On 07/22/2011 05:36 PM, Richard Guenther wrote:
> That said - I'm reasonably happy with the pass now, but it's rather large
> (this review took 40min again ...) so I appreciate a second look from
> somebody else.
>
Ian,
Do you have a moment to give a second look to a gimple CFG optimization? Th
Hi Richard,
On 07/30/2011 09:21 AM, Tom de Vries wrote:
> This is an updated version of the patch. I have 2 new patches and an updated
> testcase which I will sent out individually.
>
> Patch set was bootstrapped and reg-tested on x86_64.
>
> Ok for trunk?
>
You already approved the the 2 new
Removed extra blank lines and pass tests through "indent".
2011/8/23 Uros Bizjak :
> On Tue, Aug 23, 2011 at 4:19 PM, Ilya Tocar wrote:
>> I removed unnecessary expands/builtins and tests are now compiled with -O2.
>> Is this version ok?
>
> OK with minor comments:
>
> - Please remove extra blank
On Wed, Aug 24, 2011 at 12:48:06PM +0400, Ilya Tocar wrote:
> Removed extra blank lines and pass tests through "indent".
You haven't:
@@ -25113,6 +25125,9 @@ static const struct builtin_description
bdesc_multi_arg[] =
"__built
Hello,
it would be really nice if there can be reached an agreement here.
The bug
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38644
is open since 2008-12-27.
As a user I don't care if it is fixed in the middle or back end or where ever.
There are some duplicated bugs here, so this is not a
On 08/24/2011 12:01 AM, Mikael Morin wrote:
this is an attempt to fix my recent breakage for PR50050.
I forgot that shape can't always be known, and thus, that for some
expressions, the shape field is a NULL pointer.
Neither bootstraped, nor regression tested, but it is in progress. My machine
d
"Joseph S. Myers" writes:
> I didn't touch the USER_H definition in mips/t-sdemtk (so that target
> will not install this new header). The comment "Remove stdarg.h and
> stddef.h from USER_H." suggests that maybe it should be added to that
> definition, but stdfix.h is also missing from that defi
On Tue, Aug 23, 2011 at 7:29 PM, Michael Meissner
wrote:
> Over the years, it has been a problem for ports like the PowerPC that have
> more builtins than the standard list. Code that references the built_in_decls
> and implicit_built_in_decls arrays are supposed to check DECL_BUILT_IN_CLASS
> be
2011/8/24 Jakub Jelinek :
> On Wed, Aug 24, 2011 at 12:48:06PM +0400, Ilya Tocar wrote:
>> Removed extra blank lines and pass tests through "indent".
>
> You haven't:
Ah sorry only noticed one in sse.md.
>
> @@ -25113,6 +25125,9 @@ static const struct builtin_description
> bdesc_multi_arg[] =
>
This concludes the series to try to fixup data-dependence analysis.
The code is still "wrong", but it appears that the following split
out patch cures the testcases I have and does not degenerate any
in the testsuite. The patch also makes sense - we analyze the
MEM_REF base pointer for its evolut
Bernd Schmidt writes:
> * regrename.c (struct du_head): Remove member terminated.
> (create_new_chain): Don't initialize it.
> (scan_rtx_reg): Don't set or test it, test the open_chains_set
> bitmap instead.
> (tick, this_tick): New global variables, moved out of
>
On Tuesday 23 August 2011 14:26:59 Tobias Burnus wrote:
> The bug is a regression: An error was printed with 4.1.x but since 4.3.x
> one gets an ICE. [No idea what GCC 4.2 does.] The solution is simply:
> Returning if there is a MATCH_ERROR.
>
> See PR (esp. comment 2) for a more detailed descript
Hi Tobias,
PPS: I am looking for someone to review my simple regression fix [4.3 to
4.7] at
http://gcc.gnu.org/ml/fortran/2011-08/msg00186.html
it *should* be safe - see last comment in the PR for a bit more context.
That one is OK.
Thanks!
Thomas
On Tue, 23 Aug 2011, Uros Bizjak wrote:
> Hello!
>
> 2011-08-23 Uros Bizjak
>
> * config/i386/64/sfp-machine.h (ASM_INVALID): New define.
> (ASM_DIVZERO): Ditto.
> (FP_HANLDE_EXCEPTIONS): Use ASM_INVALID and ASM_DIVZERO.
>
> Tested on x86_64-pc-linux-gnu {,-m32}. Committed
Until very recently, I've been using the bundled hg convert extension to
convert an rsynced copy of the full gcc repo to mercurial. This has two
major disadvantages:
* The process is very slow, i.e. it can take more than half an hour even
for only a few revs.
* It's a one-way thing, so I have
Sorry, I'm find this a bit tough to review. Could you provide some
overview comments somewhere to say what the new algorithm is?
The comment at the head of regrename.c still describes the current
bb-local algorithm.
One thing though:
Bernd Schmidt writes:
> @@ -215,8 +306,9 @@ merge_overlapping
This fixes PR49396, the fix has been lying around in my tree for
some time.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
2011-08-24 Richard Guenther
PR c/49396
c-family/
* c-cppbuiltin.c (c_cpp_builtins_optimize_pragma): Fix condit
This avoids the file/location clutter in front of each line
in the vectorizer dump. While this is useful for people
requesting -fvectorizer-verbose=N in dump files this makes
you unable to compare dumps for testcases on a branch and trunk.
It also makes lines excessively long because the testsuit
On Wed, 24 Aug 2011, Richard Sandiford wrote:
> "Joseph S. Myers" writes:
> > I didn't touch the USER_H definition in mips/t-sdemtk (so that target
> > will not install this new header). The comment "Remove stdarg.h and
> > stddef.h from USER_H." suggests that maybe it should be added to that
>
On Wed, Aug 24, 2011 at 11:31 AM, Ilya Tocar wrote:
> 2011/8/24 Jakub Jelinek :
>> On Wed, Aug 24, 2011 at 12:48:06PM +0400, Ilya Tocar wrote:
>>> Removed extra blank lines and pass tests through "indent".
>>
>> You haven't:
> Ah sorry only noticed one in sse.md.
>>
>> @@ -25113,6 +25125,9 @@ stat
On 08/24/2011 12:58 PM, Rainer Orth wrote:
With both ways, the output of hg log --debug looks like this:
extra: branch=gcc-4_6-branch
extra:
convert_revision=svn:138bc75d-0d04-0410-961f-82ee72b054a4/branches/gcc-4_6-branch@177861
extra: branch=default
extra:
convert_re
I noticed we do not optimize any of the vectorizer alias tests
with -fno-strict-overflow. The following fixes this.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
2011-08-24 Richard Guenther
* fold-const.c (fold_comparison): Fold &a < &a + 4 even
This applies some TLC to the code I just changed and removes one
more bogus transform which seems to be possible without regressions
now.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
2011-08-24 Richard Guenther
* tree-data-ref.c (dr_analyze_indices
On 08/18/2011 11:37 PM, Joseph S. Myers wrote:
C1X provides a standard way of declaring non-returning functions, with
the _Noreturn keyword and a header stdnoreturn.h defining a macro
"noreturn" to expand to _Noreturn. This patch implements this syntax
and header. Bootstrapped with no regressio
Hi,
AVX2 change missed convert_to_mode. I checked in this patch as an
obvious fix.
Thanks.
H.J.
---
2011-08-24 H.J. Lu
PR target/50172
* config/i386/i386.c (ix86_expand_builtin): Convert to Pmode if
needed.
diff --git a/gcc/config/i386/i386.c b/gcc/config/i386/i386
On 08/24/2011 12:12 PM, Mikael Morin wrote:
On Tuesday 23 August 2011 14:26:59 Tobias Burnus wrote:
OK for the trunk?
OK.
Thanks for the review! Committed to the 4.7 trunk (Rev. 178038); I will
backport the patch later. (For cross-readers: That's a patch for a 4.3+
ice-on-invalid-code regre
On 08/24/2011 07:24 AM, Joseph S. Myers wrote:
On Wed, 24 Aug 2011, Richard Sandiford wrote:
"Joseph S. Myers" writes:
I didn't touch the USER_H definition in mips/t-sdemtk (so that target
will not install this new header). The comment "Remove stdarg.h and
stddef.h from USER_H." suggests tha
Hi,
On Wed, 24 Aug 2011, Richard Guenther wrote:
> This avoids the file/location clutter in front of each line in the
> vectorizer dump. While this is useful for people requesting
> -fvectorizer-verbose=N in dump files this makes you unable to compare
> dumps for testcases on a branch and tru
Tom> That exists -- -fdirectives-only.
Jakub> It isn't exactly what would be needed, as e.g. \\\n are removed from
Jakub> from #defines and thus they get different location of the tokens.
Dodji> Would it be acceptable to just change the output of -fdirective to fit?
Dodji> Or are we bound to not
> "Dodji" == Dodji Seketeli writes:
Dodji> * line-map.c (linemap_add): Assert that reason must not be
Dodji> LC_RENAME when called for the first time on a "main input file".
This is ok. I can't approve the rest but it seems reasonable.
Tom
On Fri, 19 Aug 2011, Richard Guenther wrote:
> On Fri, 19 Aug 2011, Eric Botcazou wrote:
>
> > > Looking at the Ada case I believe this happens because
> > > Ada has negative DECL_FIELD_OFFSET values (but that's
> > > again in sizetype, not ssizetype)? Other host_integerp
> > > uses in Ada opera
On Wed, 24 Aug 2011, Michael Matz wrote:
> Hi,
>
> On Wed, 24 Aug 2011, Richard Guenther wrote:
>
> > This avoids the file/location clutter in front of each line in the
> > vectorizer dump. While this is useful for people requesting
> > -fvectorizer-verbose=N in dump files this makes you unab
As I mentioned earlier, C++11 allows the memory model to be determined
at runtime. Rather than invoke a call to a runtime library or inline a
switch statement with all the valid options, just default to SEQ_CST
mode.This can easily be changed later if it is determined to be needed.
The do
On 08/24/2011 10:23 AM, Andrew MacLeod wrote:
As I mentioned earlier, C++11 allows the memory model to be determined
at runtime. Rather than invoke a call to a runtime library or inline
a switch statement with all the valid options, just default to SEQ_CST
mode.This can easily be changed l
On 08/17/2011 11:39 AM, Andrew MacLeod wrote:
Next step, change the C++ header files to use the new __sync
builtins. pretty straightforward.
mostly.
A previous patch changed and documents the behaviour at runtime which is
transparent to these template changes. Another patch fixed the
cons
> I've been trying for 2 days to replicate this with various
> configurations and none have failed.
I configure for i586 with --enable-checking=yes,rtl. And I also have a
comparison failure on x86-64 with the same configure options:
Bootstrap comparison failure!
libcpp/lex.o differs
--
Eric B
On Wed, 24 Aug 2011, Paolo Bonzini wrote:
> Would it make any sense to accept _Noreturn as an alternative spelling for the
> __noreturn__ attribute, in case someone is using the non-underscore-wrapped
> variant?
I suppose it might help some people use stdnoreturn.h, though I'm not
really sure st
On Sun, 07 Aug 2011 18:47:57 +0100
Richard Sandiford wrote:
> This patch caused several regressions on big-endian 64-bit MIPS
> targets, which now try to shift single-precision floating-point
> arguments to the top of an FPR. [...]
Sorry for the breakage!
> The patch below borrows the padding
(fortran@ readers: The full patch is at
http://gcc.gnu.org/ml/gcc-patches/2011-08/msg01927.html )
On 08/23/2011 08:43 PM, Dodji Seketeli wrote:
OK for trunk?
The Fortran part is OK. Thanks for the patch!
Tobias
gcc/fortran/
* scanner.c (load_file): Don't abuse LC_RENAME reason whi
On 08/03/11 17:38, Richard Sandiford wrote:
> Bernd Schmidt writes:
>> +@findex simple_return
>> +@item (simple_return)
>> +Like @code{(return)}, but truly represents only a function return, while
>> +@code{(return)} may represent an insn that also performs other functions
>> +of the function epil
Bernd Schmidt writes:
> On 08/03/11 17:38, Richard Sandiford wrote:
>> Bernd Schmidt writes:
>>> +@findex simple_return
>>> +@item (simple_return)
>>> +Like @code{(return)}, but truly represents only a function return, while
>>> +@code{(return)} may represent an insn that also performs other func
On 08/24/11 19:17, Richard Sandiford wrote:
> OK with that change from a MIPS and rtl and perspective.
Thanks. What else is in there? Trivial x86 changes, and a slightly less
trivial but still tiny ARM bit, I suppose. Richard/Ramana?
Bernd
This is a refactoring patch to make all pph caching function take a
pph_pickle_cache* directly as a first parameter instead of a pph_stream*.
This will allow us to have a global "pph_pickle_cache *pph_preloaded_cache" to
store the preloaded nodes instead of preload the nodes at the beginning of
On Wed, Aug 24, 2011 at 12:54 PM, Joseph S. Myers
wrote:
> On Tue, 23 Aug 2011, Uros Bizjak wrote:
>> 2011-08-23 Uros Bizjak
>>
>> * config/i386/64/sfp-machine.h (ASM_INVALID): New define.
>> (ASM_DIVZERO): Ditto.
>> (FP_HANLDE_EXCEPTIONS): Use ASM_INVALID and ASM_DIVZERO.
>>
On Wed, Aug 24, 2011 at 2:39 PM, Ilya Tocar wrote:
Also, why is fmaintrin.h including immintrin.h? You can't include
fmaintrin.h
directly and x86intrin.h has already included it before including
fmaintrin.h.
>>> Makes sense. Removed it.
>>
>> Do we need to add -mfma to gcc.
Tom de Vries writes:
> Do you have a moment to give a second look to a gimple CFG optimization? The
> optimization removes duplicate basic blocks and reduces code size by 1-2%.
>
> The latest patch is posted at
> http://gcc.gnu.org/ml/gcc-patches/2011-08/msg01602.html.
I'm not really the best
This patch from Roberto Lublinerman cleans up the AST dumping code he
added, and removes some trailing whitespace. It also adds support for
frontend specific optimization options. Bootstrapped and ran Go
testsuite on x86_64-unknown-linux-gnu. Committed to mainline.
Ian
2011-08-24 Roberto Lub
On 11-08-24 13:37 , Gabriel Charette wrote:
2011-08-24 Gabriel Charette
* pph-streamer-in.c (ALLOC_AND_REGISTER): Take a pph_pickle_cache*
as a first parameter instead of a pph_stream*.
Update all users.
(ALLOC_AND_REGISTER_ALTERNATE): Likewise.
* pph-st
> Any comments on this?
>
> I'd like to get it in; Phil found a bug in the std::tuple printer, and
> it would be nice to put in a test case along with the fix.
Hey Tom (and Phil!).
Sorry for the delay: this looks fine. Please put it in on trunk and
enjoy your vacation!
-benjamin
This patch bumps up the parameter 'hot-bb-count-fraction' from 1
to 4. This results in about a 0.5% geomean performance
improvement across internal benchmarks for x86-64 LIPO. The parameter
change effectively increases the number of functions/callsites which
are considered hot. The perfo
On Wed, 24 Aug 2011 17:04:55 +0100
Julian Brown wrote:
> On Sun, 07 Aug 2011 18:47:57 +0100
> Richard Sandiford wrote:
>
> > This patch caused several regressions on big-endian 64-bit MIPS
> > targets, which now try to shift single-precision floating-point
> > arguments to the top of an FPR. [
Good for google branches.
Need to measure the performance and size impact on SPEC2k/06 with this
patch and the previous hotcaller patch for trunk.
thanks,
David
On Wed, Aug 24, 2011 at 2:43 PM, Mark Heffernan wrote:
> This patch bumps up the parameter 'hot-bb-count-fraction' from 1
> to 40
This patch creates an independent cache, pph_preloaded_cache, which contains
all common nodes which used to be directly added to each stream's cache.
The streams' cache now only contain nodes which really belong to them.
The preloaded cache is last in the lookup on the way out (potentially allow
Wouldn't -Wshadow be more useful if it obeyed -Wno-system-headers?
For code like
#include
int foo (int atof);
int foo (int atof) { return atof; }
we currently do not warn on the prototype, but do on the function
definition, leading to reports such as
http://sourceware.org
54 matches
Mail list logo
