On 2011/3/30 上午 12:23, Richard Earnshaw wrote:
On Tue, 2011-03-29 at 22:53 +0800, Chung-Lin Tang wrote:
On 2011/3/29 下午 10:26, Richard Earnshaw wrote:
On Tue, 2011-03-29 at 18:25 +0800, Chung-Lin Tang wrote:
On 2011/3/24 06:51 PM, Richard Earnshaw wrote:
On Thu, 2011-03-24 at 12:56 +0900,
Richard Henderson r...@redhat.com writes:
On 03/29/2011 06:21 AM, Richard Sandiford wrote:
- enum machine_mode mode0 = insn_data[(int) icode].operand[1].mode;
- enum machine_mode mode1 = insn_data[(int) icode].operand[2].mode;
- enum machine_mode tmp_mode;
+ enum machine_mode xmode0 =
Richard Sandiford richard.sandif...@linaro.org writes:
The ??? is saying that the V8QI-derived MEM is really a 3-byte access,
not a 4-byte (SI) access, and so on. The comment makes the mode sound
like a representational niceity, but really, there's no such thing as
a conservatively
Hi,
With this patch a data-ref is marked as unconditionally read or
written also if its adjacent field is read or written unconditionally
in the loop.
My concern is that this is not safe enough, even though the fields
have to be non-pointers and non-aggregates, and this optimization is
applied
On Wed, 2011-03-30 at 15:35 +0800, Chung-Lin Tang wrote:
On 2011/3/30 上午 12:23, Richard Earnshaw wrote:
On Tue, 2011-03-29 at 22:53 +0800, Chung-Lin Tang wrote:
On 2011/3/29 下午 10:26, Richard Earnshaw wrote:
On Tue, 2011-03-29 at 18:25 +0800, Chung-Lin Tang wrote:
On 2011/3/24 06:51
On Tue, Mar 29, 2011 at 5:53 PM, Jakub Jelinek ja...@redhat.com wrote:
Hi!
I've left these 4 out because I thought they should be fine given
that the other operands are equal. But the testcase shows I was wrong.
Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok
for
On Tue, Mar 29, 2011 at 7:00 PM, Aldy Hernandez al...@redhat.com wrote:
[Language lawyers, please correct me if I have mis-interpreted the upcoming
standard in any way.]
In the C++ memory model, contiguous bitfields comprise a single memory
location, so it's fair game to bit twiddle them when
Hi Vlad,
First, I want to echo H-P's thanks for tackling this area. I just wondered:
Vladimir Makarov vmaka...@redhat.com writes:
The following patch is to solve PR48336, PR48342, PR48345. The
profitable hard regs exclude hard regs which are prohibited for the
corresponding allocno mode.
Hi,
I checked this pre-approved patch into trunk/4.6/4.5/4.4.
H.J.
---
Index: ChangeLog
===
--- ChangeLog (revision 171717)
+++ ChangeLog (working copy)
@@ -1,3 +1,9 @@
+2011-03-30 H.J. Lu hongjiu...@intel.com
+
+ PR
On Wed, 2011-03-30 at 12:38 +0800, Chung-Lin Tang wrote:
On 2011/3/30 06:35 AM, Ramana Radhakrishnan wrote:
Hi Carrot,
On 26/03/11 15:14, Carrot Wei wrote:
Index: arm.md
===
--- arm.md(revision 171337)
+++ arm.md
On 30 March 2011 12:59, Richard Guenther richard.guent...@gmail.com wrote:
On Wed, Mar 30, 2011 at 11:13 AM, Ira Rosen ira.ro...@linaro.org wrote:
Hi,
With this patch a data-ref is marked as unconditionally read or
written also if its adjacent field is read or written unconditionally
in the
This patch adds a NegativeAlias .opt facility, for one option to be
marked as an alias for the negative form of another option. There are
quite a lot of options like that, and this is of particular use where
it allows option handler code to be removed because the aliases can
now be handled purely
On 30 March 2011 14:41, Richard Guenther richard.guent...@gmail.com wrote:
On Wed, Mar 30, 2011 at 2:22 PM, Ira Rosen ira.ro...@linaro.org wrote:
On 30 March 2011 12:59, Richard Guenther richard.guent...@gmail.com wrote:
On Wed, Mar 30, 2011 at 11:13 AM, Ira Rosen ira.ro...@linaro.org wrote:
Hi Guys,
I am applying the patch below to sync the RX backend with the 4.6
branch. In practice this means bringing in the peepholes to combine
extending loads and simple arithmetic expressions, and adjusting the
memory move cost cost for stores. (The insn length attribute is
needed
The memory model is not implementable on strict-alignment targets
that do not have a byte store operation. But we previously said that ;)
Yes. I think we should issue an error when we have such a target and
the user tries -fmemory-model=c++0x. However, how many strict-alignment
targets
On Wed, Mar 30, 2011 at 4:11 PM, Aldy Hernandez al...@redhat.com wrote:
The memory model is not implementable on strict-alignment targets
that do not have a byte store operation. But we previously said that ;)
Yes. I think we should issue an error when we have such a target and the
user
On Tue, 2011-03-29 at 20:55 +, Joseph S. Myers wrote:
This patch continues cleaning up the toplevel configure.ac, in particular
as regards cases handling GCC libraries for targets where GCC is no longer
(or never was) supported.
The principle there, as discussed for the original
On Tue, 2011-01-18 at 16:56 +, Yufeng Zhang wrote:
This patch replaces define_constants in the ARM backend with
define_c_enum for defining the available indexes for the
unspecs/unspecvs expressions. This improves the readability of the
intermediate dumps for machine-specific operations.
The attached patch allows long jumps to __mcount, defines
PROFILE_BEFORE_PROLOGUE and ensures ASM_OUTPUT_REG_PUSH pre-decrements the
stack pointer.
2011-03-30 Stuart Henderson stuart.hender...@analog.com
From Bernd Schmidt
* config/bfin/bfin.h (FUNCTION_PROFILER): Take
While working on the dragonegg plugin I noticed that the Fortran front-end
declares frexp with the parameters the wrong way round. Instead of
double frexp(double x, int *exp);
it is declared as
double frexp(int *exp, double x);
This is fairly harmless but might as well be fixed, so here is a
On Tue, 2011-03-29 at 15:38 +0100, Richard Sandiford wrote:
This ARM patch allows *UNSPEC_* constants to be printed in dump files.
It's very much a target decision whether this is worth doing, but just
in case...
...tested on arm-linux-gnueabi. OK to install?
Richard
gcc/
*
On Mar 30, 2011, at 5:41 AM, gcchelp.5.ad...@0sg.net wrote:
Tests that now fail, but worked before:
g++.dg/warn/Wstrict-aliasing-float-ref-int-obj.C (test for warnings, line 7)
Tests that now work, but didn't before:
g++.dg/warn/Wstrict-aliasing-float-ref-int-obj.C (test for warnings,
On Wed, Mar 30, 2011 at 4:23 PM, Jeff Law l...@redhat.com wrote:
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
As originally discussed here:
http://gcc.gnu.org/ml/gcc-patches/2010-05/msg00987.html
Changes since the original submission:
Per Richi's suggestion I changed the patch to use a
Richard Earnshaw rearn...@arm.com writes:
On Tue, 2011-03-29 at 15:38 +0100, Richard Sandiford wrote:
This ARM patch allows *UNSPEC_* constants to be printed in dump files.
It's very much a target decision whether this is worth doing, but just
in case...
...tested on arm-linux-gnueabi. OK
On Fri, 2011-02-18 at 10:56 +, Ian Bolton wrote:
Is the following patch OK for trunk?
2011-01-26 Ian Bolton ian.bol...@arm.com
* testsuite/gcc.dg/20100906-1.c: Use -fno-short-enums
option for target arm_eabi.
Index: gcc/testsuite/gcc.dg/20100906-1.c
On Mar 30, 2011, at 7:40 AM, Richard Guenther wrote:
Is forcing word-alignment too big of a hammer, or will the users for these
architectures be content with having no support for the C++0x memory model?
I think a memory model that cannot be reasonably (read: also fast) implemented
on all HW
On 03/30/2011 01:53 AM, Richard Sandiford wrote:
This comes back to the point that we really should know up-front what
modes op0 and op1 actually have. (The thing I left as a future clean-up.)
Until then, the process implemented by yesterday's patch was supposed to be:
- work out what
On Mon, 2011-02-14 at 14:20 +, Andrew Stubbs wrote:
This patch is a rework of an old one:
http://gcc.gnu.org/ml/gcc-patches/2010-06/msg01080.html
The ARM parts of that patch were approved, but the target independent
parts were never reviewed (AFAICT), and the patch no longer applies.
On Thu, 2011-02-10 at 15:29 +, Andrew Stubbs wrote:
I'm posting this patch on behalf of Jie.
The patch simply prevents a test failure on non-thumb2 targets.
OK?
Andrew
2010-09-30 Jie Zhang j...@codesourcery.com
gcc/testsuite/
* gcc.target/arm/neon-thumb2-move.c:
On Wed, 30 Mar 2011, Richard Earnshaw wrote:
2011-03-29 Joseph Myers jos...@codesourcery.com
(arm-*-coff): Don't disable libgcj.
(arm*-*-linux-gnueabi): Remove useless assignment.
(arm-*-riscix*): Don't disable libgcj.
RISC iX support was removed from GCC years
As people noted one computes UIDs for PHIs and one doesn't. Now I'm
working on sth that needs it for PHIs, thus the following fix.
We have to exclude LTO as it re-builds PHIs and gets confused with
inconsistent numbering.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
Hi!
MEM_REFs which can represent type punning on lhs don't force
non-gimple types to be addressable. This causes various problems
in the expander, which wasn't prepared to handle that.
This patch tries to fix what I've found and adds a bunch of
testcases. The original report was with just -O2
The following patch is to fix an arm bootstrap failure described in
http://gcc.gnu.org/ml/gcc/2011-03/msg00499.html
when variable mode is set but not used because it is used when macro
HONOR_REG_ALLOC_ORDER is defined.
I found that 2 variables mode in different scopes is defined in function
On Mon, Mar 28, 2011 at 6:11 PM, Vladimir Makarov vmaka...@redhat.com wrote:
On 03/27/2011 07:25 PM, Vladimir Makarov wrote:
I submitted the following patch. The patch contains original
patches from ira-improv branch, changes addressing all Keneth Zadeck's
comments in
On 03/30/2011 06:19 AM, Richard Sandiford wrote:
Hi Vlad,
First, I want to echo H-P's thanks for tackling this area. I just wondered:
Vladimir Makarovvmaka...@redhat.com writes:
The following patch is to solve PR48336, PR48342, PR48345. The
profitable hard regs exclude hard regs which are
On 03/30/2011 06:21 PM, Michael Matz wrote:
Okay for trunk? (regstrapping on x86_64-linux in progress)
OK. Thanks for tracing the bug. I think the issue could be the reason
for the elusive PR 47516 - thus, you might consider adding that PR to
the ChangeLog and close the PR afterwards.
On 03/29/2011 08:49 AM, Henderson, Stuart wrote:
(match_operand:SI 2 immediate_operand )))]
{
- if (INTVAL (operands[2]) != 16)
+ if (GET_CODE (operands[2]) != CONST_INT || INTVAL (operands[2]) != 16)
FAIL;
Perhaps use const_int_operand instead of
On 03/29/2011 05:57 AM, Henderson, Stuart wrote:
- operands[1] = bfin_gen_compare (operands[1], SImode);
+ operands[1] = bfin_gen_compare (operands[1], GET_MODE (operands[0]));
FWIW, you can use MODEmode to get the proper value
without having to read it from the operand.
r~
Hi,
On Wed, 30 Mar 2011, Tobias Burnus wrote:
On 03/30/2011 06:21 PM, Michael Matz wrote:
Okay for trunk? (regstrapping on x86_64-linux in progress)
OK. Thanks for tracing the bug. I think the issue could be the reason for the
elusive PR 47516 - thus, you might consider adding that PR to
On 03/30/2011 07:33 PM, Michael Matz wrote:
Hi,
On Wed, 30 Mar 2011, Tobias Burnus wrote:
On 03/30/2011 06:21 PM, Michael Matz wrote:
Okay for trunk? (regstrapping on x86_64-linux in progress)
OK. Thanks for tracing the bug. I think the issue could be the reason for the
elusive PR 47516 -
I committed this patch to libgo to add some missing Makefile
dependencies. Bootstrapped on x86_64-unknown-linux-gnu. Committed to
mainline.
Ian
diff -r 58ba48318471 libgo/Makefile.am
--- a/libgo/Makefile.am Wed Mar 30 08:17:36 2011 -0700
+++ b/libgo/Makefile.am Wed Mar 30 10:35:55 2011 -0700
The testcase in 48281 started breaking in 4.6.0 because of my change to
finish_compound_literal to stop putting constant compound literals into
static variables, because it was interfering with constexpr.
First I noticed that the crash was due to non-constant CONSTRUCTORs with
TREE_CONSTANT
On Wed, Mar 30, 2011 at 08:02:38AM -0700, H.J. Lu wrote:
Hi,
Currently, we limit XVECEXP to 26 elements in machine description
since we use letters 'a' to 'z' to encode them. I don't see any
reason why we can't go beyond 'z'. This patch removes this restriction.
Any comments?
That was
Hello,
This is about the line program emitted by the compiler into the
.debug_line section, without optimization. In the example
accompanying the patch below, at the beginning of the function f, the
compiler emits two .loc asm directives that are identical. The first
one is right before the
I merged mainline revision 171737 onto the gccgo branch.
Ian
A couple of missed cases.
Tested x86_64-pc-linux-gnu, applying to trunk and 4.6.
commit cea17025eb232f3931ce34e16dedff7fd42d2478
Author: Jason Merrill ja...@redhat.com
Date: Wed Mar 30 15:17:27 2011 -0400
PR c++/48369
* semantics.c (potential_constant_expression_1): Handle
Hi!
We ICE on the attached testcase, because combiner changes mode of
a pseudo (which was set just once and used once plus in debug insns)
from SImode to QImode, but the uses in debug insns aren't adjusted.
Combiner has code to adjust this too (propagate_for_debug), but only updates
debug insns
Hello!
Attached ports go to ALPHA architecture.
There are however several problems with the build:
a) Bootstrap compare failure in the gcc/go directory due to binutils
bug [1], fixed in latest binutils SVN, use --disable-bootstrap
b) alpha doesn't define struct user_regs_struct from which type
On Wed, 2 Mar 2011, Richard Earnshaw wrote:
Could you remove the documentation entries for the hard/soft-float
aliases please? They're really only there for legacy reasons.
Is this the change you want here?
2011-03-30 Joseph Myers jos...@codesourcery.com
* config/arm/arm.opt
On Wed, Mar 30, 2011 at 9:58 PM, Uros Bizjak ubiz...@gmail.com wrote:
Hello!
Attached ports go to ALPHA architecture.
There are however several problems with the build:
a) Bootstrap compare failure in the gcc/go directory due to binutils
bug [1], fixed in latest binutils SVN, use
On 30 Mar 2011, at 21:11, Joseph S. Myers jos...@codesourcery.com wrote:
On Wed, 2 Mar 2011, Richard Earnshaw wrote:
Could you remove the documentation entries for the hard/soft-float
aliases please? They're really only there for legacy reasons.
Is this the change you want here?
Rainer Orth r...@cebitec.uni-bielefeld.de writes:
2011-03-24 Rainer Orth r...@cebitec.uni-bielefeld.de
go:
* go.test/go-test.exp (go-set-goarch): Use sparc64 for 64-bit SPARC.
diff -r de1b3baf021b gcc/testsuite/go.test/go-test.exp
--- a/gcc/testsuite/go.test/go-test.exp
I brought over this patch to the Go testsuite from the master
repository. This avoids an explicit list of supported architectures, so
that we don't have to add each gcc architecture to the list. Ran Go
testsuite on x86_64-unknown-linux-gnu. Committed to mainline.
Ian
Index:
This patch implements the writing of DECL_LANG_SPECIFIC fields. It's
needed to write global_namespace. The reading part is still
incomplete, but I wanted to flush this out before it got too big.
The main changes:
- We can now write out references to TEMPLATE_DECLs. They are stored
in the
The guality tests are failing/passing intermittently. This confuses
our builders which expect 0 failures and 0 xpasses. When they work,
it's not that they are working but dejagnu failed to launch gdb:
-
spawn [open
Rainer Orth r...@cebitec.uni-bielefeld.de writes:
* go-test.exp wasn't updated for the change from sparcv9 to sparc64.
While I still don't agree with the new name, at least the two should
be consistent.
I committed this earlier today.
* env.go needs to accept sparc and sparc64.
I fixed
Rainer Orth r...@cebitec.uni-bielefeld.de writes:
Apart from the lack of wait4, libgo on IRIX 6.5 contained an undefined
reference to strerror_r. This patch provides a replacement, based on
gnulib/lib/strerror_r.c, but massively simplified.
I addressed this in a different way, with a new
Hi,
This is the first cleanup for cprop.c. These cleanups are only
possible now, thanks to splitting the CPROP code out from gcse.c
The first change is to not treat unfolded conditions as constant in
gcse_constant_p(). This never happens because:
- during hash table building any such condition
Rainer Orth r...@cebitec.uni-bielefeld.de writes:
Currently, libgo uses wait4 unconditionally, which is missing on IRIX
6.5. Fortunately, the rusage * arg is used nowhere, so I've decide to
replace wait4 by waitpid, which seems to be considerably more portable.
Thanks for the patch, but I
Hi,
this is a regression present for x86-64 on mainline and 4.6 branch with the
options -Os -mpreferred-stack-boundary=5 -fstack-check -fno-omit-frame-pointer.
This improbable combination of options is necessary because you need to have
stack checking + stack realignment +
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
Sigh. I changed a reg_equiv_address to reg_equiv_mem when I changed
from explicit VEC_index references to macros. That caused all kinds of
interesting problems for ia32 :-)
I'm also fixing a comment typo that was introduced around the same time.
-BEGIN PGP SIGNED MESSAGE-
Hash: SHA1
On 03/30/11 10:51, Richard Sandiford wrote:
Nice cleanup thanks. Just noticed a couple of things:
Jeff Law l...@redhat.com writes:
*** struct reload
*** 100,106
int inc;
/* A reg for which reload_in is the equivalent.
On Tue, 29 Mar 2011, Joseph S. Myers wrote:
Other cleanups here:
cris*-*-none acts just like cris*-*-elf in
config.gcc so it's appropriate to make the * subcase of cris*-*-* act
like the -elf case;
'k.
mmix-*-* disabled libgloss, i.e. libgloss for the
host, which is never built anyway.
Hi,
this came up when looking into why the static ctors contain useless trees
(like casts). We can simply canonicalize them while varpool analyzes
pending decls. It'll look at initialzers once, where we can gimplify
them. This requires making canonicalize_constructor_val be able to be
On Wed, Mar 30, 2011 at 21:16, Michael Matz m...@suse.de wrote:
* tree.c (decl_init_priority_insert): Don't create entry for
default priority.
(decl_fini_priority_insert): Ditto.
(fields_compatible_p, find_compatible_field): Remove.
* tree.h
Hi,
I noticed this while working on early-merging LTO. The DECL_VINDEX slot
of FUNCTION_DECLs is supposed to hold the numeric index of the vtable
slot if it's a virtual function. During parsing the C++ frontend uses it
to hold a reference to itself, which then later is supposed to be
On Wed, Mar 30, 2011 at 21:32, Michael Matz m...@suse.de wrote:
* tree.c (free_lang_data_in_decl): Zero DECL_VINDEX if it's not
an integer.
* tree.h (tree_decl_non_common.vindex): Adjust comment.
OK.
Diego.
No functional changes. Just some refactoring to stop mixing up
reading and writing routines in the same file.
Tested on x86_64.
cp/ChangeLog.pph
2011-03-30 Diego Novillo dnovi...@google.com
* Make-lang.in (CXX_AND_OBJCXX_OBJS): Add cp/pph-streamer-out.o
and
Hi,
I fear I wasn't as thorough in also splitting this one into several
patches, but the different cleanups are at least mostly in different
files. They are:
* lto-lang remembers all builtin decls in a local list, to be returned
by the getdecls langhook. But as we have our own
Hi Ramana
On Wed, Mar 30, 2011 at 6:35 AM, Ramana Radhakrishnan
ramana.radhakrish...@linaro.org wrote:
Hi Carrot,
How about adding an alternative only enabled for T2 that uses the `l'
constraint and inventing new constraints for some of the constant values
that are valid for 16 bit
On 2011/3/30 05:28 PM, Richard Earnshaw wrote:
On Wed, 2011-03-30 at 15:35 +0800, Chung-Lin Tang wrote:
On 2011/3/30 上午 12:23, Richard Earnshaw wrote:
On Tue, 2011-03-29 at 22:53 +0800, Chung-Lin Tang wrote:
On 2011/3/29 下午 10:26, Richard Earnshaw wrote:
On Tue, 2011-03-29 at 18:25 +0800,
71 matches
Mail list logo