Ping
On Sat, Dec 18, 2010 at 3:30 AM, Carrot Wei car...@google.com wrote:
On Fri, Dec 17, 2010 at 4:18 AM, Richard Earnshaw rearn...@arm.com wrote:
On Thu, 2010-12-16 at 14:45 -0800, Carrot Wei wrote:
Hi
Compile the following c code with options -march=armv7-a -mthumb -Os
int foo (int s)
Hi Richard,
What is length used for in the rx port? I don't see any branch shortening
going on here; out of range branches are completely handled by the assembler.
You might be better off simply deleting the length attribute, so that the
compiler skips the bulk of the shorten_branches pass.
On Thu, Mar 17, 2011 at 7:42 PM, Uros Bizjak ubiz...@gmail.com wrote:
On Thu, Mar 17, 2011 at 6:51 PM, H.J. Lu hongjiu...@intel.com wrote:
All i386 ISA options, except for -mavx and -mfma, have Save. This patch
fixes it. OK for trunk and 4.6?
2011-03-17 H.J. Lu hongjiu...@intel.com
On Thu, Mar 17, 2011 at 11:38 PM, Jakub Jelinek ja...@redhat.com wrote:
On Thu, Mar 17, 2011 at 09:07:43PM +0100, Jakub Jelinek wrote:
The following fixes the testcase in cross to hppa-* as well.
I'll bootstrap/regtest it momentarily, ok if it passes?
Now bootstrapped/regtested on
On Fri, 11 Mar 2011, Joseph S. Myers wrote:
On Fri, 11 Mar 2011, Richard Guenther wrote:
Indeed. I tried to let the array case alone (because it's so
complicated) but failed to do so. Appearantly
if (declarator-kind == cdk_array TYPE_QUALS (element_type))
type =
Richard Guenther richard.guent...@gmail.com writes:
It seemed to have disabled linker-plugin support for old binutils
with --with-plugin-ld=/usr/local/bin/gold, explicit -fuse-linker-plugin says
it is not supported. The system linker does not have plugin support
(nor gold). /usr/local has
Richard Guenther rguent...@suse.de writes:
It's sort of historic ... with plugin support for GNU ld we can probably
Could you or someone else who understands this stuff please document it
in install.texi then? Thanks.
drop this flag for 4.7 (eventually I'd even like to enforce the use
of
On Thu, Mar 17, 2011 at 14:21, FX fxcoud...@gmail.com wrote:
Thanks for the review!
- Use the type size_t for tempdirlen as that is the return type of
strlen() and argument type for get_mem().
- You can use a const size_t variable for the length of the string
slash rather than calling
This documents --with-plugin-ld and removes the --enable-gold
documentation (which is bogus now).
Does this look ok?
Thanks,
Richard.
2011-03-18 Richard Guenther rguent...@suse.de
* doc/install.texi (--enable-gold): Remove.
(--with-plugin-ld): Document.
*
Richard Guenther rguent...@suse.de writes:
--- 1666,1676
Enable support for link-time optimization (LTO). This is enabled by
default, and may be disabled using @option{--disable-lto}.
! @item --with-plugin-ld=@var{pathname}
! Enable an alternate linker to be used at link-time
PR48170 is about an ICE in code hoisting when using non-default parameter
value. The ICE is due to an outdated assert; the attached patch simply removes
it. I will commit the patch under the trivial rule if no one objects in 24
hours.
The asserts used to verify an invariant that held true
Hi,
the attached patch uses define_c_enum for defining the unspec constant
as it recently has been changed for other targets as well.
Committed to mainline.
Bye,
-Andreas-
2011-03-18 Andreas Krebbel andreas.kreb...@de.ibm.com
* config/s390/s390.md: Use define_c_enum for the unspec
Hi,
the below is a start on the long overdue task of checking the values
produces by the random distributions. For the time being I'm simply
adapting rather elementary code in GSL, already good enough to enable
regression tests for libstdc++/48114.
I expect that testing along similar lines
On SH, there are many new warnings like
note: non-delegitimized UNSPEC 46 found in variable location
which are reported by dwarf2out.c:const_ok_for_output_1.
The attached patch is a minimal fix. It's tested with
the top level make -k check on cross sh4-unknown-linux-gnu.
Currently, it's hard
Benjamin Kosnik b...@redhat.com writes:
... and I suppose the thinko alone should be fixed in 4_6-branch too,
right?
Yeah. The versioning for the latest symbols has to match, or be done
in a compatible way. But hang on, I'll wait on Rainer and move trunk to
branch if he can confirm trunk is
A problem causing wrong code to be generated for decimal floating
point programs:
[PATCH] Fix PR46399 - missing mode promotion for libcall args - updated
http://gcc.gnu.org/ml/gcc-patches/2011-03/msg00141.html
To my understandig this is a testcase bug. Probably the warning is
expected in that
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Andrew included a patch in this PR to make
optimize_bitfield_assignment_op work again. I've fixed a couple
indention problems and added a couple comments.
Bootstrapped and regression tested on x86_64-unknown-linux-gnu. Applied
to trunk.
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 03/18/11 05:09, Maxim Kuvyrkov wrote:
PR48170 is about an ICE in code hoisting when using non-default parameter
value. The ICE is due to an outdated assert; the attached patch simply
removes it. I will commit the patch under the trivial rule
As described in the PR, gcc.dg/torture/20090618-1.c fails on IRIX 5 and
6 due to undefined references to mempcpy, which doesn't exist in libc.
This might be a target bug. Richard suggested to skip the test, which
this patch does.
Tested with the appropriate runtest invocations. Applied to
This tries to extend the previously posted CCP folding patch by
introducing a generic interface for non-tree-building, GIMPLE SSA
aware folding. The low-level interface for folding regular
operations is
/* Fold the expression composed by *CODEP, TYPE and valueized operands
*OP0P,
*OP1P and
On Fri, 18 Mar 2011, Rainer Orth wrote:
As described in the PR, gcc.dg/torture/20090618-1.c fails on IRIX 5 and
6 due to undefined references to mempcpy, which doesn't exist in libc.
This might be a target bug. Richard suggested to skip the test, which
this patch does.
Tested with the
On Fri, 18 Mar 2011, Rainer Orth wrote:
Richard Guenther rguent...@suse.de writes:
--- 1666,1676
Enable support for link-time optimization (LTO). This is enabled by
default, and may be disabled using @option{--disable-lto}.
! @item --with-plugin-ld=@var{pathname}
!
48118 turned out to be caused by the patch for 23372; if we skip the
ck_rvalue, then we never determine that there is no suitable constructor
to use for copying from a volatile variable. Removing that code fixes
48118.
But that reintroduces 23372 (redundant copying of PODs on argument
My assert that there wouldn't be any more wrappers around the CALL_EXPR
was overconfident; cp_build_function_call_vec still does argument
conversions and uses build_cxx_call, which calls build_cplus_new to wrap
a call returning a class in TARGET_EXPR. That should be fixed, but in
the meantime
I think gengtype is failing me here, but I'm not sure. In PR 48167 we
are failing to link cc1objplus because of undefined references:
cp/parser.o:(.rodata+0x2b18): undefined reference to `gt_ggc_mx_cp_parser'
cp/parser.o:(.rodata+0x2b20): undefined reference to `gt_pch_nx_cp_parser'
cp/tree.o:
I noticed that gcc.dg/vect/slp-multitypes-2.c is failing on Solaris
8/x86 with Sun as:
FAIL: gcc.dg/vect/slp-multitypes-2.c (test for excess errors)
WARNING: gcc.dg/vect/slp-multitypes-2.c compilation failed to produce executable
Excess errors:
Assembler: slp-multitypes-2.c
Hi!
The testcase below is miscompiled on x86_64-linux.
The problem is that try_head_merge_bb uses df_get_bb_dirty
to see if it can use df_get_live_out () info
(through simulate_backwards_to_point) reliably, but as at
-O1 the live problem isn't computed, only lr problem,
df_get_live_out () returns
On Fri, Mar 18, 2011 at 17:23, Kenneth Zadeck zad...@naturalbridge.com wrote:
I believe that this is not the right way to go.
if someone specifies -fcrossjumping, then the pass should turn on live for
the duration of the pass just as ifcvt does. If they ask for crossjumping
you should give
Hi,
On Fri, 18 Mar 2011, Kenneth Zadeck wrote:
I believe that this is not the right way to go.
if someone specifies -fcrossjumping, then the pass should turn on live for the
duration of the pass just as ifcvt does.If they ask for crossjumping you
should give them crossjumping and not
While looking at the check-* targets, I noticed check-consistency.
Further investigation revealed that it is completely unused. It was
introduced in 2001 as poart of the consistency.vlad testsuite (i960-only,
it seems)
2001-02-12 Geoffrey Keating geo...@redhat.com
[...]
* Makefile.in
Paolo,
Splitting this test fixes the failures I described in
http://gcc.gnu.org/ml/libstdc++/2011-03/msg00076.html
OK for trunk?
Thanks. Diego.
2011-03-18 Diego Novillo dnovi...@google.com
* mixed-mode_neg-1.cc: Rename from mixed-mode_neg.cc
* mixed-mode_neg-2.cc: Factor
On 03/18/2011 06:16 PM, dnovi...@google.com wrote:
Paolo,
Splitting this test fixes the failures I described in
http://gcc.gnu.org/ml/libstdc++/2011-03/msg00076.html
OK for trunk?
Sure, it's Ok. A nit, however: we have been using quite consistently the
*_neg suffix for such kind of testcases,
On Fri, Mar 18, 2011 at 13:23, Paolo Carlini paolo.carl...@oracle.com wrote:
On 03/18/2011 06:16 PM, dnovi...@google.com wrote:
Paolo,
Splitting this test fixes the failures I described in
http://gcc.gnu.org/ml/libstdc++/2011-03/msg00076.html
OK for trunk?
Sure, it's Ok. A nit, however:
On 03/18/2011 06:29 PM, Diego Novillo wrote:
Will do. Thanks.
Sorry again - otherwise I have to do it myself anyway: why don't you
split it a more meaningful way, like splitting out all the comparisons?
That would avoid the naming issue in the first place.
Thanks,
Paolo.
On Fri, Mar 18, 2011 at 13:31, Paolo Carlini paolo.carl...@oracle.com wrote:
On 03/18/2011 06:29 PM, Diego Novillo wrote:
Will do. Thanks.
Sorry again - otherwise I have to do it myself anyway: why don't you split
it a more meaningful way, like splitting out all the comparisons? That would
Hi,
java bootstrap currently fails on s390 with:
/build3/gcc-head/libjava/classpath/javax/swing/plaf/basic/BasicLookAndFeel.java:1649:0:
internal compiler error: in
dwarf2out_var_location, at dwarf2out.c:21972
Please submit a full bug report,
with preprocessed source if appropriate.
See
Hi,
the attached patch adds handling for GOTOFF unspecs to
delegitimize_address.
A lot of note: non-delegitimized UNSPEC 112 messages disappear with
this. However not all of them.
Committed to mainline.
Bye,
-Andreas-
2011-03-18 Andreas Krebbel andreas.kreb...@de.ibm.com
*
On Fri, Mar 18, 2011 at 17:31, Jakub Jelinek ja...@redhat.com wrote:
On Fri, Mar 18, 2011 at 12:23:11PM -0400, Kenneth Zadeck wrote:
I believe that this is not the right way to go.
if someone specifies -fcrossjumping, then the pass should turn on
live for the duration of the pass just as
On Mon, 14 Mar 2011, Jakub Jelinek wrote:
--- c99status.html6 Apr 2010 11:34:44 - 1.55
+++ c99status.html14 Mar 2011 13:52:00 - 1.56
@@ -21,6 +21,7 @@ paragraph 6) do not count as library iss
pThis page describes the C99 support in mainline GCC, not in any
I have imported the latest config.guess and config.sub from upstream
repository and committed to both GCC Subversion repository and SRC CVS
repository. Among other changes, this allows configure to recognize
AIX 7.1.
* config.guess: Update to version 2011-02-02
* config.sub:
The Obj-C++ FE is kind of weird as it shares files from cp/ and objc/,
so I'm missing some other connection I need to make to fix this.
Any ideas?
As far as I can see, the problem is that header files (such as cp/parser.h)
generate GC stuff that gets put into gtype-{lang}.h, where {lang} is
On Tue, 15 Mar 2011, Jason Merrill wrote:
+/* Likewise, for constexpr function call evaluations. N3225 specifies a
+ minimum of 512. */
+int max_constexpr_depth = 512;
Since all you're doing in the option handler is setting a global variable,
it's better to declare the option in the
2011/3/18 Georg-Johann Lay a...@gjlay.de:
Is it tested for regressions ?
Denis.
I ran tests against svn 170942 (latest 4.7.0 snapshot). Besides
timestamps, the diff looks like this:
1435a1436,1437
XPASS: gcc.dg/sibcall-3.c execution test
XPASS: gcc.dg/sibcall-4.c execution test
On Fri, Mar 18, 2011 at 06:45:57PM +0100, Paolo Bonzini wrote:
Is the live problem so much more useful for crossjumping than lr problem?
All it wants to prove is if it can safely move a couple of instructions
across some other instructions, and the live_union bitmap that is computed
using
i see your point.I had forgotten that crossjumping is not really a
pass, it is more of an infectious agent.
On 03/18/2011 01:45 PM, Paolo Bonzini wrote:
On Fri, Mar 18, 2011 at 17:31, Jakub Jelinekja...@redhat.com wrote:
On Fri, Mar 18, 2011 at 12:23:11PM -0400, Kenneth Zadeck wrote:
I
On Wed, 16 Mar 2011, Tom Tromey wrote:
It seems ok to me to accept it as a GNU extension.
But, it still doesn't warn with -std=c89 or -std=c99.
-pedantic seems too heavy for this, to me.
It is not the function of -std to enable diagnostics for extensions;
that's the function of -pedantic.
On Fri, 18 Mar 2011, Richard Guenther wrote:
What I think is safe in grokdeclarator is using TYPE_MAIN_VARIANT here if
*either* the type given in the declaration specifiers is an array type
(TREE_CODE (type) == ARRAY_TYPE, as in your previous patch) *or* the first
declarator that is
On Fri, 18 Mar 2011 18:59:47 +0100 (CET)
Nicola Pero nicola.p...@meta-innovation.com wrote:
The Obj-C++ FE is kind of weird as it shares files from cp/ and objc/,
so I'm missing some other connection I need to make to fix this.
This all looks ugly. Anyhow, until someone refactors
Hello Joseph,
* Joseph S. Myers wrote on Fri, Mar 18, 2011 at 12:12:09PM CET:
Bootstrapped with no regressions on x86_64-unknown-linux-gnu. OK to
commit?
The build system changes look OK from a build system technical POV,
with a couple of comments below.
2011-03-17 Joseph Myers
On Fri, 18 Mar 2011, Ralf Wildenhues wrote:
@@ -744,7 +744,6 @@ case ${target} in
libgloss_dir=cris
;;
crx-*-*)
-noconfigdirs=$noconfigdirs target-libstdc++-v3 target-mudflap
${libgcj}
;;
Why not also remove the line before and after this one?
Is that
Hi,
Like this? Tested on my insanely deep tree and on mainline on x86_64.
Sure.
Thanks,
Paolo
* Joseph S. Myers wrote on Fri, Mar 18, 2011 at 08:46:56PM CET:
On Fri, 18 Mar 2011, Ralf Wildenhues wrote:
@@ -744,7 +744,6 @@ case ${target} in
libgloss_dir=cris
;;
crx-*-*)
-noconfigdirs=$noconfigdirs target-libstdc++-v3 target-mudflap
${libgcj}
;;
I have cherry picked rev 171161
(http://codereview.appspot.com/4273077/) from trunk to fix the last
testsuite failure we were getting in libstdc++. I've committed the
change to google/main. I will commit the same change to
google/integration and google/gcc-4_6.
Diego.
+ /* cp/parser.h gives gt-cp-parser.h for cp/parser.c ! */
+ { DIR_PREFIX_REGEX cp/parser\\.h$,
+ REG_EXTENDED, NULL_REGEX,
+ gt-cp-parser.h, cp/parser.c, NULL_FRULACT },
But cp/parser.c also gets its own gt-cp-parser.h. Won't they
conflict? Doesn't this mean that cp/parser.h
Hello!
Just a trivial cleanup, no functional changes.
2011-03-18 Uros Bizjak ubiz...@gmail.com
* config/i386/i386.md (floatSSEMODEI24:modeX87MODEF:mode2):
Rewrite using indirect functions.
(lwp_slwpcb): Ditto.
(avx_vextractf128mode): Ditto.
On Fri, Mar 18, 2011 at 12:51 PM, Eric Botcazou ebotca...@adacore.com wrote:
Is this patch OK for trunk?
This has always worked for the other ports AFAIK so I don't think we should
disable it without evaluating the impact on them. If reload has so many
problems with x32, maybe more
On 03/17/2011 03:28 PM, H.J. Lu wrote:
It will require very extensive changes to x86 backend since stack pointer
and frame pointer are really in DImode for x32 as x32 process runs in
64bit mode. If you can make it to work, please let me know.
Yes, the stack pointer uses all 64 bits for e.g.
On 03/18/2011 01:56 PM, H.J. Lu wrote:
On Fri, Mar 18, 2011 at 12:51 PM, Eric Botcazou ebotca...@adacore.com wrote:
Is this patch OK for trunk?
This has always worked for the other ports AFAIK so I don't think we should
disable it without evaluating the impact on them. If reload has so many
On 03/18/2011 01:40 PM, Uros Bizjak wrote:
if (X87MODEF:MODEmode == SFmode)
- insn = gen_truncxfsf2 (operands[0], reg);
+ insn = gen_truncxfsf2;
else if (X87MODEF:MODEmode == DFmode)
- insn = gen_truncxfdf2 (operands[0], reg);
+ insn = gen_truncxfdf2;
On Fri, 18 Mar 2011, Ralf Wildenhues wrote:
* Joseph S. Myers wrote on Fri, Mar 18, 2011 at 08:46:56PM CET:
On Fri, 18 Mar 2011, Ralf Wildenhues wrote:
@@ -744,7 +744,6 @@ case ${target} in
libgloss_dir=cris
;;
crx-*-*)
-noconfigdirs=$noconfigdirs
On Fri, Mar 18, 2011 at 16:17, Nicola Pero
nicola.p...@meta-innovation.com wrote:
+ /* cp/parser.h gives gt-cp-parser.h for cp/parser.c ! */
+ { DIR_PREFIX_REGEX cp/parser\\.h$,
+ REG_EXTENDED, NULL_REGEX,
+ gt-cp-parser.h, cp/parser.c, NULL_FRULACT },
But cp/parser.c also gets its
This week's London WG14 meeting agreed to disallow the use of typedefs
in declaring anonymous structure and union fields, as per N1549. This
patch implements this, restricting the use of typedefs in this case to
-fms-extensions and -fplan9-extensions as in 4.5 and earlier releases.
Bootstrapped
OK.
Jason
On Fri, Mar 18, 2011 at 09:16:55PM +, Joseph S. Myers wrote:
This week's London WG14 meeting agreed to disallow the use of typedefs
in declaring anonymous structure and union fields, as per N1549. This
patch implements this, restricting the use of typedefs in this case to
-fms-extensions
On Fri, Mar 18, 2011 at 2:03 PM, Richard Henderson r...@redhat.com wrote:
On 03/17/2011 03:28 PM, H.J. Lu wrote:
It will require very extensive changes to x86 backend since stack pointer
and frame pointer are really in DImode for x32 as x32 process runs in
64bit mode. If you can make it to
On 03/18/2011 02:56 PM, H.J. Lu wrote:
X86 backend uses Pmode for hardware pointer size. Changes
it to 32bit for x32, which is really 64bit process, breaks many
assumptions of x86 backend. push/pop/call is just the tip of
the iceberg.
Please enumerate many assumptions. I can't think of any
On 3/18/2011 2:52 PM, Jakub Jelinek wrote:
This week's London WG14 meeting agreed to disallow the use of typedefs
in declaring anonymous structure and union fields, as per N1549. This
patch implements this, restricting the use of typedefs in this case to
-fms-extensions and
* Joseph S. Myers wrote on Fri, Mar 18, 2011 at 10:09:04PM CET:
On Fri, 18 Mar 2011, Ralf Wildenhues wrote:
* Joseph S. Myers wrote on Fri, Mar 18, 2011 at 08:46:56PM CET:
On Fri, 18 Mar 2011, Ralf Wildenhues wrote:
Why not also remove the line before and after this one?
Is that
On 03/18/2011 02:51 PM, H.J. Lu wrote:
See analysis in:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47502
You're patching the wrong place. See can_combine_p,
which can test for specific sources, rather than
cant_combine_insn_p which has no access to sources.
r~
On Fri, Mar 18, 2011 at 03:32:45PM -0700, H.J. Lu wrote:
On Fri, Mar 18, 2011 at 3:18 PM, Richard Henderson r...@redhat.com wrote:
On 03/18/2011 02:56 PM, H.J. Lu wrote:
X86 backend uses Pmode for hardware pointer size. Changes
it to 32bit for x32, which is really 64bit process, breaks many
On Fri, Mar 18, 2011 at 06:37:38PM +0100, Andreas Krebbel wrote:
Hi,
the attached patch adds handling for GOTOFF unspecs to
delegitimize_address.
A lot of note: non-delegitimized UNSPEC 112 messages disappear with
this. However not all of them.
Please look at
On Fri, 18 Mar 2011, Jakub Jelinek wrote:
On Fri, Mar 18, 2011 at 09:16:55PM +, Joseph S. Myers wrote:
This week's London WG14 meeting agreed to disallow the use of typedefs
in declaring anonymous structure and union fields, as per N1549. This
patch implements this, restricting the
On Fri, 18 Mar 2011, Ralf Wildenhues wrote:
- and the general reason for not
removing the empty cases is that there are plenty of such cases already
there, some for truly ancient targets.
I can understand keeping them when there are later case matches that
might match the
Thomas Koenig wrote:
+ if (!(*e)-value.function.esym-attr.pure
+ !(*e)-value.function.esym-attr.implicit_pure
+ !(*e)-value.function.esym-attr.elemental)
+ return 0;
I have not followed the discussion nor have I fully read the patch, but
what's the reason for
This patch makes conditional macros that are used in the PowerPC and SPU ports
not be 'defined' for the #ifdef, #ifndef, and #if defined(...) tests. The way
the PowerPC and SPU uses conditional macros is provide conditional keywords to
support the vector systen (i.e. if the next token after
This week's London WG14 meeting decided that typedef redefinition
should not be allowed in the case of variably modified types. This
patch implements this, giving an error for such redefinitions.
Bootstrapped with no regressions on x86_64-unknown-linux-gnu. Applied
to mainline. Any comments
Hi,
On Fri, Mar 11, 2011 at 03:30:53PM +0100, Jan Hubicka wrote:
Index: src/gcc/cgraph.c
===
--- src.orig/gcc/cgraph.c
+++ src/gcc/cgraph.c
@@ -2495,11 +2495,11 @@ cgraph_add_new_function (tree fndecl, bo
case
On Mar 18, 2011, at 10:59 AM, Nicola Pero wrote:
This all looks ugly. Anyhow, until someone refactors everything, the
following
patch fixes ObjC++ bootstrap in trunk by processing cp/parser.h in the same
way
as the other C++ headers are processed. :-)
Ok to commit ?
Ok for the
78 matches
Mail list logo
