Ramana Radhakrishnan writes:
> On 14 September 2011 13:30, Richard Sandiford
> wrote:
>> neon_vget_high and neon_vget_low extract one half of a vector.
>> The patterns look like:
>>
>> (define_insn "neon_vget_highv16qi"
>> [(set (match_operand:V8QI 0 "s_register_operand" "=w")
>> (vec_sel
Just realized ChangeLog needs to be changed as well.
ChangeLog:
* config/i386/i386.c (ix86_using_red_zone): Remove inline.
(TARGET_STACK_USING_RED_ZONE): New.
* config/rs6000/rs6000.c (rs6000_stack_using_red_zone): New.
(TARGET_STACK_USING_RED_ZONE): New.
(
Backport mainline r173371 to ARM/embedded-4_6-branch.
Committed.
2011-09-28 Terry Guo
Backport r173371 from mainline
2011-05-04 Andreas Krebbel
* calls.c (emit_library_call_value_1): Invoke
promote_function_mode hook on libcall arguments.
* explow.c
> > -static inline bool
> > +extern bool
>
> static bool
>
> > ix86_using_red_zone (void)
> > {
> > return TARGET_RED_ZONE && !TARGET_64BIT_MS_ABI;
> > @@ -35958,6 +35958,9 @@ ix86_autovectorize_vector_sizes (void)
> > #define TARGET_STACK_PROTECT_FAIL ix86_stack_protect_fail
> > #endif
> >
This is okay
On Mon, Sep 26, 2011 at 6:14 PM, Guozhi Wei wrote:
> Hi
>
> This patch removes the XPASS of test case 20040204-1.c on arm.
>
> Tested by
> on arm: make check-gcc
> RUNTESTFLAGS="--target_board=arm-sim/thumb/arch=armv7-a
> tree-ssa.exp=20040204-1.c"
> on x86_64: make check-gcc RUNTE
Jan,
The following patch starts a series of patches which improve devirtualization
optimizations in GCC.
This patch builds on ipa-cp.c and ipa-prop.c infrastructure for analyzing
parameters and jump functions and adds basic estimation of devirtualization
benefit from inlining an edge. E.g., i
uple? I kind of like
> it seems simple. It's what's in N2965, so let's start there.
>
> Since it's the first TR2 header, I did the build bits for you.
>
> Also, I cc'd the library list.
>
> best,
> benjamin
> <20110927-4.patch>
Oops, this is the right patch.
-benjamindiff --git a/libstdc++-v3/ChangeLog b/libstdc++-v3/ChangeLog
index a59a0b6..e1176ee 100644
--- a/libstdc++-v3/ChangeLog
+++ b/libstdc++-v3/ChangeLog
@@ -1,5 +1,14 @@
2011-09-27 Benjamin Kosnik
+ * doc/Makefile.am: Add tr2 support.
+ * doc/Makefile.in
> This patch consists intrinsics to properly create the bases and
> direct_bases of a class in the correct order (including multiple
> nested ambiguous virtual and non-virtual classes) for N2965
> (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2009/n2965.html).
> This allows you to create
> -Original Message-
> From: Richard Guenther [mailto:richard.guent...@gmail.com]
> Sent: Tuesday, September 27, 2011 3:41 PM
> To: Jiangning Liu
> Cc: Andrew Pinski; gcc-patches@gcc.gnu.org
> Subject: Re: [PATCH] Fix stack red zone bug (PR38644)
>
> On Tue, Sep 27, 2011 at 5:32 AM, Jian
On 9/27/2011 4:24 PM, Joseph S. Myers wrote:
On Tue, 27 Sep 2011, Andi Kleen wrote:
if (!flag_random_seed)
{
+ /* Try urandom first. Time of day is too likely to collide. */
+
+ int fd = open("/dev/urandom", O_RDONLY);
+ read(fd,&random_seed, sizeof(random_seed));
+
On Tue, 27 Sep 2011, Basile Starynkevitch wrote:
> Gerald, if the patch is OK, could you please apply it?
> I am very uneasy about CVS for the GCC website.
Sure, happy to Basile. I made some minor changes ( instead of
, instead of , this kind) and committed the updated
patch below.
Thanks,
Ge
This new test works already, but we had no coverage for inline assembly.
Committed to branch.
* g++.dg/pph/x0asm1.h: New.
* g++.dg/pph/x1asm1.cc: New.
diff --git a/gcc/testsuite/g++.dg/pph/x0asm1.h
b/gcc/testsuite/g++.dg/pph/x0asm1.h
new file mode 100644
index 000..82e191f
> Hi,
> I build Mozilla without debug info with slim LTO and non-LTO.
> slim LTO build is:
>
> real14m26.882s
> user78m47.547s
> sys 6m58.870s
> jh@evans:/abuild/jh/build-mozilla-new14> du -s .
> 1056120 .
> jh@evans:/abuild/jh/build-mozilla-new14> size toolkit/library/libxul.so
>t
On Tue, 2011-09-27 at 22:36 +0900, Kaz Kojima wrote:
> Thanks for this work! A few minor style issues:
>
Thanks for checking and sorry for the trouble.
The attached patch and ChangeLog below should fix it.
I have also added a test case for SI mode abs.
Cheers,
Oleg
ChangeLog:
2011-09-28 O
OK for google/gcc-4_6 and google/main branches.
-Easwaran
>
> On Tue, Sep 27, 2011 at 4:07 PM, Sriraman Tallam wrote:
>>
>> Made all the changes. Attaching new patch of updated files.
>>
>> On Tue, Sep 27, 2011 at 11:26 AM, Easwaran Raman wrote:
>> >
>> > +static inline hashval_t
>> > +edge_hash_
On Tue, 27 Sep 2011, Andi Kleen wrote:
>if (!flag_random_seed)
> {
> + /* Try urandom first. Time of day is too likely to collide. */
> +
> + int fd = open("/dev/urandom", O_RDONLY);
> + read(fd, &random_seed, sizeof(random_seed));
> + close(fd);
There appear to be no
This patch fixes a bug in the parser which cause an internal compiler
error when copying attributes from cloned methods. The bug occurs
when a class has both an annotated constructor and a template method.
Bootstrapped and passed gcc regression testsuite on
x86_64-unknown-linux-gnu. Okay for goo
Made all the changes. Attaching new patch of updated files.
On Tue, Sep 27, 2011 at 11:26 AM, Easwaran Raman wrote:
>
> +static inline hashval_t
> +edge_hash_function (unsigned int id1, unsigned int id2)
> +{
> + /* If the number of functions is less than 1000, this gives a unique value
> +
Oleg Endo wrote:
> The gcc.target/sh/mfmovd.c test was ran even for targets that can't
> support it at all. While I was at it, I've also noticed that for some
> reason SH4A specific tests were ran for other targets and reported a
> fake pass, although the tests are actually unsupported.
The patc
> > I believe the assembler code only works with HJ's version of binutils
> > and his BFD ld (not gold) currently. HJ can you supply the minimal version?
> > Non assembler/LTO should work always.
>
> The Linux binutils 2.21.51.0.3 is the first Linux binutils which
> supports "ld -r" on mixed IR/no
Hi,
I build Mozilla without debug info with slim LTO and non-LTO.
slim LTO build is:
real14m26.882s
user78m47.547s
sys 6m58.870s
jh@evans:/abuild/jh/build-mozilla-new14> du -s .
1056120 .
jh@evans:/abuild/jh/build-mozilla-new14> size toolkit/library/libxul.so
textdata bss
On Mon, Sep 26, 2011 at 7:59 AM, Andi Kleen wrote:
>> > ld -r is now supported with LTO
>>
>> Thanks, forgot about this one. This also needs support at linker side,
>> right?
>
> Only if you include assembler or non LTO code.
> Without that it should work with any linker.
>
>> Do you know minima
From: Andi Kleen
I had some trouble with random build failures in a large LTO project
and it turned out to be random seed collisions in a highly parallel build
(thanks to Honza for suggesting that)
There were multiple problems:
- The way to generate the random seed is pure (milli seconds plus pi
From: Andi Kleen
When available use /dev/urandom to get the random seem. This will lower the
probability
of collisions.
On other systems it will fallback to the old methods.
Passes bootstrap + testsuite on x86_64. Ok?
gcc/:
* 2011-09-26 Andi Kleen
* toplev.c (init_local_tick): Tr
On 09/15/11 00:51, Richard Henderson wrote:
> On 09/13/2011 08:36 AM, Bernd Schmidt wrote:
>> On 09/13/11 15:05, Richard Sandiford wrote:
>>> It just feels like checking for trap_if or turning off cross-jumping
>>> are working around problems in the representation of shrink-wrapped
>>> functions.
Installed.
Gerald
Index: extensions.html
===
RCS file: /cvs/gcc/wwwdocs/htdocs/extensions.html,v
retrieving revision 1.49
diff -u -r1.49 extensions.html
--- extensions.html 17 Sep 2011 22:24:10 - 1.49
+++ extensions.html
2011/9/27 Richard Henderson :
> On 09/27/2011 12:58 PM, Kai Tietz wrote:
>> +#ifdef HAVE_GAS_SECTION_EXCLUDE
>> + if ((flags & SECTION_EXCLUDE) != 0)
>> + *f++ = 'e';
>> +#endif
>
> If 'e' is not supported, I wonder if we should emit 'n'?
>
> Otherwise ok.
>
>
> r~
This sounds fair. At least
On 09/27/2011 12:58 PM, Kai Tietz wrote:
> +#ifdef HAVE_GAS_SECTION_EXCLUDE
> + if ((flags & SECTION_EXCLUDE) != 0)
> +*f++ = 'e';
> +#endif
If 'e' is not supported, I wonder if we should emit 'n'?
Otherwise ok.
r~
Hi,
this patch enables same feature as in thread "[google] Add
SECTION_EXCLUDE flag and exclude .gnu.callgraph sections
(issue5126041)" for pe-coff x86 and x64 targets.
ChangeLog
2011-09-27 Kai Tietz
* configure.ac: Add test for new section attribute
specifier "e" via define
On Sep 20, 2011, Jakub Jelinek wrote:
> For NOTE_INSN_CALL_ARG_LOCATION, the locations aren't location lists, but
> a single location at the point of the call. They are independent of
> all other locations, so any kind of caching only decreases the chance
> that a suitable location
With the pro
Richard Henderson schrieb:
On 09/27/2011 12:56 AM, Georg-Johann Lay wrote:
* config/avr/avr-protos.h (avr_out_addto_sp): New prototype.
* config/avr/avr.c (avr_out_addto_sp): New function.
(adjust_insn_length): Handle ADJUST_LEN_ADDTO_SP.
* config/avr/avr.md (adj
I committed the patch to google/gcc-4_6 branch.
Thanks,
-Sri.
* output.h (SECTION_EXCLUDE): New flag for exclude sections.
* varasm.c (default_elf_asm_named_section): Add "e" to section flags
marked as SECTION_EXCLUDE.
* final.c (rest_of_handle_final): Exclude .gn
On 11-09-27 14:21 , Cary Coutant wrote:
OK to cherry-pick r179209 from gcc-4_6-branch to google/gcc-4_6?
http://gcc.gnu.org/ml/gcc-cvs/2011-09/msg00828.html
Sure. Assuming you've validated it.
Diego.
On Tue, 27 Sep 2011, Jan Hubicka wrote:
> thanks for corrections. This is what I've comitted now.
Here is another small update I just crafted, that makes a stylistic
change here and there, introduces LTO as an abbreviation, adds some
missing articles and spaces,...
Installed.
Gerald
Index: cha
+static inline hashval_t
+edge_hash_function (unsigned int id1, unsigned int id2)
+{
+ /* If the number of functions is less than 1000, this gives a unique value
+ for every function id combination. */
+ const int MULTIPLIER = 1000;
+ return id1* MULTIPLIER + id2;
Change to id1 << 16 | id2
OK to cherry-pick r179209 from gcc-4_6-branch to google/gcc-4_6?
http://gcc.gnu.org/ml/gcc-cvs/2011-09/msg00828.html
-cary
include/ChangeLog:
PR 40831
* demangle.h (enum demangle_component_type): Add
DEMANGLE_COMPONENT_CLONE.
libiberty/ChangeLog:
PR 40831
On 09/23/2011 02:10 AM, Paolo Bonzini wrote:
> On 09/23/2011 10:56 AM, Paolo Bonzini wrote:
>> Also, I am curious about one thing: while this is of course a very
>> pragmatic solution, you could also convert AVR to get rid of CC0, do
>> this at expansion time, and get split-wide-types to work as in
Dear Jakub,
This is, of course, OK for trunk. In fact, I would say that it verges
on obvious.
Thanks for taking care of it.
Paul
On Fri, Sep 23, 2011 at 4:44 PM, Jakub Jelinek wrote:
> Hi!
>
> I've noticed with the
> 2009-05-29 Eric Botcazou
>
> * tree-ssa-loop-ivopts.c (strip_offse
> Index: final.c
> ===
> --- final.c (revision 179104)
> +++ final.c (working copy)
> @@ -4428,7 +4428,7 @@ rest_of_handle_final (void)
> && cgraph_node (current_function_decl) != NULL
> && (cgraph_node (current_fun
Submitted the SECTION_EXCLUDE part to trunk.
Thanks,
-Sri.
On Tue, Sep 27, 2011 at 10:51 AM, Kai Tietz wrote:
>
> 2011/9/27 Richard Henderson :
> > On 09/23/2011 04:45 PM, Sriraman Tallam wrote:
> >> I also want the SECTION_EXCLUDE part alone to be considered for trunk.
> >
> > This is ok for tr
Submitted to trunk after approval from Richard :
http://gcc.gnu.org/ml/gcc-patches/2011-09/msg01747.html
Thanks,
-Sri.
On Mon, Sep 26, 2011 at 2:09 PM, Sriraman Tallam wrote:
> Hi,
>
> Here is a simple patch to add SECTION_EXCLUDE to the list of
> section flag macros. Is this ok for trunk?
>
>
2011/9/27 Richard Henderson :
> On 09/23/2011 04:45 PM, Sriraman Tallam wrote:
>> I also want the SECTION_EXCLUDE part alone to be considered for trunk.
>
> This is ok for trunk.
>
> Kai, is there a similar flag for pe-coff? I.e. is there something
> reasonable that we can add to i386_pe_asm_named
>2. Abandon using a header file at all. Instead use a configure test
> to see if we are using an assembler that supports textual names in a
> .eabi_attribute directive and if so use the names rather than the numbers.
I'm not sure this is mutually exclusive with using a header file, but using
> On 11-09-27 13:00 , Andi Kleen wrote:
>
>> The other obvious way to make it faster would be to skip the assembler
>> for slim LTO and just output the data directly. There's not much need
>> for it with LTO.
>
> Agreed. That's what we do with pph images.
>
>> But that's a larger project.
>
> Inde
On 09/23/2011 04:45 PM, Sriraman Tallam wrote:
> I also want the SECTION_EXCLUDE part alone to be considered for trunk.
This is ok for trunk.
Kai, is there a similar flag for pe-coff? I.e. is there something
reasonable that we can add to i386_pe_asm_named_section?
r~
Hi,
this patch updates testsuite to cover both fat and slim LTO when linker plugin
is used and also both linker plugin and collect2 paths. I didn't wanted to
slow down testing too much so I just distributes the flags across existing runs
with aim to maximize the coverage of testing matrix that is
On 11-09-12 10:50 , Gabriel Charette wrote:
Oops forgot to reply all the first time...
On Fri, Sep 9, 2011 at 4:54 PM, Diego Novillo wrote:
This was not causing any failures, but it is pretty wasteful to read
the same PPH more than once.
We cannot just skip them, however. We need to read the
On 11-09-27 13:00 , Andi Kleen wrote:
The other obvious way to make it faster would be to skip the assembler
for slim LTO and just output the data directly. There's not much need
for it with LTO.
Agreed. That's what we do with pph images.
But that's a larger project.
Indeed.
Diego.
This finishes removing constants and builtins out of the cache.
This time in a slightly more elegant way.
The patch introduces a new version of pph_out_start_record exclusively
for trees (pph_out_start_tree_record). If the tree is not cacheable
then it emits a PPH_RECORD_START_NO_CACHE record to
Diego Novillo writes:
> On Mon, Sep 26, 2011 at 14:42, Jan Hubicka wrote:
>> Hi,
>> this patch implements slim LTO. It is updated version of Andi's patch. It is
>> done by terminating IPA optimization after analyzing and outputting LTO and
>> in
>> compile_file skipping stuff that outputs asse
On Tue, 27 Sep 2011, Jan Hubicka wrote:
> thanks for corrections. This is what I've comitted now.
And here are some markup fixes on top, that address complaints by
the validator.
Gerald
Index: changes.html
===
RCS file: /cvs/gcc/w
The second patch re-factors pph_cache_get to remove the cache
selection logic into a separate function. I needed this to implement
pph_cache_sign in terms of pph_cache_get. Before this, pph_cache_get
tried to decide what cache to use. After this patch, a separate
pph_cache_select() function mak
Hi Richard,
You should only emit the comment for -dA, I think.
Really ? I thought that it was kind of nice to have the comments in the
default output.
If I go with -dA as the trigger then I assume that it is also OK to add
the comments if --verbose-asm is enabled, yes ?
Cheers
Nick
Hi Ramana,
+/* Get the definitions of the ARM EABI Attribute tag values. */
+#define BFD_ARCH_SIZE
+#include "elf/arm.h"
Defining BFD_ARCH_SIZE appears to be a bit of a hack. I would also
ifdef this inclusion on TARGET_AAPCS since we shouldn't really be
caring about object attributes for non
This series of patches prepare support for detecting state mutation
when reading tree nodes from different PPH images
(http://gcc.gnu.org/ml/gcc-patches/2011-09/msg00658.html).
This first patch introduces a new data structure to represent entries
in the pickle cache (pph_cache_entry). The data as
On Tue, 27 Sep 2011, Kirill Yukhin wrote:
> So, if you are ok, let's wait a couple of days for maintainers inputs.
Yep, looks good. Unless you hear to the contrary from one of the
x86 maintainers, I suggest you go ahead and commit in two days.
Gerald
On 27 September 2011 16:55, Paul Brook wrote:
>> Hi Nick,
>>
>> On 27 September 2011 15:38, Nick Clifton wrote:
>> > Any comments or objections to this patch ? If not, I will apply it
>> > next week.
>>
>> I've got a few objections to this patch as it stands today while I
>> don't object to th
> Hi Nick,
>
> On 27 September 2011 15:38, Nick Clifton wrote:
> > Any comments or objections to this patch ? If not, I will apply it
> > next week.
>
> I've got a few objections to this patch as it stands today while I
> don't object to the motivation for it.
>
> > +/* Get the definitions o
On 09/27/2011 12:56 AM, Georg-Johann Lay wrote:
> * config/avr/avr-protos.h (avr_out_addto_sp): New prototype.
> * config/avr/avr.c (avr_out_addto_sp): New function.
> (adjust_insn_length): Handle ADJUST_LEN_ADDTO_SP.
> * config/avr/avr.md (adjust_len): Add "addto_sp".
>
On 09/27/2011 07:38 AM, Nick Clifton wrote:
> +#define EMIT_EABI_ATTRIBUTE(name,val)
> \
> + do \
> +{
> \
> + as
Hi Nick,
On 27 September 2011 15:38, Nick Clifton wrote:
> Any comments or objections to this patch ? If not, I will apply it
> next week.
I've got a few objections to this patch as it stands today while I
don't object to the motivation for it.
> +/* Get the definitions of the ARM EABI Attri
Hi,
On Tue, Sep 27, 2011 at 03:26:03PM +0100, Richard Sandiford wrote:
> This patch fixes a miscompilation of stage1 c-parser.o in an ARM bootstrap.
> When an access to an enum field was SRAed, a component ref used the type
> of the integer temporary variable instead of the type of the enum.
> It
> > This caused:
> >
> > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49179
> >
> >
>
> This also caused:
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49091
>
Hi,
the problem is sign overflow in time computation. Time should be capped by
MAX_TIME
and we compute MAX_TIME * INLINE_SIZE_SCALE *
On Tue, Sep 27, 2011 at 4:26 PM, Richard Sandiford
wrote:
> This patch fixes a miscompilation of stage1 c-parser.o in an ARM bootstrap.
> When an access to an enum field was SRAed, a component ref used the type
> of the integer temporary variable instead of the type of the enum.
> It therefore did
Hi,
it has turned out that I was bit too optimistic in my expectations that
improved ipa-inline-analysis on scalar parameters will be enough to solve needs
of tramp3d. In tramp3d as in many of C++ programs most code is passed by
reference and while I have some patches in this area, the tracking is
Hi Richard, Hi Paul, Hi Ramana,
I am planning to check in the patch below to add comments to the
.eabi_attribute directives emitted by the ARM GCC backend. So for
example a simple hello world source file might compile to something
like this:
.cpu arm7tdmi
.fpu softvfp
This patch fixes a miscompilation of stage1 c-parser.o in an ARM bootstrap.
When an access to an enum field was SRAed, a component ref used the type
of the integer temporary variable instead of the type of the enum.
It therefore didn't alias other accesses to the same structure,
and was scheduled a
Hello,
> Ok, this does have two anti-dep edges. But still, only a single true
> dependence(?) ... can you see why?
The intra edge [3681 -(T,2,0)-> 3682] was created by haifa-sched and I guess
that because both of the expected true-dep edges (one for the target
and one for the address) are identic
Gerald, Andi,
thanks for corrections. This is what I've comitted now.
Index: changes.html
===
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-4.7/changes.html,v
retrieving revision 1.33
diff -u -r1.33 changes.html
--- changes.html25 Se
Oleg Endo wrote:
> The attached patch improves the generated code for integer abs
> operations on SH, in particular SH4. There was already some code that
> was supposed to utilize SH's conditional execution it but it was never
> triggered, because the standard branch-free abs code was generated at
Rainer Orth writes:
> Solaris 8 and 9 suffer from the same problem. The following patch
> allowed the bootstrap to complete. An IRIX bootstrap is currently
> running, but will take some time to complete.
>
> Rainer
>
>
> 2011-09-23 Rainer Orth
>
> * mksysinfo.sh: Provide TIOCSCTT
> On Mon, Sep 26, 2011 at 14:42, Jan Hubicka wrote:
> > Hi,
> > this patch implements slim LTO. It is updated version of Andi's patch. It
> > is
> > done by terminating IPA optimization after analyzing and outputting LTO and
> > in
> > compile_file skipping stuff that outputs assembly.
>
> Nic
Hi,
Gerald, thanks for fixing my "excellent" English :)
Here is updated patch:
Index: htdocs/gcc-4.7/changes.html
===
RCS file: /cvs/gcc/wwwdocs/htdocs/gcc-4.7/changes.html,v
retrieving revision 1.28
diff -p -r1.28 changes.html
*** ht
On Mon, Sep 26, 2011 at 14:42, Jan Hubicka wrote:
> Hi,
> this patch implements slim LTO. It is updated version of Andi's patch. It is
> done by terminating IPA optimization after analyzing and outputting LTO and in
> compile_file skipping stuff that outputs assembly.
Nice! Have you measured ob
An --enable-werror cross build (but apparently not a normal C++ bootstrap)
showed up a signed/unsigned warning in ipa-inline-analysis.c.
Tested on arm-linux-gnueabi. Applied as obvious.
Richard
gcc/
* ipa-inline-analysis.c (predicate_probability): Avoid comparison
between signe
On Tue, Sep 27, 2011 at 12:30 AM, Jeff Law wrote:
> -BEGIN PGP SIGNED MESSAGE-
> Hash: SHA1
>
> On 08/18/11 15:59, Richard Henderson wrote:
>> On 08/17/2011 12:21 AM, Richard Guenther wrote:
>>> The patch itself looks sensible, though I am surprised ifcvt
>>> doesn't run in cfglayout mode
On Mon, Sep 26, 2011 at 8:42 PM, Jan Hubicka wrote:
> Hi,
> this patch implements slim LTO. It is updated version of Andi's patch. It is
> done by terminating IPA optimization after analyzing and outputting LTO and in
> compile_file skipping stuff that outputs assembly.
>
> After some considerati
On Tue, Sep 27, 2011 at 8:32 AM, Ira Rosen wrote:
> On 26 September 2011 17:12, Richard Guenther
> wrote:
>> On Mon, Sep 19, 2011 at 9:54 AM, Ira Rosen wrote:
>>> Hi,
>>>
>>> This patch adds a support of widening shift left. The following
>>> pattern is detected:
>>>
>>> type a_t;
>>> TYPE a_T,
OK.
Jason
Ping:
http://gcc.gnu.org/ml/gcc-patches/2011-09/msg00811.html
Bernd
Hi,
This patch adds a support of multiple types (in the same SLP instance)
in basic block vectorization.
Bootstrapped and tested on powerpc64-suse-linux.
Applied to trunk.
Ira
ChangeLog:
* tree-vect-stmts.c (vectorizable_type_demotion): Handle basic block
vectorization.
On Tue, Sep 27, 2011 at 9:47 AM, Revital Eres wrote:
> Hello,
>
>> ok, so if we have an auto-inc'ing insn which defines (auto-inc's) an
>> addr register and another (say, result) register, we want to allow the
>> result register to have life ranges in excess of ii (by eliminating
>> anti-dep edges
On 09/25/11 19:27, Richard Sandiford wrote:
> The store itself can still be a single SDC1 instruction, so we should
> generate the same notes regardless of mips_split_64bit_move_p.
>
> If that's right, then how about the patch below (tested on
> mips64-linux-gnu, but without the shrink-wrap patche
On Tue, Sep 27, 2011 at 10:47 AM, Revital Eres wrote:
> Hello,
>
>> This
>> + /* Skip instructions that do not set a register. */
>> + if (set && !REG_P (SET_DEST (set)))
>> + continue;
>> is ok. Can you also prevent !set insns from having reg_moves? (To be updated
>> once auto_i
Hello All
Gerald suggested me to submit a patch to the http://gcc.gnu.org/extensions.html
page about GCC MELT.
I am attaching it.
Gerald, if the patch is OK, could you please apply it?
I am very uneasy about CVS for the GCC website.
Regards.
--
Basile STARYNKEVITCH http://starynkevi
Hi,
submitter asks us to be more accurate in some error messages about
naming 'struct' vs 'class'.
Turns out - IMHO the issue is a bit nitpicking - that we do have in
place machinery for this, already used by cp_parser_class_head, which
sets CLASSTYPE_DECLARED_CLASS appropriately. I'm simply
I noticed we dump the original statement twice since tuples.
Fixed as obvious.
Richard.
2011-09-27 Richard Guenther
* tree-object-size.c (compute_object_sizes): Fix dumping of
folded statement.
Index: gcc/tree-object-size.c
==
In Alfa mode for formal verification, the default is now to interpret compiler
permissions like GNAT does, with an option to have a stricter standard-only
interpretation. This is currently used for sizes of implicit base types of
integer types.
Tested on x86_64-pc-linux-gnu, committed on trunk
20
This turns the call to Directly_Designated_Type into a call to Designated_Type
in the recently applied change aimed at improving the handling of the indirect
array case. It is OK to look at the actual designated type here.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-27 Eric Botcaz
If the prefix is a single protected object and the selector is a discriminant
or an entry family, this is a non-overloaded candidate interpretation, and
possible primitive operations of the type must not be examined.
The following must compile and execute quietly in Ada 2005 mode:
---
with FooBar
This fixes a few things in Safe_Prefixed_Reference that are hard to grasp by
the reader and restores the original semantics of the function with regard to
the Variable_Ref parameter.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-27 Eric Botcazou
* exp_util.adb (Safe_Prefix
According to some experiments, the guard area size is 16KB on Windows 64.
(Stack size must be a multiple of the allocation granularity - 64KB - to
get better guard size estimation).
No OS independant testcase.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-27 Tristan Gingold
This new implementation do not introduced a new API in s-taprop. It is
then simpler and will be nicer with cross platforms. It will also be
easier to introduce new locking policies if needed.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-27 Pascal Obry
* s-osinte-hpux.ads,
Continuation of previous changes:
Protected object are now using rwlock. This is to enable multiple
readers (functions) to enter the PO at the same time. The rwlock
is based on GNU/Linux pthread implementation.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-27 Pascal Obry
*
Both types and corresponding routines are defined for GNU/Linux.
Those definitions are made aliases for pthread_mutex_t and
pthread_mutexattr_t on platforms where read/write lock is
not supported by the GNAT runtime.
Still work in progress, more changes to follow.
Tested on x86_64-pc-linux-gnu,
Ian Lance Taylor writes:
>> * IRIX will be worse: while it has TIOCNOTTY, it completely lacks
>> TIOCSCTTY.
>>
>> Suggestions?
>
> For a missing TIOCSCTTY I think the simplest solution will be to set it
> in syscall_irix.go to some innocuous value if there is one. E.g.,
> TIOCNXCL if Irix supp
The Concurrent_Readers_Locking policy is added.
This policy is needed to implement concurrent readers on PO
functions. this is a work in progress, more changes to enable this
capability will follow
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-27 Pascal Obry
* rtsfind.ads:
This fixes PR50363.
Bootstrapped and tested on x86_64-unknown-linux-gnu.
Richard.
2011-09-27 Richard Guenther
PR tree-optimization/50363
* tree-ssa-pre.c (create_expression_by_pieces): Handle
pointer conversions in POINTER_PLUS_EXPRs properly.
* gcc.dg/tortu
Dinar Temirbulatov writes:
> I prepared the new version of patch, re-tested new version on again on
> mipsel-unknown-linux-gnu with no new regressions.
Thanks, looks good. I'll apply once the copyright assignment comes through.
Richard
1 - 100 of 110 matches
Mail list logo
