Quoting Ralf Wildenhues :
Can you separate the configure and the build steps into separate make
targets so that one can run one but not the other? Or would that have
adverse effects on the load pattern (it might, due to the way make
orders scheduling of jobs)?
I've added a separate target 'co
On Thu, 2011-04-14 at 00:48 +0200, Paolo Carlini wrote:
> Hi!
>
> over the last 2 days or so, Daniel Krugler - having filed in due course
> his Copyright Assignment - finally kindly contributed to the project
> shiny new std::is_constructible, std::is_default_constructible and
> std::is_destruc
On 04/14/2011 10:32 AM, Magnus Fromreide wrote:
This is just a nit, but why do you remove the /// is_function comments?
Because was on the forward declaration, not on the implementation
itself, where I wanted it.
Paolo.
On Wed, Apr 13, 2011 at 8:43 PM, Nathan Froyd wrote:
> On Wed, Apr 13, 2011 at 11:07:15AM +0200, Richard Guenther wrote:
>> On Tue, Apr 12, 2011 at 5:09 PM, Nathan Froyd
>> wrote:
>> > Granted, but that fact should still be recorded. The situation we have
>> > today, for something like:
>> >
>>
On Thu, Apr 14, 2011 at 12:20 AM, Jan Hubicka wrote:
> Hi,
> this patch moves inline_summary from field in cgraph_node into its own on side
> datastructure. This moves it from arcane decision of mine to split all IPA
> data
> into global/local datas stored in common datastructure into the scheme
> This is the "real" fix for PR46076 that I wanted to persue. Make
> function pointer type conversions useless as to more aggressively
> be able to turn indirect into direct calls. This requires that we
> preserve the original type signature of the called function as
> presented by the frontend.
These test cases
gcc.target/avr/trivial.c
gcc.target/avr/torture/trivial.c
put non-const variables in read-only section .progmem.data which
causes an error. The patch rewrites the files to make respective
variables real-only (and besides that fixes line endings).
Johann
./testsuite
2011-04-14
On Thu, 14 Apr 2011, Eric Botcazou wrote:
> > This is the "real" fix for PR46076 that I wanted to persue. Make
> > function pointer type conversions useless as to more aggressively
> > be able to turn indirect into direct calls. This requires that we
> > preserve the original type signature of t
2011/4/14 Georg-Johann Lay :
> These test cases
>
> gcc.target/avr/trivial.c
> gcc.target/avr/torture/trivial.c
>
> put non-const variables in read-only section .progmem.data which
> causes an error. The patch rewrites the files to make respective
> variables real-only (and besides that fixes line
The "rotl3" expanders (mode \in {HI,SI,DI}) violates synopsis by
using 4 operands instead of 3. This runs in ICE in top of
optabs.c:maybe_gen_insn.
The right way to do this is to use match_dup, not match_operand. So
the fix is obvious.
Regenerated avr-gcc and run against avr testsuite:
gcc.targe
On Fri, Apr 08, 2011 at 01:50:24PM -0400, Jason Merrill wrote:
> On 03/24/2011 09:15 AM, Nathan Froyd wrote:
>> + tree t = make_node (CASE_LABEL_EXPR);
>> +
>> + TREE_TYPE (t) = void_type_node;
>> + SET_EXPR_LOCATION (t, input_location);
>
> As jsm and richi said, using input_location like this
On Thu, 14 Apr 2011, Nathan Froyd wrote:
> On Fri, Apr 08, 2011 at 01:50:24PM -0400, Jason Merrill wrote:
> > On 03/24/2011 09:15 AM, Nathan Froyd wrote:
> >> + tree t = make_node (CASE_LABEL_EXPR);
> >> +
> >> + TREE_TYPE (t) = void_type_node;
> >> + SET_EXPR_LOCATION (t, input_location);
> >
This tests are intended to reveal the respective PRs because the test
case is more stable under slight variations in code (both of
application or compiler).
This test case migh also be helpful for older versions of avr-gcc, in
particular if PR41894 is actually fixed.
2011-04-14 Georg-Johann Lay
When using the C++ frontend to build gcc.dg/fold-bitand-4.c it currently
fails to optimize because we don't handle &a[i] properly in
get_pointer_modulus_and_residue. We do have nice conservative code
in get_object_alignment that just needs splitting out. Which is
what the following patch does.
We are currently not special casing alloca related builtins in the
alias oracle while we can very easily do better, similar to how
we handle malloc and free. The exception is BUILT_IN_STACK_RESTORE
which is not tied to a particular allocation.
Bootstrapped and tested on x86_64-unknown-linux-gnu,
Hi,
On Wed, 13 Apr 2011, Bernd Schmidt wrote:
> > I'll also note that the first hunk of your change is in a loop commented
> > with "Compute upper bound, bla ...", meaning to be a heuristic, and your
> > second change is this:
>
> It's not a heuristic. Both of these loops are necessary to comp
Ping. Contains only changes outside config/arm.
http://gcc.gnu.org/ml/gcc-patches/2011-03/msg01509.html
Bernd
On 03/23/2011 04:46 PM, Bernd Schmidt wrote:
> I've discovered a problem with -mlong-calls on ARM. The bug was first
> reported against a new target, but I'd copied the relevant code f
On Thu, Apr 14, 2011 at 2:40 PM, Bernd Schmidt wrote:
> Ping. Contains only changes outside config/arm.
>
> http://gcc.gnu.org/m/gcc-patches/2011-03/msg01509.html
Ok.
Thanks,
Richard.
>
> Bernd
>
> On 03/23/2011 04:46 PM, Bernd Schmidt wrote:
>> I've discovered a problem with -mlong-calls on A
On Fri, Apr 8, 2011 at 6:51 PM, Ramana Radhakrishnan
wrote:
> On 08/04/11 10:57, Carrot Wei wrote:
>>
>> Hi
>>
>> This is the second part of the fixing for
>>
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47855
>>
>> This patch contains the length computation for insn patterns
>> "*arm_movqi_insn
> I have opened PR48590 for at least one issue that I see. ...
The fix for pr48590 (r172427) does not speedup fatigue.f90 compiled
with -fstack-arrays.
Dominique
We have
static unsigned int
get_decl_align_unit (tree decl)
{
unsigned int align = LOCAL_DECL_ALIGNMENT (decl);
return align / BITS_PER_UNIT;
}
LOCAL_DECL_ALIGNMENT may increase alignment for local variable. But it is
never saved. DECL_ALIGN (decl) returns the old alignment. This patch
upd
Currently DSE does not allow for a call to be the redundant killing
stmt. The following patch fixes that and also simplifies the DSE
code by removing most of the domwalk stuff which wasn't necessary
anymore.
The alias-oracle is extended to handle some builtin calls as killing
stmts so that we ca
On Thu, 14 Apr 2011, Eric Botcazou wrote:
> > This is the "real" fix for PR46076 that I wanted to persue. Make
> > function pointer type conversions useless as to more aggressively
> > be able to turn indirect into direct calls. This requires that we
> > preserve the original type signature of t
Richard Henderson writes:
> On 03/31/2011 03:19 AM, Richard Sandiford wrote:
>> * recog.h (insn_operand_data): Add an "allows_mem" field.
>> * genoutput.c (output_operand_data): Initialize it.
>> * optabs.c (maybe_legitimize_operand_same_code): New function.
>> (maybe_legitimiz
Richard Sandiford writes:
> I'm implementing the multi-vector load/store support described here:
>
> http://gcc.gnu.org/ml/gcc/2011-03/msg00322.html
>
> For the purposes of this message, the main thing is that the losds
> and stores are more naturally represented as GIMPLE_CALLs rather
> than
On Thu, Apr 14, 2011 at 09:43, Richard Sandiford
wrote:
> +/* This file specifies a list of internal "functions". These functions
> + differ from built-in functions in that they have no linkage and cannot
> + be called directly by the user. They represent operations that are only
> + synt
On Thu, Apr 14, 2011 at 3:34 PM, H.J. Lu wrote:
> We have
>
> static unsigned int
> get_decl_align_unit (tree decl)
> {
> unsigned int align = LOCAL_DECL_ALIGNMENT (decl);
> return align / BITS_PER_UNIT;
> }
>
> LOCAL_DECL_ALIGNMENT may increase alignment for local variable. But it is
> never s
Hi,
On Tue, 12 Apr 2011, Dominique Dhumieres wrote:
> > The resulting speed up for nf.f90 is rather remarkable. What specific
> > feature of the fortran leads to a 30=>15s ?
>
> I think it is the automatic array in the subroutine trisolve. Note that the
> speedup is rather 27->19s and may be d
On Thu, Apr 14, 2011 at 6:57 AM, Richard Guenther
wrote:
> On Thu, Apr 14, 2011 at 3:34 PM, H.J. Lu wrote:
>> We have
>>
>> static unsigned int
>> get_decl_align_unit (tree decl)
>> {
>> unsigned int align = LOCAL_DECL_ALIGNMENT (decl);
>> return align / BITS_PER_UNIT;
>> }
>>
>> LOCAL_DECL_ALI
On Thu, Apr 14, 2011 at 4:01 PM, H.J. Lu wrote:
> On Thu, Apr 14, 2011 at 6:57 AM, Richard Guenther
> wrote:
>> On Thu, Apr 14, 2011 at 3:34 PM, H.J. Lu wrote:
>>> We have
>>>
>>> static unsigned int
>>> get_decl_align_unit (tree decl)
>>> {
>>> unsigned int align = LOCAL_DECL_ALIGNMENT (decl);
On Thu, 2011-04-14 04:12:31 -0400, Joern Rennecke
wrote:
> --- contrib/config-list.mk(revision 0)
> +++ contrib/config-list.mk(revision 0)
> @@ -0,0 +1,98 @@
[...]
> +$(LIST): make-log-dir make-script-dir
> + -mkdir $@
> + (cd $@ && \
> + ../../gcc/configure \
> + --target
On 14/02/11 18:20, Joseph S. Myers wrote:
Is there a reason you didn't add these functions to the shared libgcc
(adjust t-bpabi and t-symbian accordingly, add them to libgcc-bpabi.ver at
version GCC_4.6.0)? The GCC-specific names were deliberately made
static-only in the expectation that they wo
On Thu, Apr 14, 2011 at 7:09 AM, Richard Guenther
wrote:
> On Thu, Apr 14, 2011 at 4:01 PM, H.J. Lu wrote:
>> On Thu, Apr 14, 2011 at 6:57 AM, Richard Guenther
>> wrote:
>>> On Thu, Apr 14, 2011 at 3:34 PM, H.J. Lu wrote:
We have
static unsigned int
get_decl_align_unit (tree
On Thu, Apr 14, 2011 at 4:27 PM, H.J. Lu wrote:
> On Thu, Apr 14, 2011 at 7:09 AM, Richard Guenther
> wrote:
>> On Thu, Apr 14, 2011 at 4:01 PM, H.J. Lu wrote:
>>> On Thu, Apr 14, 2011 at 6:57 AM, Richard Guenther
>>> wrote:
On Thu, Apr 14, 2011 at 3:34 PM, H.J. Lu wrote:
> We have
>>
On Wed, Apr 13, 2011 at 3:46 PM, Jason Merrill wrote:
> On 04/11/2011 04:50 PM, Rodrigo Rivas wrote:
>>
>> Because the type of the expression must have complete type *only* if
>> it is an array.
>
> Actually, if it has class type, it must also have a complete type or the
> class member lookup is i
> "Dodji" == Dodji Seketeli writes:
Tom> One thing to note is that there may be some code that assumes an
Tom> ordering of location values. If you hand out ordinary and macro
Tom> locations separately (which I think is what is going on), then code
Tom> doing this may break.
[...]
Dodji> Now
The issue here was that with a complex expression in DECL_SIZE for z,
when wrap_cleanups tried to protect the initializer of m it walked into
the sizeof and thus the bounds expression, which it really shouldn't be
messing with.
This patch avoids this issue by saving bounds values which have si
We were just failing to check complain for this one error.
Tested x86_64-pc-linux-gnu, applying to trunk.
commit e807a327e3e667c8d79356cf644e493fbb82033f
Author: Jason Merrill
Date: Wed Apr 13 18:14:45 2011 -0400
PR c++/48557
* typeck.c (cp_build_binary_op): Don't decay void o
On 04/14/2011 10:57 AM, Jason Merrill wrote:
We were just failing to check complain for this one error.
Er, actually that's the next patch. For this one, we were complaining
in decay_conversion about misuse of void. I noticed that some other
places avoid calling that on void operands, so th
Here's the one where we just need to check complain before giving an error.
Tested x86_64-pc-linux-gnu, applying to trunk.
commit a9fa1134b357ad03ac5279c926d67af75728062d
Author: Jason Merrill
Date: Wed Apr 13 18:17:17 2011 -0400
PR c++/48531
* init.c (build_value_init_noctor)
When reviewing Rodrigo's patch for range-based for, I noticed that the
flags for some calls to build_new_method_call were wrong. Fixed thus.
Tested x86_64-pc-linux-gnu, applied to trunk.
commit ea51123767c2512840f39002d2efac52111d1a7c
Author: Jason Merrill
Date: Tue Apr 12 22:41:48 2011 -040
This patch adjusts the C frontend to retain array-refs in
address-taking operations instead of expanding them via
pointer arithmetic. As seen in PR48571 we can't re-construct
array-refs in the middle-end, thus we should keep information
as specified by the source level.
For
int a[10];
int *foo
On Thu, Apr 07, 2011 at 09:37:17AM -0400, Nathan Froyd wrote:
> My recent patch removing DECL_RTL from CONST_DECLs caused regressions in
> the ObjC++ testsuite on Darwin targets. The problem is that
> DECL_REGISTER was being called on CONST_DECLs; DECL_REGISTER says:
>
> /* In VAR_DECL and PARM_D
Hello,
as my patch for stack-arrays in fortran reveals we a problem that VLA
objects basically prevent all inlining to happen. They're transformed
into alloca calls and those are considered to disable inlining. The
(correct) fear being that inlining a bare alloca call into a loop leads to
un
On 04/14/2011 06:42 AM, Richard Sandiford wrote:
> The !side_effects_p check is the only change from last time.
>
> Tested on x86_64-linux-gnu, this time on a machine set up for building
> the 32-bit multilibs. Also tested on arm-linux-gnueabi. OK to install?
>
> Richard
>
>
> gcc/
> *
On Thu, Apr 14, 2011 at 5:18 PM, Michael Matz wrote:
> Hello,
>
> as my patch for stack-arrays in fortran reveals we a problem that VLA
> objects basically prevent all inlining to happen. They're transformed
> into alloca calls and those are considered to disable inlining. The
> (correct) fear b
Hi,
On Thu, 14 Apr 2011, Richard Guenther wrote:
> > + if (align > DECL_ALIGN (decl))
> > + DECL_ALIGN (decl) = align;
>
> Shouldn't this unconditionally set DECL_ALIGN in case
> LOCAL_DECL_ALINGMENT returns something smaller?
Decreasing alignment of DECLs points to a problem elsewhere, so
On Mon, Apr 11, 2011 at 11:18:53PM +0300, Janne Blomqvist wrote:
>
> the attached patch hopefully fixes the bootstrap failure on
> alpha*-dec-osf* due to said platform having clock_gettime() in librt,
> supporting weak symbols but not weak undefined symbols. The patch is
> the same as #38 in the P
Hi,
On Thu, 14 Apr 2011, Michael Matz wrote:
> no stack-arrayswith stack-arrays
> no addtional options: 10.2s 8.8s
> + -fwhole-program: 7.1s 8.8s
> + -fwhole-program -flto: 10.1s
On Thu, Apr 14, 2011 at 8:26 AM, Michael Matz wrote:
> Hi,
>
> On Thu, 14 Apr 2011, Richard Guenther wrote:
>
>> > + if (align > DECL_ALIGN (decl))
>> > + DECL_ALIGN (decl) = align;
>>
>> Shouldn't this unconditionally set DECL_ALIGN in case
>> LOCAL_DECL_ALINGMENT returns something smaller?
>
> We have a flag for this already on the CALL_EXPR. But we don't retain it
> over tuples, and hence we also don't look at it in
> inline_forbidden_p_stmt. This patch fixes both. (The strange testing of
> builtin-ness is because in CALL_EXPR the ALLOCA_FOR_VAR_P and
> CALL_FROM_THUNK_P flags are
On Thu, Apr 14, 2011 at 5:48 PM, H.J. Lu wrote:
> On Thu, Apr 14, 2011 at 8:26 AM, Michael Matz wrote:
>> Hi,
>>
>> On Thu, 14 Apr 2011, Richard Guenther wrote:
>>
>>> > + if (align > DECL_ALIGN (decl))
>>> > + DECL_ALIGN (decl) = align;
>>>
>>> Shouldn't this unconditionally set DECL_ALIGN i
> I'm working on a fix along this line, it needs some tweaks elsewhere.
Indeed, here's what I needed to bootstrap on x86-64/Linux.
* cfgexpand.c (expand_call_stmt): Rematerialize original function type.
* config/i386/i386.c (ix86_expand_builtin): Use get_callee_fndecl.
--
Eric
On Thu, 14 Apr 2011, Eric Botcazou wrote:
> > I'm working on a fix along this line, it needs some tweaks elsewhere.
>
> Indeed, here's what I needed to bootstrap on x86-64/Linux.
>
>
> * cfgexpand.c (expand_call_stmt): Rematerialize original function type.
> * config/i386/i386.c (ix
Hi,
On Thu, 14 Apr 2011, Eric Botcazou wrote:
> > We have a flag for this already on the CALL_EXPR. But we don't retain
> > it over tuples, and hence we also don't look at it in
> > inline_forbidden_p_stmt. This patch fixes both. (The strange testing
> > of builtin-ness is because in CALL_E
2011/4/14 Georg-Johann Lay :
> The "rotl3" expanders (mode \in {HI,SI,DI}) violates synopsis by
> using 4 operands instead of 3. This runs in ICE in top of
> optabs.c:maybe_gen_insn.
>
> The right way to do this is to use match_dup, not match_operand. So
> the fix is obvious.
>
> Regenerated avr-gc
2011/4/14 Georg-Johann Lay :
> This tests are intended to reveal the respective PRs because the test
> case is more stable under slight variations in code (both of
> application or compiler).
>
> This test case migh also be helpful for older versions of avr-gcc, in
> particular if PR41894 is actual
> I have the following, the gimple_call_chain check is to prevent
> Ada bootstrap from failing.
On x86?
> I suppose your patch is ok, maybe with simply not wrapping the NOP_EXPR
> around builtins.
I can try that indeed...
--
Eric Botcazou
This patchlet adds -finline-limit=0 to dg-options in
testsuite/gcc.target/avr/torture/pr41885.c
because otherwise optimizers will fold all tests and actually no test
function is called when optimization is on. The test case still passes
all tests.
testsuite/
2011-04-14 Georg-Johann Lay
OK.
Jason
I recently tried to build with --enable-libobjc-gc. While this worked
out of the box on Solaris, I ran intro trouble on both Tru64 UNIX and
Darwin.
On Tru64 UNIX, one needs to compile with -pthread since the boehm-gc
headers include which error out if -D_REENTRANT is missing.
On Darwin, the bui
Denis Chertykov schrieb:
> 2011/4/14 Georg-Johann Lay :
>> This tests are intended to reveal the respective PRs because the test
>> case is more stable under slight variations in code (both of
>> application or compiler).
>>
>> This test case migh also be helpful for older versions of avr-gcc, in
>
Denis Chertykov schrieb:
> 2011/4/14 Georg-Johann Lay :
>> The "rotl3" expanders (mode \in {HI,SI,DI}) violates synopsis by
>> using 4 operands instead of 3. This runs in ICE in top of
>> optabs.c:maybe_gen_insn.
>>
>> The right way to do this is to use match_dup, not match_operand. So
>> the fix i
Hello!
Attached patch converts "Intel SSE4.1 instructions" and remaining sections.
2011-04-14 Uros Bizjak
* config/i386/sse.md (sse4_1): New mode attribute.
(_blend): Macroize from
avx_blend and
sse4_1_blend using VF mode iterator.
(_blendv): Macroize f
Hello!
Attached patch converts "Intel SSE4.1 instructions" and remaining sections.
2011-04-14 Uros Bizjak
* config/i386/sse.md (sse4_1): New mode attribute.
(_blend): Macroize from
avx_blend and
sse4_1_blend using VF mode iterator.
(_blendv): Macroize f
2011/4/14 Georg-Johann Lay :
> Denis Chertykov schrieb:
>> 2011/4/14 Georg-Johann Lay :
>>> The "rotl3" expanders (mode \in {HI,SI,DI}) violates synopsis by
>>> using 4 operands instead of 3. This runs in ICE in top of
>>> optabs.c:maybe_gen_insn.
>>>
>>> The right way to do this is to use match_du
2011/4/14 Georg-Johann Lay :
> Denis Chertykov schrieb:
>> 2011/4/14 Georg-Johann Lay :
>>> This tests are intended to reveal the respective PRs because the test
>>> case is more stable under slight variations in code (both of
>>> application or compiler).
>>>
>>> This test case migh also be helpfu
2011/4/14 Georg-Johann Lay :
> This patchlet adds -finline-limit=0 to dg-options in
>
> testsuite/gcc.target/avr/torture/pr41885.c
>
> because otherwise optimizers will fold all tests and actually no test
> function is called when optimization is on. The test case still passes
> all tests.
>
>
> te
Joseph S. Myers wrote:
>
> m68k-aout was obsoleted in 4.4 and removed in 4.5 - while some OSes
> with various odd object file formats are still supported, various
> newer features such as LTO may not work so well with them.
I am aware that generic m68k-aout as target was deprecated and is gone
no
Joseph S. Myers wrote:
> On Thu, 7 Apr 2011, Gunther Nikl wrote:
>
>> Is there a (recommended) way how a subtarget can override hook
>> functions?
>
> The implementation of the hook should contain "if" (preferred over #if)
> conditions on the particular ABI in use, which could be determined by a
>> This patch reintroduce the -mflat option on SPARC. The -mfalt option was
>> deprecated in february 2004 with GCC 3.4.6. Now, with the support of
>> LEON processor on GCC, this option has found a new interest.
> Just a couple of remarks:
> - the epilogue isn't fully RTL-ized,
> - delay slot fi
This is a corner case that happens not to affect LTO because the "bad"
value is the last one packed.
In pack_ts_type_value_fields, the last thing we pack is
TYPE_ALIAS_SET, which in many cases ends up being -1. However, we
request bp_pack_value to pack HOST_BITS_PER_INT (32), whereas -1 is 64
bi
> I recently tried to build with --enable-libobjc-gc. While this worked
> out of the box on Solaris, I ran intro trouble on both Tru64 UNIX and
> Darwin.
>
> [...]
>
> The following patch fixes both issues. [...]
>
> Ok for mainline if both pass?
Yes.
[and by the way, I think you're fixing PR l
Hi!
void
foo (void)
{
register long s __asm ("r13");
register long t __asm ("r14");
register long u __asm ("r15");
asm volatile ("xorq %%r12, %%r12" : : : "r12");
asm volatile ("xorq %0, %0" : "=r" (s));
asm volatile ("xorq %0, %0" : "=r" (t));
asm volat
On Thu, Mar 31, 2011 at 7:57 AM, H.J. Lu wrote:
> On Mon, Mar 21, 2011 at 11:40 AM, H.J. Lu wrote:
>> On Mon, Mar 14, 2011 at 12:28 PM, H.J. Lu wrote:
>>> On Thu, Jan 27, 2011 at 2:40 AM, Richard Guenther
>>> wrote:
On Thu, Jan 27, 2011 at 12:12 AM, H.J. Lu wrote:
> On Tue, Dec 14, 20
Hi!
The insertps and vinsertps insns work differently if %2 is a register
resp. memory. The register is 128-bit XMM register and the upper 2 bits
of the 8 bit immediate then select which of the 4 parts of that register to
choose. If it is a memory though, the upper 2 bits of the 8 bit immediate
> See http://gcc.gnu.org/ml/gcc-patches/2011-04/msg01087.html . ...
With this patch on top of revision 172429 (and the second patch for
-fstack-arrays) I now get
Date & Time : 14 Apr 2011 20:48:24
Test Name
David Miller writes:
> Thanks Rainer, if you cons up a patch to add the ".align" directive I'll ACK
> that as well.
That's my plan: I'll include such a patch in my S8 to S11 bootstraps
over the weekend to make sure nothing breaks.
Rainer
--
On Thu, Apr 14, 2011 at 02:50:53PM -0400, Diego Novillo wrote:
> @@ -1190,8 +1190,19 @@ bitpack_create (struct lto_output_stream *s)
> static inline void
> bp_pack_value (struct bitpack_d *bp, bitpack_word_t val, unsigned nbits)
> {
> - bitpack_word_t word = bp->word;
> + bitpack_word_t mask,
This patch is the first of a series of 7 patches to reconstruct the
compilation context out of a PPH image.
Currently, there are 32 failures in the PPH testsuite. Most are
assembly miscomparisons. There are still missing bits in the images
that we are not pickling properly, but I think it makes
This test case is getting the compiler into an infinite recursive
loop. I'm disabling it for now so we can keep working on the other
PPH failures.
Diego.
* g++.dg/pph/p1mean.cc: Force syntax error.
diff --git a/gcc/testsuite/ChangeLog.pph b/gcc/testsuite/ChangeLog.pph
index 806e96a..a5
When pickling type bitfields, we were not processing TYPE_ADDR_SPACE.
* lto-streamer-in.c (unpack_ts_base_value_fields): Handle
TYPE_ADDR_SPACE.
* lto-streamer-out.c (pack_ts_base_value_fields): Likewise.
diff --git a/gcc/ChangeLog.pph b/gcc/ChangeLog.pph
index 6cd6bf0..6
We were not handling TYPE_DECLs properly. They were being handled by
the first if, but the second one was an 'else if'. Sigh.
Tested on x86_64.
Diego.
* pph-streamer-in.c (pph_stream_read_tree): Fix handling of
TYPE_DECLs.
* pph-streamer-out.c (pph_stream_write_tree):
This patch handles the reading of TYPE_LANG_SPECIFIC fields.
Tested on x86_64.
Diego.
* name-lookup.c (pph_stream_write_binding_table): Handle
NULL BT->CHAIN entries.
(pph_stream_read_binding_table): New.
* pph-streamer-in.c (pph_stream_read_lang_specific): Renam
This patch implements the writing of TYPE_LANG_SPECIFIC fields.
Tested on x86_64.
* Make-lang.in (cp/name-lookup.o): Add dependency on
CXX_PPH_STREAMER_H.
* name-lookup.c: Include pph-streamer.h.
(pph_stream_write_binding_table): New.
* pph-streamer-out.c (
This patch factors out the allocation of sorted_fields_type into a new
function so it can be called from the PPH reader.
* class.c (sorted_fields_type_new): Factor out of ...
(finish_struct_1): ... here.
diff --git a/gcc/cp/class.c b/gcc/cp/class.c
index 1325260..b040449 100
I noticed that there were a couple of references to mips-tfile on MIPS,
but this is only used on Tru64 UNIX these days. The following patch
corrects this.
Included in an i386-pc-solaris2.11 bootstrap to ensure the doc changes
are syntactically correct. The MIPS changes should be obvious, though
Michael Matz wrote:
Try this patch. I've verified that capacita and nf work with it and
-march=native -ffast-math -funroll-loops -fstack-arrays -O3 . In fact all
of polyhedron works for me on these flags. (I've set a ulimit -s of
512MB, but I don't know if such a large amount is required).
T
Tobias Burnus wrote:
no stack-arrayswith stack-arrays
+ -fwhole-program -flto: 10.1s 8.9s
+ -fwhole-program -flto -finline-limit=600 4.8s 3.6s
I wonder whether the following is special to my system* or generally
true. I
Hi,
I am checking this patch into x32 branch.
H.J.
---
commit 9f59060d93b78a252212c621dea82b458001c6c5
Author: H.J. Lu
Date: Sat Apr 9 08:18:42 2011 -0700
Require ia32 instead of ilp32 on pr48389.c.
diff --git a/gcc/testsuite/ChangeLog.x32 b/gcc/testsuite/ChangeLog.x32
index 0081f4c..46
On Thu, Apr 14, 2011 at 9:11 PM, Jakub Jelinek wrote:
> The insertps and vinsertps insns work differently if %2 is a register
> resp. memory. The register is 128-bit XMM register and the upper 2 bits
> of the 8 bit immediate then select which of the 4 parts of that register to
> choose. If it i
On Thu, Apr 14, 2011 at 15:28, Jakub Jelinek wrote:
> If bitpack_word_t has BITS_PER_BITPACK_WORD bits, then for
> nbits = BITS_PER_BITPACK_WORD this will be undefined.
> Use say
> mask = ((bitpack_word_t) 2 << (nbits - 1)) - 1;
> or something similar (assertion ensures that nbits isn't 0).
Quit
Rainer Orth writes:
> Ok for mainline if that passes?
Definitely, thanks.
Richard
> 2011-04-08 Rainer Orth
>
> * config/mips/mips.h (DBX_OUTPUT_SOURCE_LINE): Remove.
> * config/mips/mips.opt (mmips-tfile): Remove.
>
> * doc/install.texi (Specific, mips-*-*): Move mips-tfile
I have forgotten to mentionned that I have a variant of fatigue
in which I have done the inlining manually along with few other
optimizations and the timing for it is
[macbook] lin/test% gfc -Ofast fatigue_v8.f90
[macbook] lin/test% time a.out > /dev/null
2.793u 0.002s 0:02.79 100.0%0+0k 0+1io
> I suppose your patch is ok, maybe with simply not wrapping the NOP_EXPR
> around builtins.
Here's what I've installed after bootstrapping/regtesting on x86_64-suse-linux.
2011-04-14 Eric Botcazou
* cfgexpand.c (expand_call_stmt): Rematerialize the original function
type if
An earlier change reduced the number of bits in cpp_hashnode.directive_index
from 7 to 5, as that was sufficient for indexing the directives. Tom Tromey
asked for a static check on the size. This patch adds that check.
Unfortunately, five bits are not sufficient for the alternate use of
cpp_has
Ping.
On 2011/3/31 10:57 PM, Chung-Lin Tang wrote:
> This PR doesn't exactly trigger currently on trunk; a REG_DEAD note that
> occurs in trunk, but not in the 4.5-based compilers which this bug was
> reported for, currently blocks auto-inc-dec from doing its job, and just
> happens to avoid this
2011/4/11 Basile Starynkevitch :
> On Fri, 8 Apr 2011 19:55:51 +0200
> Basile Starynkevitch wrote:
>
>> Committed revision 172203 (on trunk).
>>
>> Actually, the above committed patch is better than nothing, but not
>> perfect. It happens to generate ggc_alloc macros for things which are
>> not de
This patch to the Go frontend uses the backend interface for select
statements. Bootstrapped and ran Go testsuite on
x86_64-unknown-linux-gnu. Committed to mainline.
Ian
2011-04-14 Ian Lance Taylor
* go-gcc.cc (Backend::error_statement): New function.
Index: gcc/go/go-gcc.cc
On Thu, Apr 14, 2011 at 18:27, Steve Kargl
wrote:
> On Mon, Apr 11, 2011 at 11:18:53PM +0300, Janne Blomqvist wrote:
>>
>> the attached patch hopefully fixes the bootstrap failure on
>> alpha*-dec-osf* due to said platform having clock_gettime() in librt,
>> supporting weak symbols but not weak un
1 - 100 of 102 matches
Mail list logo
