2014-12-16 7:58 GMT+01:00 Tobias Burnus bur...@net-b.de:
Hi Janus, hi all,
Janus Weil wrote:
here is a regression fix for a problem with the NON_OVERRIDABLE
attribute. For non-overridable type-bound procedures we do not have to
generate a call to the vtable, but can just translate it to a
Jan Hubicka hubi...@ucw.cz writes:
this should be fixed by patch to handle extern aliases correctly I commited
later that day.
Does the problem still reproduce for you?
It works again.
Andreas.
--
Andreas Schwab, SUSE Labs, sch...@suse.de
GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE
Built and currently regtested on x86-64-gnu-linux.
OK for the trunk?
OK.
I’m not sure exactly why we have a hardcoded minimal value of (2^16-1) for “max
array constructor”, or a default max stack of 2^15. But that’s a separate issue.
Thanks,
FX
On Thu, Dec 11, 2014 at 8:08 PM, Richard Biener
richard.guent...@gmail.com wrote:
On Thu, Dec 11, 2014 at 10:56 AM, Bin.Cheng amker.ch...@gmail.com wrote:
On Wed, Dec 10, 2014 at 9:47 PM, Richard Biener
richard.guent...@gmail.com wrote:
On Fri, Dec 5, 2014 at 1:15 PM, Bin Cheng
CCing Sebastian.
Thanks,
bin
-- Forwarded message --
From: Bin.Cheng amker.ch...@gmail.com
Date: Tue, Dec 16, 2014 at 4:42 PM
Subject: Re: [PATCH PR62178]Improve candidate selecting in IVOPT, 2nd try.
To: Richard Biener richard.guent...@gmail.com
Cc: Bin Cheng bin.ch...@arm.com,
Hi,
This patch fixes an obvious typo which may affect the DDG creation of SMS and
make this optimization produce buggy code.
Bootstrapped on x86_64-suse-linux. Also passed check-gcc test for
aarch64-linux-gnu.
OK for the trunk?
Index: gcc/ddg.c
On December 16, 2014 9:51:25 AM CET, Yangfei (Felix) felix.y...@huawei.com
wrote:
Hi,
This patch fixes an obvious typo which may affect the DDG creation of
SMS and make this optimization produce buggy code.
Bootstrapped on x86_64-suse-linux. Also passed check-gcc test for
aarch64-linux-gnu.
I am not qualified to review the actual code changes, but from the
description it looks good to me.
It adds a EH frame to every function, right? In 64-bit mode there is
no runtime penalty, right? Do you have any idea about binary size
increase? Does gcc build in C++ mode nowadays? It can be a
Cross referenced this patch from
https://code.google.com/p/thread-sanitizer/issues/detail?id=78
On Tue, Dec 16, 2014 at 12:25 PM, Dmitry Vyukov dvyu...@google.com wrote:
I am not qualified to review the actual code changes, but from the
description it looks good to me.
It adds a EH frame to
On Mon, 15 Dec 2014, Jan Hubicka wrote:
On Mon, 15 Dec 2014, Jan Hubicka wrote:
Hi, actually this patch break fortran, I get streaming error in: lto1:
internal compiler error: in streamer_get_pickled_tree apparently picking
error_mark_node for variable constructor results in
On 12/15/2014 11:00 PM, Sergio Durigan Junior wrote:
+# Check if grep supports the '--text' option.
+
+GREP_TEXT_OPT=--text
+if grep --text 21 | grep unrecognized option /dev/null 21 ; then
+ GREP_TEXT_OPT=
+fi
+
That assumes all greps output unrecognized option on
unrecognized options.
On Mon, Dec 15, 2014 at 7:50 PM, Jakub Jelinek ja...@redhat.com wrote:
Hi!
As discussed in the PR, to support exceptions in -fsanitize=thread code,
it is desirable to call __tsan_func_exit also when leaving functions by
means of exceptions.
Adding EH too late sounds too hard to me, so this
On Tue, Dec 16, 2014 at 09:36:33AM +, Pedro Alves wrote:
On 12/15/2014 11:00 PM, Sergio Durigan Junior wrote:
+# Check if grep supports the '--text' option.
+
+GREP_TEXT_OPT=--text
+if grep --text 21 | grep unrecognized option /dev/null 21 ; then
+ GREP_TEXT_OPT=
+fi
+
That
On Dec 15, 2014, at 2:37 PM, Sergio Durigan Junior sergi...@redhat.com wrote:
This weekend I was running GDB's testsuite with many options enabled,
and I noticed that, for some specific configurations (specifically
when testing gdbserver), I was getting the following error:
Binary file
spawn /daten/aranym/gcc/gcc-20141216/Build/gcc/testsuite/g++/../../xg++
-B/daten/aranym/gcc/gcc-20141216/Build/gcc/testsuite/g++/../../
/daten/aranym/gcc/gcc-20141216/gcc/testsuite/g++.dg/abi/covariant4.C
-fno-diagnostics-show-caret -fdiagnostics-color=never -nostdinc++
-I/daten/aranym/gcc/gcc
Hi,
On 12/15/2014 11:25 PM, Jason Merrill wrote:
Why does error recovery fail? I would expect to be able to just drop
the 'friend' and treat it as a normal non-static data member.
I agree, that was my first thought too. Unfortunately we do non-trivial
preparatory work *before* calling
Hi all,
This is a simple patch to add another two ACLE 2.0 predefined macros
into aarch64 backend.
They are __ARM_ALIGN_MAX_PWR and __ARM_ALIGN_MAX_STACK_PWR. Currently,
those two values are hard-wired to 16.
The following clauses from ACLE 2.0 documentation indicate the meaning
of those
On Tue, Dec 16, 2014 at 10:47:06AM +0100, Richard Biener wrote:
On Mon, Dec 15, 2014 at 7:50 PM, Jakub Jelinek ja...@redhat.com wrote:
Hi!
As discussed in the PR, to support exceptions in -fsanitize=thread code,
it is desirable to call __tsan_func_exit also when leaving functions by
On 12/16/2014 10:03 AM, Mike Stump wrote:
input_interrupt, count = 0 c = 0 ('^@')
(The ^@ above is the NUL byte.)
So, either, the tool should not generate 0 in the output, which, is rather
anti-social,
Yeah, this is actually a gdbserver debug output that misses an if debugging
guard;
On 12 December 2014 at 14:48, Wilco Dijkstra wdijk...@arm.com wrote:
This patch generalizes the code alignment and lets each CPU set function,
jump and loop alignment
independently. The defaults for A53/A57 are based the original patch by James
Greenhalgh.
OK for trunk?
ChangeLog:
On 12 December 2014 at 15:19, Wilco Dijkstra wdijk...@arm.com wrote:
Add an override for TARGET_MIN_DIVISIONS_FOR_RECIP_MUL and set the minimum
number of divisions to 2.
This gives ~0.5% speedup on SPECFP2000/2006.
OK for trunk?
ChangeLog:
2014-12-13 Wilco Dijkstra wdijk...@arm.com
On Thu, Nov 13, 2014 at 1:54 PM, Bin Cheng bin.ch...@arm.com wrote:
Hi,
As commented at https://gcc.gnu.org/ml/gcc-patches/2014-09/msg00684.html,
this is a simple patch enabling neon memset inlining on
cortex-a53/cortex-a57 in AArch32 mode.
Test on
Hi again,
On 12/16/2014 11:17 AM, Paolo Carlini wrote:
Hi,
On 12/15/2014 11:25 PM, Jason Merrill wrote:
Why does error recovery fail? I would expect to be able to just drop
the 'friend' and treat it as a normal non-static data member.
I agree, that was my first thought too. Unfortunately we
On December 16, 2014 9:51:25 AM CET, Yangfei (Felix) felix.y...@huawei.com
wrote:
Hi,
This patch fixes an obvious typo which may affect the DDG creation of
SMS and make this optimization produce buggy code.
Bootstrapped on x86_64-suse-linux. Also passed check-gcc test for
On 15/12/2014 17:24, Segher Boessenkool wrote:
On Mon, Dec 15, 2014 at 04:51:14PM +0100, Paolo Bonzini wrote:
Random questions:
1) did you check that it never triggers on e.g. an x86 bootstrap, and
that it doesn't trigger too often on PPC64?
I have checked on my largish connection of
On Tue, Dec 16, 2014 at 01:25:54PM +0400, Dmitry Vyukov wrote:
I am not qualified to review the actual code changes, but from the
description it looks good to me.
It adds a EH frame to every function, right? In 64-bit mode there is
no runtime penalty, right? Do you have any idea about binary
#define DECL_VABD_VAR(VAR) \
be careful with your cut and paste. VABD should probably be VFMA_N here,
although it's purely a naming convention :-)
The v3 patch attached fixed this minor issue. Thanks.
It's OK for me with that change, but I'm not a maintainer.
One
This issue is already fixed by your commit r218760.
Thanks.
Hale.
-Original Message-
From: gcc-patches-ow...@gcc.gnu.org [mailto:gcc-patches-
ow...@gcc.gnu.org] On Behalf Of Hale Wang
Sent: Tuesday, December 16, 2014 10:17 AM
To: 'Vladimir Makarov'; GCC Patches
Subject: RE: patch
Hi!
I have posted yesterday a patch to fix data symbolization using
libbacktrace, this patch is a backport of the change I've sent upstream
that has been committed today.
2014-12-16 Jakub Jelinek ja...@redhat.com
* sanitizer_common/sanitizer_symbolizer_libbacktrace.cc,
Please ignore this one, I will further refine it. Sorry for disturbing!
Thanks,
bin
On Tue, Dec 16, 2014 at 4:42 PM, Bin.Cheng amker.ch...@gmail.com wrote:
On Thu, Dec 11, 2014 at 8:08 PM, Richard Biener
richard.guent...@gmail.com wrote:
On Thu, Dec 11, 2014 at 10:56 AM, Bin.Cheng
As discussed in the PR, this adds folding of (1 A) 1) if in
a eq/ne comparison. The assembly diff on my x86_64 box is
movq%rsp, %rbp
.cfi_def_cfa_register 6
movl%edi, -4(%rbp)
- movl-4(%rbp), %eax
- movl$1, %edx
- movl%eax, %ecx
-
On Mon, Dec 15, 2014 at 10:24:47AM -0600, Segher Boessenkool wrote:
On Mon, Dec 15, 2014 at 04:51:14PM +0100, Paolo Bonzini wrote:
1) did you check that it never triggers on e.g. an x86 bootstrap, and
that it doesn't trigger too often on PPC64?
I have checked on my largish connection of
Hi,
The cprop_hardreg pass does not consider REG_DEAD notes when
propagating, and this causes issues if target specific code uses
dead_or_set_regno_p to know if it can clobber registers.
For example, regcrop transforms
(insn 7 4 8 2 (set (reg:SI 16 r16 [orig:43 D.1617 ] [43])
On Tue, 16 Dec 2014, Marek Polacek wrote:
As discussed in the PR, this adds folding of (1 A) 1) if in
a eq/ne comparison. The assembly diff on my x86_64 box is
movq%rsp, %rbp
.cfi_def_cfa_register 6
movl%edi, -4(%rbp)
- movl-4(%rbp), %eax
- movl
This patch rejects 't' outside user-defined predicates.
2014-12-16 Prathamesh Kulkarni prathamesh.kulka...@linaro.org
* genmatch.c (parser::parsing_match): New.
(parser::parser): Initialize parsing_match to false.
(parser::parse_pattern): Reset parsing_match when
sorry for the noise. I sent it just before our conversation on IRC.
On 16 December 2014 at 19:58, Prathamesh Kulkarni
prathamesh.kulka...@linaro.org wrote:
This patch rejects 't' outside user-defined predicates.
2014-12-16 Prathamesh Kulkarni prathamesh.kulka...@linaro.org
*
On Tue, Dec 16, 2014 at 11:41 AM, Yangfei (Felix) felix.y...@huawei.com wrote:
On December 16, 2014 9:51:25 AM CET, Yangfei (Felix)
felix.y...@huawei.com
wrote:
Hi,
This patch fixes an obvious typo which may affect the DDG creation of
SMS and make this optimization produce buggy code.
I am testing the following patch to properly setup capture_ids for
parsing (match ...
Bootstrap and regtest running on x86_64-unknown-linux-gnu.
Richard.
2014-12-16 Richard Biener rguent...@suse.de
* genmatch.c (parser::parser): Initialize capture_ids.
On Tue, Dec 16, 2014 at 11:23 AM, Jakub Jelinek ja...@redhat.com wrote:
On Tue, Dec 16, 2014 at 10:47:06AM +0100, Richard Biener wrote:
On Mon, Dec 15, 2014 at 7:50 PM, Jakub Jelinek ja...@redhat.com wrote:
Hi!
As discussed in the PR, to support exceptions in -fsanitize=thread code,
it
On 12/16/2014 05:40 AM, Paolo Carlini wrote:
In better detail: grokdeclarator is called, via grokfield, by
cp_parser_member_declaration. The latter stores the friendship
information in a friend_p local flag, which remains true when
grokdeclarator returns.
Maybe check function_declarator_p in
Hi Richard,
Here is updated patch which includes
(1) split critical edges for aggressive if conversion.
(2) delete all stuff related to support of critical edge predication.
(3) only one function - predicate_scalar_phi performs predication.
(4) function find_phi_replacement_condition was deleted
On Tue, Dec 16, 2014 at 03:08:23PM +0100, Richard Biener wrote:
You can use
(for cmp (ne eq)
icmp (eq ne)
(simplify
(cmp (bit_and (lshift integer_onep @0) integer_onep) integer_zerop)
(icmp @0 { build_zero_cst (TREE_TYPE (@0)); })))
to combine both patterns. Btw, shoudln't it
On 12/16/2014 05:09 AM, Andreas Schwab wrote:
covariant4.C:(.text._ZN5ModelD2Ev[_ZN5ModelD5Ev]+0x1e): undefined reference to
`operator delete(void*, unsigned int)'.
Can you determine why this reference isn't being satisfied by libstdc++?
Jason
Hi all,
According to ACLE 2.0, the value of __ARM_SIZEOF_WCHAR_T should be
defined in terms of byte, which means it should be 2 or 4. This patch
corrects the error in arm backend.
arm-none-eabi regression test has been done, no new issues.
Okay for trunk?
Regards,
Renlin Li
On 12/12/2014 06:21 PM, Dominique Dhumieres wrote:
Martin,
Your test g++.dg/ipa/pr64146.C fails on darwin:
grep bind pr64146.C.051i.icf
returns nothing, so the first scan fails, while the second one succeeds.
Dominique
Hello.
You are right, I forgot to decorate test case with:
+/* {
Recent commit 218765 adding sreal::to_double() breaks on AIX due to math.h
being included before _LARGE_FILES and __STDC_FORMAT_MACROS being defined
later in config.h and system.h, respectively.
2014-12-16 Michael Haubenwallner michael.haubenwall...@ssi-schaefer.com
Both config.h and
On 12/11/2014 03:03 PM, Richard Biener wrote:
On Thu, Dec 11, 2014 at 2:49 PM, Martin Liška mli...@suse.cz wrote:
Hello.
In PR64146, for position independent code IPA ICF should be more careful
about thunk creation.
Patch can bootstrap on x86_64-linux-pc and no new regression was seen.
Ready
Hi!
If for RSHIFT_EXPR vr0 is not VR_RANGE or is symbolic, currently we make
the result VARYING, even when we can do much better just by trying to
shift the min and max values down.
Divisions/modulo already handles it similarly, and +/-/ also handle it.
Bootstrapped/regtested on x86_64-linux and
On Tuesday, December 16 2014, Jakub Jelinek wrote:
On Tue, Dec 16, 2014 at 09:36:33AM +, Pedro Alves wrote:
On 12/15/2014 11:00 PM, Sergio Durigan Junior wrote:
+# Check if grep supports the '--text' option.
+
+GREP_TEXT_OPT=--text
+if grep --text 21 | grep unrecognized option
The cprop_hardreg pass does not consider REG_DEAD notes when
propagating, and this causes issues if target specific code uses
dead_or_set_regno_p to know if it can clobber registers.
As explained in the audit trail, it doesn't have to.
Killing regs that are marked REG_DEAD fixes the
Jason Merrill ja...@redhat.com writes:
On 12/16/2014 05:09 AM, Andreas Schwab wrote:
covariant4.C:(.text._ZN5ModelD2Ev[_ZN5ModelD5Ev]+0x1e): undefined reference
to `operator delete(void*, unsigned int)'.
Can you determine why this reference isn't being satisfied by libstdc++?
$ objdump -tC
On Tue, Dec 16, 2014 at 10:16:42AM -0500, Jason Merrill wrote:
On 12/16/2014 05:09 AM, Andreas Schwab wrote:
covariant4.C:(.text._ZN5ModelD2Ev[_ZN5ModelD5Ev]+0x1e): undefined reference
to `operator delete(void*, unsigned int)'.
Can you determine why this reference isn't being satisfied by
On Tue, Dec 16, 2014 at 05:54:04PM +0100, Jakub Jelinek wrote:
On Tue, Dec 16, 2014 at 10:16:42AM -0500, Jason Merrill wrote:
On 12/16/2014 05:09 AM, Andreas Schwab wrote:
covariant4.C:(.text._ZN5ModelD2Ev[_ZN5ModelD5Ev]+0x1e): undefined
reference to `operator delete(void*, unsigned int)'.
OK, thanks.
Jason
Hi,
On 12/16/2014 04:10 PM, Jason Merrill wrote:
On 12/16/2014 05:40 AM, Paolo Carlini wrote:
In better detail: grokdeclarator is called, via grokfield, by
cp_parser_member_declaration. The latter stores the friendship
information in a friend_p local flag, which remains true when
Bin.Cheng wrote:
Multisource/Benchmarks/mafft/pairlocalalign is regressed but I can't
reproduce it in cmd. The running time of compilation of
pairlocalalign.c is too small comparing to the results. I also tried
to invoke it by using RunSafely.sh but no lucky either. So any
documentation on
On 12/16/2014 12:49 PM, Paolo Carlini wrote:
I see what you mean: try to somehow realize that grokdeclarator issued
an error and we are in error recovery. What about directly addressing
NSDMIs, the specific case at issue, thus the below? A bit ad-hoc-ish but
on the other hand should be lighter
On Tue, 16 Dec 2014, Marek Polacek wrote:
On Tue, Dec 16, 2014 at 03:08:23PM +0100, Richard Biener wrote:
You can use
(for cmp (ne eq)
icmp (eq ne)
(simplify
(cmp (bit_and (lshift integer_onep @0) integer_onep) integer_zerop)
(icmp @0 { build_zero_cst (TREE_TYPE (@0));
On Tue, Dec 16, 2014 at 5:04 PM, Michael Haubenwallner
michael.haubenwall...@ssi-schaefer.com wrote:
Recent commit 218765 adding sreal::to_double() breaks on AIX due to math.h
being included before _LARGE_FILES and __STDC_FORMAT_MACROS being defined
later in config.h and system.h, respectively.
On Tue, 16 Dec 2014, Jakub Jelinek wrote:
Hi!
If for RSHIFT_EXPR vr0 is not VR_RANGE or is symbolic, currently we make
the result VARYING, even when we can do much better just by trying to
shift the min and max values down.
Divisions/modulo already handles it similarly, and +/-/ also
PR 61273 points out that for ; false; c - false { doesn't parse
correctly. It's because the false { is incorrectly interpreted as
being a potential composite literal. This patch from Chris Manghane
fixes the parsing bug. Bootstrapped and ran Go testsuite on
x86_64-unknown-linux-gnu. Committed
On Fri, 12 Dec 2014, Jakub Jelinek wrote:
Hi!
-fsanitize=float-cast-overflow sanitization is done in convert.c and calls
there save_expr. Unfortunately, save_expr is a no-go for the C FE, we need
c_save_expr, but as convert.c is shared by all FEs, the only way to arrange
that would be a
On Tue, Dec 16, 2014 at 07:28:22AM -0600, Segher Boessenkool wrote:
and type0 is when it didn't
do either but still ended up with I1 and I2 the same UID (I think it
might be called with the same insn twice; not a good thing, it does
not know how to handle this; and it is really worrisome that
Hello!
When building libgo on CentOS 5.11, following warnings appear:
In file included from /usr/include/fcntl.h:38:0,
from sysinfo.c:6:
/home/uros/gcc-build/gcc/include-fixed/sys/stat.h:317:16: warning:
inline function ‘lstat64’ declared but never defined
__inline__ int
On Tue, Dec 16, 2014 at 5:04 PM, Michael Haubenwallner
michael.haubenwall...@ssi-schaefer.com wrote:
Recent commit 218765 adding sreal::to_double() breaks on AIX due to math.h
being included before _LARGE_FILES and __STDC_FORMAT_MACROS being defined
later in config.h and system.h,
This patch by Chris Manghane fixes a compiler crash when copying an
empty composite literal. This fixes GCC PR 61264. Bootstrapped and
ran Go testsuite on x86_64-unknown-linux-gnu. Committed to mainline.
Ian
diff -r 22a19b12e472 go/expressions.cc
--- a/go/expressions.cc Tue Dec 16 10:50:55
On Tue, Dec 16, 2014 at 11:05 AM, Uros Bizjak ubiz...@gmail.com wrote:
When building libgo on CentOS 5.11, following warnings appear:
In file included from /usr/include/fcntl.h:38:0,
from sysinfo.c:6:
/home/uros/gcc-build/gcc/include-fixed/sys/stat.h:317:16: warning:
inline
Hi
In Google application we hit the same problem as
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63341, so we also need
the patch r215585 for Google/4.9 branch.
It passed following tests:
bootstrap and regression test on x86-64.
regression test on ppc.
Google reference 18687126.
OK for
The fix is already in upstream gcc-4.9 branch? If yes, we just need a merge.
David
On Tue, Dec 16, 2014 at 11:30 AM, Carrot Wei car...@google.com wrote:
Hi
In Google application we hit the same problem as
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63341, so we also need
the patch r215585
Yes, it has been long time since last merge, so it is good idea to do
another merge.
On Tue, Dec 16, 2014 at 11:32 AM, Xinliang David Li davi...@google.com wrote:
The fix is already in upstream gcc-4.9 branch? If yes, we just need a merge.
David
On Tue, Dec 16, 2014 at 11:30 AM, Carrot Wei
On Tue, Dec 16, 2014 at 8:22 PM, Ian Lance Taylor i...@golang.org wrote:
On Tue, Dec 16, 2014 at 11:05 AM, Uros Bizjak ubiz...@gmail.com wrote:
When building libgo on CentOS 5.11, following warnings appear:
In file included from /usr/include/fcntl.h:38:0,
from sysinfo.c:6:
Hi,
while looking on profiles of firefox linktime I noticed that HWI abs functions
are implemented offline that slows down the normalize function.
Those are always win when implemented inline.
Bootstrapped/regtested x86_64-linux, comitted as obvious.
* hwint.c (abs_hwi, absu_hwi): Move
Hi,
On 12/16/2014 07:18 PM, Jason Merrill wrote:
On 12/16/2014 12:49 PM, Paolo Carlini wrote:
I see what you mean: try to somehow realize that grokdeclarator issued
an error and we are in error recovery. What about directly addressing
NSDMIs, the specific case at issue, thus the below? A bit
Hi,
I have optimized inliner to take advantage that the heap is now in sreal and
found that using small numbers leads to excessive time (7% of WPA) spent by
sreal::normalize. This is because the normalization is implemented by loops
that are unnecesary and can account considerable time when
Tobias Burnus wrote:
The Fortran part of the .opt changes has been approved, but for
Enum(), some bits had to be moved to gcc/flag-types.h – and review for
that file is still missing.
https://gcc.gnu.org/ml/gcc-patches/2014-12/msg01068.html
Still pinging the gcc/flag-types.h change.
Hi,
The attached patch fixes the gcc.target/sh/fpchg.c test case. The test
case actually never worked since it requires at least -O option to
output the expected fpchg insn. Moreover, scan-assembler fpchg would
match on the file name string in the asm output. Fixed and tested with
make
This patch by Chris Manghane fixes a Go compiler crash when using a
switch statement where the switch value is an untyped constant. This
is not normally a useful thing to do, but of course the compiler
should not crash. Bootstrapped and ran Go testsuite on
x86_64-unknown-linux-gnu. Committed to
Ping^5 for: https://gcc.gnu.org/ml/gcc-patches/2014-11/msg00519.html
On Tue, Nov 25, 2014 at 6:54 PM, Nathaniel Smith n...@pobox.com wrote:
Ping^4 for: https://gcc.gnu.org/ml/gcc-patches/2014-11/msg00519.html
On Tue, Nov 18, 2014 at 12:53 AM, Nathaniel Smith n...@pobox.com wrote:
Hello,
Hi,
The gcc.target/sh/pr54089-1.c test case started to fail for SH2A. Due
to different combine paths the expected rotcr insn is sometimes not
generated as expected at -O1. Changing it to -O2 produces the expected
insn sequences. Tested with
make -k check-gcc RUNTESTFLAGS=sh.exp=pr54089-1.c
On Tuesday, December 16 2014, Mike Stump wrote:
So, either, the tool should not generate 0 in the output, which, is rather
anti-social, or one should strip the funny characters in a more portable
fashion.
tr and cat -v come to mind; both should be pretty portable.
OK to apply?
Try cat
Hi,
while looking into inliner's behaviour on sreal I noticed that const bulitins
arenot predicted to become constant when their parameters are. This patch
implement that. It is not fully optimal, because cost of the const call will
be still accounted, but it enables better propagation of
Hi,
conversion to sreal makes it possible to compute badness in more streamlined
manner. Together with the sreal::normalize change this patch finally makes
fibheap badness calcualtion to be out of radar in profiling and I hope it
makes it more maintainable by eliminating the issues from roundoff
OK, thanks.
Jason
This patch syncs google-4_9 autofdo implementation to trunk (as much
as possible).
Bootstrapped and passed regression test and performance test.
OK for google-4_9?
Thanks,
Dehao
Index: gcc/auto-profile.c
===
--- gcc/auto-profile.c
ok.
David
On Tue, Dec 16, 2014 at 2:15 PM, Dehao Chen de...@google.com wrote:
This patch syncs google-4_9 autofdo implementation to trunk (as much
as possible).
Bootstrapped and passed regression test and performance test.
OK for google-4_9?
Thanks,
Dehao
Hi,
using min function of fibheap template results in compile error on non-existent
data fields in fibheap node.
Fixed thus.
Bootstrapped/regtested x86_64-linux, comitted.
Honza
* fibonacci_heap.h (min): Return m_data instead of non-existing data.
Index: fibonacci_heap.h
This patch fixes the bug for undefined symbol in AutoFDO build.
Testing on going. OK for google-4_9 branch?
Thanks,
Dehao
Index: gcc/auto-profile.c
===
--- gcc/auto-profile.c (revision 218784)
+++ gcc/auto-profile.c (working copy)
ping...
Thanks,
Dehao
On Tue, Nov 18, 2014 at 2:29 PM, Dehao Chen de...@google.com wrote:
This patch updates ssa and inline summary in the correct location for AutoFDO.
Bootstrapped and passed regression test. OK for trunk?
Thanks,
Dehao
gcc/ChangeLog:
2014-11-18 Dehao Chen
Does it paper over the real bug?
David
On Tue, Dec 16, 2014 at 2:38 PM, Dehao Chen de...@google.com wrote:
This patch fixes the bug for undefined symbol in AutoFDO build.
Testing on going. OK for google-4_9 branch?
Thanks,
Dehao
Index: gcc/auto-profile.c
gcc/ChangeLog:
2014-11-18 Dehao Chen de...@google.com
* auto-profile.c (afdo_annotate_cfg): Invoke update_ssa in the right
place.
(auto_profile): Recompute inline summary after processing cgraph
node.
Index: gcc/auto-profile.c
This patch to the Go frontend changes it to never build hash/equality
functions for structures created for thunks (go and defer statements).
These functions are never needed, and they can cause problems when a
thunk is used to pass an unexported type from a different package to a
function defined
The real bug is in debug info. The callgraph should look like:
foo-D1EV-D4EV
While everything is inlined into foo, but in the inline stack D1EV is missing.
This patch does not fix that problem, but is also reasonable because
if callee's definition is not available, we should not promote the
I committed this to the 4.9 branch.
Ian
On Tue, Dec 16, 2014 at 1:45 PM, Lynn A. Boger
labo...@linux.vnet.ibm.com wrote:
Hi,
Please backport the attached patches to gcc 4.9 based on
https://gcc.gnu.org/ml/gcc-patches/2014-10/msg02994.html.
2014-12-16 Lynn Boger labo...@linux.vnet.ibm.com
On Tue, Dec 16, 2014 at 3:07 PM, Dehao Chen de...@google.com wrote:
The real bug is in debug info. The callgraph should look like:
foo-D1EV-D4EV
While everything is inlined into foo, but in the inline stack D1EV is missing.
Is it DFEed?
This patch does not fix that problem, but is also
On Tue, Dec 16, 2014 at 07:28:22AM -0600, Segher Boessenkool wrote:
I did a run for powerpc64, one for powerpc, and one for x86-64.
The powerpc64 bootstrap was with pre-installed GMP etc.; the others
had those libraries in-tree.
type1 is when try_combine used the ancient combine code to
I have committed this to the 4.9 branch.
Ian
On Thu, Dec 11, 2014 at 12:54 PM, Lynn A. Boger
labo...@linux.vnet.ibm.com wrote:
Hi all,
Please backport the following to gcc 4.9
https://gcc.gnu.org/ml/gcc-patches/2014-10/msg02980.html.
There has been a request to get the fixes that went into
When looking deeper into the problem I discovered that very sometimes
try_combine is called with the same insn for I1 and I2. This is quite
bad since try_combine does not expect that at all. I expect this is
caused by my patches adding a LOG_LINK per register, which can mean
a pair of insns has
On Mon, 15 Dec 2014, Tobias Burnus wrote:
The Fortran part of the .opt changes has been approved, but for Enum(), some
bits had to be moved to gcc/flag-types.h – and review for that file is still
missing.
https://gcc.gnu.org/ml/gcc-patches/2014-12/msg01068.html
The flag-types.h changes are
On Mon, 15 Dec 2014, Eric Botcazou wrote:
Do you have a reason for using fp-bit instead of soft-fp?
Apart from the obvious historical reason, probably not, but recently added
ports (Blackfin, Epiphany) also use it so I'm not sure we want to change it.
I doubt they have any good reason for
On Mon, 15 Dec 2014, Eric Botcazou wrote:
(and you should verify that the port builds cleanly with --enable-werror
-always, for both 32-bit and 64-bit hosts, when building using current
trunk
GCC).
Do you mean a bootstrap of the cross-compiler with --enable-werror-always on
a
1 - 100 of 107 matches
Mail list logo
