Il 24/07/2012 22:17, Sandra Loosemore ha scritto:
> I was looking to see what needs to be done to un-stick this previously
> submitted patch:
>
> http://gcc.gnu.org/ml/gcc-patches/2012-05/msg01419.html
>
> Paolo's suggestion was to re-write this to use a "tortoise-and-hare"
> algorithm to detect
Hello,
The attached patch reorganizes the SH left shift patterns. The final
generated shift insns don't differ much, but it seems to have some
positive side effects on register allocation in some cases.
Tested with
make -k check RUNTESTFLAGS="--target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a/-mb,-m2
Gerald,
it is ok ?
Tristan.
On Jul 19, 2012, at 12:15 PM, Tristan Gingold wrote:
>
> On Jul 19, 2012, at 11:37 AM, Pedro Alves wrote:
>
>> On 07/19/2012 08:30 AM, Tristan Gingold wrote:
>>
>>> No regression on i386 GNU/Linux.
>>> Committed.
>>
>> Nice. Is this NEWS / docs --or wherever rel
Hello,
here is a slight improvement to VRP for sum and difference of intervals.
There are some things I left because I didn't understand them enough:
* range_int_cst_p (&vr0): I thought it was always true by that time, but
it isn't obvious
* TYPE_PRECISION (expr_type) <= HOST_BITS_PER_DOUBL
Thanks!
I've removed declarations. New patch attached.
testsuite/ChangeLog:
2012-07-20 Anna Tikhonova
* gcc.dg/20020201-1.c: Remove declarations for exit, abort,
rand, srand. Include .
2012/7/24 Janis Johnson :
> On 07/20/2012 05:23 AM, Anna Tikhonova wrote:
>> Hi all,
>>
>> I've f
On Tue, Jul 24, 2012 at 5:50 PM, Andrew Pinski
wrote:
> Hi,
> Before tuples was introduced, VN used to lookup the simplified
> expression to see if it was available already and use that instead of
> the non simplified one. This patch adds the support back to VN to do
> exactly that.
>
> OK? Boo
On Tue, Jul 24, 2012 at 9:27 PM, Marek Polacek wrote:
> Ping.
Ok.
Thanks,
Richard.
> On Fri, Jul 13, 2012 at 02:00:54PM +0200, Marek Polacek wrote:
>> I think the comment at the start of the file is wrong, since
>> it speaks about varpool_finalize_variable, but there's no such
>> function (not
On Tue, Jul 24, 2012 at 9:32 PM, Uros Bizjak wrote:
> Hello!
>
> Attached patch backports recent LEA improvements to 4.7 branch.
As in regression fixes, or wrong-code fixes? This looks suspiciously not
appropriate for 4.7 ...
Richard.
> The
> patch unifies handling of LEA addresses with generi
On Tue, Jul 24, 2012 at 7:13 PM, Steven Bosscher wrote:
>> AFAIR the qsort is just for getting a stable ordering over two pools to find
>> the leaked regsets afterwards, the type of ordering doesn't matter.
>
> What matters is that the compare function gives a reliable result. You
> can't subtract
> Please see attached patch that does all this with minimum surgery and
> also sets missing SSE prefetch for x86_32 with -mprfchw
Thanks, applied!
> +++ b/gcc/config/i386/prfchwintrin.h
>
> +#if !defined _X86INTRIN_H_INCLUDED && !defined _MM3DNOW_H_INCLUDED
> +# error "Never use directly; include
On Wed, Jul 25, 2012 at 1:49 PM, Richard Guenther
wrote:
> Oh, bigger bitmap_head? That's bad ... :/ So much for '#ifdefs are bad' :/
Bigger bitmap_head isn't a problem. A bigger bitmap_element would be bad.
For GGC allocated bitmaps, nothing changed (rounding, etc.).
If you think it's a prob
On Wed, Jul 25, 2012 at 1:45 PM, Richard Guenther
wrote:
> On Tue, Jul 24, 2012 at 9:32 PM, Uros Bizjak wrote:
>> Hello!
>>
>> Attached patch backports recent LEA improvements to 4.7 branch.
>
> As in regression fixes, or wrong-code fixes? This looks suspiciously not
> appropriate for 4.7 ...
I
On Wed, Jul 25, 2012 at 1:51 PM, Kirill Yukhin wrote:
>> Please see attached patch that does all this with minimum surgery and
>> also sets missing SSE prefetch for x86_32 with -mprfchw
> Thanks, applied!
>
>> +++ b/gcc/config/i386/prfchwintrin.h
>>
>> +#if !defined _X86INTRIN_H_INCLUDED && !defin
>> Is it OK for trunk?
>
> OK.
Thanks!
Checked in.
http://gcc.gnu.org/viewcvs?view=revision&revision=189844
Next I think would be rdseed* insns.
Thanks, K
This fixes a typo in a comment.
Applied to trunk as obvious.
gcc/
* basic-block.c: Fix typo in comment.
diff --git a/gcc/basic-block.h b/gcc/basic-block.h
index 15d34e0..dff06e4 100644
--- a/gcc/basic-block.h
+++ b/gcc/basic-block.h
@@ -108,7 +108,7 @@ struct GTY(()) gimple_bb_info {
On Wed, Jul 25, 2012 at 12:21 PM, Marc Glisse wrote:
> Hello,
>
> here is a slight improvement to VRP for sum and difference of intervals.
> There are some things I left because I didn't understand them enough:
>
> * range_int_cst_p (&vr0): I thought it was always true by that time, but it
> isn't
On Wed, Jul 25, 2012 at 1:59 PM, Steven Bosscher wrote:
> On Wed, Jul 25, 2012 at 1:49 PM, Richard Guenther
> wrote:
>> Oh, bigger bitmap_head? That's bad ... :/ So much for '#ifdefs are bad' :/
>
> Bigger bitmap_head isn't a problem. A bigger bitmap_element would be bad.
>
> For GGC allocated
Hi Sandra,
One suggestion - rather than having architecture specific test files,
why not just have a single generic test case with a new
dg-require-naked-attribute qualifier. That way the mcore port would be
tested as well as the ARM port.
Something like this? The code part of the patch is un
The optabs rewrite removed the special handling for FIXUNS_TRUNC_LIKE_FIX_TRUNC.
This was a minor short-hand only used by two targets, and can be replaced by a
grand total of two patterns in each.
Cross-compiled to vax-linux and v850-elf and committed.
r~
* system.h (FIXUNS_TRUNC_LIKE_FI
Hi again,
Here is second patch which adds support of rdseed[16|32|64] insn.
Changelog entry:
2012-07-25 Kirill Yukhin
Michael Zolotukhin
* common/config/i386/i386-common.c (OPTION_MASK_ISA_RDSEED_SET): New.
(OPTION_MASK_ISA_RDSEED_UNSET): Likewise.
(ix86_h
Now that access control is subject to SFINAE, I was worried that
tf_no_access_control might cause wrong overload resolution. On further
investigation I think that it wasn't a problem, but it is unnecessary.
To preserve the access control checking on default arguments, I changed
check_default_
(CC gcc-patches)
On 25 July 2012 10:26, François Dumont wrote:
> Hi
>
> Here is a patch proposal for PR 54075. I also took the occasion to fix
> something that has been delay so far which is usage of std::max to get the
> number of buckets to use. The problem of using std::max when using the h
54020 notes that we were accepting a constexpr function that is
ill-formed, no diagnostic required, but it would be easy to give a
diagnostic for.
54086 suggests that the standard seems to allow constexpr and const
together.
Tested x86_64-pc-linux-gnu, applying to trunk.
commit 92be1e0a23c73
Hi,
While reviewing code to fix a previous patch submission around assembly
dialect handling[1], I found that the dialect handling in asm_fprintf
is different from that in output_asm_insn and it might be broken too,
since encountering a '|' simply leads to skipping all of the string
until a '}' is
That's what I get for assuming front-ends don't touch back-end things
like optabs and libcalls...
I'm reasonably certain that a better solution would be to move the
set_stack_check_libfunc declaration elsewhere and/or totally rearrange
that specific interface: c.f. lang_hooks.eh_protect_cleanup_ac
On 07/25/2012 08:10 AM, Siddhesh Poyarekar wrote:
> 2012-07-25 Siddhesh Poyarekar
>
> * final.c [ASSEMBLER_DIALECT](do_assembler_dialects): New
> function to implement assembler dialects.
> (output_asm_insn): Use do_assembler_dialects.
> (asm_fprintf): Likewise.
Ok ...
On Wed, 25 Jul 2012 08:31:14 -0700, Richard wrote:
> >
> > * gcc.dg/asm-dialect-1.c: New test case.
>
> ... except this should go in gcc.target/i386/ without the { target }
> qualifier.
>
Thanks, here's the updated version.
Regards,
Siddhesh
gcc/ChangeLog:
2012-07-25 Siddhesh Poyarekar
On Wed, Jul 25, 2012 at 4:07 PM, Kirill Yukhin wrote:
> Here is second patch which adds support of rdseed[16|32|64] insn.
>
> Changelog entry:
> 2012-07-25 Kirill Yukhin
> Michael Zolotukhin
>
> * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDSEED_SET): New.
>
On 07/24/2012 09:40 PM, Sandra Loosemore wrote:
> PR target/53633
>
> gcc/
> * target.def (warn_func_return): New hook.
> * doc/tm.texi.in (TARGET_WARN_FUNC_RETURN): New hook.
> * doc/tm.texi: Regenerate.
> * doc/sourcebuild.texi (Effective-Target Keywords): Doc
On Tue, Jul 24, 2012 at 5:51 PM, Cary Coutant wrote:
> This patch is for the google/gcc-4_7 branch.
>
> Add -gfission as alias for -gsplit-dwarf.
>
> Tested by bootstrap and regression testing.
>
>
> * common.opt (-gfission): Alias for -gsplit-dwarf.
> (-gno-fission): Alias for -gn
Per Richard Henderson's suggestion
(http://gcc.gnu.org/ml/gcc-patches/2012-06/msg01370.html), this patch
changes the IVOPTS and straight-line strength reduction passes to make
use of data computed by init_expmed. This required adding a new
convert_cost array in expmed to store the costs of convert
On Wed, Jul 25, 2012 at 5:20 PM, Richard Henderson wrote:
> That's what I get for assuming front-ends don't touch back-end things
> like optabs and libcalls...
That is _almost_ the case. In fact, this is one of the last problems...
$ egrep
"\"optabs\.h|\"libfuncs\.h|\"expr\.h|\"rtl|\"recog|\"ba
As documented in tree.def and checked in tree-cfg.c, the second and third
operands of BIT_FIELD_REF are INTEGER_CSTs so there is no need to do fancy
things on them. The patch also removes the now unused build6 stuff and fixes a
buglet in the folder.
Tested on x86_64-suse-linux, OK for the main
On 07/25/2012 09:13 AM, William J. Schmidt wrote:
> Per Richard Henderson's suggestion
> (http://gcc.gnu.org/ml/gcc-patches/2012-06/msg01370.html), this patch
> changes the IVOPTS and straight-line strength reduction passes to make
> use of data computed by init_expmed. This required adding a new
This fixes a mixup of kind and size when storing a real(kind=10) number
for read, causing a loss of two bytes. Tested on m68k-linux and checked
in as obvious.
Andreas.
* io/list_read.c (list_formatted_read_scalar): Fix copying real
value back to temporary.
diff --git a/libgfortr
On 24/07/12 23:26, Bharathi Seshadri (bseshadr) wrote:
> Hi,
>
> Attached is a patch that modifies BE8_LINK_SPEC to pass -be8 option to the
> linker by default if the target is big endian, and to not pass -be8 option if
> -mlittle-endian is used. It also preserves the existing usage whereby -be
On 07/25/2012 09:30 AM, Eric Botcazou wrote:
> 2012-07-25 Eric Botcazou
>
> * expr.c (expand_expr_real_1): Do not expand operand #1 and #2
> of BIT_FIELD_REF for ignored results.
> * fold-const.c (fold_ternary_loc) : Check that the
> CONSTRUCTOR has vector type before us
On 07/23/2012 08:13 AM, Geert Bosch wrote:
> IIUC, all ports are supposed to implement the atomic built-ins. If they are
> not
> supported in hardware, there should be a library function for it that uses
> locking.
> The problem we're trying to address is builds failing because of undefined
> re
Committed as obvious (Rev. 189859).
Tobias
2012-07-25 Tobias Burnus
* trans-types.c (gfc_real16_is_float128): Fix spelling
in a comment.
* trans.h (struct gfc_array_info): Ditto.
* gfortran.h (gfc_expr): Ditto.
* simplify.c (gfc_count): Ditto.
On 07/25/2012 09:57 AM, Richard Henderson wrote:
>
> I'll echo Nick's comments about arm asm in a common test.
> There's no need to have anything but __asm__(""); there.
>
> Ok with that change.
Thanks! Here's the version I committed.
-Sandra
2012-07-25 Sandra Loosemore
Paul B
On 07/18/2012 05:48 AM, Eric Botcazou wrote:
> 2012-07-18 Eric Botcazou
>
> * gimple-low.c (lower_try_catch): New function.
> (lower_stmt) : Use it to lower GIMPLE_TRY_CATCH.
> : Delete.
> : Likewise.
>
>
> 2012-07-18 Eric Botcazou
>
> * gnat.dg/noreturn5.ad[
On 07/05/2012 07:29 AM, Nick Clifton wrote:
> static void
> -mn10300_encode_section_info (tree decl, rtx rtl, int first ATTRIBUTE_UNUSED)
> +mn10300_encode_section_info (tree decl, rtx rtl, int first)
> {
>rtx symbol;
>
> @@ -2480,6 +2488,8 @@
>
>if (flag_pic)
> SYMBOL_REF_FLAG (
Resending in plain text mode so that it goes through to mailing list...
On Wed, Jul 25, 2012 at 11:32 AM, Teresa Johnson wrote:
>
>
>
>
> On Tue, Jul 24, 2012 at 3:03 PM, Chris Manghane wrote:
>>
>> This patch modifies pmu-profile to allow gooda_feedback access to
>>
>>
> Needs some more contex
On 06/29/2012 02:03 AM, Nick Clifton wrote:
> 2012-06-29 Nick Clifton
>
> * config/mn10300/mn10300.c (REG_SAVE_BYTES): Delete.
> (mn10300_get_live_callee_saved_regs): If requested return a count
> of the number of bytes in the mask.
> (mn10300_expand_prologue): Add argum
This patch is for the google/gcc-4_7 branch.
Backport pending upstream patch to output DW_AT_high_pc as a constant
instead of a relocated address. This significantly reduces the number
of entries used in the .debug_addr table with -gfission.
Tested: Ran regression tests and validate-failures.py.
On Wed, Jul 25, 2012 at 11:45 AM, Cary Coutant wrote:
> This patch is for the google/gcc-4_7 branch.
>
> Backport pending upstream patch to output DW_AT_high_pc as a constant
> instead of a relocated address. This significantly reduces the number
> of entries used in the .debug_addr table with -g
On 07/18/2012 05:44 PM, Sterling Augustine wrote:
+/* Remove an entry from the addr table. Since we have already numbered
+ all the entries, the best we can do here is null it out. */
This sounds like a design flaw. Since we have everything in memory,
there's no reason we shouldn't be abl
On 07/17/2012 05:22 AM, Richard Guenther wrote:
On Wed, Jul 4, 2012 at 6:35 PM, Sandra Loosemore
wrote:
Ping? Original post with patch is here:
http://gcc.gnu.org/ml/gcc-patches/2012-06/msg00319.html
Can you update the patch and numbers based on what Bill did for
straight-line strength re
On Wed, 2012-07-25 at 09:59 -0700, Richard Henderson wrote:
> On 07/25/2012 09:13 AM, William J. Schmidt wrote:
> > Per Richard Henderson's suggestion
> > (http://gcc.gnu.org/ml/gcc-patches/2012-06/msg01370.html), this patch
> > changes the IVOPTS and straight-line strength reduction passes to make
On Wed, 2012-07-25 at 13:39 -0600, Sandra Loosemore wrote:
> On 07/17/2012 05:22 AM, Richard Guenther wrote:
> > On Wed, Jul 4, 2012 at 6:35 PM, Sandra Loosemore
> > wrote:
> >>
> >> Ping? Original post with patch is here:
> >>
> >> http://gcc.gnu.org/ml/gcc-patches/2012-06/msg00319.html
> >
> >
The following issue was found by Alessandro. (It got triggered by a
larger test case, which is required for a larger patch by Alessandro,
which is not yet finished.)
Accessing the "lower[-1]" is probably not the best idea …
Build on x86-64-linux.
OK for the trunk?
Tobias
2012-07-25 Alessandr
Attached patch applied to trunk. I am building 4.7 branch to also apply
the patch to this branch.
2012-07-25 François Dumont
PR libstdc++/54075
* include/bits/hashtable.h
(_Hashtable<>::_Hashtable(_InputIterator, _InputIterator,
size_type, ...): Remove std::ma
On 07/25/2012 12:40 PM, William J. Schmidt wrote:
> Thanks for the quick review! Excellent point about the array size. The
> attached revised patch follows your suggestion to limit the size.
>
> I only did this for the new field, as changing all the existing
> accessors to inline functions is mo
Tobias Burnus wrote:
The following issue was found by Alessandro. (It got triggered by a
larger test case, which is required for a larger patch by Alessandro,
which is not yet finished.)
Accessing the "lower[-1]" is probably not the best idea …
Build on x86-64-linux.
OK for the trunk?
I att
This patch changes test g++.dg/cpp0x/nullptr21.C to be self-checking
instead of examining printf output via dg-output. Now it can be run
reliably on bare-metal platforms. I couldn't see any indication in the
test itself or in the gcc-patches archives to show that the printfs
were significant.
Ch
Hello Everyone,
This patch is for the Cilk Plus branch mainly affecting the LTO flag. This
patch will allow the flag_enable_cilk to carry through into the LTO code. Also,
I added one small case statement for __cilkrts_detach.
Thanks,
Balaji V. Iyer.
Index: gcc/c-family/c.opt
===
Hello,
This adds a new SH option -mzdcbranch / -mnozdcbranch to tell the
compiler whether zero displacement conditional branches are fast on the
target CPU. It is enabled by default for SH4* and thus does not change
any of the established behavior.
Currently this does affect only one pattern (neg
Oleg Endo wrote:
> The attached patch reorganizes the SH left shift patterns. The final
> generated shift insns don't differ much, but it seems to have some
> positive side effects on register allocation in some cases.
>
> Tested with
> make -k check RUNTESTFLAGS="--target_board=sh-sim
> \{-m2/
>> +/* Remove an entry from the addr table. Since we have already numbered
>> + all the entries, the best we can do here is null it out. */
>
> This sounds like a design flaw. Since we have everything in memory, there's
> no reason we shouldn't be able to pack the table appropriately. Why can
Oleg Endo wrote:
> This adds a new SH option -mzdcbranch / -mnozdcbranch to tell the
> compiler whether zero displacement conditional branches are fast on the
> target CPU. It is enabled by default for SH4* and thus does not change
> any of the established behavior.
> Currently this does affect o
On Tue, 24 Jul 2012, Richard Henderson wrote:
> On 07/21/2012 06:10 AM, Oleg Endo wrote:
> > I think on SH the cost test in lshift_cheap_p with
> > gen_rtx_ASHIFT (word_mode, const1_rtx, reg), speed_p);
> >
> > will always 'fail', because of sh.c (shiftcosts):
> > /* There is no pattern for co
Applied.
Gerald
Index: htdocs/svn.html
===
RCS file: /cvs/gcc/wwwdocs/htdocs/svn.html,v
retrieving revision 1.175
diff -u -3 -p -r1.175 svn.html
--- htdocs/svn.html 18 Jul 2012 01:59:05 - 1.175
+++ htdocs/svn.html 25
On Wed, Jul 25, 2012 at 4:00 PM, Cary Coutant wrote:
>>> +/* Remove an entry from the addr table. Since we have already numbered
>>> + all the entries, the best we can do here is null it out. */
>>
>> This sounds like a design flaw. Since we have everything in memory, there's
>> no reason we
On Wed, 25 Jul 2012, Tristan Gingold wrote:
> it is ok ?
Two minor changes, fine to commit with those.
1. Full stop at the end of the sentence.
2. Avoid extra ... pair, rather add this to the existing.
Thanks,
Gerald
This patch is for the google/gcc-4_7 branch.
Compute a dwo_id and write the DW_AT_GNU_dwo_id attribute to both the
skeleton compile unit DIE and the DWO compile unit DIE.
Tested: passed regression tests.
2012-07-25 Cary Coutant
* gcc/dwarf2out.c (dwo_id_placeholder): Delete.
> From: Ulrich Weigand
> Date: Tue, 24 Jul 2012 19:38:15 +0200
> > > I've implemented this as a separate hook, rather than using the existing
> > > hooks because there's a strong likelihood of breaking some existing ABIs
> > > if I did it another way.
> > >
> > > There are a couple of tests that
+ if (dwarf_version >= 4 || dwarf_split_debug_info)
>>>
>>> Shouldn't -gsplit-debug require -gdwarf-4+ anyway?
>>
>> Seems better to check explicitly than to rely on an implied relationship.
>
> I think Jason is right here. Although they are theoretically separate,
> there has been no tes
Hi,
The Linux kernel already supports non-executable stack since around
February 2010. This patch has GCC emit the notes that are associated
with non-executable stack.
OK? Bootstrapped and tested on mips64-linux-gnu with kernel already
supporting the non-executable stack with no regressions.
This patch to libgo fixes the runtime/pprof test to pass, and enables
it. it cleans up the runtime interface for profiling a little bit.
Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu.
Committed to mainline and 4.7 branch.
Ian
diff -r 55f2af449cde libgo/Makefile.am
--- a/libgo/Ma
On 07/25/2012 01:27 AM, Paolo Bonzini wrote:
>
> What I'm worried about is the extra cost of malloc-ing and free-ing the
> pointer set. Perhaps you can skip the pointer set creation in the common
> case where find_comparison_args does not iterate? Something like this:
>
> [snip]
I think this v
70 matches
Mail list logo
