This is a follow-up to:
http://gcc.gnu.org/ml/gcc/2013-10/msg00075.html
The outcome on IRC was that __asm ought to behave like aliases eventually,
so we should abandom the renaming-in-C thing. One way would be to code
the functions directly as asm files, or use inline asm, but it seems a
-Original Message-
From: Kyrill Tkachov [mailto:kyrylo.tkac...@arm.com]
Sent: 10 October 2013 17:15
To: Paulo Matos
Cc: gcc-patches@gcc.gnu.org
Subject: Re: [PATCH] Fix PR58682
Hi Paul,
On 10/10/13 17:10, Paulo Matos wrote:
- inline_call (edge, true, new_indirect_edges,
On Thu, Oct 10, 2013 at 7:12 PM, Andrew MacLeod amacl...@redhat.com wrote:
On 10/10/2013 10:35 AM, Richard Biener wrote:
On Thu, Oct 10, 2013 at 4:30 PM, Richard Biener
richard.guent...@gmail.com wrote:
On Thu, Oct 10, 2013 at 3:10 PM, Andrew MacLeod amacl...@redhat.com
wrote:
More fun
On Thu, Oct 10, 2013 at 8:07 PM, tsaund...@mozilla.com wrote:
From: Trevor Saunders tsaund...@mozilla.com
Hi,
This makes the implementation of stack vectors simpler and easier to use.
This
On Fri, Oct 11, 2013 at 5:31 AM, Andrew MacLeod amacl...@redhat.com wrote:
Missed a bit in tree-flow.h.. I mistakenly assumed omp_region belonged
there :-P
Anyway by moving struct omp_region into omp_low.h, along with the prototypes
from tree-flow.h, gimple.h and tree.h. Everything works
On Fri, Oct 11, 2013 at 10:12:17AM +0200, Richard Biener wrote:
I can spot the end of a comment containing OpenMP Region Tree
that you didn't move - probably an oversight.
Otherwise ok.
Please wait for the gomp4 merge (which I'm going to commit today).
Jakub
On Fri, Oct 11, 2013 at 9:29 AM, Richard Sandiford
rdsandif...@googlemail.com wrote:
This is a follow-up to:
http://gcc.gnu.org/ml/gcc/2013-10/msg00075.html
The outcome on IRC was that __asm ought to behave like aliases eventually,
so we should abandom the renaming-in-C thing. One way
On Fri, Oct 11, 2013 at 10:17:41AM +0200, Richard Biener wrote:
asm(.alias __sync_synchronize sync_synchronize);
It is .set, but not everywhere.
/* The MIPS assembler has different syntax for .set. We set it to
.dummy to trap any errors. */
#undef SET_ASM_OP
#define SET_ASM_OP \t.dummy\t
But
Hi,
On 10/11/2013 05:55 AM, Jason Merrill wrote:
On 10/10/2013 03:31 PM, Paolo Carlini wrote:
On 10/10/2013 08:26 PM, Jason Merrill wrote:
On 10/10/2013 08:33 AM, Paolo Carlini wrote:
+ expr_type = TREE_TYPE (expr) = cp_build_qualified_type
+(TREE_TYPE (expr), cp_type_quals
On 2013-10-11 3:26, Jason Merrill wrote:
Can we put this in cp_parser, invert the sense of the flag, and
only
set it during cp_parser_parameter_declaration_clause?
Done. But it still needs a flag to disable in the case of explicit
template specialization.
Can't you just check
Jakub Jelinek ja...@redhat.com writes:
On Fri, Oct 11, 2013 at 10:17:41AM +0200, Richard Biener wrote:
asm(.alias __sync_synchronize sync_synchronize);
It is .set, but not everywhere.
/* The MIPS assembler has different syntax for .set. We set it to
.dummy to trap any errors. */
#undef
On Fri, Oct 11, 2013 at 11:43 AM, Richard Sandiford
rdsandif...@googlemail.com wrote:
Jakub Jelinek ja...@redhat.com writes:
On Fri, Oct 11, 2013 at 10:17:41AM +0200, Richard Biener wrote:
asm(.alias __sync_synchronize sync_synchronize);
It is .set, but not everywhere.
/* The MIPS assembler
On Tue, Oct 08, 2013 at 09:50:55PM +0200, Jakub Jelinek wrote:
I'd like to merge (most of) gomp-4_0-branch to trunk.
I've incorporated the feedback as r203339, r203391 and r203392
and committed the whole merge including those 3 commits to trunk
as r203408. I've merged then trunk up to r203408
Hi,
On 10/10/2013 08:48 PM, Jason Merrill wrote:
I think that committing was important for two primary reasons:
1. We should not have irreversible side-effects while in a tentative
state. For example, we shouldn't create a permanent entry in the
symbol table, or issue an error
Richard Biener richard.guent...@gmail.com writes:
On Fri, Oct 11, 2013 at 11:43 AM, Richard Sandiford
rdsandif...@googlemail.com wrote:
Jakub Jelinek ja...@redhat.com writes:
On Fri, Oct 11, 2013 at 10:17:41AM +0200, Richard Biener wrote:
asm(.alias __sync_synchronize sync_synchronize);
It
On Fri, Oct 11, 2013 at 12:48 PM, Richard Sandiford
rdsandif...@googlemail.com wrote:
Richard Biener richard.guent...@gmail.com writes:
On Fri, Oct 11, 2013 at 11:43 AM, Richard Sandiford
rdsandif...@googlemail.com wrote:
Jakub Jelinek ja...@redhat.com writes:
On Fri, Oct 11, 2013 at
This adjusts dumping to show differences in TREE_OVERFLOW on
INTEGER_CSTs in the IL (to make the issue causing VRP differences
visible) and to also dump whether an abnormal SSA name is
a default def.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
2013-10-11 Richard
On Wed, Oct 9, 2013 at 7:11 PM, Bill Schmidt
wschm...@linux.vnet.ibm.com wrote:
Hi,
This is a follow-up to the recent patch that fixed constant permute
control vectors for little endian. When the control vector is constant,
we can adjust the constant and use a vperm without increasing code
Hi,
In my previous patch to split LE VSX loads and stores to introduce
permutes, I managed to miss the vector float modes. This patch corrects
the oversight, fixing up a few more test failures.
Bootstrapped and tested on both powerpc64-unknown-linux-gnu and
powerpc64le-unknown-linux-gnu with no
On Mon, Oct 07, 2013 at 03:52:25PM +0200, Marc Glisse wrote:
2013-10-08 Marc Glisse marc.gli...@inria.fr
PR tree-optimization/58480
gcc/
* tree-vrp.c (infer_nonnull_range): New function.
(infer_value_range): Call infer_nonnull_range.
This broke whole bunch of OpenMP
On Fri, Oct 11, 2013 at 8:48 AM, Bill Schmidt
wschm...@linux.vnet.ibm.com wrote:
Hi,
In my previous patch to split LE VSX loads and stores to introduce
permutes, I managed to miss the vector float modes. This patch corrects
the oversight, fixing up a few more test failures.
Bootstrapped
Jakub Jelinek ja...@redhat.com writes:
On Fri, Oct 11, 2013 at 10:17:41AM +0200, Richard Biener wrote:
asm(.alias __sync_synchronize sync_synchronize);
It is .set, but not everywhere.
/* The MIPS assembler has different syntax for .set. We set it to
.dummy to trap any errors. */
#undef
Hello,
On 09 Oct 14:20, Richard Henderson wrote:
On 10/09/2013 03:24 AM, Kirill Yukhin wrote:
Here's 2nd subpatch. It fixes missing `v' constraints.
And one v constraint that shouldn't have been.
Exactly!
Ok.
Checked into main trunk: http://gcc.gnu.org/ml/gcc-cvs/2013-10/msg00379.html
Hello,
On 09 Oct 14:25, Richard Henderson wrote:
On 10/09/2013 03:24 AM, Kirill Yukhin wrote:
Here's 3rd subpatch. It extends VF1 and VI iterators.
Ok.
Thanks, checked into main trunk:
http://gcc.gnu.org/ml/gcc-cvs/2013-10/msg00380.html
K
On Fri, 11 Oct 2013, Jakub Jelinek wrote:
On Mon, Oct 07, 2013 at 03:52:25PM +0200, Marc Glisse wrote:
2013-10-08 Marc Glisse marc.gli...@inria.fr
PR tree-optimization/58480
gcc/
* tree-vrp.c (infer_nonnull_range): New function.
(infer_value_range): Call
On 10/11/2013 04:16 AM, Jakub Jelinek wrote:
On Fri, Oct 11, 2013 at 10:12:17AM +0200, Richard Biener wrote:
I can spot the end of a comment containing OpenMP Region Tree
that you didn't move - probably an oversight.
Otherwise ok.
Please wait for the gomp4 merge (which I'm going to commit
Richard Biener richard.guent...@gmail.com writes:
On Fri, Oct 11, 2013 at 12:48 PM, Richard Sandiford
rdsandif...@googlemail.com wrote:
Richard Biener richard.guent...@gmail.com writes:
On Fri, Oct 11, 2013 at 11:43 AM, Richard Sandiford
rdsandif...@googlemail.com wrote:
Jakub Jelinek
Hello,
On 09 Oct 14:32, Richard Henderson wrote:
On 10/09/2013 03:25 AM, Kirill Yukhin wrote:
Here's 4th subpatch. It extends V iterator.
Ok.
Thanks, checked into main trunk:
http://gcc.gnu.org/ml/gcc-cvs/2013-10/msg00382.html
K
Hello,
On 09 Oct 14:34, Richard Henderson wrote:
On 10/09/2013 03:25 AM, Kirill Yukhin wrote:
Here's 5th subpatch. It introduces `multdiv' code iterator.
This is the sort of patch I like to see. It's the first one
you've sent that's done exactly one thing. Congratulations.
Ok.
Thanks,
Hello
On 09 Oct 14:35, Richard Henderson wrote:
On 10/09/2013 03:26 AM, Kirill Yukhin wrote:
Here's 6th subpatch. It extends VI2 and VI124 iterators.
Ok.
Thanks, checked into main trunk:
http://gcc.gnu.org/ml/gcc-cvs/2013-10/msg00384.html
K
On Fri, Oct 11, 2013 at 05:39:05PM +0400, Kirill Yukhin wrote:
Hello,
On 09 Oct 14:34, Richard Henderson wrote:
On 10/09/2013 03:25 AM, Kirill Yukhin wrote:
Here's 5th subpatch. It introduces `multdiv' code iterator.
This is the sort of patch I like to see. It's the first one
you've
Hello,
On 09 Oct 14:37, Richard Henderson wrote:
On 10/09/2013 03:26 AM, Kirill Yukhin wrote:
Here's 7th subpatch. It extends VI4 and VI8 iterators.
Ok.
Thanks, checked into main trunk:
http://gcc.gnu.org/ml/gcc-cvs/2013-10/msg00385.html
K
On 11 Oct 15:43, Jakub Jelinek wrote:
On Fri, Oct 11, 2013 at 05:39:05PM +0400, Kirill Yukhin wrote:
Thanks, checked into main trunk:
http://gcc.gnu.org/ml/gcc-cvs/2013-10/msg00383.html
Everybody can read gcc-cvs mailing list, it's archives or svn log
or git log, there is no need to
Hi!
On Fri, Oct 11, 2013 at 02:44:16PM +0200, Jan-Benedict Glaw wrote:
The recent change probably gave us this[1]:
g++ -c -DIN_GCC_FRONTEND -DIN_GCC_FRONTEND -g -O2 -DIN_GCC
-DCROSS_DIRECTORY_STRUCTURE -fno-exceptions -fno-rtti
-fasynchronous-unwind-tables -W -Wall -Wno-narrowing
On 10/11/2013 04:51 AM, Paolo Carlini wrote:
The 6th time we get there when compiling the testcase, that is when we
are converting to int, at that line this is expr_type:
reference_type 0x768587e0
type integer_type 0x76858738 int readonly type_6 SI
Aha. The patch is OK then.
On Mon, 7 Oct 2013, Marc Glisse wrote:
2013-10-08 Marc Glisse marc.gli...@inria.fr
[...]
gcc/
* doc/extend.texi (returns_nonnull): New function attribute.
By the way, I'll commit this obvious doc fix next chance I get:
2013-10-12 Marc Glisse marc.gli...@inria.fr
*
On Fri, 11 Oct 2013, Marc Glisse wrote:
On Fri, 11 Oct 2013, Jakub Jelinek wrote:
OT, do you plan to define ATTRIBUTE_RETURNS_NONNULL for
GCC_VERSION = 4009 in ansidecl.h and use it on the various xmalloc
etc. prototypes?
I was planning to at least have a look at some point. I can do that
On 10/11/2013 09:56 AM, Jakub Jelinek wrote:
With the operator bool (), there is ambiguity in the
if (((mask something) 1) == 0)
tests (so had to use OMP_CLAUSE_MASK_{1,0} instead of {1,0})
This is an example of why operator bool is a bad idea in general. If we
were using C++11, we could
On Fri, 11 Oct 2013, Jakub Jelinek wrote:
OT, do you plan to define ATTRIBUTE_RETURNS_NONNULL for
GCC_VERSION = 4009 in ansidecl.h and use it on the various xmalloc
etc. prototypes?
I just read this note in libiberty/concat.c:
This function uses xmalloc() which is expected to be a
On Fri, Oct 11, 2013 at 04:21:04PM +0200, Marc Glisse wrote:
On Fri, 11 Oct 2013, Jakub Jelinek wrote:
OT, do you plan to define ATTRIBUTE_RETURNS_NONNULL for
GCC_VERSION = 4009 in ansidecl.h and use it on the various xmalloc
etc. prototypes?
I just read this note in libiberty/concat.c:
On Fri, Oct 11, 2013 at 10:11:21AM -0400, Jason Merrill wrote:
another
possibility is not to add operator bool () overload that introduces that
ambiguity, but then if (mask something) needs to be replaced with
if ((mask something) != 0) and operator != (int) added.
I guess I slightly prefer
On 10/11/2013 06:28 AM, Paolo Carlini wrote:
The issue is a regression in 4_7/4_8 too, what should we do in those
branches? I'm thinking applying the change to 4_8 too and either not
fixing in 4_7 or just reverting the cp_parser_commit_to_tentative_parse
change which improved the diagnostic for
On 10/11/2013 03:59 PM, Jason Merrill wrote:
On 10/11/2013 06:28 AM, Paolo Carlini wrote:
The issue is a regression in 4_7/4_8 too, what should we do in those
branches? I'm thinking applying the change to 4_8 too and either not
fixing in 4_7 or just reverting the
On 11 October 2013 04:53, Jeff Law l...@redhat.com wrote:
With your change it seems to me that we do a single round of spilling
caller-save setup, then align the stack, then restart. The net result being
we align the stack a lot more often.
Yes, but AFAICT, that should not result in more
Hello,
With -mno-ieee-fp, or when that flag is implied by other flags such as
-ffast-math, i386 backend will use ordered rather than unordered comparisons
(fcom* instead of fucom*, or their SSE counterparts).
Expansion of several rounding builtins currently uses unordered comparisons
always,
On 10/10/13 9:25 AM, Jakub Jelinek wrote:
That looks broken. You leave everything from the last size till the current
one uninitialized, so it would only work if combine_split_insns
always increments max_reg_num () by at most one.
Good catch.
Furthermore, there are macros which should be
On Thu, 10 Oct 2013, Tobias Burnus wrote:
Joseph: Could you have a look at the C part? Thanks!
Your error missing loop condition loop with IVDEP pragma should, I
think, say loop condition in loop. Since the pragma is ivdep
(case-sensitive), it should also say %ivdep% not IVDEP. This
Hi,
The vtbx intrinsics are implemented in assembly without noting
that their tmp1 operand is early-clobber. This can, when the
wind blows the wrong way, result in us making a total mess of
the state of registers.
Fix by marking the required operands as early-clobber.
Regression tested against
Hi,
this is just an ICE on invalid, but it's a 4_8/4_9 Regression, and, more
importantly, we don't emit any sensible error message before ICEing. It
manifests itself as TEMPLATE_PARM_INDEX unhandled by the main
cxx_eval_constant_expression switch, and just adding the code to the
bunch of
Hi,
the attached patch introduces a new target hook
(targetm.calls.special_function_flags) which can be used by a backend
to add specific call properties to builtins.
On S/390 this is used for the *tbegin* and *tabort builtins whose
control flow otherwise is not modelled correctly.
On S/390
This patch improves the handling of invalid ASCII characters in the Go
frontend when they appear in identifiers. Note that Go input is by
definition always UTF-8, so the ASCII-specific assumptions here are OK.
Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu.
Committed to mainline.
On Oct 11, 2013, at 7:21 AM, Marc Glisse marc.gli...@inria.fr wrote:
I just read this note in libiberty/concat.c:
I am afraid that if I add the returns_nonnull attribute to xmalloc in
include/libiberty.h, it will break this supported use, no? Or is this comment
out-of-date?
I think the
That would definitely be a good move. Maybe someone should approve it?
Yes, but I agree that we ought to restrict it to trailing zero-sized arrays.
That would help to allay Jakub's concerns about possible ABI fallouts.
--
Eric Botcazou
On 10/10/13 18:41, Jeff Law wrote:
On 10/10/13 04:00, Andreas Krebbel wrote:
On 09/10/13 21:46, Jeff Law wrote:
On 08/21/13 03:21, Andreas Krebbel wrote:
[RFC] Allow functions calling mcount before prologue to be leaf functions
http://gcc.gnu.org/ml/gcc-patches/2013-04/msg00993.html
I don't
Hi Honza!
I will give it a try.
I understand how the proposed patch would be. That is OK to me.
OK, my undertanding always was, that the decoders works sort of sequentially.
One decoder of bdver1 can do
vector,
double, single,
single, single, signle
where decoder 1 is somehow special but
On 10/09/2013 03:29 AM, Kirill Yukhin wrote:
+(define_insn avx512f_vec_dup_memmode
+ [(set (match_operand:VI48F_512 0 register_operand =x)
+ (vec_duplicate:VI48F_512
+ (match_operand:ssescalarmode 1 nonimmediate_operand xm)))]
+ TARGET_AVX512F
+
Ping.
Thanks,
Martin
On Wed, Sep 11, 2013 at 03:02:02PM +0200, Martin Jambor wrote:
Hi,
edge_within_scc should really be a part of API accompanying
ipa_reduced_postorder just like ipa_get_nodes_in_cycle and so this
patch moves it to ipa-utils.c and gives it the ipa_ prefix.
On 10/11/2013 10:36 AM, Jakub Jelinek wrote:
Since the coding standards say Conversion operators should be
avoided (because they can't be explicit), I think this is the way
to go.
We then violate the coding standard in vec.h:
/* Type to provide NULL values for vecT, A, L. This is used to
This patch from Rémy Oudompheng fixes the Go frontend to accept an
integral float constant as an index into a string. That was already
fixes for arrays, but not for strings. Bootstrapped and ran Go
testsuite on x86_64-unknown-linux-gnu. Committed to mainline. Will
commit to 4.8 branch when the
How does a TEMPLATE_PARM_INDEX get this far? It should have been
detected as dependent before this.
Jason
Hello,
here are some uses for returns_nonnull in libiberty.
Bootstrap+testsuite (default languages) on x86_64-unknown-linux-gnu.
2013-10-11 Marc Glisse marc.gli...@inria.fr
PR tree-optimization/58689
include/
* ansidecl.h (ATTRIBUTE_RETURNS_NONNULL): New macro.
*
With the patch now...
On Fri, 11 Oct 2013, Marc Glisse wrote:
Hello,
here are some uses for returns_nonnull in libiberty.
Bootstrap+testsuite (default languages) on x86_64-unknown-linux-gnu.
2013-10-11 Marc Glisse marc.gli...@inria.fr
PR tree-optimization/58689
include/
*
This needs a testcase (compile with -dA and use scan-assembler; see
other tests in g++.dg/debug/dwarf2).
Jason
It's hard to get a testcase without
http://gcc.gnu.org/viewcvs/gcc?view=revisionrevision=201856 because
none of these *INTERNAL* symbols will be emitted in debug info.
Thanks,
Dehao
On Fri, Oct 11, 2013 at 10:55 AM, Jason Merrill ja...@redhat.com wrote:
This needs a testcase (compile with -dA
This patch to the Go frontend gives a more useful error message for an
invalid call to unsafe.Offsetof on a method value (method values look
like field references, which are valid for unsafe.offsetof).
Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu.
Committed to mainline. Will
On Fri, 2013-10-11 at 19:53 +0200, Marc Glisse wrote:
With the patch now...
[...]
-extern char *concat_copy (char *, const char *, ...) ATTRIBUTE_SENTINEL;
+extern char *concat_copy (char *, const char *, ...) ATTRIBUTE_SENTINEL
ATTRIBUTE_NONNULL(1) ATTRIBUTE_RETURNS_NONNULL;
An aesthetic
On Fri, 11 Oct 2013, David Malcolm wrote:
On Fri, 2013-10-11 at 19:53 +0200, Marc Glisse wrote:
With the patch now...
[...]
-extern char *concat_copy (char *, const char *, ...) ATTRIBUTE_SENTINEL;
+extern char *concat_copy (char *, const char *, ...) ATTRIBUTE_SENTINEL
On 10/09/2013 03:30 AM, Kirill Yukhin wrote:
+;; Return true if OP is either -1 constant or stored in register.
+(define_predicate register_or_constm1_operand
+ (ior (match_operand 0 register_operand)
+ (match_test op == constm1_rtx)))
This won't do the right thing, because you're not
When I patched the compiler for PR 58587 to not set -mvsx-timode automatically
for ISA 2.06 sytems (power7/power8), there was an unexpected problem between
ISA 2.07 quad memory support and not allowing TImode variables in VSX
registers.
The movti insn in the -mno-vsx-timode case, did not have
This patch to the Go compiler makes it give an error if the same name is
used for a method receiver and some parameter to the method.
Bootstrapped and ran Go testsuite on x86_64-unknown-linux-gnu.
Committed to mainline. Will commit to 4.8 branch when the branch
reopens.
Ian
diff -r 09a1bcee0ad3
Your patch changes the rule that the application can override
xmalloc() and get functions that return NULL...
Hi,
On 10/11/2013 07:46 PM, Jason Merrill wrote:
How does a TEMPLATE_PARM_INDEX get this far? It should have been
detected as dependent before this.
Thanks. Interesting. We get there from the convert_nontype_argument call
at line 6453 of pt.c (in convert_template_argument) , which is
On Fri, 11 Oct 2013, DJ Delorie wrote:
Your patch changes the rule that the application can override
xmalloc() and get functions that return NULL...
Only after Jakub and Mike told me, earlier today, that this rule didn't
exist anymore. If it does, we can just forget about this patch.
--
On Fri, Oct 11, 2013 at 02:30:13PM -0400, DJ Delorie wrote:
Your patch changes the rule that the application can override
xmalloc() and get functions that return NULL...
Why would you want to do that? Big part of libiberty itself
wouldn't allow such an implementation.
E.g. concat.c, argv.c,
Hi,
In x32, when a TLS address is in DImode and Pmode is SImode,
copy_addr_to_reg will fail. This patch adds ix86_copy_addr_to_reg
to first copy DImode address into a DImode register and then to generate
SImode SUBREG in this case. Tested on x32. OK for trunk?
Thanks.
H.J.
---
gcc/
Whoops, I didn't notice I had left in old code for lq/stq that I had commented
out. This patch replaces the previous patch, and eliminates the commented out
code. Sorry about that.
[gcc]
2013-10-11 Michael Meissner meiss...@linux.vnet.ibm.com
PR target/58673
*
Why would you want to do that?
I wouldn't, but gcc isn't the only user of libiberty.
Hi,
I checked in this patch to update baseline_symbols.txt for x32.
H.J.
---
Index: ChangeLog
===
--- ChangeLog (revision 203455)
+++ ChangeLog (working copy)
@@ -1,3 +1,7 @@
+2013-10-11 H.J. Lu hongjiu...@intel.com
+
+
On 11 October 2013 17:45, James Greenhalgh james.greenha...@arm.com wrote:
Hi,
The vtbx intrinsics are implemented in assembly without noting
that their tmp1 operand is early-clobber. This can, when the
wind blows the wrong way, result in us making a total mess of
the state of registers.
Hi, all -
We recently built a compiler with a default -fabi-version value that's
different from the 2 that's used as default on trunk. The result of
this was a test failure in g++.dg/ext/altivec-7.C, which compiles a
bunch of empty functions with vector arguments and inspects the
generated
On 10/11/2013 08:29 PM, Paolo Carlini wrote:
Are we maybe failing to bump processing_template_decl somewhere while
processing the specialization?
... I'm finishing testing this, already past g++.dg/dg.exp...
Paolo.
Index: cp/pt.c
On Thu, 2013-10-10 at 01:05 +, Joseph S. Myers wrote:
On Wed, 9 Oct 2013, David Malcolm wrote:
This patch adds an --enable-host-shared option throughout the various
configure/Make machinery for host code, adding -fPIC where appropriate
when enabled.
Please document this in
The problem shown by pr58640 is the more aggressive jump threading is
recording a jump threading opportunity that crosses through two loop
headers.
The updating code is not prepared to update the CFG and loop structures
in that situation. The net result is we have two latch edges for a
On Fri, 11 Oct 2013, DJ Delorie wrote:
Your patch changes the rule that the application can override
xmalloc() and get functions that return NULL...
How about this more limited version? I'll see about following Jakub's
advice to apply the attributes to the other functions only in gcc.
On Fri, 11 Oct 2013, David Malcolm wrote:
On Thu, 2013-10-10 at 01:05 +, Joseph S. Myers wrote:
On Wed, 9 Oct 2013, David Malcolm wrote:
This patch adds an --enable-host-shared option throughout the various
configure/Make machinery for host code, adding -fPIC where appropriate
PR tree-optimization/58689
* ansidecl.h (ATTRIBUTE_RETURNS_NONNULL): New macro.
* libiberty.h (basename, lbasename, dos_lbasename, unix_lbasename,
concat_copy): Mark with attributes nonnull(1) and returns_nonnull.
(concat_copy2, xstrerror): Mark with
On Fri, 2013-10-11 at 20:45 +, Joseph S. Myers wrote:
On Fri, 11 Oct 2013, David Malcolm wrote:
On Thu, 2013-10-10 at 01:05 +, Joseph S. Myers wrote:
On Wed, 9 Oct 2013, David Malcolm wrote:
This patch adds an --enable-host-shared option throughout the various
Hi,
An earlier patch (r203167) changed the default probability for
builtin_expect to 90%.
It does not work properly for the following case:
while (__builin_expect (expr, 1)) { }
W/o builtin_expect, the exit probability is 9% while w/
builtin_expect, the exit probability is 10%.
It seems
The trunk version of this patch is submitted for review.
David: can we have this patch for google/gcc-4_8 branch first?
It tested with regression and google internal benchmarks.
Thanks,
-Rong
2013-10-11 Rong Xu x...@google.com
* predict.c (tree_predict_by_opcode): Bump the
On Oct 11, 2013, at 11:55 AM, Brooks Moses bmo...@google.com wrote:
This patch fixes the failure by adjusting the test to look for the
names using the standard mangling. It passes with all ABI versions;
the compiler always emits the standard symbols, and with versions 1,
2, and 3 it also
Why this 'percent += 4' instead of taking the max?
David
On Fri, Oct 11, 2013 at 2:10 PM, Rong Xu x...@google.com wrote:
The trunk version of this patch is submitted for review.
David: can we have this patch for google/gcc-4_8 branch first?
It tested with regression and google internal
I want to differentiate the cases w/o and w/ builtin.
If I take the max, they will be the same (91%).
-Rong
On Fri, Oct 11, 2013 at 2:26 PM, Xinliang David Li davi...@google.com wrote:
Why this 'percent += 4' instead of taking the max?
David
On Fri, Oct 11, 2013 at 2:10 PM, Rong Xu
Should it be max + some_delta then?
David
On Fri, Oct 11, 2013 at 2:32 PM, Rong Xu x...@google.com wrote:
I want to differentiate the cases w/o and w/ builtin.
If I take the max, they will be the same (91%).
-Rong
On Fri, Oct 11, 2013 at 2:26 PM, Xinliang David Li davi...@google.com wrote:
ok. that makes sense.
On Fri, Oct 11, 2013 at 2:41 PM, Xinliang David Li davi...@google.com wrote:
Should it be max + some_delta then?
David
On Fri, Oct 11, 2013 at 2:32 PM, Rong Xu x...@google.com wrote:
I want to differentiate the cases w/o and w/ builtin.
If I take the max, they will
This patch by Chris Manghane changes the Go frontend to use the backend
interface for function code expressions. Bootstrapped and ran Go
testsuite on x86_64-unknown-linux-gnu. Committed to mainline.
Ian
2013-10-11 Chris Manghane cm...@google.com
* go-gcc.cc
here is the new patch. Note that the hitrate won't change by this
patch for the case of
while (__builtin_expect (exp, 0))
Index: predict.c
===
--- predict.c (revision 203462)
+++ predict.c (working copy)
@@ -1951,11 +1951,42 @@
ok for google branch.
David
On Fri, Oct 11, 2013 at 3:17 PM, Rong Xu x...@google.com wrote:
here is the new patch. Note that the hitrate won't change by this
patch for the case of
while (__builtin_expect (exp, 0))
Index: predict.c
It's hard to get a testcase without
http://gcc.gnu.org/viewcvs/gcc?view=revisionrevision=201856 because
none of these *INTERNAL* symbols will be emitted in debug info.
The original code was in there as a form of assembly-time assertion
that the symbol would never get output. Now that you want
This patch to the Go frontend fixes the handling of hidden methods for
unnamed types. If an interface has hidden methods, we must make the
interface table comdat if it is for an unnamed type. When we create a
stub method for an unnamed type, we don't make it publically visible.
Bootstrapped and
On Fri, 11 Oct 2013, David Malcolm wrote:
Thanks. Presumably the initially posted configure/make patch still
needs review, right?
Yes (by a build system maintainer, probably).
--
Joseph S. Myers
jos...@codesourcery.com
1 - 100 of 108 matches
Mail list logo