The patch is bootstrapping now on x86_64-pc-linux-gnu.
It very likely breaks bootstrap with RTL checking enabled:
/sil.a/gnatmail/gnatmail-x/build-sil/x86-linux/gnat/obj/./gcc/xgcc
-B/sil.a/gnatmail/gnatmail-x/build-sil/x86-linux/gnat/obj/./gcc/
-B/usr/gnat/i686-pc-linux-gnu/bin/
On Tue, Mar 20, 2012 at 09:51:07AM +0100, Eric Botcazou wrote:
The patch is bootstrapping now on x86_64-pc-linux-gnu.
It very likely breaks bootstrap with RTL checking enabled:
/sil.a/gnatmail/gnatmail-x/build-sil/x86-linux/gnat/obj/./gcc/xgcc
Hi,
I am starting to build fortran for VMS.
The first serious issue was with libgfortran/configure. It checks for several
math functions, but directly in libm using AC_CHECK_LIB.
But there is on such things as libm on VMS systems (thus requiring to define
MATH_LIBRARY as ). Therefore all
Il 19/03/2012 13:32, Paolo Carlini ha scritto:
Should the addition be \$$ to escape it for the shell as well as for
make?
(I know it works, but that might not be true for all shells.)
i don't think that $, could be expaneded by any shell.
I'm not worried about it not expanding, but rather
Il 14/03/2012 15:53, Richard Guenther ha scritto:
I noticed when trying to fix PR52584 that vector lowering always
creates unsigned vector types. The following fixes that, also
getting rid of the weird use of a langhook.
Bootstrapped and tested on x86_64-unknown-linux-gnu.
Richard.
On Mon, 19 Mar 2012, Richard Guenther wrote:
On Mon, 19 Mar 2012, Eric Botcazou wrote:
But it's only ever computed for RECORD_TYPEs where DECL_QUALIFIER is
unused.
OK, that could work indeed.
For now giving up seems to be easiest (just give up when
DECL_FIELD_OFFSET is not
Il 19/03/2012 20:13, Uros Bizjak ha scritto:
2012-03-19 Uros Bizjak ubiz...@gmail.com
* config/i386/i386.c (get_thread_pointer): Add tp_mode argument.
Generate ZERO_EXTEND in place if GET_MODE (tp) != tp_mode.
(legitimize_tls_address) TLS_MODEL_INITIAL_EXEC: Always
Hi,
ggc-page.c uses size_t to cast pointers to an integer type. Unfortunately,
this isn't portable for systems (such as … VMS) where size_t precision is less
than pointers precision.
Fortunately, thanks to configure, uintptr_t type is always present, so this
path simply replaces size_t by
Hi Tristan,
I am starting to build fortran for VMS.
Thanks for the patch and this endeavor.
The first serious issue was with libgfortran/configure. It checks for
several math functions, but directly in libm using AC_CHECK_LIB.
But there is on such things as libm on VMS systems
I think
On Mar 20, 2012, at 10:48 AM, Tobias Burnus wrote:
Hi Tristan,
I am starting to build fortran for VMS.
Thanks for the patch and this endeavor.
The first serious issue was with libgfortran/configure. It checks for
several math functions, but directly in libm using AC_CHECK_LIB.
But
On 03/20/2012 10:22 AM, Paolo Bonzini wrote:
I think \$$ is better. Besides that it looks good.
Thanks for the review. Yesterday, when Benjamin had a look, I decided to
go ahead and just commit the patch as posted (after having double
checked that indeed it worked for me), thus no escaping.
Richard Guenther richard.guent...@gmail.com writes:
On Tue, Mar 20, 2012 at 12:31 AM, Mike Stump mikest...@comcast.net wrote:
On Mar 19, 2012, at 2:44 PM, Richard Sandiford wrote:
Mike Stump mikest...@comcast.net writes:
If we're going to remove the assert, we need to define stuff like
that.
This should fix PR52627.
Applied as obvious.
Richard.
2012-03-20 Richard Guenther rguent...@suse.de
PR gcov-profile/52627
* libgcov.c (init_mx): Fix mutex name.
Index: libgcc/libgcov.c
===
*** libgcc/libgcov.c
On Tue, Mar 20, 2012 at 09:58:29AM +0100, Jakub Jelinek wrote:
Yeah, my bootstrap just failed the same. Will test:
2012-03-20 Jakub Jelinek ja...@redhat.com
* config/i386/i386.c (ix86_decompose_address) case ZERO_EXTEND:
If operand isn't UNSPEC, return 0.
Committed as
Hi!
This patch improves register - register broadcast AVX2 permutations
and also starts using vpermps where possible for V8SFmode
permutations. Bootstrapped/regtested on x86_64-linux and i686-linux,
ok for trunk?
2012-03-20 Jakub Jelinek ja...@redhat.com
PR target/52607
*
Hi!
On a testcase with 256 __builtin_shuffle V4DFmode calls I've counted
17 calls to expand_vselect{,_vconcat} on average for each __builtin_shuffle
call (some during testing, some during actual expansion, but that is
also often not successful). This patch adjusts the code that for testing
it
Hi!
This patch implements expansion of 32-byte shuffles using vperm2[fi]128
insn followed by some single insn two operand shuffle (can be e.g. vunpck*,
vshuf* or vblend*). Bootstrapped/regtested on x86_64-linux and i686-linux,
ok for trunk?
2012-03-20 Jakub Jelinek ja...@redhat.com
On Mar 15, 2012, at 10:37 AM, Richard Guenther wrote:
On Wed, 14 Mar 2012, Tristan Gingold wrote:
[…]
Well. To make this work in LTO the main function (thus, the program
entry point) should be marked at cgraph level and all users of
MAIN_NAME_P should instead check a flag on the cgraph
Il 20/03/2012 11:33, Paolo Carlini ha scritto:
On 03/20/2012 10:22 AM, Paolo Bonzini wrote:
I think \$$ is better. Besides that it looks good.
Thanks for the review. Yesterday, when Benjamin had a look, I decided to
go ahead and just commit the patch as posted (after having double
checked
On Tue, 20 Mar 2012, Tristan Gingold wrote:
On Mar 15, 2012, at 10:37 AM, Richard Guenther wrote:
On Wed, 14 Mar 2012, Tristan Gingold wrote:
[…]
Well. To make this work in LTO the main function (thus, the program
entry point) should be marked at cgraph level and all users of
Richard Guenther richard.guent...@gmail.com writes:
I've no objection to moving the assert down to after the GEN_INT.
But it sounds like I'm on my own with the whole CONST_DOUBLE sign thing.
(That is, if we remove the assert altogether, we effectively treat the
number as sign-extended if it
On Tue, Mar 20, 2012 at 1:26 PM, Richard Sandiford
rdsandif...@googlemail.com wrote:
Richard Guenther richard.guent...@gmail.com writes:
I've no objection to moving the assert down to after the GEN_INT.
But it sounds like I'm on my own with the whole CONST_DOUBLE sign thing.
(That is, if we
Hi,
On Tue, 20 Mar 2012, Richard Sandiford wrote:
If Mike or anyone is up to doing that, then great. But if instead it's
just a case of handling zero correctly, moving rather than removing the
assert seems safer.
I'm obviously not explaining this well :-)
Actually you did. I've tried
The Neon vector-compare operations allow a register or zero (vector of)
as the operands. However, we currently permit any immediate. This can
allow the optimizer to sink a non-trivial constant into a loop when it
was better left in a register. Further, by hiding the register needed
to
On Mar 20, 2012, at 1:21 PM, Richard Guenther wrote:
On Tue, 20 Mar 2012, Tristan Gingold wrote:
On Mar 15, 2012, at 10:37 AM, Richard Guenther wrote:
On Wed, 14 Mar 2012, Tristan Gingold wrote:
[…]
Well. To make this work in LTO the main function (thus, the program
entry point)
On Tue, 20 Mar 2012, Tristan Gingold wrote:
On Mar 20, 2012, at 1:21 PM, Richard Guenther wrote:
On Tue, 20 Mar 2012, Tristan Gingold wrote:
On Mar 15, 2012, at 10:37 AM, Richard Guenther wrote:
On Wed, 14 Mar 2012, Tristan Gingold wrote:
[…]
Well. To make this work
On 01/02/12 13:23, Richard Earnshaw wrote:
On 31/01/12 05:15, Richard Henderson wrote:
Despite how trivial this is, I assume this must wait for stage1.
Ok?
r~
* longlong.h [arm] (umul_ppmm): Use umull.
[arm] (count_trailing_zeros): Use __builtin_ctz.
armv3m also has the
Hi,
this patch which removes one of only two FIXMEs in tree-sra.c has been
sitting in my patch queue for over a year. Yesterday I noticed it
there, bootstrapped and tested it on x86_64-linux and it passed.
I'd like to either commit it or just remove the comment, if there
likely still are size
On 03/20/12 04:27, Jakub Jelinek wrote:
2012-03-20 Jakub Jelinek ja...@redhat.com
* config/i386/i386.c (vselect_insn): New variable.
(init_vselect_insn): New function.
(expand_vselect, expand_vselect_insn): Add testing_p argument.
Call init_vselect_insn if
On Tue, 20 Mar 2012, Martin Jambor wrote:
Hi,
this patch which removes one of only two FIXMEs in tree-sra.c has been
sitting in my patch queue for over a year. Yesterday I noticed it
there, bootstrapped and tested it on x86_64-linux and it passed.
I'd like to either commit it or just
Hi,
this regression is about literal_type_p ICEing for types which cannot be
completed. Indeed, for the testcase, complete_type cannot complete the
type but doesn't error out either, just returns the type as-is, and the
gcc_assert triggers. We could imagine handling such types in the caller
On Tue, Mar 20, 2012 at 4:19 AM, Jakub Jelinek ja...@redhat.com wrote:
On Tue, Mar 20, 2012 at 09:58:29AM +0100, Jakub Jelinek wrote:
Yeah, my bootstrap just failed the same. Will test:
2012-03-20 Jakub Jelinek ja...@redhat.com
* config/i386/i386.c (ix86_decompose_address) case
On 03/20/12 04:22, Jakub Jelinek wrote:
2012-03-20 Jakub Jelinek ja...@redhat.com
PR target/52607
* config/i386/i386.md (isa attribute): Add avx2 and noavx2.
(enabled attribute): Handle avx2 and noavx2 isas.
* config/i386/sse.md (avx2_vec_dupv8sf_1,
On Mar 20, 2012, at 3:19 PM, Richard Guenther wrote:
[…]
I'd rather get away from using a global main_identifier_node, instead
make that frontend specific, and introduce targetm.main_assembler_name
which the assembler-name creating langhook would make sure to use
when mangling what the FE
On Tue, 20 Mar 2012, Tristan Gingold wrote:
On Mar 20, 2012, at 3:19 PM, Richard Guenther wrote:
[…]
I'd rather get away from using a global main_identifier_node, instead
make that frontend specific, and introduce targetm.main_assembler_name
which the assembler-name creating
On Mar 20, 2012, at 5:01 PM, Richard Guenther wrote:
On Tue, 20 Mar 2012, Tristan Gingold wrote:
On Mar 20, 2012, at 3:19 PM, Richard Guenther wrote:
[…]
I'd rather get away from using a global main_identifier_node, instead
make that frontend specific, and introduce
This is the followup I promised to do on the expand and rtl_eh pass
merge. The following patch merges up to unshare_all_rtl, after which
we should be able to pass all RTL testing(?)
Bootstrapped and tested on x86_64-unknown-linux-gnu.
Honza, does this look like what you had in mind?
On 03/20/2012 04:31 AM, Jakub Jelinek wrote:
2012-03-20 Jakub Jelinek ja...@redhat.com
PR target/52607
* config/i386/i386.c (expand_vec_perm_vperm2f128): New function.
(ix86_expand_vec_perm_const_1): Call it.
Ok.
r~
On 18 March 2012 23:27, Jonathan Wakely wrote:
Now that G++ supports it we can use a NSDMI for std::list::_M_size to
avoid needing conditional compilation to set it in the constructors.
I think the attached patch is an improvement so I plan to commit it to
trunk soon unless I hear objections.
Add reduced test case for a common lookup failure.
This is reduced from the standard library. The type mbstate_t is
declared in the global scope and then incorporated into 'std' via
using.
When the header doing this is converted into a PPH image, the lookup
for identifier 'mbstate_t' produces
Hi,
On Tue, Mar 20, 2012 at 04:08:31PM +0100, Richard Guenther wrote:
On Tue, 20 Mar 2012, Martin Jambor wrote:
Hi,
this patch which removes one of only two FIXMEs in tree-sra.c has been
sitting in my patch queue for over a year. Yesterday I noticed it
there, bootstrapped and
On 18 March 2012 14:56, Gerald Pfeifer wrote:
On Thu, 8 Mar 2012, Jonathan Wakely wrote:
The manual claims a future version of G++ will support a hybrid
instantiation model, which I don't think is still planned, and
describes extern templates as an extension when they are in C++11.
*
On Tue, Mar 20, 2012 at 4:52 PM, H.J. Lu hjl.to...@gmail.com wrote:
Yeah, my bootstrap just failed the same. Will test:
2012-03-20 Jakub Jelinek ja...@redhat.com
* config/i386/i386.c (ix86_decompose_address) case ZERO_EXTEND:
If operand isn't UNSPEC, return 0.
Committed as
This patch fixes bugs in section_ordering linker plugin to correctly handle
functions that are file local or comdat. Such functions could be
defined more than once in different object files and can have the same name.
The callgraph edge profiles should be attributed to the right local function.
On Tue, Mar 20, 2012 at 10:54 AM, Uros Bizjak ubiz...@gmail.com wrote:
On Tue, Mar 20, 2012 at 4:52 PM, H.J. Lu hjl.to...@gmail.com wrote:
Yeah, my bootstrap just failed the same. Will test:
2012-03-20 Jakub Jelinek ja...@redhat.com
* config/i386/i386.c (ix86_decompose_address)
On Tue, Mar 20, 2012 at 7:27 PM, H.J. Lu hjl.to...@gmail.com wrote:
I think use the OS provided instruction to load TP into DImode register
could simplify the code.
Which OS provided instruction?
Please see how TP is defined in get_thread_pointer, it is in ptr_mode:
rtx tp =
The _stat allocation functions aren't supposed to be called directly.
Tested x86_64-pc-linux-gnu, applying as obvious.
commit f0e560b78591a50e39bf3ccf41bc3d87f43927e8
Author: Jason Merrill ja...@redhat.com
Date: Mon Mar 19 16:35:05 2012 -0400
* tree-streamer-in.c (streamer_alloc_tree):
Martin Jambor wrote:
Hi,
this is another iteration of my attempts to fix expansion of
misaligned memory accesses on strict-alignment platforms (which was
suggested by Richi in
http://gcc.gnu.org/ml/gcc-patches/2011-08/msg00931.html and my first
attempt was posted as
My change for core issue 1270 to allow brace elision in all
list-initialization resulted in multiple calls to reshape_init for the
same CONSTRUCTOR, which we weren't handling properly.
Tested x86_64-pc-linux-gnu, applying to trunk.
commit d7b5c07030b11599b846342b19ff50306043965f
Author: Jason
That assert is there to make sure that we don't try to test for
literality of an incomplete type. We should check for completeness
before trying to check for literality.
Jason
It would be nice to add some unit/regression test cases of some sort.
David
http://codereview.appspot.com/5851044/diff/1/callgraph.c
File callgraph.c (right):
http://codereview.appspot.com/5851044/diff/1/callgraph.c#newcode309
callgraph.c:309: if (!is_prefix_of (_ZL, name))
How about static
On Tue, Mar 20, 2012 at 11:43 AM, Uros Bizjak ubiz...@gmail.com wrote:
On Tue, Mar 20, 2012 at 7:27 PM, H.J. Lu hjl.to...@gmail.com wrote:
I think use the OS provided instruction to load TP into DImode register
could simplify the code.
Which OS provided instruction?
Please see how TP is
http://codereview.appspot.com/5851044/diff/1/callgraph.c
File callgraph.c (right):
http://codereview.appspot.com/5851044/diff/1/callgraph.c#newcode513
callgraph.c:513: const int section_priority[] = {0, 3, 4, 2, 1};
Add a comment about section_priority
http://gcc.gnu.org/viewcvs?view=revisionrevision=185588
Johann
Jakub Jelinek wrote:
Hi!
This patch adds ASSERT_EXPRs derived from
tmp_N = var_M cst2;
if (tmp_N cmp val)
where cst2 and val are constants. Bootstrapped/regtested
on x86_64-linux and i686-linux, ok for trunk?
2012-03-14 Jakub Jelinek ja...@redhat.com
PR
Dropping the first patch which does not work because at expand-time there
must not be pre-/post-modify addressing :-(
This solutions turns completely away from MEM and addressing modes:
It represents loads from the 16-bits address-spaces as UNSPEC.
The code is as expected now with the additional
And here is the patch...
Georg-Johann Lay wrote:
Dropping the first patch which does not work because at expand-time there
must not be pre-/post-modify addressing :-(
This solutions turns completely away from MEM and addressing modes:
It represents loads from the 16-bits address-spaces as
On Tue, Mar 20, 2012 at 8:54 PM, Georg-Johann Lay a...@gjlay.de wrote:
Dropping the first patch which does not work because at expand-time there
must not be pre-/post-modify addressing :-(
Have you tried to fix that, instead? Or at least ask around a bit to
see what people would think about
On 03/20/2012 08:22 PM, Jason Merrill wrote:
That assert is there to make sure that we don't try to test for
literality of an incomplete type. We should check for completeness
before trying to check for literality.
You mean, in the relevant caller, here in check_field_decls:
/* If at
Hi,
The attached patch removes the fsca instruction support for DFmode on
SH4A when -ffast-math is enabled.
Tested against rev 18 with
make -k check RUNTESTFLAGS=--target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a-single/-mb,
-m4-single/-ml,-m4-single/-mb,
-m4a-single/-ml,-m4a-single/-mb}
and no
... this simple also passes testing.
Paolo.
/cp
2012-03-20 Paolo Carlini paolo.carl...@oracle.com
PR c++/52487
* class.c (check_field_decls): Call literal_type_p only
on complete types.
/testsuite
2012-03-20 Paolo Carlini
Hi,
Attached is Kaz's patch from the PR.
Tested against rev 18 with
make -k check RUNTESTFLAGS=--target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a-single/-mb,
-m4-single/-ml,-m4-single/-mb,
-m4a-single/-ml,-m4a-single/-mb}
It fixes a few sometimes failing libstdc++ failures.
No new failures
On Mar 20, 2012, at 6:55 AM, Michael Matz wrote:
Actually you did. I've tried yesterday to come up with a text that would
do the same (because I agree with you that deleting the assert changes
the spec of the function,
The spec of the function is the text above the definition of the
This patch adds support to version for corei7 with -mvarch option. The
versioning supported is in the case where a loop generates a LCP stalling
instruction in corei7. In such cases, on corei7, limiting the unroll factor to
try to keep the unrolled loop body small enough to fit in the Corei7's
Hello,
This patch splits a couple of pieces of cse_insn out to new functions.
There are no functional changes, and no code generation differences as
far as I could tell on x86_64 (-m64 and -m32).
The purpose of the patch is and, loto hopefully make cse_insn easier
to understand. In a follow-up
Hi,
The attached patch adds some test cases for PR 50751 to check whether
mov.b insns are generated.
Tested on sh-sim with
make check-gcc RUNTESTFLAGS=sh.exp=pr50751* --target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a-single/-mb,-m4-single/-ml,
-m4-single/-mb,-m4a-single/-ml,-m4a-single/-mb}
to
On Tue, Mar 20, 2012 at 2:04 PM, Sriraman Tallam tmsri...@google.com wrote:
This patch adds support to version for corei7 with -mvarch option. The
versioning supported is in the case where a loop generates a LCP stalling
instruction in corei7. In such cases, on corei7, limiting the unroll
For google/gcc-4_6 branch.
Fix bug where we were outputting a garbage value for the index operand
of DW_OP_GNU_addr_index.
Tested: incremental remake in GCC build directory and hand tested.
2012-03-20 Cary Coutant ccout...@google.com
* dwarf2out.c (size_of_loc_descr): Use val_index
I've merged revision 185588 of the 4.7 branch to the gccgo branch.
This is a switch from merging from mainline.
Ian
On Tue, Mar 20, 2012 at 2:06 PM, Steven Bosscher stevenb@gmail.com wrote:
This patch splits a couple of pieces of cse_insn out to new functions.
There are no functional changes, and no code generation differences as
far as I could tell on x86_64 (-m64 and -m32).
The purpose of the patch
LGTM
http://codereview.appspot.com/5866047/
On Mar 20, 2012, at 5:26 AM, Richard Sandiford wrote:
So what I was trying to say was that if we remove the assert
altogether, and allow CONST_DOUBLEs to be wider than 2 HWIs,
we need to define what the implicit high-order HWIs of a
CONST_DOUBLE are, just like we already do for CONST_INT.
Oleg Endo oleg.e...@t-online.de wrote:
It fixes a few sometimes failing libstdc++ failures.
No new failures otherwise.
OK to apply?
OK.
Maybe this one should be backported to 4.7.x, too?
Sure.
Regards,
kaz
Oleg Endo oleg.e...@t-online.de wrote:
The attached patch removes the fsca instruction support for DFmode on
SH4A when -ffast-math is enabled.
Tested against rev 18 with
make -k check RUNTESTFLAGS=--target_board=sh-sim
\{-m2/-ml,-m2/-mb,-m2a-single/-mb,
-m4-single/-ml,-m4-single/-mb,
Oleg Endo oleg.e...@t-online.de wrote:
The attached patch adds some test cases for PR 50751 to check whether
mov.b insns are generated.
OK.
Regards,
kaz
GCC 4.7 adds mangling for new-expressions in a function signature, but I
now notice it produces wrong mangling for 'new auto' in simple cases.
This patch fixes it.
Tested x86_64-pc-linux-gnu, applying to trunk.
This also seems like it might be a candidate for 4.7.0. What do you
think,
Uploaded new patch.
On 2012/03/20 19:25:38, davidxl wrote:
It would be nice to add some unit/regression test cases of some sort.
Made the existing unit test case check the final layout.
David
http://codereview.appspot.com/5851044/diff/1/callgraph.c
File callgraph.c (right):
On 2012/03/20 19:26:26, eraman wrote:
http://codereview.appspot.com/5851044/diff/1/callgraph.c
File callgraph.c (right):
http://codereview.appspot.com/5851044/diff/1/callgraph.c#newcode513
callgraph.c:513: const int section_priority[] = {0, 3, 4, 2, 1};
Add a comment about section_priority
ok for google branches after checkin validation.
David
http://codereview.appspot.com/5851044/
I've been working on a proposal for return type deduction for normal
functions for the next C++ standard, and so I'm adding -std=c++1y to
control it.
Tested x86_64-pc-linux-gnu, applying to trunk.
commit ef9ed182fec9249e396fdb0ed2bd0cc4d725e956
Author: Jason Merrill ja...@redhat.com
Date:
80 matches
Mail list logo