A child unit given by a subprogram body without a declaration is given a
declaration when it is compiled, to simplify code generation and tools. If
the child unit appears in several with-clauses for the same unit (for example
through with_clauses in various subunits) all with_clauses must refer to
This patch modifies the behavior of Getopt in GNAT.Command_Line to not call
the Callback function in argument on switches that have been dealt with
automatically.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-02 Johannes Kanig
* g-comlin.adb (Getopt): Return when switch is
This patch improves compilation speed (in particular, speed of generating cross
reference information) when compiling packages with huge numbers of tagged
types and interfaces, with complicated inheritance patterns. No test is
available -- the problem only occurred for enormous packages.
Tested o
The high-level procedure Getopt did not expose a feature of its lowlevel
counterpart, namely the Concatenate option which modifies the behavior of the
command line parser. This patch adds this option to the high-level interface.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-02 Johann
when rewriting X'range (N) into X'First (N) .. X'Last (N), the dimension
indicator N must not be shared, if present in the original range. Even though
it is a static constant, its source location may be modified when printing
expanded code under -gnatDL, and node sharing leads to a double modifica
When overflow checking is enabled, the front end generates range checks
on the indexing of a slice with a bound of the form Boolean'Pos (not X)
that get mistranslated by gigi when optimization is enabled. The checks
should not have been generated in the first place (one is a tautological
full range
In the case a class-wide limited (build-in-place) function that returns
a call to another build-in-place function with a controlled result, the
compiler passes null as the value for the implicit finalization master
parameter rather than passing along its own finalization master formal.
This fails a
This restriction, which is not required to be partition-wide consistent,
requires an explicit aliased keyword for an object to which 'Access,
'Unchecked_Access, or 'Address is applied, and forbids entirely the
use of the 'Unrestricted_Access attribute for objects.
The following test program tests
If either the target or the expression in an array or record ssignment may be
bit-aligned, the assignment must be expanded into component assignments. The
object (or subcomponent) may be given by an unchecked conversion, in which case
we must examine its expression to determine potential mis-alignm
This patch avoids reporting a redundant error on primitives inherited
from primitives.
package Test_Pkg is
type I1 is Interface;
procedure P (A : I1) is abstract;
type I2 is interface;
procedure Q (B : I2) is abstract;
type DT2 is new I1 and I2 with null record;
On Thu, Sep 1, 2011 at 2:41 PM, Michael Matz wrote:
> Hi,
>
> the last change in ipa-split generated a new use of a previously unused
> PARM_DECL. When one does this one has to call add_referenced_var. Not
> doing so can cause segfault when accessing the (not initialized) var
> annotation. So,
> Well, the comment for that folding is totally odd - of _course_
> unsigned sizetype things can overflow (we hid that issue merely
> by pretending all unsigned sizetype constants (yes, only constants)
> are signed. Huh.)
It's again the special semantics of sizetypes whereby we pretend that they
On Thu, Sep 1, 2011 at 4:16 PM, Aldy Hernandez wrote:
>
>> My point is, the middle-end infrastructure makes it possible for this
>> case to appear, and it seems to be easy to handle conservatively.
>> There isn't a need to wait for users to run into an ICE or an assert we
>> put
>> there IMHO. If
On Thu, Sep 1, 2011 at 5:19 PM, Jason Merrill wrote:
> On 09/01/2011 11:10 AM, Aldy Hernandez wrote:
>>>
>>> Basically you can only touch the size of the CLASSTYPE_AS_BASE variant.
>>> For many classes this will be the same as the size of the class itself.
>>
>> All this code is in the middle end,
On Thu, Sep 1, 2011 at 8:52 PM, Jan Hubicka wrote:
>> - Nevertheless, this method of devirtualization cannot automatically
>> de-thunkize this-adjusting thunks and newly direct calls to them
>> cannot be inlined because the inliner does not have this capability
>> now. This is in fact a reg
On Fri, 2 Sep 2011, Eric Botcazou wrote:
> > Well, the comment for that folding is totally odd - of _course_
> > unsigned sizetype things can overflow (we hid that issue merely
> > by pretending all unsigned sizetype constants (yes, only constants)
> > are signed. Huh.)
>
> It's again the specia
References to IN parameters are not considered in Alfa section, as these will
be translated as constants in the intermediate language for formal
verification.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-02 Yannick Moy
* lib-xref-alfa.adb (Is_Alfa_Reference): Ignore IN pa
A contract, either in GNAT syntax (pragma precondition and postcondition) or
Ada 2012 aspect syntax, is suspicious when: for a function, it does not mention
the result; for a function or procedure, it only refers to the pre-state. GNAT
now detects these cases on the following code:
$ gcc -c -gnatc
Pragmas on generic units, if they are not library unit pragmas, are inherited
by each instantiation of the generic. Pragma Convention was omitted from this
processing.
The following must execute quietly:
gnatmake -q address_test.adb
address_test
---
with Ada.Text_Io; use Ada.Text_IO;
with
Ada 2012 AI05-0192 is a binding interpretation that makes it clear that when
reading a discriminated value using the Input stream attribute, discriminants
values read from the stream must match any constraint imposed by the subtype
given as the attribute's prefix. These checks are now done by check
The usage of GNAT tools has been enhanced to include lines for options
--version and --help. The test for this is to invoke a GNAT tool with --help
and to check that there are lines for --version and --help in the output.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-02 Vincent Celie
On Tue, 30 Aug 2011, Richard Guenther wrote:
> On Tue, 30 Aug 2011, Richard Guenther wrote:
>
> > On Tue, 30 Aug 2011, Uros Bizjak wrote:
> >
> > > On Tue, Aug 30, 2011 at 11:15 AM, Richard Guenther
> > > wrote:
> > >
> > > >> >> > Hmm. But then I'd have to try emit an insn, right? Currentl
The Project Manager is now checking that the names of Stand-Alone
libraries have the syntax of Ada identifiers, so that the binder
generated files that contains unit and subprogram names derived from the
library name will compile. If this check fails, an error is reported and
the tool terminates it
Hello, gentle maintainer.
This is a message from the Translation Project robot.
A revised PO file for textual domain 'gcc' has been submitted
by the Japanese team of translators. The file is available at:
http://translationproject.org/latest/gcc/ja.po
(This file, 'gcc-4.6.1.ja.po', has jus
When a subunit source file in a project being extended is replaced with
one in the extending project with a different source file name, the
subunit source in the project being extended is still used.
This patch fixes this. The test is to use a subunit source
pkg-execute.adb in a project being exten
Class-wide pre- and postconditions of a primitive operation of some type T
apply to the overriding operation of any descendant of T. Therefore, any
mention of of a formal of type T in the expression for the condition must be
interpreted as a being of type T'class. A similar rule applies to access
The code generated by the compiler to handle the initialization of
limited class-wide interface objects initialized by means of an
aggregate erroneously generates a copy of the object (which causes
a runtime exception in the application). After this patch the
following test compiles and executes we
The function Unique_Name did not actually compute a unique name for
enumeration literals, because the type to which the literal belongs was not
taken into account. This patch fixes the issue.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-02 Johannes Kanig
* sem_util.adb (U
a GNAT artifact: Limited_Controlled is declared as an extension of Root_
Controlled, and thus has a useless Adjust operation. This operation should not
be inherited by other limited controlled types. An explicit Adjust for them is
not overriding.
gcc -c -gnat05 lib2.ads
must yield
lib2.ad
The msghdr structure has changed between vms 7 and 8.
This patch works arounds the incompatibility.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-02 Tristan Gingold
* g-socthi-vms.adb (c_sendmsg, c_recvmsg): Use unpacked msg if on vms 7
Index: g-socthi-vms.adb
===
This patch fixes an omission in the code that resolves actuals in a call.
Previous to this patch, and actual in a call that is an overloaded function
call, one of whose interpretations returns an unconstrained limited type may
be resolved incorrectly.
The command
gnatmake -q -gnat05 main
This variable is used in Hi-Lite to group effects that cannot be described
more precisely. The new version 3 of Why, the backend of Hi-Lite, does not
accept variables to start with uppercase letters, so we choose a value which
works with both versions.
Tested on x86_64-pc-linux-gnu, committed on t
This patch suppresses these false, which can occur when a postcondition
contains a quantified expression.
The following test should compile quietly.
gnatmake -f -gnat2012 -gnatwa t.adb
package T is
procedure P (S : in out String) with
Post => (for all C of S => C = C);
end T;
package body T
This patch adds information to illegal uss of incomplete types. If the prefix
of a selected component is an incomplete type, and the completion is found
later in the same unit, an additional error indicates the position of the
full declaration.
gcc -c -gnat05 obj.adb
must yield:
obj.adb:6:
When the SPARK restriction mode was set, check that a declaration of
unconstrained type is allowed only for constants of type string.
Tested on x86_64-pc-linux-gnu, committed on trunk
2011-09-02 Marc Sango
* sem_ch3.adb (Analyze_Object_Declaration): Change
comment and add add
On 1 September 2011 12:50, Bernd Schmidt wrote:
> Shrink-wrapping tests on ARM had one additional failure, which I could
> track down to a stmfd instruction being emitted where an stmhifd was
> intended. The following patch fixes the testcase; full tests running
> now. Ok?
IIUC this should have b
Hi,
tested x86_64-linux multilib, will go in 4_6-branch too.
See audit trail for details: in short when std::bitset has been updated
for constexpr, thus the constructor from unsigned long long marked as
such, the call to _M_do_sanitize in the body was inadvertently
completely removed without
Hello,
Here is a patch which adds few more splits for AGU stalls avoidance on
Atom. It also fixes cost model and detects AGU stalls more
efficiently.
Bootstrapped and checked on x86_64-linux.
Thanks,
Ilya
---
gcc/
2011-09-02 Enkovich Ilya
* config/i386/i386-protos.h (ix86_lea_outper
> While it could be possible to output_set_got such that we can
> individually annotate the instructions, it's simpler to simply
> admit that all processors currently being manufactured do want
> deep branch prediction. At which point all of the complication
> simply goes away.
Note that most of
On 09/02/11 12:35, Ramana Radhakrishnan wrote:
> On 1 September 2011 12:50, Bernd Schmidt wrote:
>> Shrink-wrapping tests on ARM had one additional failure, which I could
>> track down to a stmfd instruction being emitted where an stmhifd was
>> intended. The following patch fixes the testcase; fu
This fixes the Ada bootstrap failure introduced by alloca folding.
We now fold alloca (0) to &auto-with-size-zero which confuses us.
I didn't exactly investigate but what I think happens is that we
expand that &auto-with-size-zero to NULL instead of
virtual_stack_dynamic_rtx (see zero-size special
On Fri, Sep 2, 2011 at 11:43 AM, Richard Guenther wrote:
>> > > >> >> > Hmm. But then I'd have to try emit an insn, right? Currently
>> > > >> >> > the vectorizer simply looks for an optab handler ... the
>> > > >> >> > operands are not readily available (but their mode is known).
>> > > >> >>
Hi,
currently IPA-CP switches itself off on functions with are called with
variable number of arguments or on those whose signature cannot be
changed because of type attributes. This patch turns it on for both,
not changing the parameters of its clones in the latter case.
Whether the function's
Hi,
On Wed, Aug 31, 2011 at 09:33:03AM +0200, Richard Guenther wrote:
> On Tue, Aug 30, 2011 at 6:50 PM, Martin Jambor wrote:
> > Ping. Re-bootstrapped and re-tested yesterday on x86_64-linux.
>
> Ok. Does this also apply (maybe in modifed form) to the 4.6 branch?
>
Below is a modified patch
This patch makes us no longer treat unsigned sizetypes as sign-extending
which has caused trouble in the past (see the tree-ssa-ccp.c workarounds
removed below).
Most frontend specific changes below deal with the fact that overflow
behavior, as reported from size_binop, changes as unsigned sizety
Note that with all this mess I'll re-iterate some of my initial thoughts.
1) why not do this C++ (or C) specific stuff in the frontends, maybe
at gimplifying/genericization time? That way you wouldn't need to
worry about middle-end features but you could rely solely on what
C/C++ permit. It is
> This fixes the Ada bootstrap failure introduced by alloca folding.
> We now fold alloca (0) to &auto-with-size-zero which confuses us.
> I didn't exactly investigate but what I think happens is that we
> expand that &auto-with-size-zero to NULL instead of
> virtual_stack_dynamic_rtx (see zero-siz
On Thu, Sep 1, 2011 at 12:27 PM, Ilya Enkovich wrote:
>>
>> this seems to not allow cycles_best to drop with lower width, but
>> that it can't should be an implementation detail of get_required_cycles.
>> To make it not so, can you add a comment before the loop, like
>>
>> /* get_required_cycles
On Fri, 2 Sep 2011, Arnaud Charlet wrote:
> > This fixes the Ada bootstrap failure introduced by alloca folding.
> > We now fold alloca (0) to &auto-with-size-zero which confuses us.
> > I didn't exactly investigate but what I think happens is that we
> > expand that &auto-with-size-zero to NULL i
> > In Ada, it's quite natural to end up with a dynamically sized object of
> > size 0. For instance, if you declare an array with a dynamic bound:
> >
> >Table : Unit_Table (1 .. Last_Unit);
> >
> > and Last_Unit happens to be 0 at run-time
>
> But are we expected to read/store from the sto
On Fri, Sep 2, 2011 at 2:42 PM, Martin Jambor wrote:
> Hi,
>
> On Wed, Aug 31, 2011 at 09:33:03AM +0200, Richard Guenther wrote:
>> On Tue, Aug 30, 2011 at 6:50 PM, Martin Jambor wrote:
>> > Ping. Re-bootstrapped and re-tested yesterday on x86_64-linux.
>>
>> Ok. Does this also apply (maybe in
On Fri, Sep 2, 2011 at 2:49 PM, Aldy Hernandez wrote:
>
>> Note that with all this mess I'll re-iterate some of my initial thoughts.
>> 1) why not do this C++ (or C) specific stuff in the frontends, maybe
>> at gimplifying/genericization time? That way you wouldn't need to
>> worry about middle-e
On Fri, Sep 2, 2011 at 2:52 PM, Uros Bizjak wrote:
> On Thu, Sep 1, 2011 at 12:27 PM, Ilya Enkovich wrote:
>>>
>>> this seems to not allow cycles_best to drop with lower width, but
>>> that it can't should be an implementation detail of get_required_cycles.
>>> To make it not so, can you add a co
On 9/2/2011 8:52 AM, Arnaud Charlet wrote:
Thanks!
In Ada, it's quite natural to end up with a dynamically sized object of
size 0. For instance, if you declare an array with a dynamic bound:
Table : Unit_Table (1 .. Last_Unit);
and Last_Unit happens to be 0 at run-time
Arno
But isn't i
On Fri, 2 Sep 2011, Arnaud Charlet wrote:
> > > In Ada, it's quite natural to end up with a dynamically sized object of
> > > size 0. For instance, if you declare an array with a dynamic bound:
> > >
> > >Table : Unit_Table (1 .. Last_Unit);
> > >
> > > and Last_Unit happens to be 0 at run-t
On 9/2/2011 8:58 AM, Arnaud Charlet wrote:
In Ada, it's quite natural to end up with a dynamically sized object of
size 0. For instance, if you declare an array with a dynamic bound:
Table : Unit_Table (1 .. Last_Unit);
and Last_Unit happens to be 0 at run-time
But are we expected to read
On Fri, 2 Sep 2011, Robert Dewar wrote:
> On 9/2/2011 8:58 AM, Arnaud Charlet wrote:
> > > > In Ada, it's quite natural to end up with a dynamically sized object of
> > > > size 0. For instance, if you declare an array with a dynamic bound:
> > > >
> > > > Table : Unit_Table (1 .. Last_Unit);
> But they _do_ overflow as my debugging showed, caused by that exact
> same extract_muldiv_1 function here:
>
> case PLUS_EXPR: case MINUS_EXPR:
> /* See if we can eliminate the operation on both sides. If we can,
> we
> can return a new PLUS or MINUS. If we can't, the only r
Hi,
tested x86_64-linux multilib, a nice improvement joint work with Marc.
Mainline only of course.
Thanks,
Paolo.
/
2011-09-02 Paolo Carlini
Marc Glisse
* include/std/bitset (_Base_bitset<>::_M_are_all_aux): Remove.
(_Base_bitset<>::_
On Fri, 2 Sep 2011, Eric Botcazou wrote:
> > But they _do_ overflow as my debugging showed, caused by that exact
> > same extract_muldiv_1 function here:
> >
> > case PLUS_EXPR: case MINUS_EXPR:
> > /* See if we can eliminate the operation on both sides. If we can,
> > we
> >
Hi,
On Thu, Sep 01, 2011 at 08:52:30PM +0200, Jan Hubicka wrote:
> > - Nevertheless, this method of devirtualization cannot automatically
> > de-thunkize this-adjusting thunks and newly direct calls to them
> > cannot be inlined because the inliner does not have this capability
> > now. Thi
On Thu, 1 Sep 2011, Gary Funck wrote:
> +# MAX_LANG is the maximum number of languages that can be defined.
> +# Its value is extracted from the value of CL_PARAMS in opts.h
> +# and is passed on the command line as '-v max_lang=...'.
> +if (n_langs > max_lang) {
> + print "Error: the number of d
> Hi,
>
> On Thu, Sep 01, 2011 at 08:52:30PM +0200, Jan Hubicka wrote:
> > > - Nevertheless, this method of devirtualization cannot automatically
> > > de-thunkize this-adjusting thunks and newly direct calls to them
> > > cannot be inlined because the inliner does not have this capability
> >
2011/9/2 Richard Guenther :
>On Fri, Sep 2, 2011 at 2:52 PM, Uros Bizjak wrote:
>>
>> I assume that you need to split tune attribute to int and FP part to
>> handle reassociation for other targets, since Atom handles both in the
>> same way.
>>
>> Please also describe function return value in the
without resorting to hex printing of both high/low values.
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied to trunk.
Richard.
2011-09-02 Richard Guenther
* pretty-print.h (pp_unsigned_wide_integer): New.
* tree-pretty-print.c (dump_generic_node): Print unsigned
Hi,
noticed over the last day or so, committed to mainline.
Paolo.
//
2011-09-02 Paolo Carlini
* include/std/bitset: Trivial formatting fixes.
Index: include/std/bitset
===
--- include/std/bitset (revis
On Fri, 2 Sep 2011, Richard Guenther wrote:
> * c-decl.c (grokdeclarator): Properly check for sizes that
> cover more than half of the address-space.
The C front-end changes are OK.
--
Joseph S. Myers
jos...@codesourcery.com
On 9/2/2011 9:16 AM, Richard Guenther wrote:
The bootstrap failure showed NULL pointer dereferences (which
probably easily points to the affected part of the RTS).
Might be interesting to pursue, but we don't know that the null
pointers being dereferenced are in fact the ones returned by
alloc
On Fri, Sep 2, 2011 at 3:45 PM, Ilya Enkovich wrote:
> 2011/9/2 Richard Guenther :
>>On Fri, Sep 2, 2011 at 2:52 PM, Uros Bizjak wrote:
>>>
>>> I assume that you need to split tune attribute to int and FP part to
>>> handle reassociation for other targets, since Atom handles both in the
>>> same
On Fri, Sep 2, 2011 at 2:27 PM, Martin Jambor wrote:
> Hi,
>
> currently IPA-CP switches itself off on functions with are called with
> variable number of arguments or on those whose signature cannot be
> changed because of type attributes. This patch turns it on for both,
> not changing the para
On 09/02/2011 04:53 AM, Richard Guenther wrote:
On Thu, Sep 1, 2011 at 5:19 PM, Jason Merrill wrote:
I think it would make sense to expose this information to the back end
somehow. A hook would do the trick: call it type_data_size or type_min_size
or some such, which in the C++ front end would
2011/9/2 Richard Guenther :
> On Fri, Sep 2, 2011 at 3:45 PM, Ilya Enkovich wrote:
>> 2011/9/2 Richard Guenther :
>>>On Fri, Sep 2, 2011 at 2:52 PM, Uros Bizjak wrote:
I assume that you need to split tune attribute to int and FP part to
handle reassociation for other targets, since
> On Fri, Sep 2, 2011 at 2:27 PM, Martin Jambor wrote:
> > Hi,
> >
> > currently IPA-CP switches itself off on functions with are called with
> > variable number of arguments or on those whose signature cannot be
> > changed because of type attributes. This patch turns it on for both,
> > not cha
On Fri, 2 Sep 2011, Richard Guenther wrote:
>
> This patch makes us no longer treat unsigned sizetypes as sign-extending
> which has caused trouble in the past (see the tree-ssa-ccp.c workarounds
> removed below).
>
> Most frontend specific changes below deal with the fact that overflow
> behavi
On Fri, Sep 2, 2011 at 4:10 PM, Jason Merrill wrote:
> On 09/02/2011 04:53 AM, Richard Guenther wrote:
>>
>> On Thu, Sep 1, 2011 at 5:19 PM, Jason Merrill wrote:
>>>
>>> I think it would make sense to expose this information to the back end
>>> somehow. A hook would do the trick: call it type_da
CC_NCV rightly only allows GE(U) and LT(U). GT(U) and LE(U) have to
implemented by reversing the condition. This is handled correctly when
the condition is first expanded, but nothing stops later optimisers from
producing invalid forms.
This patch makes arm_comparison_operator check that the con
Ping for this patch to emit_library_call_value_1:
http://gcc.gnu.org/ml/gcc-patches/2011-08/msg00735.html
which fixes a bootstrap failure on MIPS since:
http://gcc.gnu.org/ml/gcc-patches/2011-06/msg02341.html
Tested on mips64-linux-gnu, mips-sgi-irix6.5 (by Rainer) and
on both big and l
Hi,
when I submitted the new IPA-CP a few months ago Honza requested that
I store the jump_functions in a VECtor rather than an array in which
they are now. The patch below does exactly that.
The last remaining such request is to rename
ipa_check_create_node_params to "something else." I guess
On Fri, 2 Sep 2011, Artem Shinkarov wrote:
> + /* Avoid C_MAYBE_CONST_EXPRs inside VEC_SHUFFLE_EXPR. */
> + tmp = c_fully_fold (v0, false, &maybe_const);
> + v0 = save_expr (tmp);
> + wrap &= maybe_const;
I suppose you need this save_expr because of the two-argument case, but
shouldn't need
Hi,
On Fri, 2 Sep 2011, Robert Dewar wrote:
> On 9/2/2011 9:16 AM, Richard Guenther wrote:
>
> Might be interesting to pursue, but we don't know that the null pointers
> being dereferenced are in fact the ones returned by alloca. May not be
> worth the effort.
Given the nature of the work-aro
> Given the nature of the work-around which makes Ada work again it's fairly
> sure that the Ada frontend does emit accesses to an alloca'ed area of
> memory even if its size is zero. I.e. definitely a real bug.
Well, it's not clear whether it's the Ada frontend or the middle which is
emitting th
On 9/2/2011 11:47 AM, Michael Matz wrote:
Hi,
On Fri, 2 Sep 2011, Robert Dewar wrote:
On 9/2/2011 9:16 AM, Richard Guenther wrote:
Might be interesting to pursue, but we don't know that the null pointers
being dereferenced are in fact the ones returned by alloca. May not be
worth the effort.
Hi Richard, this looks very similar to this patch, originally for LP:689887:
http://gcc.gnu.org/ml/gcc-patches/2011-01/msg00794.html
Apart from your additional handling in the dominance modes cases.
I remember that last patch was held down because Thumb-2 native
bootstrap failed. Did you try that
Hi,
On Fri, 2 Sep 2011, Arnaud Charlet wrote:
> > Given the nature of the work-around which makes Ada work again it's fairly
> > sure that the Ada frontend does emit accesses to an alloca'ed area of
> > memory even if its size is zero. I.e. definitely a real bug.
>
> Well, it's not clear whethe
> This fixes the Ada bootstrap failure introduced by alloca folding.
> We now fold alloca (0) to &auto-with-size-zero which confuses us.
> I didn't exactly investigate but what I think happens is that we
> expand that &auto-with-size-zero to NULL instead of
> virtual_stack_dynamic_rtx (see zero-siz
On Fri, Sep 2, 2011 at 4:41 PM, Joseph S. Myers wrote:
> On Fri, 2 Sep 2011, Artem Shinkarov wrote:
>
>> + /* Avoid C_MAYBE_CONST_EXPRs inside VEC_SHUFFLE_EXPR. */
>> + tmp = c_fully_fold (v0, false, &maybe_const);
>> + v0 = save_expr (tmp);
>> + wrap &= maybe_const;
>
> I suppose you need th
> 2011-09-02 Martin Jambor
>
> * ipa-prop.h (ipa_jump_func_t): New typedef.
> (struct ipa_edge_args): Removed field argument_count, field
> jump_functions turned into a vector.
> (ipa_set_cs_argument_count): Removed.
> (ipa_get_cs_argument_count): Updated to work o
On 09/02/11 13:42:32, Joseph S. Myers wrote:
> [..] you should just generate #if/#error in the output [...]
OK, take two, attached. (Confirmed that the #if works for
the (<, ==, >) relationships between n_langs and the max.
number of languages supported.)
- Gary
2011-09-02 Gary Funck
Hi,
On Fri, 2 Sep 2011, Richard Guenther wrote:
> > Currently regstrapping on x86_64-linux (without Ada). Okay for trunk?
>
> Ok. Time to make get_var_ann private?
Yes. Should the regstrap succeed on x86_64-linux I'll commit this
variant of the patch (hunks in tree-flow.h, tree-flow-inline.
> so accessing the With_Flags array (which is not empty) yields a SEGV
> because the base pointer is equal to Last_Unit (i.e. 2). In other words,
> the GIMPLE code looks legitimate and the bug is very likely in the stack
> slot allocation code (maybe triggered by the newly created zero-sized
> arr
Hi!
The following patch contains a WIP implementation of a new pass,
which attempts to track C string lengths and perform various
optimizations using that information.
Optimizations it currently performs:
1) optimizing away strlen calls, if the string length is known
at that point already (eit
> And this is the real fix. Richard, do you want me to apply (part of it)?
In fact I'd even propose to revert the fold_builtin_alloca_for_var part of
your patch and apply mine entirely, as eliminating alloca (0) early looks a
interesting simplification. What do you think?
--
Eric Botcazou
On Fri, Sep 2, 2011 at 9:40 AM, Eric Botcazou wrote:
>> so accessing the With_Flags array (which is not empty) yields a SEGV
>> because the base pointer is equal to Last_Unit (i.e. 2). In other words,
>> the GIMPLE code looks legitimate and the bug is very likely in the stack
>> slot allocation c
Hi,
Any comments related to the implementation, or examples of real-world
lame C string length code sequences that would be nice to optimize
will be greatly appreciated.
I'm only wondering how hard would be taking care more consistently of
the wchar_t counterparts of all these library functions.
My previous patch for preference of smaller # of registers involved
in reloads broke some tests from spec2000 on i686 because LRA started to
use MMX registers. I am working on a fix for this but unfortunately
fixing it is not easy and will take some time, so I am reverting my
previous patch.
On 09/02/2011 07:13 PM, H.J. Lu wrote:
> On Fri, Sep 2, 2011 at 9:40 AM, Eric Botcazou wrote:
>>> so accessing the With_Flags array (which is not empty) yields a SEGV
>>> because the base pointer is equal to Last_Unit (i.e. 2). In other words,
>>> the GIMPLE code looks legitimate and the bug is v
Jan Hubicka wrote:
> (edge_execution_predicate): Rewrite as...
> (set_cond_stmt_execution_predicate): ... this function; handle
> __builtin_constant_p.
This causes ICEs when building recent Linux kernels with the
CONFIG_TRACE_BRANCH_PROFLING option. This reduced test case:
sta
Scheduling floating point insns is broken on C6X; unfortunately in a way
that the simulator doesn't detect.
We let the sched DFA pick a valid schedule, and later on we try to
reconstruct a valid unit assignment. This does not work if an insn has
multiple units it can choose, and that insn's reserv
On Fri, Sep 02, 2011 at 07:14:58PM +0200, Paolo Carlini wrote:
> >Any comments related to the implementation, or examples of real-world
> >lame C string length code sequences that would be nice to optimize
> >will be greatly appreciated.
> I'm only wondering how hard would be taking care more consi
Hi,
On Fri, 2 Sep 2011, Michael Matz wrote:
> Hi,
>
> On Fri, 2 Sep 2011, Richard Guenther wrote:
>
> > > Currently regstrapping on x86_64-linux (without Ada). Okay for trunk?
> >
> > Ok. Time to make get_var_ann private?
>
> Yes. Should the regstrap succeed on x86_64-linux I'll commit thi
1 - 100 of 105 matches
Mail list logo
