Hello,
The attached patch fixes violation of memory dependencies. The
problematic scenario happens when -fmodulo-sched-allow-regmoves flag
is set and certain anti-dep edges are not created.
For example, consider the following three instructions and the edges
between them. When -fmodulo-sched-all
On 9 June 2011 13:00, Ira Rosen wrote:
> Hi,
>
> This patch fixes several vectorizer testsuite failures on ARM:
> - vect-16.c checks that the vectorization fails without -ffast-math,
> but -ffast-math is a default flag for vector tests on ARM. I renamed
> the test to no-fast-math-vect-16.c to avoi
On 12/06/11 16:45, Jie Liu wrote:
> Hi,
>
> Java HelloWorld compiled by cross gcj for RTEMS can run without
> problem now, I think it's time for get the boehm-gc patch reviewed and
> merged. :) Here is the patch:
>
> Index: mach_dep.c
> ==
Hi,
tramp3d and some other benchmarks regressed as a result of switch to new
represnetaiton of same body aliases. The reason is that inliner doesn't handle
functions with aliases so well as it used to. The problems are:
1) functions with aliases are not inlined when called once
2) we now have mor
Hi Jeff,
Thanks for the quick review.
>> the right test is rtx_equal_p(operands[0], operands[1])
Committed with above changes to the bsetqi_msx, bclrqi_msx and bnotqi_msx
patterns.
Thanks & Regards,
Kaushik Phatak
www.kpitgnutools.com
-Original Message-
From: Jeff Law [mailto:l...@redh
On Mon, Jun 13, 2011 at 12:10 AM, Jason Merrill wrote:
> On 06/12/2011 06:59 AM, Richard Guenther wrote:
>>
>> The please provide a specification on what a VIEW_CONVERT_EXPR does
>> to type-based alias analysis.
>
> If the alias set of the VIEW_CONVERT_EXPR type the same as the set for the
> opera
On Mon, Jun 13, 2011 at 5:17 AM, Mike Stump wrote:
> On Jun 12, 2011, at 3:55 AM, Richard Guenther wrote:
>> In almost all cases(*) the need for a lvalue VIEW_CONVERT_EXPR can be avoided
>> by moving the VIEW_CONVERT_EXPR to the rvalue assigned too it. Remember that
>> VIEW_CONVERT_EXPR always co
On 10 June 2011 12:14, Richard Guenther wrote:
> In the end I think we should not generate the pattern stmt during
> pattern matching but only mark the relevant statements with a
> pattern kind. Say, for each pattern we have a "main" statement
> that has related stmts belonging to the pattern tha
Ping #1 for:
http://gcc.gnu.org/ml/gcc-patches/2011-06/msg00746.html
Georg-Johann Lay:
This patch fixes testsuite failures because the testcases assume
sizeof(int) >= 4.
* gcc.c-torture/compile/pr49029.c: Add dg-require-effective-target
int32plus
* gcc.c-torture/compile
Hi,
by accident I noticed that BINFO_VIRTUALs streaming is really expensive. It
about doubles amount of IL and types streamed by Mozilla.
One obvious optimization is to not stream into ltrans unit where it is
too late to do any useful devirtualization anyway.
Doing so reduces /tmp usage from 1.7GB
> 2011-06-09 Kaushik Phatak
>
> * config/h8300/h8300.md (bsetqi_msx, bclrqi_msx, bnotqi_msx): Added
> condition to disallow non-identical memory locations.
> (*andqi3_2, andqi3_1, iorqi3_1, xorqi3_1): Reorder insn to give
> preference to bit manipulation instructions.
>
On Mon, Jun 13, 2011 at 2:40 AM, Kaushik Phatak
wrote:
> Hi Jeff,
> Thanks for the quick review.
>
>>> the right test is rtx_equal_p(operands[0], operands[1])
> Committed with above changes to the bsetqi_msx, bclrqi_msx and bnotqi_msx
> patterns.
>
Commit is incorrect. Testsuite ChangeLog entri
Hi,
this patch makes inliner to handle removal of offline copies of functions in
the same comdat groups. Since aliases are now functions and a lot of comdat
functions do have aliases, we end up having a lot more comdat groups than
before.
Bootstrapped/regtested x86_64-linux, comitted.
Honza
Hi Guys,
I am checking in the patch below to add recognition of the Tilera TILE
architectures to the top level configure script. A patch to support
these architectures has just been committed to the binutils sources.
Cheers
Nick
./ChangeLog
2011-06-10 Walter Lee
* configure.ac (t
> "Jakub" == Jakub Jelinek writes:
Jakub> I'd like to propose convert to untyped operation, e.g.
Jakub> DW_OP_GNU_convert <0> could do it (and maybe DW_OP_GNU_reinterpret <0>),
Jakub> these would convert to an integral value of the same size as DWARF
Jakub> address and make it untyped. As DW
David,
I realized I did not "reply to all" when I first got your email.
You may have my email in your spam filter.
The issue is that I don't have WAA, and as Richard Guenther suggestion,
I want to ask if you could to do the commit for Freescale.
Thanks in advance for your patience.
In referenc
Hi.
this patch makes ipa_reverse_postorder to walk the alias references improving
effectivit of early opts on units with aliases.
Bootstrapped/regtested x86_64-linux, comitted.
Honza
* ipa-utils.c (postorder_stack): New structure.
(ipa_reverse_postorder): Handle aliases.
Index:
Hi,
On Fri, Jun 10, 2011 at 04:55:43PM +0200, Jan Hubicka wrote:
> Hi,
> this patch updated ipa-cp and ipa-prop for aliases. It is basically
> easy - we don't analyze nodes represneting aliases and when
> propagating we skip them, like everywhere else.
>
> There are two problems I noticed. Fir
On 06/13/2011 06:51 AM, Richard Guenther wrote:
But I suppose you want the array-ref be folded to a constant eventually?
Right.
I'm not going to keep arguing about VIEW_CONVERT_EXPR, but that brings
me back to my original question: is it OK to add a permissive mode to
the function, or should
On Jun 9, 2011, at 9:27 AM, Georg-Johann Lay wrote:
> This patch fixes testsuite failures because the testcases assume
> sizeof(int) >= 4.
>
> * gcc.c-torture/compile/pr49029.c: Add dg-require-effective-target
> int32plus
> * gcc.c-torture/compile/pr49163.c: Ditto.
Well, since n
The test g++.dg/cpp0x/nullptr21.C fails on IA64 HP-UX in 32 bit mode and
in looking at the problem I tracked it down to some code in
cxx_init_decl_processing that was added in r159131 to create nullptr_node:
2010-05-06 Magnus Fromreide
Jason Merrill
Add support for C++0x
OK.
Jason
On Thu, Jun 2, 2011 at 7:15 AM, Jan Kratochvil
wrote:
>
> libiberty/
> 2011-05-24 Jan Kratochvil
>
> * cp-demangle.c (struct d_print_info): Remove field options.
> (d_print_init): Remove parameter options.
> (cplus_demangle_print_callback): Update all the callers.
>
[In CCing Richard Henderson]
Denis Chertykov schrieb:
2011/6/10 Georg-Johann Lay :
Then I have a question on spill failures:
There is PR46278, an optimization flaw that goes as follows:
The avr BE defines fake addressing mode X+const that has to be written
down in asm as
X += const
a = *X
X
On Thu, Jun 2, 2011 at 7:16 AM, Jan Kratochvil
wrote:
>
> include/
> 2011-05-24 Jan Kratochvil
>
> * demangle.h (DMGL_RET_POSTFIX): Extend the comment.
> (DMGL_RET_DROP): New.
>
> libiberty/
> 2011-05-24 Jan Kratochvil
>
> * cp-demangle.c (d_print_comp) : Do
> not
On Thu, Jun 2, 2011 at 7:17 AM, Jan Kratochvil
wrote:
>
> libiberty/
> 2011-05-24 Jan Kratochvil
>
> * cp-demangle.c (d_print_comp) :
> Suppress d_print_mod for DMGL_RET_POSTFIX.
> * testsuite/demangle-expected: New testcases for --ret-postfix.
This is OK.
Thanks.
Ian
Mike Stump schrieb:
On Jun 9, 2011, at 9:27 AM, Georg-Johann Lay wrote:
This patch fixes testsuite failures because the testcases assume
sizeof(int) >= 4.
* gcc.c-torture/compile/pr49029.c: Add dg-require-effective-target
int32plus
* gcc.c-torture/compile/pr49163.c: Dit
On Mon, Jun 13, 2011 at 08:18:52PM +0200, Georg-Johann Lay wrote:
> For exammple, I added this line to, e.g.
> * gcc.c-torture/execute/cmpsi-2.c
> * gcc.c-torture/execute/pr45262.c
> in trunk r172757
> http://gcc.gnu.org/viewcvs?view=revision&revision=172757
That was a mistake.
gcc.c-torture/exec
Hi!
As the testcase shows, the
http://gcc.gnu.org/ml/gcc-patches/2010-06/msg02945.html
patch looks wrong, MEM_ATTRS matters quite a lot for the
alias oracle, so ignoring it leads to miscompilations.
Instead of just reverting the patch, this patch attempts to add
some exceptions, notably when MEM_
On Jun 13, 2011, at 11:18 AM, Georg-Johann Lay wrote:
> Who is a "specific maintainer" here?
I'd be happy to have the author of the testcase weigh in, or someone that cares
about int32plus, or the the person that fixed the bug in the compiler for which
the testcase was written...
> I found you
On Thu, Jun 09, 2011 at 06:27:56PM +0200, Georg-Johann Lay wrote:
> This patch fixes testsuite failures because the testcases assume
> sizeof(int) >= 4.
>
> * gcc.c-torture/compile/pr49029.c: Add dg-require-effective-target
> int32plus
> * gcc.c-torture/compile/pr49163.c: Ditto.
> Hi,
>
> On Fri, Jun 10, 2011 at 04:55:43PM +0200, Jan Hubicka wrote:
> > Hi,
>
> > this patch updated ipa-cp and ipa-prop for aliases. It is basically
> > easy - we don't analyze nodes represneting aliases and when
> > propagating we skip them, like everywhere else.
> >
> > There are two prob
We were running into a GC ICE because when reading a sorted_fields_type
field we were never setting the lenght of the vector.
This patch moves the setting of the vector length to the actual
allocation function (sorted_fields_type_new). It was being done in
finish_struct_1 before, so vectors read
http://codereview.appspot.com/4591066/diff/3001/gcc/tree-threadsafe-analyze.c
File gcc/tree-threadsafe-analyze.c (right):
http://codereview.appspot.com/4591066/diff/3001/gcc/tree-threadsafe-analyze.c#newcode1159
gcc/tree-threadsafe-analyze.c:1159: gcc_assert (false);
On 2011/06/11 17:52:51, Dieg
On Mon, Jun 13, 2011 at 12:44, wrote:
>> could
>> 2149 be optimized away. */
>> 2150 if (TREE_CODE (TREE_TYPE (DECL_ORIGIN (fdecl))) == METHOD_TYPE
>> 2151 && gimple_call_num_args(call) > 0)
>
>> Wouldn't it be easier to make fdecl == DECL_ORIGIN (fdecl) earlier in
>
> the
>>
>>
2011/6/13 Georg-Johann Lay :
>
> So you think is is pointless/discouraged to give a more realistic
> description of AVR addressing be means of MODE_CODE_BASE_REG_CLASS (instead
> of BASE_REG_CLASS) resp. REGNO_MODE_CODE_OK_FOR_BASE_P?
>
>> Look carefully at `out_movqi_r_mr'.
>> There are even two f
On Tue, Jun 07, 2011 at 12:24:06PM +0200, Richard Guenther wrote:
> Probably easier and more complete to do
>
> if (lhs && TREE_CODE (lhs) != SSA_NAME)
> {
> tree base = get_base_address (lhs);
Done in the patch below, bootstrapped/regtested again on x86
On Mon, 13 Jun 2011 20:06:21 +0200, Ian Lance Taylor wrote:
> The new options parameter should be the second parameter, not the last.
OK, I agree.
> This is OK with those changes.
Checked in.
Thanks,
Jan
http://gcc.gnu.org/viewcvs?view=revision&revision=174999
--- trunk/libiberty/ChangeLog
On Mon, 13 Jun 2011 20:08:40 +0200, Ian Lance Taylor wrote:
> On Thu, Jun 2, 2011 at 7:16 AM, Jan Kratochvil
> wrote:
> >
> > include/
> > 2011-05-24 Jan Kratochvil
> >
> > * demangle.h (DMGL_RET_POSTFIX): Extend the comment.
> > (DMGL_RET_DROP): New.
> >
> > libiberty/
> > 2011-0
On Mon, 13 Jun 2011 20:09:57 +0200, Ian Lance Taylor wrote:
> On Thu, Jun 2, 2011 at 7:17 AM, Jan Kratochvil
> wrote:
> >
> > libiberty/
> > 2011-05-24 Jan Kratochvil
> >
> > * cp-demangle.c (d_print_comp) :
> > Suppress d_print_mod for DMGL_RET_POSTFIX.
> > * testsuite/dem
Hi,
The patch ( http://gcc.gnu.org/ml/gcc-patches/2011-02/txt00059.txt ) which
introduces splitting avx256 unaligned loads.
However, we found that it causes significant regressions for cpu2006 (
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49089 ).
In this work, we introduce a tune option that
This caused http://gcc.gnu.org/PR49398.
You might also consider putting the PR numbers in your commit log in
the future that these patches attempt to fix so that the PR's in
bugzilla get the associated commit message. Please add links to this
commit to the individual PRs this commit attempts to fi
>
> I'm re-testing the attached version.
Shouldn't this be backported to release branches - specifically 4.6
(and 4.5 since this is where TARGET_VFP_SINGLE was introduced) ?
cheers
Ramana
The patch correct statement about CFG. According to the comment on
the function gimple_expand_cfg in gcc/cfgexpand.c, the tree CFG now
is preserved and expanded to RTL representation.
2011-06-14 Wei-Ren Chen
* gcc/doc/cfg.texi: Correct statement about CFG.
Index: cfg.texi
===
On 06/11/2011 12:23 AM, Thomas Henlich wrote:
I don't agree with this; with the patch we now output 10 significant
digits, whereas 9 is sufficient for a binary->ascii->binary roundtrip.
So please retain the "reduce d by one when E editing is used" thing
for list format and G0. This is just a side
This patch lowers calls to the builtin make function to runtime calls at
lowering time, rather than at backend conversion time. This patch also
adjusts type descriptors to use kindNoPointers aka GO_NO_POINTERS as the
6g library does. This lets the runtime memory allocation code quickly
check whet
This patch adds an intermediate gcov text format which does not require
source code. This format can be used by lcov or other tools.
I have bootstrapped it on x86 and all tests pass. Okay for main?
Thanks,
Sharad
2011-06-13 Sharad Singhai
Google Ref 3
* doc/gcov.texi: D
Hello,
> Please add the following:
> o A clarification that rows_length is used only (as an optimization) to
> back off quickly from trying to schedule a node in a full row; that is, to
> avoid running through futile DFA state transitions.
> o An assert that ps->rows_length[i] equals the number of
48 matches
Mail list logo
