From: Andi Kleen
This implements the freeing back of large chunks in the ggc madvise path
Richard Guenther asked for. This way on systems with limited
address space malloc() and other allocators still have
a chance to get back at some of the memory ggc freed. The
fragmented pages are still just
From: Andi Kleen
Jakub had some concerns about the performance of page alignments in
ggc-page, which use a hardware division instructions currently.
This patch changes them all to use a new PAGE_ALIGN macro, which
exploits that pages are a power of two.
2011-10-21 Andi Kleen
* ggc-pa
This version addresses all earlier review comments. Passes bootstrap
and testing on x86-64. Ok?
-Andi
From: Andi Kleen
There were some concerns that the earlier munmap patch could lead
to address space being freed that cannot be allocated again
by ggc due to fragmentation. This patch adds a fragmentation
fallback to solve this: when a GGC_QUIRE_SIZE sized allocation fails,
try again with a page s
From: Andi Kleen
This one place in ggc forgot to round page_entry->bytes to the
next page boundary, which lead to all the heuristics in freeing to
check for continuous memory failing. Round here too, like all other
allocators already do. The memory consumed should be the same
for MMAP because the
Marc Glisse writes:
>> _Z1aIFYviEEvPT_
>> void a(void (*)(int) extern "C")
Does g++ ever actually generate a symbol name like this? Is it worth
worrying about?
Personally if we are going to print this at all I think the extern "C"
should be in square brackets. E.g.,
void a(void (*)(int) [ex
On 10/21/2011 06:22 PM, Mike Stump wrote:
Is there a plan on what the default language will be? C is stuck in 1980, and
will be forever. Personally, I'd rather have g++ move up language standards by
default, I'd rather make the move before 2023.
I think it makes sense to change the default
Committed. Thanks!
On Fri, Oct 21, 2011 at 5:22 PM, Mike Stump wrote:
> On Oct 21, 2011, at 12:52 PM, Jason Merrill wrote:
>> On 10/21/2011 03:11 PM, Marc Glisse wrote:
>>> Note that at least clang now defines __cplusplus to its new C++11 value
>>> (in experimental C++0X mode only). Apparently they switched around l
On Fri, Oct 21, 2011 at 12:52 PM, Paolo Carlini
wrote:
> Hi again,
>>>
>>> Another acceptable fix is to
>>> -- leave the current diagnostic as is if -fms-extensions
>>> -- suggest '()' is member function
>>> -- otherwise suggest '&'.
>>
>> Thanks for your help Gaby, in particular about the M
This is the right patch that I bootstrapped and tested.
* line-map.c (linemap_macro_map_lookup): Make sure to always pick
the map with the highest MAP_START_LOCATION.
---
libcpp/line-map.c |9 +++--
1 files changed, 7 insertions(+), 2 deletions(-)
diff --git a/libcpp/line
Hello,
I noticed that linemap_macro_map_lookup can yield the wrong map when
at the end of the lookup, we end up with a pair of maps whose indexes
are (mn,mx) with mx-mn == 1. In that case, if the map we want is mn,
we'd wrongly yield mx.
Fixed thus, bootstrapped and tested on x86_64-unknown-linu
Hello,
A virtual location can now "wrap" a reserved location. By reserved
location, I mean a location that is less than RESERVED_LOCATION_COUNT
(i.e UNKNOWN_LOCATION or BUILTINS_LOCATION).
When such a virtual location is passed to linemap_resolve_location,
that function returns the bare reserved
Hello,
The following two patches fixes virtual locations handling to progress
toward bootstrapping the compiler with CFLAGS=-ftrack-macro-expansion
CXXFLAGS=-ftrack-macro-expansion
After these two patches I'll still need to update many c++ test cases
and probably exercise some dg-prune-fu to cope
I've committed the attached patch to address a
regression caused by my recent change to argument
checking in ISHFT[C]. Both my regression test
and Dominiq's regression test passed. Sorry
about the breakage.
2011-10-20 Steven G. Kargl
PR fortran/50821
* check.c (gfc_check_ish
In:
r37370 | bernds | 2000-11-10 09:10:29 -0800 (Fri, 10 Nov 2000) | 2 lines
Several fixes to make reload handle POST_MODIFY correctly.
We see:
-regno_clobbered_p (regno, insn, mode)
+regno_clobbered_p (regno, insn, mode, sets)
unsigned int regno;
rtx insn;
enum machine_mode m
On Oct 21, 2011, at 12:52 PM, Jason Merrill wrote:
> On 10/21/2011 03:11 PM, Marc Glisse wrote:
>> Note that at least clang now defines __cplusplus to its new C++11 value
>> (in experimental C++0X mode only). Apparently they switched around last
>> June and say they are not the only ones. So if you
The calling convention with i?86 (32bits) is wrong for _ITM_free (and
_ITM_malloc/_ITM_calloc).
Here an example:
0x08049fd2 <+251>: mov-0x14(%ebp),%eax
0x08049fd5 <+254>: call 0x804b150 <_ITM_free(void*)>
Dump of assembler code for function _ITM_free(void*):
0x0804b150 <+0>:
I think we're down to minor cosmetic issues:
On 10/21/2011 03:55 PM, Tom Tromey wrote:
There are a few spots like this that are missing a space before an open
paren.
+ if (DECL_LANGUAGE(decl) == lang_c)
Another one.
- if (warn_cxx0x_compat
- && C_RID_CODE (token
On 10/03/11 17:57, Paul Brook wrote:
> 2011-10-03 Paul Brook
>
> * config/c6x/c6x.c (c6x_asm_emit_except_personality,
> c6x_asm_init_sections): New functions.
> (TARGET_ASM_EMIT_EXCEPT_PERSONALITY, TARGET_ASM_INIT_SECTIONS):
> Define.
This was also never applied. I've c
On 10/03/11 17:52, Paul Brook wrote:
> The C6XABI defined persoality routines ID 3 and 4 use a single 24-bit block
> word of unwinding data. Patch below makes sure this is preserved, rather
> than
> treating it as a set of unwinding opcode bytes.
>
> I seem to have lost this bit of code when I
On Fri, 2011-10-21 at 21:25 +0200, Jakub Jelinek wrote:
> My guess would be
> http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=180194
> DEBUG_EXPR certainly shouldn't make it through into mem_loc_descriptor,
> var-tracking is supposed to resolve that.
>
> Jakub
You are right, the bug start
On Fri, 21 Oct 2011, Aldy Hernandez wrote:
> > > X32 uses x86-64 instruction set with 32bit pointers. It has the same
> > > atomic support as x86-64 and has atomic support for int128.
> >
> > Oh, you aren't talking about 32 bit, but a 32 bit abi on a 64 bit machine.
>
> Thanks for pointing this
> "Ed" == Ed Smith-Rowland <3dw...@verizon.net> writes:
Ed> + /* Nonzero for the 2011 C++ Standard. */
Ed> + unsigned char cxx11;
I think it would be better if the new field name reflected its purpose,
so something like "user_literals".
Ed> + if (ISIDST(*cur))
Ed> + {
Ed> + typ
Julian Brown writes:
> gcc/
> * config/m68k/m68k.c (notice_update_cc): Tighten condition for
> setting CC_REVERSED for FP comparisons.
Ok.
Andreas.
--
Andreas Schwab, sch...@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756 01D3 44D5 214B 8276 4ED5
"And now for someth
On 10/21/2011 03:11 PM, Marc Glisse wrote:
Note that at least clang now defines __cplusplus to its new C++11 value
(in experimental C++0X mode only). Apparently they switched around last
June and say they are not the only ones. So if you want to follow their
lead...
Hmm, between that and the fa
On Fri, 21 Oct 2011 21:11:16 +0200, Jakub Jelinek wrote:
> Well, you calculate the sizes multiple times anyway, so I don't see why you
> during the size calculations you couldn't start with DW_FORM_ref_udata
> as first guess and compute on the side also total sizes of those
> DW_FORM_ref_udata byte
On 21 October 2011 09:15, Jonathan Wakely wrote:
> On 21 October 2011 00:43, Jonathan Wakely wrote:
>> This patch should enable macosx support for and partial
>> support for , by defining _GLIBCXX_HAS_GTHREADS on POSIX
>> systems without the _POSIX_TIMEOUTS option, and only disabling the
>> types
> Date: Fri, 21 Oct 2011 20:34:05 +0200
> From: Jan Hubicka
> I guess we could make ipa-dump/rtl-dump/tree-dump scanning to disable fat lto
> and introduce variants intended to scan late tree dumps and ipa execution
> dumps...
Ok, sounds like a plan. Are there any such
scan-tests-with-late-thi
On 10/21/2011 11:42 AM, Jakub Jelinek wrote:
> PR target/50813
> * config/i386/i386.c (expand_vec_perm_even_odd_1): Handle
> V4DImode and V8SImode for !TARGET_AVX2.
>
> * gcc.dg/torture/vshuf-32.inc: Add broadcast permutation
> from element other than first and revers
On Fri, Oct 21, 2011 at 09:54:56AM -0700, Steve Ellcey wrote:
> FYI: I am seeing this same ICE on the hppa64-hp-hpux11.11 bootstrap.
>
> (debug_expr:DI D#49)
> /ctires/gcc/nightly/src/trunk/gcc/cselib.c: In function 'void
> cselib_record_sets(rtx)':
> /ctires/gcc/nightly/src/trunk/gcc/cselib.c:24
On Tue, 9 Aug 2011, Jason Merrill wrote:
On 08/09/2011 09:14 AM, Marc Glisse wrote:
I don't think we should define the C++ 2011 value yet. In my opinion, we
should wait until:
1) the standard is official
2) gcc implements most of it: people will want to use __cplusplus as a
test to know if they
On Fri, Oct 21, 2011 at 08:01:29PM +0200, Jan Kratochvil wrote:
> I had a draft patch to use DW_FORM_ref1, DW_FORM_ref2, DW_FORM_ref_udata and
> DW_FORM_ref4 which should be slightly smaller than just DW_FORM_ref_udata.
> But it produced larger files than just DW_FORM_ref_udata. Assuming it was du
> -Original Message-
> From: Georg-Johann Lay [mailto:a...@gjlay.de]
> Sent: Friday, October 21, 2011 12:52 PM
> To: gcc-patches@gcc.gnu.org
> Cc: Weddington, Eric; Denis Chertykov
> Subject: [Patch,AVR] Clean-up avr.c: Break long lines, move targetm to
end
> of file.
>
> No functional c
Ping (changing the Cc to bother a different person...).
On Sat, 3 Sep 2011, Marc Glisse wrote:
Hello,
this patch is obviously related to PR c++/2316 ("g++ fails to overload on
language linkage") but seems fairly independent. Currently, the demangler
recognizes 'Y' for extern "C" functions a
Hello,
cpp_peek_token can fail to peek tokens sometimes because
_cpp_remaining_tokens_num_in_context counts tokens only from the current
context; worse, _cpp_token_from_context_at also get tokens only from the
context context. They should both operate on the context given by
cpp_peek_token. I di
No functional change, just a bit of clean-up.
Definition of targetm is moved towards end of the file.
Ok for trunk?
Johann
* config/avr/avr.c: Break long lines.
Define target hooks on the fly if applicable.
(TARGET_ASM_FUNCTION_RODATA_SECTION): Remove first definition
Hello,
anyone willing to commit this?
On Sat, 24 Sep 2011, Marc Glisse wrote:
On Sat, 17 Sep 2011, Joseph S. Myers wrote:
These are OK (with ChangeLog entries properly omitting the "include/",
since they go in include/ChangeLog) in the absence of libiberty maintainer
objections within 72 ho
On 10/20/11, Gabriel Charette wrote:
> I just thought about something..
>
> Earlier I said that ALL line_table issues were resolved after this
> patch (as it ignores the re-included headers that were guarded, as the
> non-pph compiler does naturally).
>
> One problem remains however, I'm pretty su
Hi!
No idea how I've missed this. With -mavx V8SImode and V4DImode
are valid, but the choice of shuffle insns for them is limited.
This patch will just pay the reinterpretation penalty and reshuffle
them as corresponding V4DFmode resp. V8SFmode instead of ICE.
Additionally, I've added two new (
On 10/21/11 15:42, Bernd Schmidt wrote:
> On 10/14/11 17:35, Vladimir Makarov wrote:
>> The scheduler part of the patch is ok for me (other part changes are
>> obvious). Could you only commit it at the beginning of the next week.
>
> I've committed this variant. It's updated for some recent chang
>
> Meh... Please no, this was the kind of scatter-patches my patch
> aimed to avoid... for example, easy to miss some tests.
>
> Instead, on top of my patch, just copy the
> scan-assembler_required_options proc to a
> scan-tree-dump_required_options. ...no wait, should forcing
> fat-lto be don
> > diff --git a/gcc/ggc-page.c b/gcc/ggc-page.c
> > index ba88e3f..eb0eeef 100644
> > --- a/gcc/ggc-page.c
> > +++ b/gcc/ggc-page.c
> > @@ -972,6 +972,54 @@ release_pages (void)
> > page_entry *p, *start_p;
> > char *start;
> > size_t len;
> > + size_t mapped_len;
> > + page_entry *next, *
On 10/21/2011 12:20 PM, Paolo Carlini wrote:
+ || tree_int_cst_equal (maxindex, integer_minus_one_node))
Use integer_all_onesp instead. OK with that change.
Jason
OK.
Jason
On Fri, 21 Oct 2011 20:01:29 +0200, Jan Kratochvil wrote:
> No regressions on {x86_64,x86_64-m32,i686}-fedora16pre-linux-gnu
> (4.7.0 20111002).
A typo, only tested for x86_64-fedora16pre-linux-gnu, sorry.
Jan
Hi,
this is a standalone patch to replace DW_FORM_ref4 by DW_FORM_ref_udata
(uleb128). It saves for libstdc++.so.debug: 5254792 -> 4859136 = 7.53%
Tested with: -O2 -gdwarf-4 -fdebug-types-section
The references are already intra-CU ones, they are never relocated, they are
pre-computed by dwarf2o
Hi again,
Another acceptable fix is to
-- leave the current diagnostic as is if -fms-extensions
-- suggest '()' is member function
-- otherwise suggest '&'.
Thanks for your help Gaby, in particular about the MS extension which
I had overlooked completely (as any hard-code Linux guy woul
On 10/21/11 11:08, Andrew MacLeod wrote:
On 10/21/2011 11:28 AM, H.J. Lu wrote:
On Fri, Oct 21, 2011 at 5:11 AM, Andrew MacLeod
wrote:
X32 has native int64 and int128.
I presume there is no atomic support for int128 though, and thats what
'condition check_effective_target_sync_int_128' is te
Jakub Jelinek wrote:
Though, what could be done is just special case OpenMP workshare regions,
insert everything into BLOCK local vars unless in OpenMP workshare, in that
case put the BLOCK with the temporary around the workshare rather than
inside of it. In the case of omp parallel workshare it
On Fri, Oct 21, 2011 at 11:52:44AM +0200, Jakub Jelinek wrote:
> On Fri, Oct 21, 2011 at 11:42:26AM +0200, Richard Guenther wrote:
> > On Fri, Oct 21, 2011 at 7:52 AM, Andi Kleen wrote:
> > > From: Andi Kleen
> > >
> > > This one place in ggc forgot to round page_entry->bytes to the
> > > next pa
FYI: I am seeing this same ICE on the hppa64-hp-hpux11.11 bootstrap.
(debug_expr:DI D#49)
/ctires/gcc/nightly/src/trunk/gcc/cselib.c: In function 'void
cselib_record_sets(rtx)':
/ctires/gcc/nightly/src/trunk/gcc/cselib.c:2424:1: internal compiler error: in
mem_loc_descriptor, at dwarf2out.c:1237
There are two proposals here. One is -fopt-info which prints out
informational notes to stderr, and the other is -fopt-report which is
more elaborate form of dump files. Are you object to both or just the
opt-report one? The former is no different from any other
informational notes we already have
On 10/21/2011 04:56 PM, Jason Merrill wrote:
I think the fix for 35602 was wrong; instead of trying to suppress the
warning, we should avoid building expressions that trip it. In this
case, the problem is a type mismatch in build_vec_init between
maxindex/iterator (ptrdiff_type_node) and array
> Date: Fri, 21 Oct 2011 17:44:15 +0200
> From: Jan Hubicka
> I also noticed that tests scanning output of late optimization passes are
> now getting UNRESOLVED state with slim LTO. We don't really lose coverage
> here because we test fat LTO with the other compilation, but probably easiest
> is
On Fri, Oct 21, 2011 at 9:08 AM, Andrew MacLeod wrote:
> On 10/21/2011 11:28 AM, H.J. Lu wrote:
>>
>> On Fri, Oct 21, 2011 at 5:11 AM, Andrew MacLeod
>> wrote:
X32 has native int64 and int128.
>>> I presume there is no atomic support for int128 though, and thats what
>>> 'condition
Jason,
I split the changelog into four parts and tweaked some of the content and
formatting.
Ed
CL_udlit_gcc_c-family
Description: Binary data
CL_udlit_gcc_cp
Description: Binary data
CL_udlit_gcc_testsuite
Description: Binary data
CL_udlit_libcpp
Description: Binary data
On 10/21/2011 11:28 AM, H.J. Lu wrote:
On Fri, Oct 21, 2011 at 5:11 AM, Andrew MacLeod wrote:
X32 has native int64 and int128.
I presume there is no atomic support for int128 though, and thats what
'condition check_effective_target_sync_int_128' is testing for.
X32 uses x86-64 instruction
On Fri, Oct 21, 2011 at 6:39 AM, Tristan Gingold wrote:
> Hi,
>
> the AIX linker supports response files with the '-fFILE' command line option.
> With this patch, collect2 reads the content of the response files and listed
> object files are handled. Therefore these files are not forgotten by th
> If running the gnat.dg testsuite, lib/gcc-dg.exp is now calling
> check_linker_plugin_available early, which ultimately calls
> ${tool}_target_compile. For all languages but Ada,
> ${tool}_target_compile can compile .c files just fine, but
> gnat_target_compile (which uses gnatmake) cannot, so i
2011/10/21 Georg-Johann Lay :
> This fixes avr_legitimize_reload_address:
>
> Since breaking out the code from LEGITIMIZE_RELOAD_ADDRESS, protiype of above
> is
> avr_legitimize_reload_address (rtx x, ...
> but must be
> avr_legitimize_reload_address (rtx *px, ...
> because at one place &x is
This fixes avr_legitimize_reload_address:
Since breaking out the code from LEGITIMIZE_RELOAD_ADDRESS, protiype of above is
avr_legitimize_reload_address (rtx x, ...
but must be
avr_legitimize_reload_address (rtx *px, ...
because at one place &x is used as input to push_reload which is now px
On Fri, Oct 21, 2011 at 5:11 AM, Andrew MacLeod wrote:
> On 10/20/2011 06:50 PM, H.J. Lu wrote:
>>
>> On Thu, Oct 20, 2011 at 3:38 PM, Joseph S. Myers
>> wrote:
>>>
>>> Do these operations exist for x32 as well as for -m64? If they do, then
>>> lp64 isn't the right test either; if not, then it
On 21 October 2011 16:25, Jakub Jelinek wrote:
> On Fri, Oct 21, 2011 at 03:44:11PM +0200, Ira Rosen wrote:
>> But it's OK to allow modifier != NONE if it's not SLP, so we need &&, no?
>
> Well, in my patch that check was guarded by the if (slp_node ...),
> so presumably it would allow modifier ==
Hi,
committed.
Thanks,
Paolo.
//
2011-10-21 Roland Stigge
PR translation/47064
* params.def: Fix typo "compilatoin" -> "compilation".
Index: params.def
===
--- params.def (revision 180288)
+
On Friday, October 21, 2011 4:42 PM, Joseph S. Myers wrote:
> On Fri, 21 Oct 2011, Andy Gibbs wrote:
>
>> Hi,
>>
>> Please can I "bump" this patch and ask for it to be approved and
>> committed:
>> http://gcc.gnu.org/ml/gcc-patches/2011-10/msg01711.html
>
> Have you sent in your copyright assignme
I think the fix for 35602 was wrong; instead of trying to suppress the
warning, we should avoid building expressions that trip it. In this
case, the problem is a type mismatch in build_vec_init between
maxindex/iterator (ptrdiff_type_node) and array_type_nelts_total
(sizetype). And indeed, co
On Fri, 21 Oct 2011, Andy Gibbs wrote:
> Hi,
>
> Please can I "bump" this patch and ask for it to be approved and committed:
> http://gcc.gnu.org/ml/gcc-patches/2011-10/msg01711.html
Have you sent in your copyright assignment papers to the FSF? The patch
is large enough to need them.
--
Jose
Thanks!
K
On Fri, Oct 21, 2011 at 6:34 PM, Uros Bizjak wrote:
> On Fri, Oct 21, 2011 at 3:58 PM, Kirill Yukhin
> wrote:
>
>> Updated testsuite/ChangeLog:
>> 2011-10-21 H.J. Lu
>> Kirill Yukhin
>>
>> * gcc.target/i386/avx2-check.h (main): Check CPUID level
>> corre
On Fri, Oct 21, 2011 at 3:58 PM, Kirill Yukhin wrote:
> Updated testsuite/ChangeLog:
> 2011-10-21 H.J. Lu
> Kirill Yukhin
>
> * gcc.target/i386/avx2-check.h (main): Check CPUID level
> correctly.
> * gcc.target/i386/bmi2-check.h: Ditto.
>
>
> Could anybody ple
This patch adds support to consistently use EIND.
The compiler never sets this SFR but uses it in table jumps and
EIJMP/EICALL
instructions.
Custom startup code could set EIND to an other value than 0 and the
compiler
should use EIND consistently given
On Fri, Oct 21, 2011 at 03:44:11PM +0200, Ira Rosen wrote:
> But it's OK to allow modifier != NONE if it's not SLP, so we need &&, no?
Well, in my patch that check was guarded by the if (slp_node ...),
so presumably it would allow modifier == NARROW vectorization in the loops
(otherwise some testc
On Fri, Oct 21, 2011 at 03:20:54PM +0200, Andreas Krebbel wrote:
> on s390 a strcat is already decomposed by fold_builtin_strcat into a
> strlen and a strcpy. Due to that 3 strlenopt testcases currently
> fail: strlenopt-17g.c, strlenopt-4.c, strlenopt-4g.c.
Well, fold_builtin_strcat does such ki
Need to make sure that this comment is still accurate:
/* Local statics and classes get the visibility of their
containing function by default, except that
-fvisibility-inlines-hidden doesn't affect them. */
i.e. given
inline int * f() { static int i; retu
Thanks,
Updated testsuite/ChangeLog:
2011-10-21 H.J. Lu
Kirill Yukhin
* gcc.target/i386/avx2-check.h (main): Check CPUID level
correctly.
* gcc.target/i386/bmi2-check.h: Ditto.
Could anybody please commit that?
K
On Fri, Oct 21, 2011 at 4:56 PM, Uros B
On Friday, October 21, 2011 12:04 PM, Richard Guenther wrote:
> What's the motivation for this? Why can't it be implemented using C++
> features such as templates and specialization?
Sorry for the delay in my reply.
The motivation, simply put, is to be able to create both templates and
macros t
On 21 October 2011 14:52, Jakub Jelinek wrote:
> On Fri, Oct 21, 2011 at 02:37:06PM +0200, Ira Rosen wrote:
>> > @@ -1620,7 +1615,13 @@ vectorizable_call (gimple stmt, gimple_s
>> >
>> > gcc_assert (!gimple_vuse (stmt));
>> >
>> > - if (modifier == NARROW)
>> > + if (slp_node || PURE_SLP_STMT
Applied, thanks.
Jason
2011/10/21 Georg-Johann Lay :
>>> This patch adds support to consistently use EIND.
>>>
>>> The compiler never sets this SFR but uses it in table jumps and EIJMP/EICALL
>>> instructions.
>>>
>>> Custom startup code could set EIND to an other value than 0 and the compiler
>>> should use EIND consist
Hi,
on s390 a strcat is already decomposed by fold_builtin_strcat into a
strlen and a strcpy. Due to that 3 strlenopt testcases currently
fail: strlenopt-17g.c, strlenopt-4.c, strlenopt-4g.c.
For strlenopt-4.c no optimization is expected anyway (stpcpy
disabled). So this can easily be fixed by a
Georg-Johann Lay wrote:
> Does it make sense to extend LEGITIMIZE_RELOAD_ADDRESS, too?
>
> For the target that needs your extension (AVR) there are different addressing
> capabilities depending on AS and there is an implementation of L_R_A.
I'd say that it is an independent issue. LEGITIMIZE_RE
On Fri, Oct 21, 2011 at 12:46 PM, Kirill Yukhin wrote:
> Here is the patch which checks CPUID correctly to get BMI/BMI2/AVX2 feature.
>
> ChangeLog entry is:
> 2011-10-21 H.J. Lu
> Kirill Yukhin
>
> * config/i386/driver-i386.c (host_detect_local_cpu): Do cpuid 7 only
>
On Fri, Oct 21, 2011 at 02:37:06PM +0200, Ira Rosen wrote:
> > @@ -1620,7 +1615,13 @@ vectorizable_call (gimple stmt, gimple_s
> >
> > gcc_assert (!gimple_vuse (stmt));
> >
> > - if (modifier == NARROW)
> > + if (slp_node || PURE_SLP_STMT (stmt_info))
> > + {
> > + if (modifier != NONE)
> 2011-10-11 Sameera Deshpande
>
>
> * config/arm/arm.c (arm_emit_ldrd_pop): New static function.
> (arm_expand_epilogue): Update.
> * config/arm/ldmstm.md (arm_ldrd_base): New pattern.
> (arm_ldr_with_update): Likewise.
rth's comment about REG_CFA_RESTORE applies
>+/* STRD in ARM mode needs consecutive registers to be stored. This function
>+ keeps accumulating non-consecutive registers until first consecutive
>register
numchar > 80.
>+ pair is found. It then generates multi-reg PUSH for all accumulated
>+ registers, and then generates STRD with
On 11 October 2011 10:27, Sameera Deshpande wrote:
> Hi!
>
> This patch generates STRD instruction instead of PUSH in thumb2 mode for
> A15.
>
> For optimize_size, original prologue is generated for A15.
> The work involves defining new functions, predicates and patterns.
>
> +/* Generate and em
> 2011-10-11 Sameera Deshpande
>
>
> * config/arm/arm-cores.def (cortex_a15): Update.
> * config/arm/arm-protos.h (struct tune_params): Add new field...
> (arm_gen_ldrd_strd): ... this.
> * config/arm/arm.c (arm_slowmul_tune): Add
> arm_gen_ldrd_strd field s
Hi Sameera,
The comment about REG_FRAME_RELATED_EXPR vs REG_CFA_RESTORE from one
of your later patches
applies here as well.
>diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
>index 3162b30..f86a3e6 100644
>--- a/gcc/config/arm/arm.c
>+++ b/gcc/config/arm/arm.c
>@@ -8754,6 +8754,140 @@ ne
Hi,
Some time back Michael pointed out that the ARM backend doesn't generate
vcvt.f32.s where you have a conversion from fixed to floating point
as in the example below. It should also be possible to generate the vector forms
of this which will be the subject of a follow-up patch .
I've chosen to
On 20 October 2011 23:50, Jakub Jelinek wrote:
> Hi!
Hi,
>
> While looking at *.vect dumps from Polyhedron, I've noticed the lack
> of SLP vectorization of builtin calls.
>
> This patch is an attempt to handle at least 1 and 2 operand builtin calls
> (SLP doesn't handle ternary stmts either yet)
Ulrich Weigand schrieb:
> Georg-Johann Lay wrote:
>> Ulrich Weigand schrieb:
>>> Hello,
>>>
>>> Georg-Johann Lay has proposed a patch to add named address space support
>>> to the AVR target here:
>>> http://gcc.gnu.org/ml/gcc-patches/2011-10/msg00471.html
>>>
>>> Since the target needs to make reg
>> This patch adds support to consistently use EIND.
>>
>> The compiler never sets this SFR but uses it in table jumps and EIJMP/EICALL
>> instructions.
>>
>> Custom startup code could set EIND to an other value than 0 and the compiler
>> should use EIND consistently given that EIND might not be ze
On Fri, 2011-10-21 at 11:26 +0200, Richard Guenther wrote:
> On Tue, Oct 18, 2011 at 4:14 PM, William J. Schmidt
> wrote:
> > +
> > + /* We don't use get_def_for_expr for S1 because TER doesn't forward
> > + S1 in some situations where this transform is useful, such as
> > + when S1
2011/10/21 Richard Guenther :
> On Thu, Oct 20, 2011 at 3:08 PM, Kai Tietz wrote:
>> Hello,
>>
>> this patch re-enables the branch-cost optimization on simple boolean-typed
>> operands, which are casted to a wider integral type. This happens due casts
>> from
>> boolean-types are preserved, but
On 10/20/2011 06:50 PM, H.J. Lu wrote:
On Thu, Oct 20, 2011 at 3:38 PM, Joseph S. Myers
wrote:
Do these operations exist for x32 as well as for -m64? If they do, then
lp64 isn't the right test either; if not, then it is.
X32 has native int64 and int128.
I presume there is no atomic suppor
Iain Sandoe writes:
> It looks like the gnat testsuite is also broken - but HP's fix doesn't
> recover that.
> .. will try and take a look - but short on time today,
I think I see what's going on: in gnat.log, I find
Running /vol/gcc/src/hg/trunk/local/gcc/testsuite/gnat.dg/dg.exp ...
ERROR: tc
> Otherwise the generic parts of the patch look good.
> Please get separate approval for the arm portions of the patch.
Is it just me or has no one else seen this patch on the archives at
gcc-patches@. ?
Ramana
On 21 Oct 2011, at 10:31, Jan Hubicka wrote:
Date: Fri, 21 Oct 2011 00:19:32 +0200
From: Jan Hubicka
Yes, if we scan assembler, we likely want -fno-fat-lto-objects.
then IIUC you need to patch *all* torture tests that use
scan-assembler and scan-assembler-not. Alternatively, patch
somewher
Hi,
this patch makes weakref that has no declaration they alias to to go the
callgraph
way instead of alias pair way. This should make practically all sane aliases go
the callgraph way, but we still do not handle aliases from functions to
variables
and vice versa.
They are not too hard to repre
2011/10/21 Georg-Johann Lay :
> This patch adds support to consistently use EIND.
>
> The compiler never sets this SFR but uses it in table jumps and EIJMP/EICALL
> instructions.
>
> Custom startup code could set EIND to an other value than 0 and the compiler
> should use EIND consistently given th
1 - 100 of 122 matches
Mail list logo
