): New fibonacci_node is used.
Remove old fibheap.h if it's not used anymore.
OK if/when prerequisites are approved.
jeff
On Thu, Nov 13, 2014 at 8:27 PM, Andrew Hsieh wrote:
> What about overloading the existing option -mbionic ? -mbionic=21 and
> above enable ifunc (so NDK can help enforce it)
>
> On Fri, Nov 14, 2014 at 11:51 AM, enh wrote:
>> On Thu, Nov 13, 2014 at 5:12 PM, Jeff Law wrote:
>>> On 11/13/14 10:4
On 11/13/14 03:08, mliska wrote:
gcc/ChangeLog:
2014-11-13 Martin Liska
* bt-load.c (add_btr_def): New fibonacci_heap is used.
(migrate_btr_defs): Likewise.
If you've removed last uses of the old fibheap code, then remove the
include of fibheap.h.
OK once prerequisites are
On 11/12/14 18:45, mliska wrote:
gcc/ChangeLog:
2014-11-13 Martin Liska
* bb-reorder.c (mark_bb_visited): New fibonacci_heap is used.
(find_traces): Likewise.
(find_traces_1_round): Likewise.
If this removed the last uses of the old fibheap code, then the old
fibheap
On Thu, Nov 13, 2014 at 3:47 PM, Jonathan Wakely wrote:
> Tim, I think Daniel still doesn't have commit access, so could you
> please commit this? (And CC the patch to gcc-patches).
Committed. Removed trailing spaces.
The svn diff is attached.
Thank you!
--
Regards,
Tim Shen
Index: include/b
On 11/12/14 18:54, mliska wrote:
2014-11-13 Martin Liska
* tracer.c (tail_duplicate): New fibonacci_heap class is used.
OK if/when prerequisites are approved.
jeff
What about overloading the existing option -mbionic ? -mbionic=21 and
above enable ifunc (so NDK can help enforce it)
On Fri, Nov 14, 2014 at 11:51 AM, enh wrote:
> On Thu, Nov 13, 2014 at 5:12 PM, Jeff Law wrote:
>> On 11/13/14 10:46, enh wrote:
>>>
>>> This feels like a bad idea to me simp
On 11/13/14 05:35, mliska wrote:
gcc/ChangeLog:
2014-11-13 Martin Liska
* predict.c (propagate_freq): More elegant sreal API is used.
(estimate_bb_frequencies): New static constants defined by sreal
replace precomputed ones.
* sreal.c (sreal::normalize): New f
On Thu, Nov 13, 2014 at 5:12 PM, Jeff Law wrote:
> On 11/13/14 10:46, enh wrote:
>>
>> This feels like a bad idea to me simply because a new compiler with an
>> old runtime will generate code that fails, right?
>>
>>
>> yes, but that's already true of PIE or gnu-style hash or...
>
> That d
On Fri, Nov 7, 2014 at 7:13 AM, Jeff Law wrote:
> On 11/05/14 02:30, Bin.Cheng wrote:
> [ ... ]
>>>
>>>
>>> rfs_result's signature has changed, I think you need to pass in tmp &
>>> tmp2.
>>> You'll need to make that trivial update for all the callers.
>>>
>>>
>>> Can you make those changes and re
On 11/13/2014 04:31 PM, Paolo Carlini wrote:
I think this should be replaced with fold_ if
(processing_template_decl) in build_enumerator.
Ok. The value can be NULL_TREE, thus in a straightforward change (per
the below) I have to check for that, otherwise we crash in
maybe_constant_value. Either
Hmm, I don't think so.
The pattern *aarch64_simd_ld1r is used for pure matching, and GCC will
never generate a pattern as follows:
[(set (match_operand:DI 0 "register_operand" "=w")
(vec_duplicate:DI
(match_operand:DI 1 "aarch64_simd_struct_operand" "Utv")))]
>
> Can a ne
Can a new case be rewritten then?
- Joey
On Fri, Nov 14, 2014 at 9:32 AM, Yangfei (Felix) wrote:
> No, we noticed this issue when improving the vld1(q?)_dup intrinsics. Thanks.
>
>
>> Is there a case or PR to demonstrate the issue? If yes, better to include it
>> as a test
>> case.
>>
>> Thank
On Thu, Nov 13, 2014 at 05:32:08PM -0700, Jeff Law wrote:
> On 11/12/14 22:55, tsaund...@mozilla.com wrote:
> >From: Trevor Saunders
> >
> >Hi,
> >
> >$subject.
> >
> >bootstrapped + regtested x86_64-unknown-linux-gnu, ok?
> >
> >Trev
> >
> >
> >ada/
> >
> > * gcc-interface/decl.c, gcc-interfa
On 2014-10-21 12:06 AM, Maxim Kuvyrkov wrote:
Hi,
This patch adds auto-prefetcher modeling to GCC scheduler. The auto-prefetcher
model is currently enabled only for ARM Cortex-A15, since this is the only CPU
that I know of to have the hardware auto-prefetcher unit.
The documentation on the a
On Thu, Nov 13, 2014 at 05:49:47PM -0700, Jeff Law wrote:
> On 11/12/14 22:55, tsaund...@mozilla.com wrote:
> >From: Trevor Saunders
> >
> >Hi,
> >
> >The only user of this is splay_tree. We only have a couple splay trees in
> >ggc
> >memory, and it wasn't clear to me any of them were tree based
On Thu, Nov 13, 2014 at 05:52:51PM -0700, Jeff Law wrote:
> On 11/12/14 22:55, tsaund...@mozilla.com wrote:
> >From: Trevor Saunders
> >
> >Hi,
> >
> >gengtype fails to create valid user marking functions for this type, which is
> >fixed by using vec here (which seems cleaner anyway).
> >
> >boots
On 11/13/2014 08:43 AM, Jakub Jelinek wrote:
> On Thu, Nov 06, 2014 at 02:25:52PM -0800, Cesar Philippidis wrote:
>> * cpp.c (cpp_define_builtins): Conditionally define _OPENACC.
>> * dump-parse-tree.c
>> (show_omp_node): Dump also OpenACC executable statements.
>
> Put (show_omp_n
No, we noticed this issue when improving the vld1(q?)_dup intrinsics. Thanks.
> Is there a case or PR to demonstrate the issue? If yes, better to include it
> as a test
> case.
>
> Thanks,
> Joey
>
> On Thu, Nov 13, 2014 at 2:14 PM, Yangfei (Felix)
> wrote:
> > Hi,
> >
> > We find that th
Hi,
this patch removes now completely unnecessary KNOWN_TYPE jump
functions, type_preserved fields from PASS_THROUGH and ANCESTOR jump
functions and type field from ANCESTOR jump functions because the
previous patch makes them unused.
This means most of dynamic type change detection is removed.
Hi,
this patch is that main part of the series, implementing propagation
of ipa_polymorphic_call_contexts. Although I tried to keep its size
checked, it is big because switching from binfos to contexts in ipa-cp
also requires changes in inlining heuristics. The patch also does not
do any specula
On Thu, Nov 13, 2014 at 05:27:13PM -0700, Jeff Law wrote:
> On 11/12/14 22:55, tsaund...@mozilla.com wrote:
> >From: Trevor Saunders
> >
> >Hi,
> >
> > This adds a gty cache attribute that calls user code after marking and
> > before
> >sweeping allowing user code to mark more objects or clear c
Hi,
this patch re-organizes data structures in ipa-cp so that they can be
re-used to propagate not only trees but also other kinds of values, as
a preparation for propagating ipa_polymorphic_call_contexts. However,
this patch does not change any functionality, I have split it off to
ease review a
Is there a case or PR to demonstrate the issue? If yes, better to
include it as a test case.
Thanks,
Joey
On Thu, Nov 13, 2014 at 2:14 PM, Yangfei (Felix) wrote:
> Hi,
>
> We find that the VALLDI mode iterator used in *aarch64_simd_ld1r
> pattern is not appropriate.
> The reason is that it'
On 11/13/14 16:42, Jan Hubicka wrote:
gcc/ChangeLog:
2014-11-13 Martin Liska
* bb-reorder.c (find_traces_1_round): Old fibheap_t type removed.
* bt-load.c: Include of fibheap.h is removed.
* cgraphunit.c: Likewise.
* config/i386/i386.c: Likewise.
* ipa
On 11/13/14 08:28, David Malcolm wrote:
It was pointed out to me on IRC that I could instead use RAII for this,
so here's an alternative version of the patch that puts it in a class,
so that you can put:
auto_assert_no_gc no_gc_here;
into a scope to get the assertion failure if someone use
On 11/13/2014 11:09 AM, Jakub Jelinek wrote:
> On Thu, Nov 13, 2014 at 11:03:47AM -0800, Cesar Philippidis wrote:
@@ -233,6 +242,90 @@ static tree scan_omp_1_op (tree *, int *, void *);
*handled_ops_p = false; \
break;
+/* Helper function to get the reductio
On 11/13/14 10:46, enh wrote:
This feels like a bad idea to me simply because a new compiler with an
old runtime will generate code that fails, right?
yes, but that's already true of PIE or gnu-style hash or...
That doesn't make it the right thing to do. I would argue that's a bug
tha
On 11/13/14 12:37, Mike Stump wrote:
I was doing a merge, and it failed to even compile the runtime
libraries due to checking in bitmap. bitmap goes to remove set bits
from the bitmap (the second hunk in a two hunk set), and it fails to
update the current pointer. That memory is freed and then
Index: gcc/ChangeLog
===
--- gcc/ChangeLog (revision 217506)
+++ gcc/ChangeLog (working copy)
@@ -1,3 +1,9 @@
+2014-11-13 Felix Yang
+
+ * ipa-utils.h: Fix typo in comments.
+ * ipa-profile.c: Likewise.
+
On 11/13/14 14:01, Joseph Myers wrote:
If you want to accept as much code as possible and only give errors if
floating-point code would actually be generated - for example, if you want
to be able to #include with glibc (where it defines an inline
version of atof) - then you need to look at GIMP
This adds the schedule model for ThunderX. There are a few TODOs in that
not all of the SIMD is model currently. Also the idea of a simple
shift/extend is not modeled and all cases where there is a shift/extend
is considered as non simple and take up two cycles rather than correct
value of one cyc
Hi,
This set of patches improve support for the ThunderX processor from Cavium.
The first patch adds support for macro fusion that is in the ThunderX processor.
The next patch adds the scheduler which was missing in the original addition.
The last patch adds tuning field to allow the processors s
On ThunderX, I found that aligning functions/loops/jumps to an 8 byte
boundary have a slightly better performance because the hardware issue
and dispatch matches what GCC's schedule has created.
I set generic, cortex-a53 and cortex-a57 also to be 8 byte aligned
also. Someone might want to change
In ThunderX, any 1 cycle arthemantic instruction that produces the flags
register, will be fused with a branch. This patch depends on
https://gcc.gnu.org/ml/gcc-patches/2014-11/msg01508.html.
Note I know bit 1 is going is already going to be used and that is why I
proposed this being bit 2.
Build
On 11/12/14 15:33, Jakub Jelinek wrote:
Hi!
This patch makes the -fdiagnostics-color= default configurable, and
changes the default (if no configure option is specified for it)
to --with-diagnostics-color=auto. The previous behavior can be
restored with --with-diagnostics-color=auto-if-env , th
On 11/12/14 22:55, tsaund...@mozilla.com wrote:
From: Trevor Saunders
Hi,
gengtype fails to create valid user marking functions for this type, which is
fixed by using vec here (which seems cleaner anyway).
bootstrapped + regtested powerpc64-linux (gcc 110 since gcc20 died) ok?
Trev
gcc/Cha
On 11/12/14 22:55, tsaund...@mozilla.com wrote:
From: Trevor Saunders
Hi,
The only user of this is splay_tree. We only have a couple splay trees in ggc
memory, and it wasn't clear to me any of them were tree based instead of hash
based for performance reasons, so I chose to just convert them
Hi,
here is upated version with bitfields and also tested on PPC64-linux/aix.
I hacked configury to use system awk instead of gawk, so the changes are
hopefully safe.
OK?
Honza
* optc-save-gen.awk: Output cl_target_option_eq,
cl_target_option_hash, cl_target_option_stream_out,
On 11/12/14 22:55, tsaund...@mozilla.com wrote:
From: Trevor Saunders
hi,
The problem here is that hash_table used to zero element storage, but if the
empty element is not 0 then all elements appear to be in use.
bootstrapped + regtested x86_64-unknown-linux-gnu, ok?
Trev
gcc/ChangeLog:
2
On 11/12/14 22:55, tsaund...@mozilla.com wrote:
From: Trevor Saunders
Hi,
$subject.
bootstrapped + regtested x86_64-unknown-linux-gnu, ok?
Trev
ada/
* gcc-interface/decl.c, gcc-interface/utils.c: replace htab with
hash_table.
cp/
* cp-objcp-common.c: Use hash_tab
On 11/12/14 22:55, tsaund...@mozilla.com wrote:
From: Trevor Saunders
Hi,
This adds a gty cache attribute that calls user code after marking and before
sweeping allowing user code to mark more objects or clear caches as
appropriate. User code for hash_table is set up to work similarly to if
libgomp/
* libgomp.map (GOMP_4.0.1): Add GOMP_offload_unregister.
* libgomp_target.h (struct mapping_table): Replace with addr_pair.
* target.c (struct gomp_device_descr): Remove register_image_func,
get_table_func. Add load_image_func, unload_image_func.
(g
Hi,
This patch fixes offloading from dlopened libraries, part 1 is for libgomp and
part 2 is for intelmic plugin (it's not finished yet).
How it works:
When a library is loaded it calls GOMP_offload_register as usually.
At this time some devices may already be initialized, and some may be not.
Th
> Honza,
>
> On 23 June 2014 06:24, Jan Hubicka wrote:
>
> > --- lto-cgraph.c(revision 211881)
> > +++ lto-cgraph.c(working copy)
> > @@ -614,6 +614,7 @@ lto_output_varpool_node (struct lto_simp
> > /* in_other_partition. */
> > }
> >bp_pack_value (&bp, node->
We do not do sophisticated recursive call detection in einline phase.
It only happens in ipa-inline phase.
Dehao
On Thu, Nov 13, 2014 at 3:18 PM, Xinliang David Li wrote:
> On Thu, Nov 13, 2014 at 2:57 PM, Dehao Chen wrote:
>> IIRC, AutoFDO the actual iteration for AutoFDO is mostly <3. But it
> gcc/ChangeLog:
>
> 2014-11-13 Martin Liska
>
> * bb-reorder.c (find_traces_1_round): Old fibheap_t type removed.
> * bt-load.c: Include of fibheap.h is removed.
> * cgraphunit.c: Likewise.
> * config/i386/i386.c: Likewise.
> * ipa-inline.c: Likewise.
> * v
Hello,
in general I like this addition - I was not aware that sreal has no support for
negative values.
This would be serious maintainance burden if sreals was used for profile
updating - it is very
easy to get negative temporaries while dong the updates.
> gcc/ChangeLog:
>
> 2014-11-13 Martin
> gcc/ChangeLog:
>
> 2014-11-13 Martin Liska
>
> * fibonacci_heap.h: New file.
> * ipa-inline.c (update_edge_key): New heap API is used.
> (update_caller_keys): Likewise.
> (update_callee_keys): Likewise.
> (lookup_recursive_calls): Likewise.
> (recursive_in
> Hello.
>
> Following patch set introduces new template class for fibonacci heap
> that is mainly utilized by IPA inliner. Apart from that, I also enhanced
> existing sreal implementation so that it can also handle negative numbers.
>
> I was primary motivated by Honza, because current heap impl
On Thu, Nov 13, 2014 at 2:57 PM, Dehao Chen wrote:
> IIRC, AutoFDO the actual iteration for AutoFDO is mostly <3. But it
> should not harm to set max iter as 10.
>
> On Thu, Nov 13, 2014 at 2:51 PM, Xinliang David Li wrote:
>> After inline summary is recomputed, the large code growth problem will
On Tue, Nov 11, 2014 at 8:02 AM, H.J. Lu wrote:
> On Mon, Nov 10, 2014 at 11:42 AM, H.J. Lu wrote:
>> On Mon, Nov 10, 2014 at 5:44 AM, Richard Biener
>> wrote:
>>> On Mon, Nov 10, 2014 at 2:43 PM, Jakub Jelinek wrote:
On Mon, Nov 10, 2014 at 05:32:32AM -0800, H.J. Lu wrote:
> On Mon, N
On Nov 13, 2014, at 2:33 PM, Jeff Law wrote:
>> We don’t support fprintf prologues anymore, they were removed years ago.
> Did we ever get all the ports converted?
Ah… sorry, I was wrong. We merely hookized it and the tm.h interface is gone.
TARGET_ASM_FUNCTION_PROLOGUE and TARGET_ASM_FUNCTIO
IIRC, AutoFDO the actual iteration for AutoFDO is mostly <3. But it
should not harm to set max iter as 10.
On Thu, Nov 13, 2014 at 2:51 PM, Xinliang David Li wrote:
> After inline summary is recomputed, the large code growth problem will
> also be better controlled, right?
For this case, recompu
On Thu, Nov 13, 2014 at 11:23:31PM +0100, Markus Trippelsdorf wrote:
> On 2014.11.14 at 01:19 -0500, Trevor Saunders wrote:
> > On Thu, Nov 13, 2014 at 03:48:34PM +0100, Markus Trippelsdorf wrote:
> > > On 2014.11.13 at 15:11 +0100, mliska wrote:
> > >
> > > > + /* Destructor. */
> > > > + virt
After inline summary is recomputed, the large code growth problem will
also be better controlled, right?
David
On Thu, Nov 13, 2014 at 2:48 PM, Xinliang David Li wrote:
> Is there a need to have 10 iterations of early inline for autofdo?
>
> David
>
> On Thu, Nov 13, 2014 at 2:25 PM, Dehao Chen
Is there a need to have 10 iterations of early inline for autofdo?
David
On Thu, Nov 13, 2014 at 2:25 PM, Dehao Chen wrote:
> In AutoFDO, we increase einline iterations. This could lead to
> extensive code bloat if we have recursive calls like:
>
> dtor() {
> destroy(node);
> }
>
> destroy(nod
On 11/13/2014 10:47 AM, Andrew Pinski wrote:
On Thu, Nov 13, 2014 at 9:42 AM, Sandra Loosemore
wrote:
On 11/13/2014 10:27 AM, Richard Earnshaw wrote:
On 13/11/14 17:05, Ramana Radhakrishnan wrote:
On Thu, Nov 13, 2014 at 4:55 PM, Sandra Loosemore
wrote:
This patch to the AArch64 back end
2014-11-13 23:30 GMT+01:00 Tim Shen :
> On Thu, Nov 13, 2014 at 2:11 PM, Daniel Krügler
> wrote:
>> + typedef typename sub_match<_Bi_iter>::string_type string_type;
>
> Use _String_type instead of string_type? I'm not 100% sure of this.
I thought about this first, but withdraw that for two r
On 11/13/14 15:14, Mike Stump wrote:
On Nov 13, 2014, at 9:21 AM, Jeff Law wrote:
Presumably we can get the same kinds of problems with ports that don't emit
prologues/epilogues as RTL?
I use prologue/epilogue to emit rtl in my port. I’d like this optimization to
kick on in my port, as we
On Thu, Nov 13, 2014 at 2:11 PM, Daniel Krügler
wrote:
> + typedef typename sub_match<_Bi_iter>::string_type string_type;
Use _String_type instead of string_type? I'm not 100% sure of this.
--
Regards,
Tim Shen
On Thu, Nov 13, 2014 at 7:19 AM, Kyrill Tkachov wrote:
>
> On 13/11/14 07:24, Andrew Pinski wrote:
>>
>> On Tue, Nov 11, 2014 at 3:55 AM, Kyrill Tkachov
>> wrote:
>>>
>>> Hi all,
>>>
>>> This is the aarch64 implementation of the macro fusion hook, used to fuse
>>> mov+movk instructions together.
In AutoFDO, we increase einline iterations. This could lead to
extensive code bloat if we have recursive calls like:
dtor() {
destroy(node);
}
destroy(node) {
destroy(left)
destroy(right)
}
In this case, the size growth will be around 8 which is smaller than
threshold (11). However, if we
On 2014.11.14 at 01:19 -0500, Trevor Saunders wrote:
> On Thu, Nov 13, 2014 at 03:48:34PM +0100, Markus Trippelsdorf wrote:
> > On 2014.11.13 at 15:11 +0100, mliska wrote:
> >
> > > + /* Destructor. */
> > > + virtual ~cgraph_summary ()
> > > + {
> > > +destroy ();
> > > + }
> >
> > From
On Wed, Oct 15, 2014 at 10:03 AM, Martin Liška wrote:
>
> Hello
>
> There's final version of the patch I'm going to commit tomorrow in the
> morning (CEST).
> Thank you Honza for the review.
>
> Martin
This caused:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63856
--
H.J.
On Nov 13, 2014, at 9:21 AM, Jeff Law wrote:
> Presumably we can get the same kinds of problems with ports that don't emit
> prologues/epilogues as RTL?
I use prologue/epilogue to emit rtl in my port. I’d like this optimization to
kick on in my port, as we do explain everything in rtl.
I do w
Hi,
I committed a patch from Dominique d'Humieres to restore bootsrap broken by
our offloading patches.
Thanks,
-- Ilya
2014-11-13 Dominique Dhumieres
PR bootstrap/63853
gcc/
* gcc.c (handle_foffload_option): Replace strchrnul with strchr.
* lto-wrapper.c (parse_en
On Wed, 5 Nov 2014, James Norris wrote:
> Hi!
>
> This patch represents the changes for OpenACC 2.0
> in the C front-end. At present these files will
> not compile as the changes for the middle end are
> not present.
So will things compile with the combination of this patch and the
middle-end p
On Thu, 13 Nov 2014, Thomas Schwinge wrote:
> gcc/doc/invoke.texi | 14
You're adding documentation for -fopenacc, but I don't see any .opt file
changes in this patch, and I'd expect the option to be added in the same
patch as its documentation.
--
Joseph S. Myers
Hi all,
since everyone seems to agree that it makes in principle sense to have a
reduce libgfortran (even though a full one is nicer, if the system
permits), let's look at the patch itself ;-)
Bernd Schmidt wrote:
The ptx port by its nature is lacking features that are expected on
"normal" m
Bootstrapped/regtested on power8-linux, ok for trunk?
2014-11-13 Marek Polacek
* tree.c (build_common_builtin_nodes): Remove doubled ECF_LEAF.
diff --git gcc/tree.c gcc/tree.c
index cf37a19..5c6fe0b 100644
--- gcc/tree.c
+++ gcc/tree.c
@@ -9935,7 +9935,7 @@ build_common_builtin_nodes
On Thu, Nov 13, 2014 at 1:45 PM, Jakub Jelinek wrote:
> On Thu, Nov 13, 2014 at 01:33:07PM -0800, Teresa Johnson wrote:
>> > Supposedly the printfs should have been removed and the #include
>> > isn't needed then either. No need to clutter the test output and log
>> > files.
>> > On the other s
New -mcpu= values need documenting in invoke.texi.
--
Joseph S. Myers
jos...@codesourcery.com
As Richi pointed in the pr audit trail, instrumenting via folding is
bad. In this case we changed __builtin_unreachable, created by the
inliner, into BUILT_IN_UBSAN_HANDLE_BUILTIN_UNREACHABLE which requires
VOPS, which is a no-no in folding. So this patch:
- marks BUILT_IN_UBSAN_HANDLE_BUILTIN_UN
On Thu, Nov 13, 2014 at 01:33:07PM -0800, Teresa Johnson wrote:
> > Supposedly the printfs should have been removed and the #include
> > isn't needed then either. No need to clutter the test output and log files.
> > On the other side, tests should abort (); or __builtin_abort (); on failure,
> >
On 12/11/14 11:15, Richard Biener wrote:
Please find a better way to communicate possibly_undefined_stmt than
enlarging struct loop. Like associating it with the niter bound
we record (so you can also have more than one).
Unfortunately, the bounds get regenerated frequently, but the upper
bou
1 - 100 of 263 matches
Mail list logo
