Re: [PATCH PR target/65058] AIX: missing extern decorations "[DS]" for functions and "[RW]" for variables

[Michael Haubenwallner]

Hi David,

Am 2015-02-14 um 22:05 schrieb David Edelsohn:
> Hi, Michael
> 
> Thanks for noticing this.  This patch generally seems to be on the
> right track.  The original ASM_OUTPUT_EXTERNAL code was not completely
> correct in the pedantic sense.  It should use [UA] mapping class
> instead of [RW] for all non-function descriptor symbols.

One more thought: How is that dollar_inside thing seen in
ASM_OUTPUT_EXTERNAL supposed to work/be used?

> This patch also needs a dg-final scan-assembler test to check for [DS]
> and [UA] to ensure that this does not regress again in the future.

Thanks for committing!

/haubi/ (the one you gave commit access already:)

Re: [patch] fix PR65048: check that jump-thread paths are still valid

[Jeff Law]

On 02/13/15 16:50, Sebastian Pop wrote:

Hi,

the attached patch fixes PR65048 by checking before jump-threading that a path
to be threaded is still valid: as the testcase shows, there may be paths that
are not connected anymore because the cfg has changed in a previous jump-thread.

 PR tree-optimization/65048
 * tree-ssa-threadupdate.c (valid_jump_thread_path): New.
 (thread_through_all_blocks): Call valid_jump_thread_path.
 Remove invalid FSM jump-thread paths.

 * gcc.dg/tree-ssa/ssa-dom-thread-9.c: New.

The patch passed bootstrap and regression tests on x86_64-linux.
Ok for trunk?

These kinds of situations are normally pruned out in mark_threaded_blocks.

The dumps for the FSM threads are a bit sparse -- they don't show the 
entire path.  That makes it much harder to see what's going on.


It also appears that FSM is  registering lots of duplicate paths.

grep Registering j*.dom1 | grep -v PHI | sort
  Registering FSM jump thread: (10, 12) incoming edge;  (15, 3) nocopy;
  Registering FSM jump thread: (11, 12) incoming edge;  (15, 16) nocopy;
  Registering FSM jump thread: (16, 3) incoming edge;  (15, 16) nocopy;
  Registering FSM jump thread: (5, 10) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (5, 10) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (5, 11) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (5, 11) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (6, 14) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (6, 14) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (6, 3) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (6, 3) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (7, 10) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (7, 10) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (7, 11) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (7, 11) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (8, 14) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (8, 14) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (8, 3) incoming edge;  (13, 14) nocopy;
  Registering FSM jump thread: (8, 3) incoming edge;  (13, 14) nocopy;


Something seems wrong there.


Anyway, so what node precisely is not connected?  Is that happening as a 
result of the duplicated jump threads or is it something else?


Jeff

Re: [Haifa Scheduler] Fix latent bug in macro-fusion/instruction grouping

[Jeff Law]

On 02/11/15 02:20, James Greenhalgh wrote:


On Mon, Feb 09, 2015 at 11:16:56PM +, Jeff Law wrote:

On 02/06/15 05:24, James Greenhalgh wrote:


---
2015-02-06  James Greenhalgh  

* haifa-sched.c (recompute_todo_spec): After applying a
replacement and cancelling a dependency, also clear the
SCHED_GROUP_P flag.

My worry here would be that we might be clearing a SCHED_GROUP_P that
had been set for some reason other than macro-fusion.


Yeah, I also had this worry. This patch tackles the problem from the
other direction. If we see a SCHED_GROUP_P on an insn, treat it as a
hard dependency, and don't try to rewrite it. I think this will always
be "safe" but it might pessimize if the dependency breaker would have
resulted in better code generation.

I don't think this gives you anything towards fixing your bug, but
it clears mine.
Right.  Mine was in the management of the ready queue.  We allowed 
something with SCHED_GROUP_P to get deferred for several cycles.  While 
it was deferred another insn that was previously deferred became ready 
and fired.  That messed up the scheduling group and ultimately resulted 
in incorrect code.


The fix was actually pretty simple, We just queue the SCHED_GROUP_P for 
a single cycle, then reevaluate.




I've bootstrapped and tested on x86_64-unknown-linux-gnu with no
issues and given it a quick check on the problem code from before,
where it has the desired impact.

Thanks,
James

---
2015-02-10  James Greenhalgh  

* haifa-sched.c (recompute_todo_spec): Treat SCHED_GROUP_P
as forcing a HARD_DEP between instructions, thereby
disallowing rewriting to break dependencies.

OK.
jeff

Re: [PATCH][RFA][LRA] Don't try to break down subreg expressions if insn already matches

[Jeff Law]

On 02/14/15 04:23, Maxim Kuvyrkov wrote:


FYI, (and not related to the core issue of this patch)

The use of mult vs shift by combine is a problem that Venkat is
working on, see "[RFC] Tighten memory type assumption in RTL combiner
pass" .  The combiner uses MULTs instead of SHIFTs for rtx'es that
look like addresses, even when they are, in fact, mere
logic/arithmetic operations.
Right.  I think I put that into my gcc-6 queue because we don't have a 
regression that requires this problem be fixed.


If you've got a BZ that's regressing because of this issue, don't 
hesitate to point it out and I'll move the thread into my gcc-5 queue.


jeff

Re: Ping : [PATCH] [gcc, combine] PR46164: Don't combine the insns if a volatile register is contained.

[Terry Guo]

On Sun, Feb 15, 2015 at 7:35 PM, Segher Boessenkool
 wrote:
> Hi Terry,
>
> I still think this is stage1 material.
>
>> +  /* Don't combine if dest contains a user specified register and i3 
>> contains
>> + ASM_OPERANDS, because the user specified register (same with dest) in 
>> i3
>> + would be replaced by the src of insn which might be different with
>> + the user's expectation.  */
>
> "Do not eliminate a register asm in an asm input" or similar?  Text
> explaining why REG_USERVAR_P && HARD_REGISTER_P works here would be
> good to have, too.
>
>> +  if (REG_P (dest) && REG_USERVAR_P (dest) && HARD_REGISTER_P (dest)
>> +  && (GET_CODE (PATTERN (i3)) == SET
>> +   && GET_CODE (SET_SRC (PATTERN (i3))) == ASM_OPERANDS))
>> +return 0;
>
> That works only for asms with exactly one output.  You want
> extract_asm_operands.
>
>
> Segher

Thanks Segher. Patch is updated per you suggestion. Is this one ok for stage 1?

BR,
Terry


pr64818-combine-user-specified-register.patch-4
Description: Binary data

Re: [C++ PATCH] Fix constexpr C++11 handling with lambdas (PR c++/65075)

[Jason Merrill]

OK.

Jason

RE: [PATCH, GCC, stage1] Fallback to copy-prop if constant-prop not possible

[Thomas Preud'homme]

> From: Steven Bosscher [mailto:stevenb@gmail.com]
> Sent: Tuesday, February 17, 2015 4:19 AM
> To: Thomas Preud'homme
> Cc: GCC Patches; Richard Biener
> Subject: Re: [PATCH, GCC, stage1] Fallback to copy-prop if constant-prop
> not possible
> 
> On Mon, Feb 16, 2015 at 11:26 AM, Thomas Preud'homme wrote:
> 
> >  /* Subroutine of cprop_insn that tries to propagate constants into
> > @@ -1044,40 +1042,41 @@ cprop_insn (rtx_insn *insn)
> 
> > -  /* Constant propagation.  */
> > -  if (cprop_constant_p (src))
> > -   {
> > -  if (constprop_register (reg_used, src, insn))
> > + /* Constant propagation.  */
> > + if (src_cst && cprop_constant_p (src_cst)
> > + && constprop_register (reg_used, src_cst, insn))
> > {
> >   changed_this_round = changed = 1;
> >   global_const_prop_count++;
> 
> The cprop_constant_p test is redundant, you only have non-NULL
> src_cst
> if it is a cprop_constant_p (as you test for it in find_avail_set()).

Ack.

> 
> 
> > @@ -1087,18 +1086,16 @@ retry:
> >"GLOBAL CONST-PROP: Replacing reg %d in ", 
> > regno);
> >   fprintf (dump_file, "insn %d with constant ",
> >INSN_UID (insn));
> > - print_rtl (dump_file, src);
> > + print_rtl (dump_file, src_cst);
> >   fprintf (dump_file, "\n");
> > }
> >   if (insn->deleted ())
> > return 1;
> > }
> > -   }
> > -  else if (REG_P (src)
> > -  && REGNO (src) >= FIRST_PSEUDO_REGISTER
> > -  && REGNO (src) != regno)
> > -   {
> > - if (try_replace_reg (reg_used, src, insn))
> > + else if (src_reg && REG_P (src_reg)
> > +  && REGNO (src_reg) >= FIRST_PSEUDO_REGISTER
> > +  && REGNO (src_reg) != regno
> > +  && try_replace_reg (reg_used, src_reg, insn))
> 
> Likewise for the REG_P and ">= FIRST_PSEUDO_REGISTER" tests here
> (with
> the equivalent and IMHO preferable HARD_REGISTER_P test in
> find_avail_set()).

I'm not sure I follow you here. First, it seems to me that the equivalent
test is rather REG_P && !HARD_REGISTER_P since here it checks if it's
a pseudo register.

Then, do you mean the test can be simply removed because of the
REG_P && !HARD_REGISTER_P in hash_scan_set () called indirectly by
compute_hash_table () when called in one_cprop_pass () before any
cprop_insn ()? Or do you mean I should move the check in
find_avail_set ()?

Best regards,

Thomas

Re: [PATCH][PR tree-optimization/64823] Handle threading through blocks with PHIs, but no statements V2

[Jeff Law]

On 02/16/15 14:11, Jakub Jelinek wrote:

On Mon, Feb 16, 2015 at 02:00:32PM -0700, Jeff Law wrote:

--- a/gcc/tree-vrp.c
+++ b/gcc/tree-vrp.c
@@ -10176,13 +10176,20 @@ identify_jump_threads (void)
/* We only care about blocks ending in a COND_EXPR.  While there
 may be some value in handling SWITCH_EXPR here, I doubt it's
 terribly important.  */
-  last = gsi_stmt (gsi_last_bb (bb));
+  last = gsi_stmt (gsi_last_nondebug_bb (bb));


Isn't that just
last = last_stmt (bb);
equivalent?

It is.  I'll make that change and update the comment as well.

jeff

RE: [PATCH, FT32] initial support

[James Bowman]

Aha yes. Revised attached. invoke.texi now has:

These options are defined specifically for the FT32 port.

@table @gcctabopt

@item -msim
@opindex msim
Specifies that the program will be run on the simulator.  This causes
an alternate runtime startup and library to be linked.
You must not use this option when generating programs that will run on
real hardware; you must provide your own runtime library for whatever
I/O functions are needed.

@end table

--
James Bowman
FTDI Open Source Liaison



From: Joseph Myers [jos...@codesourcery.com]
Sent: Tuesday, February 17, 2015 2:06 AM
To: James Bowman
Cc: gcc-patches@gcc.gnu.org
Subject: RE: [PATCH, FT32] initial support

On Mon, 16 Feb 2015, James Bowman wrote:

> I have updated the target options. Space-saving is now enabled by
> -Os. There is also a new option -msim to enable building for the
> simulator (the simulator is pending submission to gdb-binutils).

The documentation in this patch doesn't seem to have been updated for
those changes.

--
Joseph S. Myers
jos...@codesourcery.com

gcc-ft32.txt.gz
Description: gcc-ft32.txt.gz

Re: nvptx offloading patches [3/n], RFD

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 10:35:30PM +0100, Richard Biener wrote:
> Seeing the real format string you introduce I wonder if identifying modes
> by their names wouldn't work in 99% of all cases (apart from PSImode
> maybe).

There are various corner cases.  Plus of course sometimes insignificant, but
sometimes very significant, floating mode changes.  SFmode on one target
might be completely different from another target.

> Also for most cases we can construct the machine mode from the type.  Or
> where that is not possible stream the extra info that is necessary
> instead.

I thought we've discussed that already on IRC.  E.g. decimal modes are
identified only by mode and nothing else, and it doesn't look like it
can be easily derived from types in many cases (spent quite some time on
that).

> Overall feels like a hack BTW :)  can't we assign machine mode enum IDs in
> a target independent way?  I mean, it doesn't have to be densely
> allocated?

We iterate over modes, we have tons of tables indexed by modes, so if we
introduce gaps, we'll make the compiler bigger and slower.
If this is limited to the offloading path, like in the attached updated
patch, the overhead for native LTO should be not measurable.

--- gcc/passes.c.jj 2015-02-16 22:18:33.219702315 +0100
+++ gcc/passes.c2015-02-16 22:19:20.842917807 +0100
@@ -2460,6 +2460,7 @@ ipa_write_summaries_1 (lto_symtab_encode
   struct lto_out_decl_state *state = lto_new_out_decl_state ();
   state->symtab_node_encoder = encoder;
 
+  lto_output_init_mode_table ();
   lto_push_out_decl_state (state);
 
   gcc_assert (!flag_wpa);
@@ -2581,6 +2582,7 @@ ipa_write_optimization_summaries (lto_sy
   lto_symtab_encoder_iterator lsei;
   state->symtab_node_encoder = encoder;
 
+  lto_output_init_mode_table ();
   lto_push_out_decl_state (state);
   for (lsei = lsei_start_function_in_partition (encoder);
!lsei_end_p (lsei); lsei_next_function_in_partition (&lsei))
--- gcc/tree-streamer.h.jj  2015-02-16 22:18:33.222702266 +0100
+++ gcc/tree-streamer.h 2015-02-16 22:19:20.843917791 +0100
@@ -24,6 +24,7 @@ along with GCC; see the file COPYING3.
 
 #include "streamer-hooks.h"
 #include "lto-streamer.h"
+#include "data-streamer.h"
 #include "hash-map.h"
 
 /* Cache of pickled nodes.  Used to avoid writing the same node more
@@ -91,6 +92,7 @@ void streamer_write_integer_cst (struct
 void streamer_write_builtin (struct output_block *, tree);
 
 /* In tree-streamer.c.  */
+extern unsigned char streamer_mode_table[1 << 8];
 void streamer_check_handled_ts_structures (void);
 bool streamer_tree_cache_insert (struct streamer_tree_cache_d *, tree,
 hashval_t, unsigned *);
@@ -119,5 +121,19 @@ streamer_tree_cache_get_hash (struct str
   return cache->hashes[ix];
 }
 
+static inline void
+bp_pack_machine_mode (struct bitpack_d *bp, machine_mode mode)
+{
+  streamer_mode_table[mode] = 1;
+  bp_pack_enum (bp, machine_mode, 1 << 8, mode);
+}
+
+static inline machine_mode
+bp_unpack_machine_mode (struct bitpack_d *bp)
+{
+  return (machine_mode)
+  ((struct lto_input_block *)
+   bp->stream)->mode_table[bp_unpack_enum (bp, machine_mode, 1 << 8)];
+}
 
 #endif  /* GCC_TREE_STREAMER_H  */
--- gcc/lto-streamer-out.c.jj   2015-02-16 22:18:33.204702562 +0100
+++ gcc/lto-streamer-out.c  2015-02-16 22:20:06.659163066 +0100
@@ -2642,6 +2642,96 @@ produce_symtab (struct output_block *ob)
 }
 
 
+/* Init the streamer_mode_table for output, where we collect info on what
+   machine_mode values have been streamed.  */
+void
+lto_output_init_mode_table (void)
+{
+  memset (streamer_mode_table, '\0', MAX_MACHINE_MODE);
+}
+
+
+/* Write the mode table.  */
+static void
+lto_write_mode_table (void)
+{
+  struct output_block *ob;
+  ob = create_output_block (LTO_section_mode_table);
+  bitpack_d bp = bitpack_create (ob->main_stream);
+
+  /* Ensure that for GET_MODE_INNER (m) != VOIDmode we have
+ also the inner mode marked.  */
+  for (int i = 0; i < (int) MAX_MACHINE_MODE; i++)
+if (streamer_mode_table[i])
+  {
+   machine_mode m = (machine_mode) i;
+   if (GET_MODE_INNER (m) != VOIDmode)
+ streamer_mode_table[(int) GET_MODE_INNER (m)] = 1;
+  }
+  /* First stream modes that have GET_MODE_INNER (m) == VOIDmode,
+ so that we can refer to them afterwards.  */
+  for (int pass = 0; pass < 2; pass++)
+for (int i = 0; i < (int) MAX_MACHINE_MODE; i++)
+  if (streamer_mode_table[i] && i != (int) VOIDmode && i != (int) BLKmode)
+   {
+ machine_mode m = (machine_mode) i;
+ if ((GET_MODE_INNER (m) == VOIDmode) ^ (pass == 0))
+   continue;
+ bp_pack_value (&bp, m, 8);
+ bp_pack_enum (&bp, mode_class, MAX_MODE_CLASS, GET_MODE_CLASS (m));
+ bp_pack_value (&bp, GET_MODE_SIZE (m), 8);
+ bp_pack_value (&bp, GET_MODE_PRECISION (m), 16);
+ bp_pack_value (&bp, GET_MODE_INNER (m), 8);
+ bp_pack_value (&bp, GET_MODE_NUNITS (m), 

Re: nvptx offloading patches [3/n], RFD

[Richard Biener]

On February 16, 2015 10:08:12 PM CET, Jakub Jelinek  wrote:
>Hi!
>
>On Mon, Feb 09, 2015 at 11:20:00AM +0100, Richard Biener wrote:
>> I think (also communicated that on IRC) we should instead try not
>streaming
>> machine-modes at all but generating them at stream-in time via
>layout_type
>> or layout_decl.
>
>Here is a WIP prototype for being able to stream a machine mode
>description
>table and streaming it back in.
>In the end, I'd like to stream this out only for lto_stream_offload_p
>and
>stream it in only for ACCEL_COMPILER reading in when available, but
>wanted
>to see what it does even for native LTO.
>For that it doesn't work very well, because it seems that wpa phase
>doesn't stream in some sections and stream them out again, but instead
>somehow copies them directly to the output object, so the mode table
>isn't aware of the modes used in there that were bypassed this way.
>
>Anyway, the question is if for offloading we use wpa stage at all these
>days
>or not at all, if there is a way for ACCEL_COMPILER to differentiate
>somehow between LTO sections written by the host compiler and LTO
>sections
>perhaps created by the offloading compiler when trying to LTO the thing
>(if
>it does it at all).  Because obviously the host compiler written LTO
>(in .gnu.offload_lto_*) would need the machine modes translated, while
>LTO streamed already by the ACCEL_COMPILER (if any) generally would
>already
>use the offloading target machine modes and therefore should be treated
>as
>native lto (.gnu.lto_*). 
>
>If we don't try to write .gnu.offload_lto_* again, I think following
>patch
>with additionally not calling lto_write_mode_table for
>!lto_stream_offload_p
>and not calling lto_input_mode_table for !ACCEL_COMPILER - instead
>build
>a single shared identity table - might actually work.
>
>Thoughts on this?

Seeing the real format string you introduce I wonder if identifying modes by 
their names wouldn't work in 99% of all cases (apart from PSImode maybe).

Also for most cases we can construct the machine mode from the type.  Or where 
that is not possible stream the extra info that is necessary instead.

Overall feels like a hack BTW :)  can't we assign machine mode enum IDs in a 
target independent way?  I mean, it doesn't have to be densely allocated?

Richard.

>Bernd/Thomas, do you plan to commit the other approved patches soon?
>
>--- gcc/passes.c.jj2015-02-16 20:14:09.477345693 +0100
>+++ gcc/passes.c   2015-02-16 20:26:23.659299189 +0100
>@@ -2460,6 +2460,7 @@ ipa_write_summaries_1 (lto_symtab_encode
>   struct lto_out_decl_state *state = lto_new_out_decl_state ();
>   state->symtab_node_encoder = encoder;
> 
>+  lto_output_init_mode_table ();
>   lto_push_out_decl_state (state);
> 
>   gcc_assert (!flag_wpa);
>@@ -2581,6 +2582,7 @@ ipa_write_optimization_summaries (lto_sy
>   lto_symtab_encoder_iterator lsei;
>   state->symtab_node_encoder = encoder;
> 
>+  lto_output_init_mode_table ();
>   lto_push_out_decl_state (state);
>   for (lsei = lsei_start_function_in_partition (encoder);
>!lsei_end_p (lsei); lsei_next_function_in_partition (&lsei))
>--- gcc/tree-streamer.h.jj 2015-02-16 20:14:09.446346202 +0100
>+++ gcc/tree-streamer.h2015-02-16 21:14:50.701615850 +0100
>@@ -24,6 +24,7 @@ along with GCC; see the file COPYING3.
> 
> #include "streamer-hooks.h"
> #include "lto-streamer.h"
>+#include "data-streamer.h"
> #include "hash-map.h"
> 
> /* Cache of pickled nodes.  Used to avoid writing the same node more
>@@ -91,6 +92,7 @@ void streamer_write_integer_cst (struct
> void streamer_write_builtin (struct output_block *, tree);
> 
> /* In tree-streamer.c.  */
>+extern unsigned char streamer_mode_table[1 << 8];
> void streamer_check_handled_ts_structures (void);
> bool streamer_tree_cache_insert (struct streamer_tree_cache_d *, tree,
>hashval_t, unsigned *);
>@@ -119,5 +121,19 @@ streamer_tree_cache_get_hash (struct str
>   return cache->hashes[ix];
> }
> 
>+static inline void
>+bp_pack_machine_mode (struct bitpack_d *bp, machine_mode mode)
>+{
>+  streamer_mode_table[mode] = 1;
>+  bp_pack_enum (bp, machine_mode, 1 << 8, mode);
>+}
>+
>+static inline machine_mode
>+bp_unpack_machine_mode (struct bitpack_d *bp)
>+{
>+  return (machine_mode)
>+ ((struct lto_input_block *)
>+  bp->stream)->mode_table[bp_unpack_enum (bp, machine_mode, 1 <<
>8)];
>+}
> 
> #endif  /* GCC_TREE_STREAMER_H  */
>--- gcc/lto-streamer-out.c.jj  2015-02-16 20:14:09.046352765 +0100
>+++ gcc/lto-streamer-out.c 2015-02-16 20:26:23.665299091 +0100
>@@ -2642,6 +2642,96 @@ produce_symtab (struct output_block *ob)
> }
> 
> 
>+/* Init the streamer_mode_table for output, where we collect info on
>what
>+   machine_mode values have been streamed.  */
>+void
>+lto_output_init_mode_table (void)
>+{
>+  memset (streamer_mode_table, '\0', MAX_MACHINE_MODE);
>+}
>+
>+
>+/* Write the mode table.  */
>+static void
>+lto_write_mode_table (void)
>+{
>+  struct output_bloc

Re: [PATCH][PR tree-optimization/64823] Handle threading through blocks with PHIs, but no statements V2

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 10:20:23PM +0100, Richard Biener wrote:
> On February 16, 2015 10:11:07 PM CET, Jakub Jelinek  wrote:
> >On Mon, Feb 16, 2015 at 02:00:32PM -0700, Jeff Law wrote:
> >> --- a/gcc/tree-vrp.c
> >> +++ b/gcc/tree-vrp.c
> >> @@ -10176,13 +10176,20 @@ identify_jump_threads (void)
> >>/* We only care about blocks ending in a COND_EXPR.  While
> >there
> >> may be some value in handling SWITCH_EXPR here, I doubt it's
> >> terribly important.  */
> >> -  last = gsi_stmt (gsi_last_bb (bb));
> >> +  last = gsi_stmt (gsi_last_nondebug_bb (bb));
> 
> And if the comment is correct then it should not even matter as the condition 
> ends a basic block.

It matters, because the use is:
  if (!last
  || gimple_code (last) == GIMPLE_SWITCH
  || (gimple_code (last) == GIMPLE_COND
  && TREE_CODE (gimple_cond_lhs (last)) == SSA_NAME
  && (INTEGRAL_TYPE_P (TREE_TYPE (gimple_cond_lhs (last)))
  || POINTER_TYPE_P (TREE_TYPE (gimple_cond_lhs (last
  && (TREE_CODE (gimple_cond_rhs (last)) == SSA_NAME
  || is_gimple_min_invariant (gimple_cond_rhs (last)
thus, if a bb contains only debug statements and nothing else,
the condition is false, while if it for corresponding -g0 doesn't contain
anything, it is true (!last).

Jakub

Re: [PATCH][PR tree-optimization/64823] Handle threading through blocks with PHIs, but no statements V2

[Richard Biener]

On February 16, 2015 10:11:07 PM CET, Jakub Jelinek  wrote:
>On Mon, Feb 16, 2015 at 02:00:32PM -0700, Jeff Law wrote:
>> --- a/gcc/tree-vrp.c
>> +++ b/gcc/tree-vrp.c
>> @@ -10176,13 +10176,20 @@ identify_jump_threads (void)
>>/* We only care about blocks ending in a COND_EXPR.  While
>there
>>   may be some value in handling SWITCH_EXPR here, I doubt it's
>>   terribly important.  */
>> -  last = gsi_stmt (gsi_last_bb (bb));
>> +  last = gsi_stmt (gsi_last_nondebug_bb (bb));

And if the comment is correct then it should not even matter as the condition 
ends a basic block.

Richard.

>Isn't that just
>   last = last_stmt (bb);
>equivalent?
>
>   Jakub

Re: [PATCH][PR tree-optimization/64823] Handle threading through blocks with PHIs, but no statements V2

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 02:00:32PM -0700, Jeff Law wrote:
> --- a/gcc/tree-vrp.c
> +++ b/gcc/tree-vrp.c
> @@ -10176,13 +10176,20 @@ identify_jump_threads (void)
>/* We only care about blocks ending in a COND_EXPR.  While there
>may be some value in handling SWITCH_EXPR here, I doubt it's
>terribly important.  */
> -  last = gsi_stmt (gsi_last_bb (bb));
> +  last = gsi_stmt (gsi_last_nondebug_bb (bb));

Isn't that just
last = last_stmt (bb);
equivalent?

Jakub

Re: nvptx offloading patches [3/n], RFD

[Jakub Jelinek]

Hi!

On Mon, Feb 09, 2015 at 11:20:00AM +0100, Richard Biener wrote:
> I think (also communicated that on IRC) we should instead try not streaming
> machine-modes at all but generating them at stream-in time via layout_type
> or layout_decl.

Here is a WIP prototype for being able to stream a machine mode description
table and streaming it back in.
In the end, I'd like to stream this out only for lto_stream_offload_p and
stream it in only for ACCEL_COMPILER reading in when available, but wanted
to see what it does even for native LTO.
For that it doesn't work very well, because it seems that wpa phase
doesn't stream in some sections and stream them out again, but instead
somehow copies them directly to the output object, so the mode table
isn't aware of the modes used in there that were bypassed this way.

Anyway, the question is if for offloading we use wpa stage at all these days
or not at all, if there is a way for ACCEL_COMPILER to differentiate
somehow between LTO sections written by the host compiler and LTO sections
perhaps created by the offloading compiler when trying to LTO the thing (if
it does it at all).  Because obviously the host compiler written LTO
(in .gnu.offload_lto_*) would need the machine modes translated, while
LTO streamed already by the ACCEL_COMPILER (if any) generally would already
use the offloading target machine modes and therefore should be treated as
native lto (.gnu.lto_*). 

If we don't try to write .gnu.offload_lto_* again, I think following patch
with additionally not calling lto_write_mode_table for !lto_stream_offload_p
and not calling lto_input_mode_table for !ACCEL_COMPILER - instead build
a single shared identity table - might actually work.

Thoughts on this?

Bernd/Thomas, do you plan to commit the other approved patches soon?

--- gcc/passes.c.jj 2015-02-16 20:14:09.477345693 +0100
+++ gcc/passes.c2015-02-16 20:26:23.659299189 +0100
@@ -2460,6 +2460,7 @@ ipa_write_summaries_1 (lto_symtab_encode
   struct lto_out_decl_state *state = lto_new_out_decl_state ();
   state->symtab_node_encoder = encoder;
 
+  lto_output_init_mode_table ();
   lto_push_out_decl_state (state);
 
   gcc_assert (!flag_wpa);
@@ -2581,6 +2582,7 @@ ipa_write_optimization_summaries (lto_sy
   lto_symtab_encoder_iterator lsei;
   state->symtab_node_encoder = encoder;
 
+  lto_output_init_mode_table ();
   lto_push_out_decl_state (state);
   for (lsei = lsei_start_function_in_partition (encoder);
!lsei_end_p (lsei); lsei_next_function_in_partition (&lsei))
--- gcc/tree-streamer.h.jj  2015-02-16 20:14:09.446346202 +0100
+++ gcc/tree-streamer.h 2015-02-16 21:14:50.701615850 +0100
@@ -24,6 +24,7 @@ along with GCC; see the file COPYING3.
 
 #include "streamer-hooks.h"
 #include "lto-streamer.h"
+#include "data-streamer.h"
 #include "hash-map.h"
 
 /* Cache of pickled nodes.  Used to avoid writing the same node more
@@ -91,6 +92,7 @@ void streamer_write_integer_cst (struct
 void streamer_write_builtin (struct output_block *, tree);
 
 /* In tree-streamer.c.  */
+extern unsigned char streamer_mode_table[1 << 8];
 void streamer_check_handled_ts_structures (void);
 bool streamer_tree_cache_insert (struct streamer_tree_cache_d *, tree,
 hashval_t, unsigned *);
@@ -119,5 +121,19 @@ streamer_tree_cache_get_hash (struct str
   return cache->hashes[ix];
 }
 
+static inline void
+bp_pack_machine_mode (struct bitpack_d *bp, machine_mode mode)
+{
+  streamer_mode_table[mode] = 1;
+  bp_pack_enum (bp, machine_mode, 1 << 8, mode);
+}
+
+static inline machine_mode
+bp_unpack_machine_mode (struct bitpack_d *bp)
+{
+  return (machine_mode)
+  ((struct lto_input_block *)
+   bp->stream)->mode_table[bp_unpack_enum (bp, machine_mode, 1 << 8)];
+}
 
 #endif  /* GCC_TREE_STREAMER_H  */
--- gcc/lto-streamer-out.c.jj   2015-02-16 20:14:09.046352765 +0100
+++ gcc/lto-streamer-out.c  2015-02-16 20:26:23.665299091 +0100
@@ -2642,6 +2642,96 @@ produce_symtab (struct output_block *ob)
 }
 
 
+/* Init the streamer_mode_table for output, where we collect info on what
+   machine_mode values have been streamed.  */
+void
+lto_output_init_mode_table (void)
+{
+  memset (streamer_mode_table, '\0', MAX_MACHINE_MODE);
+}
+
+
+/* Write the mode table.  */
+static void
+lto_write_mode_table (void)
+{
+  struct output_block *ob;
+  ob = create_output_block (LTO_section_mode_table);
+  bitpack_d bp = bitpack_create (ob->main_stream);
+
+  /* Ensure that for GET_MODE_INNER (m) != VOIDmode we have
+ also the inner mode marked.  */
+  for (int i = 0; i < (int) MAX_MACHINE_MODE; i++)
+if (streamer_mode_table[i])
+  {
+   machine_mode m = (machine_mode) i;
+   if (GET_MODE_INNER (m) != VOIDmode)
+ streamer_mode_table[(int) GET_MODE_INNER (m)] = 1;
+  }
+  /* First stream modes that have GET_MODE_INNER (m) == VOIDmode,
+ so that we can refer to them afterwards.  */
+  for (int pass = 0; pass < 2; pass++)
+for (int i = 0; i < (int

[PATCH][PR tree-optimization/64823] Handle threading through blocks with PHIs, but no statements V2

[Jeff Law]

The prior version of this patch failed to bootstrap with some 
non-standard configure options on x86_64-unknown-linux-gnu.  The problem 
was existing code which looked for the last statement in a block.  It 
should have looked through non-debug insns which was a trivial change to 
use gsi_last_nondebug_bb in one more place.


Bootstrapped and regression tested on x86_64-unknown-linux-gnu with HJ's 
options.  Installed on the trunk.  Will obviously keep my eye out for 
any new issues.


Jeff
commit 4c181d63db537424b28e5d022f6cbec53594ac8f
Author: Jeff Law 
Date:   Mon Feb 16 13:54:35 2015 -0700

PR tree-optimization/64823
* tree-vrp.c (identify_jump_threads): Handle blocks with no real
statements.
* tree-ssa-threadedge.c (potentially_threadable_block): Allow
threading through blocks with PHIs, but no statements.
(thread_through_normal_block): Distinguish between blocks where
we did not process all the statements and blocks with no statements.

PR tree-optimization/64823
* gcc.dg/uninit-20.c: New test.

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 9ef0d8c..bbeee3f 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,13 @@
+2015-02-16  Jeff Law  
+
+   PR tree-optimization/64823
+   * tree-vrp.c (identify_jump_threads): Handle blocks with no real
+   statements.
+   * tree-ssa-threadedge.c (potentially_threadable_block): Allow
+   threading through blocks with PHIs, but no statements.
+   (thread_through_normal_block): Distinguish between blocks where
+   we did not process all the statements and blocks with no statements.
+
 2015-02-16  Jakub Jelinek  
James Greenhalgh  
 
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index d5769b7..06ed820 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2015-02-16  Jeff Law  
+
+   PR tree-optimization/64823
+   * gcc.dg/uninit-20.c: New test.
+
 2015-02-16  Jakub Jelinek  
James Greenhalgh  
 
diff --git a/gcc/testsuite/gcc.dg/uninit-20.c b/gcc/testsuite/gcc.dg/uninit-20.c
new file mode 100644
index 000..12001ae
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/uninit-20.c
@@ -0,0 +1,18 @@
+/* Spurious uninitialized variable warnings, from gdb */
+/* { dg-do compile } */
+/* { dg-options "-O2 -Wuninitialized" } */
+struct os { struct o *o; };
+struct o { struct o *next; struct os *se; };
+void f(struct o *o){
+  struct os *s;
+  if(o) s = o->se;
+  while(o && s == o->se){
+s++; // here `o' is non-zero and thus s is initialized
+s == o->se  // `?' is essential, `if' does not trigger the warning
+  ? (o = o->next, o ? s = o->se : 0)
+  : 0;
+  }
+}
+
+
+
diff --git a/gcc/tree-ssa-threadedge.c b/gcc/tree-ssa-threadedge.c
index 4f83991..7187d06 100644
--- a/gcc/tree-ssa-threadedge.c
+++ b/gcc/tree-ssa-threadedge.c
@@ -110,6 +110,15 @@ potentially_threadable_block (basic_block bb)
 {
   gimple_stmt_iterator gsi;
 
+  /* Special case.  We can get blocks that are forwarders, but are
+ not optimized away because they forward from outside a loop
+ to the loop header.   We want to thread through them as we can
+ sometimes thread to the loop exit, which is obviously profitable. 
+ the interesting case here is when the block has PHIs.  */
+  if (gsi_end_p (gsi_start_nondebug_bb (bb))
+  && !gsi_end_p (gsi_start_phis (bb)))
+return true;
+  
   /* If BB has a single successor or a single predecessor, then
  there is no threading opportunity.  */
   if (single_succ_p (bb) || single_pred_p (bb))
@@ -1281,16 +1290,32 @@ thread_through_normal_block (edge e,
 = record_temporary_equivalences_from_stmts_at_dest (e, stack, simplify,
*backedge_seen_p);
 
-  /* If we didn't look at all the statements, the most likely reason is
- there were too many and thus duplicating this block is not profitable.
+  /* There's two reasons STMT might be null, and distinguishing
+ between them is important.
 
- Also note if we do not look at all the statements, then we may not
- have invalidated equivalences that are no longer valid if we threaded
- around a loop.  Thus we must signal to our caller that this block
- is not suitable for use as a joiner in a threading path.  */
+ First the block may not have had any statements.  For example, it
+ might have some PHIs and unconditionally transfer control elsewhere.
+ Such blocks are suitable for jump threading, particularly as a
+ joiner block.
+
+ The second reason would be if we did not process all the statements
+ in the block (because there were too many to make duplicating the
+ block profitable.   If we did not look at all the statements, then
+ we may not have invalidated everything needing invalidation.  Thus
+ we must signal to our caller that this block is not suitable for
+ use as a joiner in a thr

Re: [debug-early] C++ clones and limbo DIEs

[Aldy Hernandez]

On 02/12/2015 11:27 AM, Jason Merrill wrote:

On 02/12/2015 01:04 PM, Aldy Hernandez wrote:

On 02/10/2015 02:52 AM, Richard Biener wrote:

On Fri, Feb 6, 2015 at 5:42 PM, Aldy Hernandez  wrote:



Of course I wonder why you need to separate handling of functions and
variables


The variables need to be handled earlier, else the call to
analyze_functions() will remove some optimized global variables away,
and we'll never see them.  I believe that Jason said they were needed
up-thread.


variables.  What breaks if you emit debug info for functions before
the first analyze_functions () call?

 >
 > I also wonder why you restrict it to functions with a GIMPLE body.

The functions, on the other hand, need to be handled after the second
call to analyze_function (and with a GIMPLE body) else we get far more
function DIEs than mainline currently does, especially wrt C++ clones.
Otherwise, we get DIEs for base constructors, complete constructors, and
what-have-yous.  Jason wanted less DIEs, more attune to what mainline is
currently doing.


I think it makes sense to generate DIEs for everything defined in the TU
if we don't have -feliminate-unused-debug-symbols.  But since clones are
artificial, emit them only if they're used.


Ok, just so we're on the same page.  I'm thinking that for 
-fNO-eliminate-unused-debug-symbols, we can iterate through 
FOR_EACH_DEFINED_FUNCTION before unreachable functions have been 
removed.  There we can output all non-clones.


Then for the -feliminate-unused-debug-symbols case, we can output 
reachable functions after the unreachable ones have been removed.  Here 
we can also dump the clones we ignored for 
-fNO-eliminate-unused-debug-symbols above, since we only want to emit 
them if they're reachable (regardless of -feliminate-unused-debug-symbols).


In either case, we always ignore those without a gimple body, otherwise 
we end up generating DIEs for the _ZN1AC2Ei constructor in the attached 
function unnecessarily.  See how the bits end up in the attached testcase:


(Oh, and we determine clonehood with DECL_ABSTRACT_ORIGIN)

Before any calls to analyze_functions()
---
Function: 'int main()' (Mangled: main) gimple_body=1 DECL_ABSTRACT_ORIGIN=0
Function: 'A::A(int)' (Mangled: _ZN1AC1Ei) gimple_body=0 
DECL_ABSTRACT_ORIGIN=1
Function: 'A::A(int)' (Mangled: _ZN1AC2Ei) gimple_body=1 
DECL_ABSTRACT_ORIGIN=1
Function: 'void foo(int)' (Mangled: _Z3fooi) gimple_body=1 
DECL_ABSTRACT_ORIGIN=0
Function: 'int bar()' (Mangled: _Z3barv) gimple_body=1 
DECL_ABSTRACT_ORIGIN=0
Function: 'void unreachable_func()' (Mangled: _ZL16unreachable_funcv) 
gimple_body=1 DECL_ABSTRACT_ORIGIN=0


After reachability analysis
(after first call to analyze_functions())
-
Function: 'int main()' (Mangled: main) gimple_body=1 DECL_ABSTRACT_ORIGIN=0
Function: 'A::A(int)' (Mangled: _ZN1AC1Ei) gimple_body=0 
DECL_ABSTRACT_ORIGIN=1
Function: 'A::A(int)' (Mangled: _ZN1AC2Ei) gimple_body=1 
DECL_ABSTRACT_ORIGIN=1
Function: 'void foo(int)' (Mangled: _Z3fooi) gimple_body=1 
DECL_ABSTRACT_ORIGIN=0
Function: 'int bar()' (Mangled: _Z3barv) gimple_body=1 
DECL_ABSTRACT_ORIGIN=0


Is this what you had in mind?  I can provide a patch to make things clearer.

Aldy
extern "C" void abort ();
struct A { A (int); int a; };

int i;

static void unreachable_func()
{
  i = 5;
}


__attribute__((noinline, noclone)) int
bar (void)
{
  return 40;
}

__attribute__((noinline, noclone)) void
foo (int x)
{
  __asm volatile ("" : : "r" (x) : "memory");
}

A::A (int x)
{
  static int p = bar ();
  foo (p);
  a = ++p;
}

int
main ()
{
  A a (42);
  if (a.a != 41)
abort ();
}

Re: [PATCH, GCC, stage1] Fallback to copy-prop if constant-prop not possible

[Steven Bosscher]

On Mon, Feb 16, 2015 at 11:26 AM, Thomas Preud'homme wrote:

>  /* Subroutine of cprop_insn that tries to propagate constants into
> @@ -1044,40 +1042,41 @@ cprop_insn (rtx_insn *insn)

> -  /* Constant propagation.  */
> -  if (cprop_constant_p (src))
> -   {
> -  if (constprop_register (reg_used, src, insn))
> + /* Constant propagation.  */
> + if (src_cst && cprop_constant_p (src_cst)
> + && constprop_register (reg_used, src_cst, insn))
> {
>   changed_this_round = changed = 1;
>   global_const_prop_count++;

The cprop_constant_p test is redundant, you only have non-NULL src_cst
if it is a cprop_constant_p (as you test for it in find_avail_set()).


> @@ -1087,18 +1086,16 @@ retry:
>"GLOBAL CONST-PROP: Replacing reg %d in ", regno);
>   fprintf (dump_file, "insn %d with constant ",
>INSN_UID (insn));
> - print_rtl (dump_file, src);
> + print_rtl (dump_file, src_cst);
>   fprintf (dump_file, "\n");
> }
>   if (insn->deleted ())
> return 1;
> }
> -   }
> -  else if (REG_P (src)
> -  && REGNO (src) >= FIRST_PSEUDO_REGISTER
> -  && REGNO (src) != regno)
> -   {
> - if (try_replace_reg (reg_used, src, insn))
> + else if (src_reg && REG_P (src_reg)
> +  && REGNO (src_reg) >= FIRST_PSEUDO_REGISTER
> +  && REGNO (src_reg) != regno
> +  && try_replace_reg (reg_used, src_reg, insn))

Likewise for the REG_P and ">= FIRST_PSEUDO_REGISTER" tests here (with
the equivalent and IMHO preferable HARD_REGISTER_P test in
find_avail_set()).


Looks good to me otherwise.

Ciao!
Steven

RE: [PATCH] Fix PR64980 and PR61960

[Bernd Edlinger]

again, with attachments,
sorry.


>
> Hi,
>
>
> this patch fixes PR64980 and PR61960 at the same time.
>
> The unreduced test case for PR64230 is also included, because a previous 
> version
> of this patch caused this test to fail but the complete test suite passed 
> without any
> indication of any problem.
>
> Boot-strapped and regression-tested on X86_64-unknown-linux-gnu.
>
> OK for trunk?
>
>
> Thanks,
> Bernd.
>
>
> PS: Special thanks to Dominique for his help with testing of this patch.
>
  2015-02-16  Bernd Edlinger  

PR fortran/64980
PR fortran/61960
* trans-expr.c (gfc_apply_interface_mapping_to_expr): Remove mapping
for component references to class objects.
(gfc_conv_procedure_call): Compare the class by name.

testsuite:
2015-02-16  Bernd Edlinger  

PR fortran/64980
PR fortran/61960
* gfortran.dg/pr61960.f90: New.
* gfortran.dg/pr64230.f90: New.
* gfortran.dg/pr64980.f03: New.



patch-pr64980.diff
Description: Binary data

[PATCH] Fix PR64980 and PR61960

[Bernd Edlinger]

Hi,


this patch fixes PR64980 and PR61960 at the same time.

The unreduced test case for PR64230 is also included, because a previous version
of this patch caused this test to fail but the complete test suite passed 
without any
indication of any problem.

Boot-strapped and regression-tested on X86_64-unknown-linux-gnu.

OK for trunk?


Thanks,
Bernd.


PS: Special thanks to Dominique for his help with testing of this patch.
  

Confidential Message. Reply ASAP ***

[Mr. David Nicodemus]

Good day,

I have been trying to reach you without success. Glad I could be able to get in 
touch with you today. Kindly reply as soon as possible in order to get back to 
you in regards to reason i have been trying to reach you.

Regards

Mr. David Nicodemus
Executive Director

Re: [PATCH][PR tree-optimization/64823] Handle threading through blocks with PHIs, but no statements

[Jeff Law]

On 02/13/15 23:12, Jack Howarth wrote:

This also breaks the bootstrap on x86_64-apple-darwin14 due to a
similar stage 2/3 comparison failure.
Thanks.  I'm pretty sure I've got the root cause of both of these 
failures.  There's a gsi_last_bb in some existing code that really needs 
to be changed into gsi_last_nondebug_bb.


It didn't matter before because a block ending in debug statements was 
never considered potentially threadable by VRP.  However it matters with 
my change because we're allowing threading through such blocks.  Strange 
that it didn't show up in my tests.  But it's definitely a real issue.


Jeff

Re: OMP builtins in offloading (was: [PATCH 1/4] Add mkoffload for Intel MIC)

[Ilya Verbin]

On Thu, Jan 08, 2015 at 16:49:40 +0100, Jakub Jelinek wrote:
> BTW, today when looking at the TARGET_OPTION_NODE streaming caused
> regressions, I've discovered that it is very hard to debug issues in the
> offloading compiler.  Would be nice if
> -save-temps -v
> printed enough information that it is actually possible to reproduce it,
> e.g. while mkoffload command is printed, one can't cut and paste it easily,
> because some env vars are required and those aren't printed in the -v dump.

Currently I see all required env vars for mkoffload in the -v dump:
COLLECT_GCC=...
COMPILER_PATH=...
.../mkoffload @...

It doesn't need anything more.

  -- Ilya

Re: [committed] Change "Q" and "T" constraints to memory constraints

[John David Anglin]

On 2015-02-16, at 11:38 AM, Richard Henderson wrote:

>> 
>> Possibly the constant can somehow be forced into the data section where the 
>> relocations
>> aren't a problem?
> 
> Hmm.  It looks like we might already do that.  See default_select_rtx_section.

Thanks, I see the problem.  default_reloc_rw_mask returns 0 when not generating 
PIC
code, so rtx went to readonly_data_section.  I was thinking that 
pa_select_section was
somehow broken.

Dave
--
John David Anglin   dave.ang...@bell.net

Re: [PATCH] Copy over section name during cloning (PR ipa/64963)

[Jan Hubicka]

> On Mon, Feb 16, 2015 at 07:23:33PM +0100, Jan Hubicka wrote:
> > > --- gcc/cgraphclones.c.jj 2015-01-09 21:59:44.0 +0100
> > > +++ gcc/cgraphclones.c2015-02-16 14:02:16.564725881 +0100
> > > @@ -577,7 +577,7 @@ cgraph_node::create_virtual_clone (vec > >char *name;
> > >  
> > >if (!in_lto_p)
> > > -gcc_checking_assert  (tree_versionable_function_p (old_decl));
> > > +gcc_checking_assert (tree_versionable_function_p (old_decl));
> > >  
> > >gcc_assert (local.can_change_signature || !args_to_skip);
> > >  
> > > @@ -617,6 +617,8 @@ cgraph_node::create_virtual_clone (vec > >   ABI support for this.  */
> > >set_new_clone_decl_and_node_flags (new_node);
> > >new_node->clone.tree_map = tree_map;
> > > +  if (!DECL_ONE_ONLY (old_decl))
> > 
> > Instead of DECL_ONE_ONLY you want to test implicit_section flag.  I think
> > resolving unique section with -ffunction-section is also needed.
> 
> DECL_ONE_ONLY was the test that 4.9 has been using here:
> 
>   /* Update the properties.
>  Make clone visible only within this translation unit.  Make sure
>  that is not weak also.
>  ??? We cannot use COMDAT linkage because there is no
>  ABI support for this.  */
>   if (DECL_ONE_ONLY (old_decl))
> DECL_SECTION_NAME (new_node->decl) = NULL;
> 
> therefore I wanted to match the 4.9 behavior, before we try something
> different incrementally.

OK. implicit_section is trye only for DECL_ONE_ONLY declarations and those
that was named by resolve_unique_section via -ffunction-sections, so the
change should be safe, but lets first go with immitating 4.9 behaviour.

Incremetnally I think we should fix this and also teach inliner to not
make code travel in between user named sections.  This may become more
issue with an LTO kernel builds.

Honza
> 
> > > +new_node->set_section (this->get_section ());
> > 
> > No need for this->...
> 
> Sure, can change that.
> 
>   Jakub

Re: [PATCH] Copy over section name during cloning (PR ipa/64963)

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 07:23:33PM +0100, Jan Hubicka wrote:
> > --- gcc/cgraphclones.c.jj   2015-01-09 21:59:44.0 +0100
> > +++ gcc/cgraphclones.c  2015-02-16 14:02:16.564725881 +0100
> > @@ -577,7 +577,7 @@ cgraph_node::create_virtual_clone (vec >char *name;
> >  
> >if (!in_lto_p)
> > -gcc_checking_assert  (tree_versionable_function_p (old_decl));
> > +gcc_checking_assert (tree_versionable_function_p (old_decl));
> >  
> >gcc_assert (local.can_change_signature || !args_to_skip);
> >  
> > @@ -617,6 +617,8 @@ cgraph_node::create_virtual_clone (vec >   ABI support for this.  */
> >set_new_clone_decl_and_node_flags (new_node);
> >new_node->clone.tree_map = tree_map;
> > +  if (!DECL_ONE_ONLY (old_decl))
> 
> Instead of DECL_ONE_ONLY you want to test implicit_section flag.  I think
> resolving unique section with -ffunction-section is also needed.

DECL_ONE_ONLY was the test that 4.9 has been using here:

  /* Update the properties.
 Make clone visible only within this translation unit.  Make sure
 that is not weak also.
 ??? We cannot use COMDAT linkage because there is no
 ABI support for this.  */
  if (DECL_ONE_ONLY (old_decl))
DECL_SECTION_NAME (new_node->decl) = NULL;

therefore I wanted to match the 4.9 behavior, before we try something
different incrementally.

> > +new_node->set_section (this->get_section ());
> 
> No need for this->...

Sure, can change that.

Jakub

[PATCH] Fix PR64748

[James Norris]

This fixes the validation of the argument to the deviceptr clause.

Bootstrapped and regtested on x86_64-unknown-linux-gnu.

OK to commit to trunk?

Jim

diff --git a/gcc/c/c-parser.c b/gcc/c/c-parser.c
index ceb9e1a..9f0d7af 100644
--- a/gcc/c/c-parser.c
+++ b/gcc/c/c-parser.c
@@ -10334,11 +10334,11 @@ c_parser_oacc_data_clause_deviceptr (c_parser *parser, tree list)
 	 c_parser_omp_var_list_parens() should construct a list of
 	 locations to go along with the var list.  */
 
-  if (TREE_CODE (v) != VAR_DECL)
-	error_at (loc, "%qD is not a variable", v);
-  else if (TREE_TYPE (v) == error_mark_node)
+  if (TREE_TYPE (v) == error_mark_node)
 	;
-  else if (!POINTER_TYPE_P (TREE_TYPE (v)))
+  else if ((TREE_CODE (v) != VAR_DECL ||
+	   TREE_CODE (v) != PARM_DECL) &&
+	   !POINTER_TYPE_P (TREE_TYPE (v)))
 	error_at (loc, "%qD is not a pointer variable", v);
 
   tree u = build_omp_clause (loc, OMP_CLAUSE_MAP);
diff --git a/gcc/cp/parser.c b/gcc/cp/parser.c
index 57dfbcc..37b4712 100644
--- a/gcc/cp/parser.c
+++ b/gcc/cp/parser.c
@@ -27988,11 +27988,11 @@ cp_parser_oacc_data_clause_deviceptr (cp_parser *parser, tree list)
 	 c_parser_omp_var_list_parens should construct a list of
 	 locations to go along with the var list.  */
 
-  if (TREE_CODE (v) != VAR_DECL)
-	error_at (loc, "%qD is not a variable", v);
-  else if (TREE_TYPE (v) == error_mark_node)
+  if (TREE_TYPE (v) == error_mark_node)
 	;
-  else if (!POINTER_TYPE_P (TREE_TYPE (v)))
+  else if ((TREE_CODE (v) != VAR_DECL ||
+	   TREE_CODE (v) != PARM_DECL) &&
+	   !POINTER_TYPE_P (TREE_TYPE (v)))
 	error_at (loc, "%qD is not a pointer variable", v);
 
   tree u = build_omp_clause (loc, OMP_CLAUSE_MAP);
diff --git a/gcc/testsuite/c-c++-common/goacc/deviceptr-1.c b/gcc/testsuite/c-c++-common/goacc/deviceptr-1.c
index 546fa82..5ec7540 100644
--- a/gcc/testsuite/c-c++-common/goacc/deviceptr-1.c
+++ b/gcc/testsuite/c-c++-common/goacc/deviceptr-1.c
@@ -8,27 +8,29 @@ fun1 (void)
 #pragma acc kernels deviceptr(u[0:4]) /* { dg-error "expected '\\\)' before '\\\[' token" } */
   ;
 
-#pragma acc data deviceptr(fun1) /* { dg-error "'fun1' is not a variable" } */
+#pragma acc data deviceptr(fun1) /* { dg-error "'fun1' is not a pointer variable" } */
+  /* { dg-error "'fun1' is not a variable in 'map' clause" "fun1 is not a varialbe in map clause" { target *-*-* } 11 } */
   ;
 #pragma acc parallel deviceptr(fun1[2:5])
-  /* { dg-error "'fun1' is not a variable" "not a variable" { target *-*-* } 13 } */
-  /* { dg-error "expected '\\\)' before '\\\[' token" "array" { target *-*-* } 13 } */
+  /* { dg-error "'fun1' is not a pointer variable" "not a pointer variable" { target *-*-* } 14 } */
+  /* { dg-error "expected '\\\)' before '\\\[' token" "array" { target *-*-* } 14 } */
+  /* { dg-error "'fun1' is not a variable in 'map' clause" "fun1 is not a varialbe in map clause" { target *-*-* } 14 } */
   ;
 
   int i;
 #pragma acc kernels deviceptr(i) /* { dg-error "'i' is not a pointer variable" } */
   ;
 #pragma acc data deviceptr(i[0:4])
-  /* { dg-error "'i' is not a pointer variable" "not a pointer variable" { target *-*-* } 21 } */
-  /* { dg-error "expected '\\\)' before '\\\[' token" "array" { target *-*-* } 21 } */
+  /* { dg-error "'i' is not a pointer variable" "not a pointer variable" { target *-*-* } 23 } */
+  /* { dg-error "expected '\\\)' before '\\\[' token" "array" { target *-*-* } 23 } */
   ;
 
   float fa[10];
 #pragma acc parallel deviceptr(fa) /* { dg-error "'fa' is not a pointer variable" } */
   ;
 #pragma acc kernels deviceptr(fa[1:5])
-  /* { dg-error "'fa' is not a pointer variable" "not a pointer variable" { target *-*-* } 29 } */
-  /* { dg-error "expected '\\\)' before '\\\[' token" "array" { target *-*-* } 29 } */
+  /* { dg-error "'fa' is not a pointer variable" "not a pointer variable" { target *-*-* } 31 } */
+  /* { dg-error "expected '\\\)' before '\\\[' token" "array" { target *-*-* } 31 } */
   ;
 
   float *fp;
@@ -44,10 +46,11 @@ fun2 (void)
   int i;
   float *fp;
 #pragma acc kernels deviceptr(fp,u,fun2,i,fp)
-  /* { dg-error "'u' undeclared" "u undeclared" { target *-*-* } 46 } */
-  /* { dg-error "'fun2' is not a variable" "fun2 not a variable" { target *-*-* } 46 } */
-  /* { dg-error "'i' is not a pointer variable" "i not a pointer variable" { target *-*-* } 46 } */
-  /* { dg-error "'fp' appears more than once in map clauses" "fp more than once" { target *-*-* } 46 } */
+  /* { dg-error "'u' undeclared" "u undeclared" { target *-*-* } 48 } */
+  /* { dg-error "'fun2' is not a pointer variable" "fun2 not a pointer variable" { target *-*-* } 48 } */
+  /* { dg-error "'i' is not a pointer variable" "i not a pointer variable" { target *-*-* } 48 } */
+  /* { dg-error "'fp' appears more than once in map clauses" "fp more than once" { target *-*-* } 48 } */
+  /* { dg-error "'fun2' is not a variable in 'map' clause" "fun2 is not a variable in map clause" { targ

Re: [PATCH] Copy over section name during cloning (PR ipa/64963)

[Jan Hubicka]

Hi,
> Hi!
> 
> As discussed in the PR, in 4.9 we used to clone DECL_SECTION_NAME
> through using copy_node on the FUNCTION_DECL, and only in selected places
> (e.g. when creating artificial_thunk.*, or when creating virtual clones
> of DECL_ONE_ONLY functions) we used to explicitly clear DECL_SECTION_NAME.
> In 5 the section name is stored in cgraph node instead, and thus not
> copied by default, so we instead need to copy it over to restore previous
> behavior, otherwise we break the Linux kernel and various other packages.
> 
> Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> trunk?
> 
> 2015-02-16  Jakub Jelinek  
>   James Greenhalgh  
> 
>   PR ipa/64963
>   * cgraphclones.c (cgraph_node::create_virtual_clone): Copy
>   section if not linkonce.  Fix up formatting.
>   (cgraph_node::create_version_clone_with_body): Copy section.
>   * trans-mem.c (ipa_tm_create_version): Likewise.
> 
>   * gcc.dg/ipa/ipa-clone-1.c: New test.

Sorry, for taking so long on this. I made similar patch yesterday just did not
get around testing it. Imissed the trans-mem case though :)
> 
> --- gcc/cgraphclones.c.jj 2015-01-09 21:59:44.0 +0100
> +++ gcc/cgraphclones.c2015-02-16 14:02:16.564725881 +0100
> @@ -577,7 +577,7 @@ cgraph_node::create_virtual_clone (vecchar *name;
>  
>if (!in_lto_p)
> -gcc_checking_assert  (tree_versionable_function_p (old_decl));
> +gcc_checking_assert (tree_versionable_function_p (old_decl));
>  
>gcc_assert (local.can_change_signature || !args_to_skip);
>  
> @@ -617,6 +617,8 @@ cgraph_node::create_virtual_clone (vec   ABI support for this.  */
>set_new_clone_decl_and_node_flags (new_node);
>new_node->clone.tree_map = tree_map;
> +  if (!DECL_ONE_ONLY (old_decl))

Instead of DECL_ONE_ONLY you want to test implicit_section flag.  I think
resolving unique section with -ffunction-section is also needed.

> +new_node->set_section (this->get_section ());

No need for this->...

OK with these changes.
>  
>/* Clones of global symbols or symbols with unique names are unique.  */
>if ((TREE_PUBLIC (old_decl)
> @@ -1009,6 +1011,7 @@ cgraph_node::create_version_clone_with_b
>new_version_node->externally_visible = 0;
>new_version_node->local.local = 1;
>new_version_node->lowered = true;
> +  new_version_node->set_section (this->get_section ());
>/* Clones of global symbols or symbols with unique names are unique.  */
>if ((TREE_PUBLIC (old_decl)
> && !DECL_EXTERNAL (old_decl)
> --- gcc/trans-mem.c.jj2015-01-14 09:55:19.0 +0100
> +++ gcc/trans-mem.c   2015-02-16 12:58:01.399808815 +0100
> @@ -4967,6 +4967,7 @@ ipa_tm_create_version (struct cgraph_nod
>new_node->externally_visible = old_node->externally_visible;
>new_node->lowered = true;
>new_node->tm_clone = 1;
> +  new_node->set_section (old_node->get_section ());
>get_cg_data (&old_node, true)->clone = new_node;
>  
>if (old_node->get_availability () >= AVAIL_INTERPOSABLE)
> --- gcc/testsuite/gcc.dg/ipa/ipa-clone-1.c.jj 2015-02-16 14:14:39.041625503 
> +0100
> +++ gcc/testsuite/gcc.dg/ipa/ipa-clone-1.c2015-02-16 14:15:31.944760949 
> +0100
> @@ -0,0 +1,20 @@
> +/* PR ipa/64693 */
> +/* { dg-do compile } */
> +/* { dg-require-named-sections "" } */
> +/* { dg-options "-O3 -fipa-cp -fipa-cp-clone -fdump-ipa-cp"  } */
> +
> +static int __attribute__ ((noinline, section ("test_section")))
> +foo (int arg)
> +{
> +  return 7 * arg;
> +}
> +
> +int
> +bar (int arg)
> +{
> +  return foo (5);
> +}
> +
> +/* { dg-final { scan-assembler "test_section" } } */
> +/* { dg-final { scan-ipa-dump "Creating a specialized node of foo" "cp" } } 
> */
> +/* { dg-final { cleanup-ipa-dump "cp" } } */
> 
>   Jakub

RE: [PATCH, FT32] initial support

[Joseph Myers]

On Mon, 16 Feb 2015, James Bowman wrote:

> I have updated the target options. Space-saving is now enabled by
> -Os. There is also a new option -msim to enable building for the
> simulator (the simulator is pending submission to gdb-binutils).

The documentation in this patch doesn't seem to have been updated for 
those changes.

-- 
Joseph S. Myers
jos...@codesourcery.com

Re: [PATCH][3/n] Fix PR65015

[Richard Biener]

On Mon, 16 Feb 2015, H.J. Lu wrote:

> On Mon, Feb 16, 2015 at 6:25 AM, Richard Biener  wrote:
> >
> > This fixes another leakage of random LTO temporary filenames into
> > executables, this time via .symtab FILE entries.  Removing it
> > doesn't work (GNU ld adds it back) and is said to be incorrect.
> 
> FWIW, ld.bfd will be fixed in 2.26.  But it is still a good idea to
> support older links.

Yeah, especiall as I plan to backport this series if the reporter is
happy.

Richard.

> > So the following patch, similar to the dwarf CU DW_AT_name uses
> > .
> >
> > Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
> >
> > Richard.
> >
> > 2015-02-16  Richard Biener  
> >
> > PR lto/65015
> > * varasm.c (default_file_start): For LTO produced units
> > emit  as file directive.
> >
> > Index: gcc/varasm.c
> > ===
> > --- gcc/varasm.c(revision 220677)
> > +++ gcc/varasm.c(working copy)
> > @@ -7043,7 +7047,13 @@ default_file_start (void)
> >  fputs (ASM_APP_OFF, asm_out_file);
> >
> >if (targetm.asm_file_start_file_directive)
> > -output_file_directive (asm_out_file, main_input_filename);
> > +{
> > +  /* LTO produced units have no meaningful main_input_filename.  */
> > +  if (in_lto_p)
> > +   output_file_directive (asm_out_file, "");
> > +  else
> > +   output_file_directive (asm_out_file, main_input_filename);
> > +}
> >  }
> >
> >  /* This is a generic routine suitable for use as TARGET_ASM_FILE_END
> 
> 
> 
> 

-- 
Richard Biener 
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Jennifer Guild,
Dilip Upmanyu, Graham Norton HRB 21284 (AG Nuernberg)

Re: [PATCH, CHKP, PR target/65044] Restrict pointer bounds checker with Sanitizer

[Ilya Enkovich]

On 16 Feb 17:01, Jakub Jelinek wrote:
> On Mon, Feb 16, 2015 at 06:56:45PM +0300, Ilya Enkovich wrote:
> > On 16 Feb 16:31, Jakub Jelinek wrote:
> > > On Mon, Feb 16, 2015 at 06:20:59PM +0300, Ilya Enkovich wrote:
> > > > This patch restricts usage of Pointer Bounds Checker with Sanitizer.  
> > > > OK for trunk?
> > > 
> > > There are many sanitizers, and for most of them I don't see why they would
> > > be in any conflict with -mmpx, it is just -fsanitize=address and
> > > -fsanitize=kernel-address.
> > > So perhaps test instead if (flag_sanitize & SANITIZE_ADDRESS) != 0, and
> > > better clear the flag_pointer_bounds after issuing the error, error () is
> > > not a fatal function, so you need something sensible for error-recovery.
> > > 
> > >   Jakub
> > 
> > I don't know all sanitizers in details.  Code generated by some of them may 
> > be incorrect from checker point of view.  Thus I just wanted to disable 
> > unexplored and untested combinations.
> 
> Shouldn't be that hard to write a testcase and test it.
> 
> Most of the sanitizers just add code like
> if (some_condition)
>   __ubsan_handle_... ();
> where from the POV of the program the __ubsan_* function reports or might
> report some problem, and optionally abort the program.
> That some_condition can be a check of the pointer value, shift count,
> divisor check, etc.
> 
>   Jakub

OK.  With no tricky memory references this should be safe.  Here is a patch to 
filter off Adress Sanitizer only.

Thanks for review!

Ilya
--
gcc/

2015-02-16  Ilya Enkovich  

PR target/65044
* toplev.c (process_options): Restrict Pointer Bounds Checker
usage with Address Sanitizer.

gcc/testsuite/

2015-02-16  Ilya Enkovich  

PR target/65044
* gcc.target/i386/pr65044.c: New.


diff --git a/gcc/testsuite/gcc.target/i386/pr65044.c 
b/gcc/testsuite/gcc.target/i386/pr65044.c
new file mode 100644
index 000..4f318d6
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr65044.c
@@ -0,0 +1,12 @@
+/* { dg-error "-fcheck-pointer-bounds is not supported with Address Sanitizer" 
} */
+/* { dg-do compile } */
+/* { dg-require-effective-target mpx } */
+/* { dg-options "-fcheck-pointer-bounds -mmpx -fsanitize=address" } */
+
+extern int x[];
+
+void
+foo ()
+{
+  x[0] = 0;
+}
diff --git a/gcc/toplev.c b/gcc/toplev.c
index 99cf180..70eb6b6 100644
--- a/gcc/toplev.c
+++ b/gcc/toplev.c
@@ -1376,6 +1376,11 @@ process_options (void)
 {
   if (targetm.chkp_bound_mode () == VOIDmode)
error ("-fcheck-pointer-bounds is not supported for this target");
+
+  if (flag_sanitize & SANITIZE_ADDRESS)
+   error ("-fcheck-pointer-bounds is not supported with Address 
Sanitizer");
+
+  flag_check_pointer_bounds = 0;
 }
 
   /* One region RA really helps to decrease the code size.  */

[PATCH] Copy over section name during cloning (PR ipa/64963)

[Jakub Jelinek]

Hi!

As discussed in the PR, in 4.9 we used to clone DECL_SECTION_NAME
through using copy_node on the FUNCTION_DECL, and only in selected places
(e.g. when creating artificial_thunk.*, or when creating virtual clones
of DECL_ONE_ONLY functions) we used to explicitly clear DECL_SECTION_NAME.
In 5 the section name is stored in cgraph node instead, and thus not
copied by default, so we instead need to copy it over to restore previous
behavior, otherwise we break the Linux kernel and various other packages.

Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
trunk?

2015-02-16  Jakub Jelinek  
James Greenhalgh  

PR ipa/64963
* cgraphclones.c (cgraph_node::create_virtual_clone): Copy
section if not linkonce.  Fix up formatting.
(cgraph_node::create_version_clone_with_body): Copy section.
* trans-mem.c (ipa_tm_create_version): Likewise.

* gcc.dg/ipa/ipa-clone-1.c: New test.

--- gcc/cgraphclones.c.jj   2015-01-09 21:59:44.0 +0100
+++ gcc/cgraphclones.c  2015-02-16 14:02:16.564725881 +0100
@@ -577,7 +577,7 @@ cgraph_node::create_virtual_clone (vecclone.tree_map = tree_map;
+  if (!DECL_ONE_ONLY (old_decl))
+new_node->set_section (this->get_section ());
 
   /* Clones of global symbols or symbols with unique names are unique.  */
   if ((TREE_PUBLIC (old_decl)
@@ -1009,6 +1011,7 @@ cgraph_node::create_version_clone_with_b
   new_version_node->externally_visible = 0;
   new_version_node->local.local = 1;
   new_version_node->lowered = true;
+  new_version_node->set_section (this->get_section ());
   /* Clones of global symbols or symbols with unique names are unique.  */
   if ((TREE_PUBLIC (old_decl)
&& !DECL_EXTERNAL (old_decl)
--- gcc/trans-mem.c.jj  2015-01-14 09:55:19.0 +0100
+++ gcc/trans-mem.c 2015-02-16 12:58:01.399808815 +0100
@@ -4967,6 +4967,7 @@ ipa_tm_create_version (struct cgraph_nod
   new_node->externally_visible = old_node->externally_visible;
   new_node->lowered = true;
   new_node->tm_clone = 1;
+  new_node->set_section (old_node->get_section ());
   get_cg_data (&old_node, true)->clone = new_node;
 
   if (old_node->get_availability () >= AVAIL_INTERPOSABLE)
--- gcc/testsuite/gcc.dg/ipa/ipa-clone-1.c.jj   2015-02-16 14:14:39.041625503 
+0100
+++ gcc/testsuite/gcc.dg/ipa/ipa-clone-1.c  2015-02-16 14:15:31.944760949 
+0100
@@ -0,0 +1,20 @@
+/* PR ipa/64693 */
+/* { dg-do compile } */
+/* { dg-require-named-sections "" } */
+/* { dg-options "-O3 -fipa-cp -fipa-cp-clone -fdump-ipa-cp"  } */
+
+static int __attribute__ ((noinline, section ("test_section")))
+foo (int arg)
+{
+  return 7 * arg;
+}
+
+int
+bar (int arg)
+{
+  return foo (5);
+}
+
+/* { dg-final { scan-assembler "test_section" } } */
+/* { dg-final { scan-ipa-dump "Creating a specialized node of foo" "cp" } } */
+/* { dg-final { cleanup-ipa-dump "cp" } } */

Jakub

[C++ PATCH] Fix constexpr C++11 handling with lambdas (PR c++/65075)

[Jakub Jelinek]

Hi!

If there are lambdas in C++11 constexpr return-stmts, we get implicit
typedefs of the lambda types, but those are artificial and we should ignore
them.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2015-02-16  Paolo Carlini  
Jakub Jelinek  

PR c++/65075
* constexpr.c (check_constexpr_bind_expr_vars): Allow
implicit typedefs for lambda types.

* g++.dg/cpp0x/pr65075.C: New test.

--- gcc/cp/constexpr.c.jj   2015-02-14 00:19:49.0 +0100
+++ gcc/cp/constexpr.c  2015-02-16 15:12:06.260262857 +0100
@@ -416,7 +416,8 @@ check_constexpr_bind_expr_vars (tree t)
 
   for (tree var = BIND_EXPR_VARS (t); var; var = DECL_CHAIN (var))
 if (TREE_CODE (var) == TYPE_DECL
-   && DECL_IMPLICIT_TYPEDEF_P (var))
+   && DECL_IMPLICIT_TYPEDEF_P (var)
+   && !LAMBDA_TYPE_P (TREE_TYPE (var)))
   return false;
   return true;
 }
--- gcc/testsuite/g++.dg/cpp0x/pr65075.C.jj 2015-02-16 15:09:48.405517062 
+0100
+++ gcc/testsuite/g++.dg/cpp0x/pr65075.C2015-02-16 15:09:23.0 
+0100
@@ -0,0 +1,17 @@
+// PR c++/65075
+// { dg-do compile { target c++11 } }
+
+typedef void (*E) ();
+template 
+constexpr E
+bar (bool a)
+{
+  return a ? []() {} : []() {};
+}
+
+void
+foo ()
+{
+  (bar (false)) ();
+  (bar (true)) ();
+}

Jakub

Re: [committed] Change "Q" and "T" constraints to memory constraints

[Richard Henderson]

On 02/14/2015 06:50 AM, John David Anglin wrote:
> Possibly the constant can somehow be forced into the data section where the 
> relocations
> aren't a problem?

Hmm.  It looks like we might already do that.  See default_select_rtx_section.


r~

Re: [PATCH] PR rtl-optimization/32219: optimizer causees wrong code in pic/hidden/weak symbol checking

[H.J. Lu]

On Mon, Feb 16, 2015 at 8:30 AM, Richard Henderson  wrote:
> On 02/16/2015 06:01 AM, H.J. Lu wrote:
>> On Mon, Feb 16, 2015 at 5:25 AM, Uros Bizjak  wrote:
>>> Hello!
>>>
 2015-02-12  H.J. Lu  
 Richard Henderson  

 PR rtl/32219
 * cgraphunit.c (cgraph_node::finalize_function): Set definition
 before notice_global_symbol.
 (varpool_node::finalize_decl): Likewise.
 * varasm.c (default_binds_local_p_2): Rename from
 default_binds_local_p_1, add weak_dominate argument.  Use direct
 returns instead of assigning to local variable.  Unify varpool and
 cgraph paths via symtab_node.  Reject undef weak variables before
 testing visibility.  Reorder tests for simplicity.
 (default_binds_local_p): Use default_binds_local_p_2.
 (default_binds_local_p_1): Likewise.
 (decl_binds_to_current_def_p): Unify varpool and cgraph paths
 via symtab_node.
 (default_elf_asm_output_external): Emit visibility when specified.
>>>
>>> It looks like this patch broke alphaev68-linux-gnu [1]. There are many
>>> failures of the type:
>>>
>>> /tmp/cck7V7MR.o: In function
>>> `__static_initialization_and_destruction_0(int, int)':^M
>>> (.text+0x3ac): relocation truncated to fit: GPRELHIGH against symbol
>>> `std::__cxx11::basic_string,
>>> std::allocator >::~basic_string()@@GLIBCXX_3.4.21' defined in
>>> .text section in
>>> /space/uros/gcc-build/alphaev68-unknown-linux-gnu/./libstdc++-v3/src/.libs/libstdc++.so^M
>>> /space/homedirs/uros/local/bin/ld: /tmp/cck7V7MR.o: gp-relative
>>> relocation against dynamic symbol
>>
>> It could be related to:
>>
>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65064
>>
>> Before this bug fix, all common symbols don't bind locally,
>> which is one of PR 32219 bugs.  After this fix, common
>> symbols bind locally.  It may cause problems on targets with
>> small data sections and common symbols aren't in small
>> data section:
>
> This is a destructor, and so obviously not a common symbol.

Then it could be:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65074


-- 
H.J.

Re: [PATCH] PR rtl-optimization/32219: optimizer causees wrong code in pic/hidden/weak symbol checking

[Richard Henderson]

On 02/16/2015 06:01 AM, H.J. Lu wrote:
> On Mon, Feb 16, 2015 at 5:25 AM, Uros Bizjak  wrote:
>> Hello!
>>
>>> 2015-02-12  H.J. Lu  
>>> Richard Henderson  
>>>
>>> PR rtl/32219
>>> * cgraphunit.c (cgraph_node::finalize_function): Set definition
>>> before notice_global_symbol.
>>> (varpool_node::finalize_decl): Likewise.
>>> * varasm.c (default_binds_local_p_2): Rename from
>>> default_binds_local_p_1, add weak_dominate argument.  Use direct
>>> returns instead of assigning to local variable.  Unify varpool and
>>> cgraph paths via symtab_node.  Reject undef weak variables before
>>> testing visibility.  Reorder tests for simplicity.
>>> (default_binds_local_p): Use default_binds_local_p_2.
>>> (default_binds_local_p_1): Likewise.
>>> (decl_binds_to_current_def_p): Unify varpool and cgraph paths
>>> via symtab_node.
>>> (default_elf_asm_output_external): Emit visibility when specified.
>>
>> It looks like this patch broke alphaev68-linux-gnu [1]. There are many
>> failures of the type:
>>
>> /tmp/cck7V7MR.o: In function
>> `__static_initialization_and_destruction_0(int, int)':^M
>> (.text+0x3ac): relocation truncated to fit: GPRELHIGH against symbol
>> `std::__cxx11::basic_string,
>> std::allocator >::~basic_string()@@GLIBCXX_3.4.21' defined in
>> .text section in
>> /space/uros/gcc-build/alphaev68-unknown-linux-gnu/./libstdc++-v3/src/.libs/libstdc++.so^M
>> /space/homedirs/uros/local/bin/ld: /tmp/cck7V7MR.o: gp-relative
>> relocation against dynamic symbol
> 
> It could be related to:
> 
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65064
> 
> Before this bug fix, all common symbols don't bind locally,
> which is one of PR 32219 bugs.  After this fix, common
> symbols bind locally.  It may cause problems on targets with
> small data sections and common symbols aren't in small
> data section:

This is a destructor, and so obviously not a common symbol.

I'll have a look.


r~

[Patch docs obvious] install.texi: Put aarch64 back in alphabetical order, add link

[James Greenhalgh]

Hi,

Looking at https://gcc.gnu.org/install/specific.html , aarch64*-*-* is
in an odd place and isn't linked to from the top of the file.

This patch fixes that by reordering the entries and adding a link
from the menu at the top of the page.

I've built the documentation with no new issues, and had a look in
firefox to ensure the changes were sane.

I've applied this patch under the obvious rule as revision r220738.

Thanks,
James

---
2015-02-16  James Greenhalgh  

* doc/install.texi (Specific): Reorder targets list to put
aarch64 in alphabetical order.  Add a link to aarch64*-*-*
from the top menu.diff --git a/gcc/doc/install.texi b/gcc/doc/install.texi
index c9e3bf1..47380a3 100644
--- a/gcc/doc/install.texi
+++ b/gcc/doc/install.texi
@@ -3238,6 +3238,8 @@ information have to.
 @ifhtml
 @itemize
 @item
+@uref{#aarch64-x-x,,aarch64*-*-*}
+@item
 @uref{#alpha-x-x,,alpha*-*-*}
 @item
 @uref{#alpha-dec-osf51,,alpha*-dec-osf5.1}
@@ -3386,6 +3388,25 @@ information have to.
 
 
 @end html
+@anchor{aarch64-x-x}
+@heading aarch64*-*-*
+Binutils pre 2.24 does not have support for selecting @option{-mabi} and
+does not support ILP32.  If it is used to build GCC 4.9 or later, GCC will
+not support option @option{-mabi=ilp32}.
+
+To enable a workaround for the Cortex-A53 erratum number 835769 by default
+(for all CPUs regardless of -mcpu option given) at configure time use the
+@option{--enable-fix-cortex-a53-835769} option.  This will enable the fix by
+default and can be explicitly disabled during during compilation by passing the
+@option{-mno-fix-cortex-a53-835769} option.  Conversely,
+@option{--disable-fix-cortex-a53-835769} will disable the workaround by
+default.  The workaround is disabled by default if neither of
+@option{--enable-fix-cortex-a53-835769} or
+@option{--disable-fix-cortex-a53-835769} is given at configure time.
+
+@html
+
+@end html
 @anchor{alpha-x-x}
 @heading alpha*-*-*
 This section contains general configuration information for all
@@ -3897,25 +3918,6 @@ removed and the system libunwind library will always be used.
 
 @html
 
-@end html
-@anchor{aarch64-x-x}
-@heading aarch64*-*-*
-Binutils pre 2.24 does not have support for selecting @option{-mabi} and
-does not support ILP32.  If it is used to build GCC 4.9 or later, GCC will
-not support option @option{-mabi=ilp32}.
-
-To enable a workaround for the Cortex-A53 erratum number 835769 by default
-(for all CPUs regardless of -mcpu option given) at configure time use the
-@option{--enable-fix-cortex-a53-835769} option.  This will enable the fix by
-default and can be explicitly disabled during during compilation by passing the
-@option{-mno-fix-cortex-a53-835769} option.  Conversely,
-@option{--disable-fix-cortex-a53-835769} will disable the workaround by
-default.  The workaround is disabled by default if neither of
-@option{--enable-fix-cortex-a53-835769} or
-@option{--disable-fix-cortex-a53-835769} is given at configure time.
-
-@html
-
 
 @end html
 @anchor{x-ibm-aix}

Re: [patch] Fix invalid attributes in libstdc++

[Jonathan Wakely]

On 03/02/15 10:37 +, Iain Sandoe wrote:

Hi Jonathan,

On 1 Feb 2015, at 15:10, Jonathan Wakely wrote:


On 01/02/15 15:08 +, Jonathan Wakely wrote:

I failed to CC gcc-patches on this patch ...

On 29/01/15 13:02 +, Jonathan Wakely wrote:

Jakub pointed out that we have some attributes that don't use the
reserved namespace, e.g. __attribute__ ((always_inline)).

This is a 4.9/5 regression and the fix was pre-approved by Jakub so
I've committed it to trunk.

When we're back in stage1 I'll fix the TODO comments in the new tests
(see PR64857) and will also rename testsuite/17_intro/headers/c++200x
to .../c++2011.



The new test fails on darwin (PR64883) and --enable-threads=single
targets (PR64885).

This is a workaround for 64883. Tested x86_64-linux, committed to
trunk.



the following additional tweaks provide further work-arounds.
... checked on darwin12 and darwin14.


OK for trunk - thanks.


I have a fixincludes patch for next stage #1.
Iain




diff --git a/libstdc++-v3/testsuite/17_intro/headers/c++1998/all_attributes.cc 
b/libstdc++-v3/testsuite/17_intro/headers/c++1998/all_attributes.cc
index 76a935e..6fc362a 100644
--- a/libstdc++-v3/testsuite/17_intro/headers/c++1998/all_attributes.cc
+++ b/libstdc++-v3/testsuite/17_intro/headers/c++1998/all_attributes.cc
@@ -26,11 +26,11 @@
// darwin headers use these, see PR 64883
# define deprecated 1
# define noreturn 1
+# define visibility 1
#endif
#define packed 1
#define pure 1
#define unused 1
-#define visibility 1

#include  // TODO: this is missing from 
#include 
diff --git a/libstdc++-v3/testsuite/17_intro/headers/c++200x/all_attributes.cc 
b/libstdc++-v3/testsuite/17_intro/headers/c++200x/all_attributes.cc
index c7ec27a..0726e3f 100644
--- a/libstdc++-v3/testsuite/17_intro/headers/c++200x/all_attributes.cc
+++ b/libstdc++-v3/testsuite/17_intro/headers/c++200x/all_attributes.cc
@@ -22,11 +22,14 @@
// Don't test 'const' and 'noreturn' because they are reserved anyway.
#define abi_tag 1
#define always_inline 1
-#define deprecated 1
+#ifndef __APPLE__
+// darwin headers use these, see PR 64883
+# define visibility 1
+# define deprecated 1
+#endif
#define packed 1
#define pure 1
#define unused 1
-#define visibility 1

#include  // TODO: this is missing from 
#include// TODO: this is missing from 
diff --git a/libstdc++-v3/testsuite/17_intro/headers/c++2014/all_attributes.cc 
b/libstdc++-v3/testsuite/17_intro/headers/c++2014/all_attributes.cc
index 533a6f1..06bcb8e 100644
--- a/libstdc++-v3/testsuite/17_intro/headers/c++2014/all_attributes.cc
+++ b/libstdc++-v3/testsuite/17_intro/headers/c++2014/all_attributes.cc
@@ -22,11 +22,14 @@
// Don't test 'const' and 'noreturn' because they are reserved anyway.
#define abi_tag 1
#define always_inline 1
-#define deprecated 1
+#ifndef __APPLE__
+// darwin headers use these, see PR 64883
+# define deprecated 1
+# define visibility 1
+#endif
#define packed 1
#define pure 1
#define unused 1
-#define visibility 1

#include  // TODO: this is missing from 
#include   // TODO: this is missing from 

Re: [PATCH, CHKP, PR target/65044] Restrict pointer bounds checker with Sanitizer

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 06:56:45PM +0300, Ilya Enkovich wrote:
> On 16 Feb 16:31, Jakub Jelinek wrote:
> > On Mon, Feb 16, 2015 at 06:20:59PM +0300, Ilya Enkovich wrote:
> > > This patch restricts usage of Pointer Bounds Checker with Sanitizer.  OK 
> > > for trunk?
> > 
> > There are many sanitizers, and for most of them I don't see why they would
> > be in any conflict with -mmpx, it is just -fsanitize=address and
> > -fsanitize=kernel-address.
> > So perhaps test instead if (flag_sanitize & SANITIZE_ADDRESS) != 0, and
> > better clear the flag_pointer_bounds after issuing the error, error () is
> > not a fatal function, so you need something sensible for error-recovery.
> > 
> > Jakub
> 
> I don't know all sanitizers in details.  Code generated by some of them may 
> be incorrect from checker point of view.  Thus I just wanted to disable 
> unexplored and untested combinations.

Shouldn't be that hard to write a testcase and test it.

Most of the sanitizers just add code like
if (some_condition)
  __ubsan_handle_... ();
where from the POV of the program the __ubsan_* function reports or might
report some problem, and optionally abort the program.
That some_condition can be a check of the pointer value, shift count,
divisor check, etc.

Jakub

Re: [patch] Fix invalid attributes in libstdc++

[Jack Howarth]

On Tue, Feb 3, 2015 at 5:37 AM, Iain Sandoe  wrote:
> Hi Jonathan,
>
> On 1 Feb 2015, at 15:10, Jonathan Wakely wrote:
>
>> On 01/02/15 15:08 +, Jonathan Wakely wrote:
>>> I failed to CC gcc-patches on this patch ...
>>>
>>> On 29/01/15 13:02 +, Jonathan Wakely wrote:
 Jakub pointed out that we have some attributes that don't use the
 reserved namespace, e.g. __attribute__ ((always_inline)).

 This is a 4.9/5 regression and the fix was pre-approved by Jakub so
 I've committed it to trunk.

 When we're back in stage1 I'll fix the TODO comments in the new tests
 (see PR64857) and will also rename testsuite/17_intro/headers/c++200x
 to .../c++2011.

>>
>> The new test fails on darwin (PR64883) and --enable-threads=single
>> targets (PR64885).
>>
>> This is a workaround for 64883. Tested x86_64-linux, committed to
>> trunk.
>> 
>
> the following additional tweaks provide further work-arounds.
> ... checked on darwin12 and darwin14.
> I have a fixincludes patch for next stage #1.
> Iain

Ping on https://gcc.gnu.org/ml/gcc-patches/2015-02/msg00122.html

>
>
>
>
>

Re: [PATCH, CHKP, PR target/65044] Restrict pointer bounds checker with Sanitizer

[Ilya Enkovich]

On 16 Feb 16:31, Jakub Jelinek wrote:
> On Mon, Feb 16, 2015 at 06:20:59PM +0300, Ilya Enkovich wrote:
> > This patch restricts usage of Pointer Bounds Checker with Sanitizer.  OK 
> > for trunk?
> 
> There are many sanitizers, and for most of them I don't see why they would
> be in any conflict with -mmpx, it is just -fsanitize=address and
> -fsanitize=kernel-address.
> So perhaps test instead if (flag_sanitize & SANITIZE_ADDRESS) != 0, and
> better clear the flag_pointer_bounds after issuing the error, error () is
> not a fatal function, so you need something sensible for error-recovery.
> 
>   Jakub

I don't know all sanitizers in details.  Code generated by some of them may be 
incorrect from checker point of view.  Thus I just wanted to disable unexplored 
and untested combinations.

Ilya

Contents of PO file 'cpplib-5.1-b20150208.sv.po'

[Translation Project Robot]

cpplib-5.1-b20150208.sv.po.gz
Description: Binary data
The Translation Project robot, in the
name of your translation coordinator.

New Swedish PO file for 'cpplib' (version 5.1-b20150208)

[Translation Project Robot]

Hello, gentle maintainer.

This is a message from the Translation Project robot.

A revised PO file for textual domain 'cpplib' has been submitted
by the Swedish team of translators.  The file is available at:

http://translationproject.org/latest/cpplib/sv.po

(This file, 'cpplib-5.1-b20150208.sv.po', has just now been sent to you in
a separate email.)

All other PO files for your package are available in:

http://translationproject.org/latest/cpplib/

Please consider including all of these in your next release, whether
official or a pretest.

Whenever you have a new distribution with a new version number ready,
containing a newer POT file, please send the URL of that distribution
tarball to the address below.  The tarball may be just a pretest or a
snapshot, it does not even have to compile.  It is just used by the
translators when they need some extra translation context.

The following HTML page has been updated:

http://translationproject.org/domain/cpplib.html

If any question arises, please contact the translation coordinator.

Thank you for all your work,

The Translation Project robot, in the
name of your translation coordinator.

Re: [PATCH, CHKP, PR target/65044] Restrict pointer bounds checker with Sanitizer

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 06:20:59PM +0300, Ilya Enkovich wrote:
> This patch restricts usage of Pointer Bounds Checker with Sanitizer.  OK for 
> trunk?

There are many sanitizers, and for most of them I don't see why they would
be in any conflict with -mmpx, it is just -fsanitize=address and
-fsanitize=kernel-address.
So perhaps test instead if (flag_sanitize & SANITIZE_ADDRESS) != 0, and
better clear the flag_pointer_bounds after issuing the error, error () is
not a fatal function, so you need something sensible for error-recovery.

Jakub

Re: [PATCH][3/n] Fix PR65015

[H.J. Lu]

On Mon, Feb 16, 2015 at 6:25 AM, Richard Biener  wrote:
>
> This fixes another leakage of random LTO temporary filenames into
> executables, this time via .symtab FILE entries.  Removing it
> doesn't work (GNU ld adds it back) and is said to be incorrect.

FWIW, ld.bfd will be fixed in 2.26.  But it is still a good idea to
support older links.

> So the following patch, similar to the dwarf CU DW_AT_name uses
> .
>
> Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.
>
> Richard.
>
> 2015-02-16  Richard Biener  
>
> PR lto/65015
> * varasm.c (default_file_start): For LTO produced units
> emit  as file directive.
>
> Index: gcc/varasm.c
> ===
> --- gcc/varasm.c(revision 220677)
> +++ gcc/varasm.c(working copy)
> @@ -7043,7 +7047,13 @@ default_file_start (void)
>  fputs (ASM_APP_OFF, asm_out_file);
>
>if (targetm.asm_file_start_file_directive)
> -output_file_directive (asm_out_file, main_input_filename);
> +{
> +  /* LTO produced units have no meaningful main_input_filename.  */
> +  if (in_lto_p)
> +   output_file_directive (asm_out_file, "");
> +  else
> +   output_file_directive (asm_out_file, main_input_filename);
> +}
>  }
>
>  /* This is a generic routine suitable for use as TARGET_ASM_FILE_END



-- 
H.J.

[PATCH, CHKP, PR target/65044] Restrict pointer bounds checker with Sanitizer

[Ilya Enkovich]

Hi,

This patch restricts usage of Pointer Bounds Checker with Sanitizer.  OK for 
trunk?

Thanks,
Ilya
--
gcc/

2015-02-16  Ilya Enkovich  

PR target/65044
* toplev.c (process_options): Restrict Pointer Bounds Checker
usage with sanitizers.

gcc/testsuite/

2015-02-16  Ilya Enkovich  

PR target/65044
* gcc.target/i386/pr65044.c: New.


diff --git a/gcc/testsuite/gcc.target/i386/pr65044.c 
b/gcc/testsuite/gcc.target/i386/pr65044.c
new file mode 100644
index 000..79ecb04
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/pr65044.c
@@ -0,0 +1,12 @@
+/* { dg-error "-fcheck-pointer-bounds is not supported with sanitizers" } */
+/* { dg-do compile } */
+/* { dg-require-effective-target mpx } */
+/* { dg-options "-fcheck-pointer-bounds -mmpx -fsanitize=address" } */
+
+extern int x[];
+
+void
+foo ()
+{
+  x[0] = 0;
+}
diff --git a/gcc/toplev.c b/gcc/toplev.c
index 99cf180..bf987c8 100644
--- a/gcc/toplev.c
+++ b/gcc/toplev.c
@@ -1376,6 +1376,9 @@ process_options (void)
 {
   if (targetm.chkp_bound_mode () == VOIDmode)
error ("-fcheck-pointer-bounds is not supported for this target");
+
+  if (flag_sanitize)
+   error ("-fcheck-pointer-bounds is not supported with sanitizers");
 }
 
   /* One region RA really helps to decrease the code size.  */

Re: [PATCH][OpenMP] Forbid usage of non-target functions in target regions

[Ilya Verbin]

On Mon, Feb 02, 2015 at 13:10:19 +0100, Jakub Jelinek wrote:
> [...]  Generally, the solution if something goes
> wrong during the offloading compilation should be just to give up on the
> offloading to the particular offloading target (i.e. fill in the sections
> libgomp reads in a way that will result in host fallback).

Do you mean something like this?
Bootstrapped/regtested on x86_64-linux and i686-linux.
Is this patch OK for stage4 or stage1?


gcc/
* collect-utils.c (do_wait, fork_execute): New argument.
* collect-utils.h: Likewise.
* collect2.c: Pass new argument to do_wait and fork_execute.
* config/i386/intelmic-mkoffload.c: Likewise.
(compile_for_target): Don't call fatal_error if compilation failed.
(generate_target_descr_file, generate_target_offloadend_file)
(prepare_target_image): Pass out_filename to compile_for_target.
* config/nvptx/mkoffload.c: Pass new argument fork_execute.
* lto-wrapper.c (num_offload_targets): New static global variable.
(compile_offload_image): Return NULL if an image was not created.
(compile_images_for_offload_targets): Call warning instead of
fatal_error if an image was not created.
(run_gcc): Do not return empty images to linker.
Pass new argument to do_wait and fork_execute.
libgomp/
* target.c (gomp_target_fallback): New static function.
(GOMP_target): Move host fallback to the new gomp_target_fallback.
Run gomp_target_fallback if tgt_fn is not present in the splay tree.


diff --git a/gcc/collect-utils.c b/gcc/collect-utils.c
index 6bbe9eb..2dce3c9 100644
--- a/gcc/collect-utils.c
+++ b/gcc/collect-utils.c
@@ -85,10 +85,10 @@ collect_wait (const char *prog, struct pex_obj *pex)
 }
 
 void
-do_wait (const char *prog, struct pex_obj *pex)
+do_wait (const char *prog, struct pex_obj *pex, bool non_fatal)
 {
   int ret = collect_wait (prog, pex);
-  if (ret != 0)
+  if (!non_fatal && ret != 0)
 fatal_error (input_location, "%s returned %d exit status", prog, ret);
 
   if (response_file && !save_temps)
@@ -201,13 +201,13 @@ collect_execute (const char *prog, char **argv, const 
char *outname,
 }
 
 void
-fork_execute (const char *prog, char **argv, bool use_atfile)
+fork_execute (const char *prog, char **argv, bool use_atfile, bool non_fatal)
 {
   struct pex_obj *pex;
 
   pex = collect_execute (prog, argv, NULL, NULL,
 PEX_LAST | PEX_SEARCH, use_atfile);
-  do_wait (prog, pex);
+  do_wait (prog, pex, non_fatal);
 }
 
 /* Delete tempfiles.  */
diff --git a/gcc/collect-utils.h b/gcc/collect-utils.h
index 2b3ed44..7b3b3da 100644
--- a/gcc/collect-utils.h
+++ b/gcc/collect-utils.h
@@ -29,8 +29,8 @@ extern struct pex_obj *collect_execute (const char *, char **,
const char *, const char *,
int, bool);
 extern int collect_wait (const char *, struct pex_obj *);
-extern void do_wait (const char *, struct pex_obj *);
-extern void fork_execute (const char *, char **, bool);
+extern void do_wait (const char *, struct pex_obj *, bool);
+extern void fork_execute (const char *, char **, bool, bool);
 extern void utils_cleanup (bool);
 
 
diff --git a/gcc/collect2.c b/gcc/collect2.c
index b53e151..f8be7da 100644
--- a/gcc/collect2.c
+++ b/gcc/collect2.c
@@ -758,7 +758,7 @@ maybe_run_lto_and_relink (char **lto_ld_argv, char 
**object_lst,
 
obstack_free (&temporary_obstack, temporary_firstobj);
   }
-  do_wait (prog, pex);
+  do_wait (prog, pex, false);
   pex = NULL;
 
   /* Compute memory needed for new LD arguments.  At most number of 
original arguemtns
@@ -803,7 +803,8 @@ maybe_run_lto_and_relink (char **lto_ld_argv, char 
**object_lst,
 
   /* Run the linker again, this time replacing the object files
  optimized by the LTO with the temporary file generated by the LTO.  */
-  fork_execute ("ld", out_lto_ld_argv, HAVE_GNU_LD && at_file_supplied);
+  fork_execute ("ld", out_lto_ld_argv, HAVE_GNU_LD && at_file_supplied,
+   false);
   post_ld_pass (true);
   free (lto_ld_argv);
 
@@ -813,7 +814,7 @@ maybe_run_lto_and_relink (char **lto_ld_argv, char 
**object_lst,
 {
   /* Our caller is relying on us to do the link
  even though there is no LTO back end work to be done.  */
-  fork_execute ("ld", lto_ld_argv, HAVE_GNU_LD && at_file_supplied);
+  fork_execute ("ld", lto_ld_argv, HAVE_GNU_LD && at_file_supplied, false);
   post_ld_pass (false);
 }
   else
@@ -1706,7 +1707,7 @@ main (int argc, char **argv)
  strip_argv[0] = strip_file_name;
  strip_argv[1] = output_file;
  strip_argv[2] = (char *) 0;
- fork_execute ("strip", real_strip_argv, false);
+ fork_execute ("strip", real_strip_argv, false, false);
}
 
 #ifdef COLLECT_EXPORT_LIST
@@ -1792,7 +1793,7 @@ main (int argc, char **argv)
   /

[PATCH] Fix PR65077

[Richard Biener]

The following removes an optimization not considering FP values
to carry pointers from PTA.  Instead to fix the underlying problem
in PR37021 this patch adds handling of the rest of handled_components_p.

Bootstrap and regtest in progress on x86_64-unknown-linux-gnu.

Richard.

2015-02-16  Richard Biener  

PR tree-optimization/65077
* tree-ssa-structalias.c (get_constraint_for_1): Handle
IMAGPART_EXPR, REALPART_EXPR and BIT_FIELD_REF.
(find_func_aliases): Allow float values to carry pointers again.

* gcc.dg/torture/pr65077.c: New testcase.

Index: gcc/tree-ssa-structalias.c
===
--- gcc/tree-ssa-structalias.c  (revision 220731)
+++ gcc/tree-ssa-structalias.c  (working copy)
@@ -3492,6 +3492,9 @@ get_constraint_for_1 (tree t, vec
  case ARRAY_REF:
  case ARRAY_RANGE_REF:
  case COMPONENT_REF:
+ case IMAGPART_EXPR:
+ case REALPART_EXPR:
+ case BIT_FIELD_REF:
get_constraint_for_component_ref (t, results, address_p, lhs_p);
return;
  case VIEW_CONVERT_EXPR:
@@ -4712,11 +4811,7 @@ find_func_aliases (struct function *fn,
 
  get_constraint_for (lhsop, &lhsc);
 
- if (FLOAT_TYPE_P (TREE_TYPE (lhsop)))
-   /* If the operation produces a floating point result then
-  assume the value is not produced to transfer a pointer.  */
-   ;
- else if (code == POINTER_PLUS_EXPR)
+ if (code == POINTER_PLUS_EXPR)
get_constraint_for_ptr_offset (gimple_assign_rhs1 (t),
   gimple_assign_rhs2 (t), &rhsc);
  else if (code == BIT_AND_EXPR
Index: gcc/testsuite/gcc.dg/torture/pr65077.c
===
--- gcc/testsuite/gcc.dg/torture/pr65077.c  (revision 0)
+++ gcc/testsuite/gcc.dg/torture/pr65077.c  (working copy)
@@ -0,0 +1,70 @@
+/* { dg-do run } */
+
+extern void abort (void);
+extern void *memcpy(void *, const void *, __SIZE_TYPE__);
+
+typedef struct {
+void *v1;
+void *v2;
+void *v3;
+union {
+   void *f1;
+   void *f2;
+} u;
+} S;
+
+
+S *getS();
+void verify_p(void *p);
+double *getP(void *p);
+
+void memcpy_bug()
+{
+  S *s;
+  double *p = getP(0);
+
+  if (p) {
+  int intSptr[sizeof(S*)/sizeof(int)];
+  unsigned i = 0;
+  for (i = 0; i < sizeof(intSptr)/sizeof(*intSptr); ++i) {
+ intSptr[i] = (int) p[i];
+  }
+  memcpy(&s, intSptr, sizeof(intSptr));
+  (s)->u.f1 = p;
+  verify_p((s)->u.f1);  
+  } else {
+  s = getS();
+  }
+  verify_p(s->u.f1);
+}
+
+double P[4];
+
+double *getP(void *p) {
+union u {
+   void *p;
+   int i[2];
+} u;
+u.p = P;
+P[0] = u.i[0];
+P[1] = u.i[1];
+return P;
+}
+
+S *getS()
+{
+  return 0;
+}
+
+void verify_p(void *p)
+{
+  if (p != P)
+abort ();
+}
+
+int main(int argc, char *argv[])
+{
+memcpy_bug();
+return 0;
+}
+

[PATCH][3/n] Fix PR65015

[Richard Biener]

This fixes another leakage of random LTO temporary filenames into
executables, this time via .symtab FILE entries.  Removing it
doesn't work (GNU ld adds it back) and is said to be incorrect.
So the following patch, similar to the dwarf CU DW_AT_name uses
.

Bootstrapped on x86_64-unknown-linux-gnu, testing in progress.

Richard.

2015-02-16  Richard Biener  

PR lto/65015
* varasm.c (default_file_start): For LTO produced units
emit  as file directive.

Index: gcc/varasm.c
===
--- gcc/varasm.c(revision 220677)
+++ gcc/varasm.c(working copy)
@@ -7043,7 +7047,13 @@ default_file_start (void)
 fputs (ASM_APP_OFF, asm_out_file);
 
   if (targetm.asm_file_start_file_directive)
-output_file_directive (asm_out_file, main_input_filename);
+{
+  /* LTO produced units have no meaningful main_input_filename.  */
+  if (in_lto_p)
+   output_file_directive (asm_out_file, "");
+  else
+   output_file_directive (asm_out_file, main_input_filename);
+}
 }
 
 /* This is a generic routine suitable for use as TARGET_ASM_FILE_END

[PATCH] Fix PRs 65063 and 63593

[Richard Biener]

Predictive commoning happens to re-use SSA names it released while
there are still uses of them (oops), confusing the hell out of
other code (expected).  Fixed thus.

Bootstrap and regtest running on x86_64-unknown-linux-gnu.

Richard.

2015-02-16  Richard Biener  

PR tree-optimization/63593
PR tree-optimization/65063
* tree-predcom.c (execute_pred_commoning_chain): Delay removing
stmts and releasing SSA names until...
(execute_pred_commoning): ... after processing all chains.

* gcc.dg/pr63593.c: New testcase.
* gcc.dg/pr65063.c: Likewise.

Index: gcc/tree-predcom.c
===
--- gcc/tree-predcom.c  (revision 220731)
+++ gcc/tree-predcom.c  (working copy)
@@ -1745,9 +1745,8 @@ execute_pred_commoning_chain (struct loo
   if (chain->combined)
 {
   /* For combined chains, just remove the statements that are used to
-compute the values of the expression (except for the root one).  */
-  for (i = 1; chain->refs.iterate (i, &a); i++)
-   remove_stmt (a->stmt);
+compute the values of the expression (except for the root one).
+We delay this until after all chains are processed.  */
 }
   else
 {
@@ -1811,6 +1810,21 @@ execute_pred_commoning (struct loop *loo
execute_pred_commoning_chain (loop, chain, tmp_vars);
 }
 
+  FOR_EACH_VEC_ELT (chains, i, chain)
+{
+  if (chain->type == CT_INVARIANT)
+   ;
+  else if (chain->combined)
+   {
+ /* For combined chains, just remove the statements that are used to
+compute the values of the expression (except for the root one).  */
+ dref a;
+ unsigned j;
+ for (j = 1; chain->refs.iterate (j, &a); j++)
+   remove_stmt (a->stmt);
+   }
+}
+
   update_ssa (TODO_update_ssa_only_virtuals);
 }
 
Index: gcc/testsuite/gcc.dg/pr63593.c
===
--- gcc/testsuite/gcc.dg/pr63593.c  (revision 0)
+++ gcc/testsuite/gcc.dg/pr63593.c  (working copy)
@@ -0,0 +1,20 @@
+/* { dg-do compile } */
+/* { dg-options "-O3 -fno-tree-vectorize" } */
+
+int in[2 * 4][4];
+int out[4];
+
+void
+foo (void)
+{
+  int sum;
+  int i, j, k;
+  for (k = 0; k < 4; k++)
+{
+  sum = 1;
+  for (j = 0; j < 4; j++)
+   for (i = 0; i < 4; i++)
+ sum *= in[i + k][j];
+  out[k] = sum;
+}
+}
Index: gcc/testsuite/gcc.dg/pr65063.c
===
--- gcc/testsuite/gcc.dg/pr65063.c  (revision 0)
+++ gcc/testsuite/gcc.dg/pr65063.c  (working copy)
@@ -0,0 +1,33 @@
+/* { dg-do run } */
+/* { dg-options "-O3 -fno-tree-loop-ivcanon -fno-tree-vectorize" } */
+
+static int in[8][4];
+static int out[4];
+static const int check_result[] = {0, 16, 256, 4096};
+
+static inline void foo ()
+{
+  int sum;
+  int i, j, k;
+  for (k = 0; k < 4; k++)
+{
+  sum = 1;
+  for (j = 0; j < 4; j++)
+   for (i = 0; i < 4; i++)
+ sum *= in[i + k][j];
+  out[k] = sum;
+}
+}
+
+int main ()
+{
+  int i, j, k;
+  for (i = 0; i < 8; i++)
+for (j = 0; j < 4; j++)
+  in[i][j] = (i + 2) / 3;
+  foo ();
+  for (k = 0; k < 4; k++)
+if (out[k] != check_result[k])
+  __builtin_abort ();
+  return 0;
+}

Re: [PATCH] PR rtl-optimization/32219: optimizer causees wrong code in pic/hidden/weak symbol checking

[H.J. Lu]

On Mon, Feb 16, 2015 at 5:25 AM, Uros Bizjak  wrote:
> Hello!
>
>> 2015-02-12  H.J. Lu  
>> Richard Henderson  
>>
>> PR rtl/32219
>> * cgraphunit.c (cgraph_node::finalize_function): Set definition
>> before notice_global_symbol.
>> (varpool_node::finalize_decl): Likewise.
>> * varasm.c (default_binds_local_p_2): Rename from
>> default_binds_local_p_1, add weak_dominate argument.  Use direct
>> returns instead of assigning to local variable.  Unify varpool and
>> cgraph paths via symtab_node.  Reject undef weak variables before
>> testing visibility.  Reorder tests for simplicity.
>> (default_binds_local_p): Use default_binds_local_p_2.
>> (default_binds_local_p_1): Likewise.
>> (decl_binds_to_current_def_p): Unify varpool and cgraph paths
>> via symtab_node.
>> (default_elf_asm_output_external): Emit visibility when specified.
>
> It looks like this patch broke alphaev68-linux-gnu [1]. There are many
> failures of the type:
>
> /tmp/cck7V7MR.o: In function
> `__static_initialization_and_destruction_0(int, int)':^M
> (.text+0x3ac): relocation truncated to fit: GPRELHIGH against symbol
> `std::__cxx11::basic_string,
> std::allocator >::~basic_string()@@GLIBCXX_3.4.21' defined in
> .text section in
> /space/uros/gcc-build/alphaev68-unknown-linux-gnu/./libstdc++-v3/src/.libs/libstdc++.so^M
> /space/homedirs/uros/local/bin/ld: /tmp/cck7V7MR.o: gp-relative
> relocation against dynamic symbol

It could be related to:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65064

Before this bug fix, all common symbols don't bind locally,
which is one of PR 32219 bugs.  After this fix, common
symbols bind locally.  It may cause problems on targets with
small data sections and common symbols aren't in small
data section:

1. Since common symbols bind locally, backend may assume
they are in small data section and lead to link-time failure.
2. Backend assume common symbols are never in small
data section.  But a definition in small data section may
override a common symbol, which still binds lcoally, and
turn a reference to common symbol to reference to small
data section.  This also may lead to link-time failure.

Those targets can't assume common symbols are in
small data section since it may change at link-time.

The most conservative solution is to make common
symbol doesn't bind locally for targets which defines
TARGET_IN_SMALL_DATA_P.

-- 
H.J.

Re: [4.8 branch] PATCH: PR middle-end/53623: [4.7/4.8 Regression] sign extension is effectively split into two x86-64 instructions

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 05:44:45AM -0800, H.J. Lu wrote:
> Should it be a concern for 4.8 backport?  Should we also backport r215205:

Probably.  But, please wait for Jeff Law's approval of all of this before
committing.

> commit b71346c449d2b4a63985a39c4c092ecdfb37b5a0
> Author: jiwang 
> Date:   Fri Sep 12 09:29:16 2014 +
> 
> [Ree] Ensure inserted copy don't change the number of hard registers
> 
> 2014-09-12  Wilco Dijkstra  
> 
>   gcc/
> * ree.c (combine_reaching_defs): Ensure inserted copy don't change the
> number of hard registers.
> 
> 
> git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@215205
> 138bc75d-0d04-0410-961f-82ee72b054a4
> 
> to 4.8 and 4.9 branches?

Jakub

Re: [4.8 branch] PATCH: PR middle-end/53623: [4.7/4.8 Regression] sign extension is effectively split into two x86-64 instructions

[H.J. Lu]

On Mon, Feb 16, 2015 at 5:24 AM, H.J. Lu  wrote:
> On Mon, Feb 16, 2015 at 5:18 AM, Jakub Jelinek  wrote:
>> On Mon, Feb 16, 2015 at 05:15:02AM -0800, H.J. Lu wrote:
>>> On Mon, Feb 16, 2015 at 4:30 AM, H.J. Lu  wrote:
>>> > On Mon, Feb 16, 2015 at 1:35 AM, Jakub Jelinek  wrote:
>>> >> On Sun, Feb 15, 2015 at 12:53:39PM -0800, H.J. Lu wrote:
>>> >>> This is a backport of the patch for PR middle-end/53623 plus all bug
>>> >>> fixes caused by it.  Tested on Linux/x86-32, Linux/x86-64 and x32.  OK
>>> >>> for 4.8 branch?
>>> >>
>>> >> What about PR64286 and PR63659, are you sure those aren't related?
>>> >> I mean, they are on the 4.9 branch and I don't see why they couldn't 
>>> >> affect
>>> >> the 4.8 backport.
>>> >>
>>> >> Jakub
>>> >
>>> > Fix for PR 63659 has been backported to 4.8 branch.  I will check if
>>> > fix for PR 64286 is needed.
>>> >
>>> > --
>>> > H.J.
>>>
>>> The fix for PR 64286 is an updated fix for PR 59754 which is caused by
>>> the fix for PR 53623.  But the testcase in the fix for PR 64286 doesn't
>>> fail on 4.8 branch + my backport of the fix for PR 53623 on Haswell.
>>> I suggest
>>>
>>> 1. We go without my current backport and backport the fix for PR 64286
>>> in a separate patch.  Or
>>> 2. We go without my backport minus the backport of the PR 59754
>>> fix and backport the fixes for PR 59754 plus PR 64286 in a separate patch
>>
>> I think keeping the branch broken is bad, even if we don't have a testcase
>> that really fails, pressumably the issue is just latent.
>> So I'd strongly prefer
>> 3. Add the PR64286 fix to the patch being tested and commit only when it as
>> whole is tested, as one commit.
>>

4.9 branch backport of  the PR64286 fix caused a regression on ARM64:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64286#c11

Should it be a concern for 4.8 backport?  Should we also backport r215205:

commit b71346c449d2b4a63985a39c4c092ecdfb37b5a0
Author: jiwang 
Date:   Fri Sep 12 09:29:16 2014 +

[Ree] Ensure inserted copy don't change the number of hard registers

2014-09-12  Wilco Dijkstra  

  gcc/
* ree.c (combine_reaching_defs): Ensure inserted copy don't change the
number of hard registers.


git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@215205
138bc75d-0d04-0410-961f-82ee72b054a4

to 4.8 and 4.9 branches?

> I will do that and restart the testing.
>
> BTW, PR 53623 was a missed optimization bug originally.  Now
> it turns out that it fixed a wrong code bug.  We are trying to extract
> a run-time testcase from PR 64941 for trunk and branches.
>
> Thanks.
>
> --
> H.J.



-- 
H.J.

Re: [PATCH] PR rtl-optimization/32219: optimizer causees wrong code in pic/hidden/weak symbol checking

[Uros Bizjak]

Hello!

> 2015-02-12  H.J. Lu  
> Richard Henderson  
>
> PR rtl/32219
> * cgraphunit.c (cgraph_node::finalize_function): Set definition
> before notice_global_symbol.
> (varpool_node::finalize_decl): Likewise.
> * varasm.c (default_binds_local_p_2): Rename from
> default_binds_local_p_1, add weak_dominate argument.  Use direct
> returns instead of assigning to local variable.  Unify varpool and
> cgraph paths via symtab_node.  Reject undef weak variables before
> testing visibility.  Reorder tests for simplicity.
> (default_binds_local_p): Use default_binds_local_p_2.
> (default_binds_local_p_1): Likewise.
> (decl_binds_to_current_def_p): Unify varpool and cgraph paths
> via symtab_node.
> (default_elf_asm_output_external): Emit visibility when specified.

It looks like this patch broke alphaev68-linux-gnu [1]. There are many
failures of the type:

/tmp/cck7V7MR.o: In function
`__static_initialization_and_destruction_0(int, int)':^M
(.text+0x3ac): relocation truncated to fit: GPRELHIGH against symbol
`std::__cxx11::basic_string,
std::allocator >::~basic_string()@@GLIBCXX_3.4.21' defined in
.text section in
/space/uros/gcc-build/alphaev68-unknown-linux-gnu/./libstdc++-v3/src/.libs/libstdc++.so^M
/space/homedirs/uros/local/bin/ld: /tmp/cck7V7MR.o: gp-relative
relocation against dynamic symbol
_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED1Ev@@GLIBCXX_3.4.21^M
/space/homedirs/uros/local/bin/ld: final link failed: Nonrepresentable
section on output^M

An example is g++.dg/torture/pr60750.C :

/space/uros/gcc-build/gcc/testsuite/g++/../../xg++
-B/space/uros/gcc-build/gcc/testsuite/g++/../../
/space/homedirs/uros/gcc-svn/trunk/gcc/testsuite/g++.dg/torture/pr60750.C
-fno-diagnostics-show-caret -fdiagnostics-color=never -nostdinc++
-I/space/homedirs/uros/gcc-build/alphaev68-unknown-linux-gnu/libstdc++-v3/include/alphaev68-unknown-linux-gnu
-I/space/homedirs/uros/gcc-build/alphaev68-unknown-linux-gnu/libstdc++-v3/include
-I/space/homedirs/uros/gcc-svn/trunk/libstdc++-v3/libsupc++
-I/space/homedirs/uros/gcc-svn/trunk/libstdc++-v3/include/backward
-I/space/homedirs/uros/gcc-svn/trunk/libstdc++-v3/testsuite/util
-fmessage-length=0 -O0 -std=c++11
-L/space/uros/gcc-build/alphaev68-unknown-linux-gnu/./libstdc++-v3/src/.libs
-B/space/uros/gcc-build/alphaev68-unknown-linux-gnu/./libstdc++-v3/src/.libs
-L/space/uros/gcc-build/alphaev68-unknown-linux-gnu/./libstdc++-v3/src/.libs
-lm -o ./pr60750.exe^M
/space/homedirs/uros/local/bin/ld: /tmp/cck7V7MR.o: gp-relative
relocation against dynamic symbol
_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED1Ev@@GLIBCXX_3.4.21^M
/tmp/cck7V7MR.o: In function
`__static_initialization_and_destruction_0(int, int)':^M
(.text+0x3ac): relocation truncated to fit: GPRELHIGH against symbol
`std::__cxx11::basic_string,
std::allocator >::~basic_string()@@GLIBCXX_3.4.21' defined in
.text section in
/space/uros/gcc-build/alphaev68-unknown-linux-gnu/./libstdc++-v3/src/.libs/libstdc++.so^M
/space/homedirs/uros/local/bin/ld: /tmp/cck7V7MR.o: gp-relative
relocation against dynamic symbol
_ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED1Ev@@GLIBCXX_3.4.21^M
/space/homedirs/uros/local/bin/ld: final link failed: Nonrepresentable
section on output^M
collect2: error: ld returned 1 exit status^M
compiler exited with status 1

[1] https://gcc.gnu.org/ml/gcc-testresults/2015-02/msg01867.html

Uros.

Re: [4.8 branch] PATCH: PR middle-end/53623: [4.7/4.8 Regression] sign extension is effectively split into two x86-64 instructions

[H.J. Lu]

On Mon, Feb 16, 2015 at 5:18 AM, Jakub Jelinek  wrote:
> On Mon, Feb 16, 2015 at 05:15:02AM -0800, H.J. Lu wrote:
>> On Mon, Feb 16, 2015 at 4:30 AM, H.J. Lu  wrote:
>> > On Mon, Feb 16, 2015 at 1:35 AM, Jakub Jelinek  wrote:
>> >> On Sun, Feb 15, 2015 at 12:53:39PM -0800, H.J. Lu wrote:
>> >>> This is a backport of the patch for PR middle-end/53623 plus all bug
>> >>> fixes caused by it.  Tested on Linux/x86-32, Linux/x86-64 and x32.  OK
>> >>> for 4.8 branch?
>> >>
>> >> What about PR64286 and PR63659, are you sure those aren't related?
>> >> I mean, they are on the 4.9 branch and I don't see why they couldn't 
>> >> affect
>> >> the 4.8 backport.
>> >>
>> >> Jakub
>> >
>> > Fix for PR 63659 has been backported to 4.8 branch.  I will check if
>> > fix for PR 64286 is needed.
>> >
>> > --
>> > H.J.
>>
>> The fix for PR 64286 is an updated fix for PR 59754 which is caused by
>> the fix for PR 53623.  But the testcase in the fix for PR 64286 doesn't
>> fail on 4.8 branch + my backport of the fix for PR 53623 on Haswell.
>> I suggest
>>
>> 1. We go without my current backport and backport the fix for PR 64286
>> in a separate patch.  Or
>> 2. We go without my backport minus the backport of the PR 59754
>> fix and backport the fixes for PR 59754 plus PR 64286 in a separate patch
>
> I think keeping the branch broken is bad, even if we don't have a testcase
> that really fails, pressumably the issue is just latent.
> So I'd strongly prefer
> 3. Add the PR64286 fix to the patch being tested and commit only when it as
> whole is tested, as one commit.
>

I will do that and restart the testing.

BTW, PR 53623 was a missed optimization bug originally.  Now
it turns out that it fixed a wrong code bug.  We are trying to extract
a run-time testcase from PR 64941 for trunk and branches.

Thanks.

-- 
H.J.

Re: [4.8 branch] PATCH: PR middle-end/53623: [4.7/4.8 Regression] sign extension is effectively split into two x86-64 instructions

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 05:15:02AM -0800, H.J. Lu wrote:
> On Mon, Feb 16, 2015 at 4:30 AM, H.J. Lu  wrote:
> > On Mon, Feb 16, 2015 at 1:35 AM, Jakub Jelinek  wrote:
> >> On Sun, Feb 15, 2015 at 12:53:39PM -0800, H.J. Lu wrote:
> >>> This is a backport of the patch for PR middle-end/53623 plus all bug
> >>> fixes caused by it.  Tested on Linux/x86-32, Linux/x86-64 and x32.  OK
> >>> for 4.8 branch?
> >>
> >> What about PR64286 and PR63659, are you sure those aren't related?
> >> I mean, they are on the 4.9 branch and I don't see why they couldn't affect
> >> the 4.8 backport.
> >>
> >> Jakub
> >
> > Fix for PR 63659 has been backported to 4.8 branch.  I will check if
> > fix for PR 64286 is needed.
> >
> > --
> > H.J.
> 
> The fix for PR 64286 is an updated fix for PR 59754 which is caused by
> the fix for PR 53623.  But the testcase in the fix for PR 64286 doesn't
> fail on 4.8 branch + my backport of the fix for PR 53623 on Haswell.
> I suggest
> 
> 1. We go without my current backport and backport the fix for PR 64286
> in a separate patch.  Or
> 2. We go without my backport minus the backport of the PR 59754
> fix and backport the fixes for PR 59754 plus PR 64286 in a separate patch

I think keeping the branch broken is bad, even if we don't have a testcase
that really fails, pressumably the issue is just latent.
So I'd strongly prefer
3. Add the PR64286 fix to the patch being tested and commit only when it as
whole is tested, as one commit.

Jakub

Re: [4.8 branch] PATCH: PR middle-end/53623: [4.7/4.8 Regression] sign extension is effectively split into two x86-64 instructions

[H.J. Lu]

On Mon, Feb 16, 2015 at 4:30 AM, H.J. Lu  wrote:
> On Mon, Feb 16, 2015 at 1:35 AM, Jakub Jelinek  wrote:
>> On Sun, Feb 15, 2015 at 12:53:39PM -0800, H.J. Lu wrote:
>>> This is a backport of the patch for PR middle-end/53623 plus all bug
>>> fixes caused by it.  Tested on Linux/x86-32, Linux/x86-64 and x32.  OK
>>> for 4.8 branch?
>>
>> What about PR64286 and PR63659, are you sure those aren't related?
>> I mean, they are on the 4.9 branch and I don't see why they couldn't affect
>> the 4.8 backport.
>>
>> Jakub
>
> Fix for PR 63659 has been backported to 4.8 branch.  I will check if
> fix for PR 64286 is needed.
>
> --
> H.J.

The fix for PR 64286 is an updated fix for PR 59754 which is caused by
the fix for PR 53623.  But the testcase in the fix for PR 64286 doesn't
fail on 4.8 branch + my backport of the fix for PR 53623 on Haswell.
I suggest

1. We go without my current backport and backport the fix for PR 64286
in a separate patch.  Or
2. We go without my backport minus the backport of the PR 59754
fix and backport the fixes for PR 59754 plus PR 64286 in a separate patch


-- 
H.J.

Re: [4.8 branch] PATCH: PR middle-end/53623: [4.7/4.8 Regression] sign extension is effectively split into two x86-64 instructions

[H.J. Lu]

On Mon, Feb 16, 2015 at 1:35 AM, Jakub Jelinek  wrote:
> On Sun, Feb 15, 2015 at 12:53:39PM -0800, H.J. Lu wrote:
>> This is a backport of the patch for PR middle-end/53623 plus all bug
>> fixes caused by it.  Tested on Linux/x86-32, Linux/x86-64 and x32.  OK
>> for 4.8 branch?
>
> What about PR64286 and PR63659, are you sure those aren't related?
> I mean, they are on the 4.9 branch and I don't see why they couldn't affect
> the 4.8 backport.
>
> Jakub

Fix for PR 63659 has been backported to 4.8 branch.  I will check if
fix for PR 64286 is needed.

-- 
H.J.

[PATCH] Fix PR ipa/65059

[Martin Liška]

Hello.

This patch is fix which was attached by Honza to the PR. Patch was tested on 
x86_64-linux-pc and no new
regression is introduced. Patch is pre-approved by Honza and I'm going to 
install the patch.

Martin
>From cfe7bd6b57cc6e0768fd72d27a7b222ab1136b32 Mon Sep 17 00:00:00 2001
From: mliska 
Date: Mon, 16 Feb 2015 11:37:29 +0100
Subject: [PATCH] Fix PR ipa/65059.

gcc/ChangeLog:

2015-02-16  Jan Hubicka  

	PR ipa/65059
	* ipa-comdats.c (ipa_comdats): Do not categorize thunks to
	external functions.
---
 gcc/ipa-comdats.c | 5 +
 1 file changed, 5 insertions(+)

diff --git a/gcc/ipa-comdats.c b/gcc/ipa-comdats.c
index ad5945f..9f43f29 100644
--- a/gcc/ipa-comdats.c
+++ b/gcc/ipa-comdats.c
@@ -328,9 +328,14 @@ ipa_comdats (void)
 
   FOR_EACH_DEFINED_SYMBOL (symbol)
 {
+  struct cgraph_node *fun;
   symbol->aux = NULL; 
   if (!symbol->get_comdat_group ()
 	  && !symbol->alias
+	  /* Thunks to external functions do not need to be categorized.  */
+	  && (!(fun = dyn_cast  (symbol))
+	  || !fun->thunk.thunk_p
+	  || fun->function_symbol ()->definition)
 	  && symbol->real_symbol_p ())
 	{
 	  tree *val = map.get (symbol);
-- 
2.1.2

Re: [PATCH, GCC, stage1] Fallback to copy-prop if constant-prop not possible

[Steven Bosscher]

On Mon, Feb 16, 2015 at 11:26 AM, Thomas Preud'homme
 wrote:
> Hi,
>
> The RTL cprop pass in GCC operates by doing a local constant/copy propagation 
> first and then a global one. In the local one, if a constant cannot be 
> propagated (eg. due to constraints of the destination instruction) a copy 
> propagation is done instead. However, at the global level copy propagation is 
> only tried if no constant can be propagated, ie. if a constant can be 
> propagated but the constraints of the destination instruction forbids it, no 
> copy propagation will be tried. This patch fixes this issue. This solves the 
> redundant ldr problem in GCC32RM-439.
>

This would address https://gcc.gnu.org/bugzilla/show_bug.cgi?id=34503#c4

I'll have a look at the patch tonight.

Ciao!
Seven

Re: [Patch, fortran] PR60898 premature release of entry symbols

[Mikael Morin]

Le 15/02/2015 19:00, Jerry DeLisle a écrit :
> On 02/15/2015 09:48 AM, Mikael Morin wrote:
> 
>> [*] I have a few failing testcases (also without the patch), namely the
>> following; does this ring a bell ?
>> FAIL: gfortran.dg/erf_3.F90
>> FAIL: gfortran.dg/fmt_g0_7.f08
>> FAIL: gfortran.dg/fmt_en.f90
>> FAIL: gfortran.dg/nan_7.f90
>> FAIL: gfortran.dg/quad_2.f90
>> FAIL: gfortran.dg/quad_3.f90
>> FAIL: gfortran.dg/round_4.f90
>>
> 
> fmt_g0_7.f08 is a new test that should be passing on x86-64 unless you
> have not updated scanner.c.  Are these fails on x86-64? I do not see
> them here on mine.
> 
On x86_64, yes.  I cleared my build tree, bootstrapped again, and the
failures are gone. :-)

Mikael

Re: [C PATCH] Don't crash on null param (PR c/65066)

[Jakub Jelinek]

On Mon, Feb 16, 2015 at 12:05:06PM +0100, Marek Polacek wrote:
> The CUR_PARAM can be null at this place, so check for that.
> 
> I had hoped that extra testing the original patch by running the C testsuite
> with -Wformat=2 enabled would detect such a case, but apparently not. :(
> 
> Bootstrapped/regtested on x86_64-linux, ok for trunk?
> 
> 2015-02-16  Marek Polacek  
> 
>   PR c/65066
>   * c-format.c (check_format_types): Handle null param.
> 
>   * gcc.dg/pr65066.c: New test.

Ok, thanks.

Jakub

[C PATCH] Don't crash on null param (PR c/65066)

[Marek Polacek]

The CUR_PARAM can be null at this place, so check for that.

I had hoped that extra testing the original patch by running the C testsuite
with -Wformat=2 enabled would detect such a case, but apparently not. :(

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2015-02-16  Marek Polacek  

PR c/65066
* c-format.c (check_format_types): Handle null param.

* gcc.dg/pr65066.c: New test.

diff --git gcc/c-family/c-format.c gcc/c-family/c-format.c
index 2f49b2d..9d03ff0 100644
--- gcc/c-family/c-format.c
+++ gcc/c-family/c-format.c
@@ -2492,6 +2492,7 @@ check_format_types (location_t loc, format_wanted_type 
*types)
  && TREE_CODE (cur_type) == INTEGER_TYPE
  && warn_format_signedness
  && TYPE_UNSIGNED (wanted_type)
+ && cur_param != NULL_TREE
  && TREE_CODE (cur_param) == NOP_EXPR)
{
  tree t = TREE_TYPE (TREE_OPERAND (cur_param, 0));
diff --git gcc/testsuite/gcc.dg/pr65066.c gcc/testsuite/gcc.dg/pr65066.c
index e69de29..883a87d 100644
--- gcc/testsuite/gcc.dg/pr65066.c
+++ gcc/testsuite/gcc.dg/pr65066.c
@@ -0,0 +1,12 @@
+/* PR c/65066 */
+/* { dg-do compile } */
+/* { dg-options "-Wformat=2" } */
+
+extern int sscanf (const char *restrict, const char *restrict, ...);
+int *a;
+
+void
+foo ()
+{
+  sscanf (0, "0x%x  #", a); /* { dg-warning "expects argument of type" } */
+}

Marek

Re: [PATCH, GCC, stage1] Fallback to copy-prop if constant-prop not possible

[Richard Biener]

On Mon, 16 Feb 2015, Thomas Preud'homme wrote:

> Hi,
> 
> The RTL cprop pass in GCC operates by doing a local constant/copy 
> propagation first and then a global one. In the local one, if a constant 
> cannot be propagated (eg. due to constraints of the destination 
> instruction) a copy propagation is done instead. However, at the global 
> level copy propagation is only tried if no constant can be propagated, 
> ie. if a constant can be propagated but the constraints of the 
> destination instruction forbids it, no copy propagation will be tried. 
> This patch fixes this issue. This solves the redundant ldr problem in 
> GCC32RM-439.

I think Steven is more familiar with this code.

Richard.

> ChangeLog entries are as follows:
> 
> *** gcc/ChangeLog ***
> 
> 2015-01-21 Thomas Preud'homme thomas.preudho...@arm.com
> 
> * cprop.c (find_avail_set): Return up to two sets, one whose source is
> a register and one whose source is a constant.  Sets are returned in
> an array passed as parameter rather than as a return value.
> (cprop_insn): Use a do while loop rather than a goto.  Try each of the
> sets returned by find_avail_set, starting with the one whose source is
> a constant.
> 
> 
> *** gcc/testsuite/ChangeLog ***
> 
> 2015-01-21 Thomas Preud'homme thomas.preudho...@arm.com
> 
> * gcc.target/arm/pr64616.c: New file.
> 
> Following testing was done:
> 
> Bootstrapped on x86_64 and ran the testsuite without regression
> Build an arm-none-eabi cross-compilers and ran the testsuite without 
> regression with QEMU emulating a Cortex-M3
> Compiled CSiBE targeting x86_64 and Cortex-M3 arm-none-eabi without 
> regression
> 
> diff --git a/gcc/cprop.c b/gcc/cprop.c
> index 4538291..c246d4b 100644
> --- a/gcc/cprop.c
> +++ b/gcc/cprop.c
> @@ -815,15 +815,15 @@ try_replace_reg (rtx from, rtx to, rtx_insn *insn)
>return success;
>  }
>  
> -/* Find a set of REGNOs that are available on entry to INSN's block.  Return
> -   NULL no such set is found.  */
> +/* Find a set of REGNOs that are available on entry to INSN's block.  If 
> found,
> +   SET_RET[0] will be assigned a set with a register source and SET_RET[1] a
> +   set with a constant source.  If not found the corresponding entry is set 
> to
> +   NULL.  */
>  
> -static struct cprop_expr *
> -find_avail_set (int regno, rtx_insn *insn)
> +static void
> +find_avail_set (int regno, rtx_insn *insn, struct cprop_expr *set_ret[2])
>  {
> -  /* SET1 contains the last set found that can be returned to the caller for
> - use in a substitution.  */
> -  struct cprop_expr *set1 = 0;
> +  set_ret[0] = set_ret[1] = NULL;
>  
>/* Loops are not possible here.  To get a loop we would need two sets
>   available at the start of the block containing INSN.  i.e. we would
> @@ -863,8 +863,10 @@ find_avail_set (int regno, rtx_insn *insn)
>   If the source operand changed, we may still use it for the next
>   iteration of this loop, but we may not use it for substitutions.  */
>  
> -  if (cprop_constant_p (src) || reg_not_set_p (src, insn))
> - set1 = set;
> +  if (cprop_constant_p (src))
> + set_ret[1] = set;
> +  else if (reg_not_set_p (src, insn))
> + set_ret[0] = set;
>  
>/* If the source of the set is anything except a register, then
>we have reached the end of the copy chain.  */
> @@ -875,10 +877,6 @@ find_avail_set (int regno, rtx_insn *insn)
>and see if we have an available copy into SRC.  */
>regno = REGNO (src);
>  }
> -
> -  /* SET1 holds the last set that was available and anticipatable at
> - INSN.  */
> -  return set1;
>  }
>  
>  /* Subroutine of cprop_insn that tries to propagate constants into
> @@ -1044,40 +1042,41 @@ cprop_insn (rtx_insn *insn)
>int changed = 0, changed_this_round;
>rtx note;
>  
> -retry:
> -  changed_this_round = 0;
> -  reg_use_count = 0;
> -  note_uses (&PATTERN (insn), find_used_regs, NULL);
> +  do
> +{
> +  changed_this_round = 0;
> +  reg_use_count = 0;
> +  note_uses (&PATTERN (insn), find_used_regs, NULL);
>  
> -  /* We may win even when propagating constants into notes.  */
> -  note = find_reg_equal_equiv_note (insn);
> -  if (note)
> -find_used_regs (&XEXP (note, 0), NULL);
> +  /* We may win even when propagating constants into notes.  */
> +  note = find_reg_equal_equiv_note (insn);
> +  if (note)
> + find_used_regs (&XEXP (note, 0), NULL);
>  
> -  for (i = 0; i < reg_use_count; i++)
> -{
> -  rtx reg_used = reg_use_table[i];
> -  unsigned int regno = REGNO (reg_used);
> -  rtx src;
> -  struct cprop_expr *set;
> +  for (i = 0; i < reg_use_count; i++)
> + {
> +   rtx reg_used = reg_use_table[i];
> +   unsigned int regno = REGNO (reg_used);
> +   rtx src_cst = NULL, src_reg = NULL;
> +   struct cprop_expr *set[2];
>  
> -  /* If the register has already been set in this block, there's
> -  nothi

[PATCH, GCC, stage1] Fallback to copy-prop if constant-prop not possible

[Thomas Preud'homme]

Hi,

The RTL cprop pass in GCC operates by doing a local constant/copy propagation 
first and then a global one. In the local one, if a constant cannot be 
propagated (eg. due to constraints of the destination instruction) a copy 
propagation is done instead. However, at the global level copy propagation is 
only tried if no constant can be propagated, ie. if a constant can be 
propagated but the constraints of the destination instruction forbids it, no 
copy propagation will be tried. This patch fixes this issue. This solves the 
redundant ldr problem in GCC32RM-439.

ChangeLog entries are as follows:

*** gcc/ChangeLog ***

2015-01-21 Thomas Preud'homme thomas.preudho...@arm.com

* cprop.c (find_avail_set): Return up to two sets, one whose source is
a register and one whose source is a constant.  Sets are returned in
an array passed as parameter rather than as a return value.
(cprop_insn): Use a do while loop rather than a goto.  Try each of the
sets returned by find_avail_set, starting with the one whose source is
a constant.


*** gcc/testsuite/ChangeLog ***

2015-01-21 Thomas Preud'homme thomas.preudho...@arm.com

* gcc.target/arm/pr64616.c: New file.

Following testing was done:

Bootstrapped on x86_64 and ran the testsuite without regression
Build an arm-none-eabi cross-compilers and ran the testsuite without 
regression with QEMU emulating a Cortex-M3
Compiled CSiBE targeting x86_64 and Cortex-M3 arm-none-eabi without 
regression

diff --git a/gcc/cprop.c b/gcc/cprop.c
index 4538291..c246d4b 100644
--- a/gcc/cprop.c
+++ b/gcc/cprop.c
@@ -815,15 +815,15 @@ try_replace_reg (rtx from, rtx to, rtx_insn *insn)
   return success;
 }
 
-/* Find a set of REGNOs that are available on entry to INSN's block.  Return
-   NULL no such set is found.  */
+/* Find a set of REGNOs that are available on entry to INSN's block.  If found,
+   SET_RET[0] will be assigned a set with a register source and SET_RET[1] a
+   set with a constant source.  If not found the corresponding entry is set to
+   NULL.  */
 
-static struct cprop_expr *
-find_avail_set (int regno, rtx_insn *insn)
+static void
+find_avail_set (int regno, rtx_insn *insn, struct cprop_expr *set_ret[2])
 {
-  /* SET1 contains the last set found that can be returned to the caller for
- use in a substitution.  */
-  struct cprop_expr *set1 = 0;
+  set_ret[0] = set_ret[1] = NULL;
 
   /* Loops are not possible here.  To get a loop we would need two sets
  available at the start of the block containing INSN.  i.e. we would
@@ -863,8 +863,10 @@ find_avail_set (int regno, rtx_insn *insn)
  If the source operand changed, we may still use it for the next
  iteration of this loop, but we may not use it for substitutions.  */
 
-  if (cprop_constant_p (src) || reg_not_set_p (src, insn))
-   set1 = set;
+  if (cprop_constant_p (src))
+   set_ret[1] = set;
+  else if (reg_not_set_p (src, insn))
+   set_ret[0] = set;
 
   /* If the source of the set is anything except a register, then
 we have reached the end of the copy chain.  */
@@ -875,10 +877,6 @@ find_avail_set (int regno, rtx_insn *insn)
 and see if we have an available copy into SRC.  */
   regno = REGNO (src);
 }
-
-  /* SET1 holds the last set that was available and anticipatable at
- INSN.  */
-  return set1;
 }
 
 /* Subroutine of cprop_insn that tries to propagate constants into
@@ -1044,40 +1042,41 @@ cprop_insn (rtx_insn *insn)
   int changed = 0, changed_this_round;
   rtx note;
 
-retry:
-  changed_this_round = 0;
-  reg_use_count = 0;
-  note_uses (&PATTERN (insn), find_used_regs, NULL);
+  do
+{
+  changed_this_round = 0;
+  reg_use_count = 0;
+  note_uses (&PATTERN (insn), find_used_regs, NULL);
 
-  /* We may win even when propagating constants into notes.  */
-  note = find_reg_equal_equiv_note (insn);
-  if (note)
-find_used_regs (&XEXP (note, 0), NULL);
+  /* We may win even when propagating constants into notes.  */
+  note = find_reg_equal_equiv_note (insn);
+  if (note)
+   find_used_regs (&XEXP (note, 0), NULL);
 
-  for (i = 0; i < reg_use_count; i++)
-{
-  rtx reg_used = reg_use_table[i];
-  unsigned int regno = REGNO (reg_used);
-  rtx src;
-  struct cprop_expr *set;
+  for (i = 0; i < reg_use_count; i++)
+   {
+ rtx reg_used = reg_use_table[i];
+ unsigned int regno = REGNO (reg_used);
+ rtx src_cst = NULL, src_reg = NULL;
+ struct cprop_expr *set[2];
 
-  /* If the register has already been set in this block, there's
-nothing we can do.  */
-  if (! reg_not_set_p (reg_used, insn))
-   continue;
+ /* If the register has already been set in this block, there's
+nothing we can do.  */
+ if (! reg_not_set_p (reg_used, insn))
+   continue;
 
-  /* Find an assignment that sets reg_used and is available
-at the sta

Re: [PATCH, PR tree-optimization/65002] Disable SRA for functions wrongly marked as read-only

[Ilya Enkovich]

2015-02-15 20:25 GMT+03:00 Mike Stump :
> On Feb 13, 2015, at 11:25 AM, Jakub Jelinek  wrote:
 2015-02-12  Ilya Enkovich  

 PR tree-optimization/65002
 * gcc.dg/pr65002.C: New.
>>>
>>> This test should have gone into g++.dg.
>>
>> Into g++.dg/opt or g++.dg/ipa in particular.
>
> Pre-approved if someone wants to svn mv it.

Moved it.

2015-02-16  Ilya Enkovich  

* gcc.dg/pr65002.C: Move ...
* g++.dg/ipa/pr65002.C: ... here.



Thanks,
Ilya

RE: [PATCH, FT32] initial support

[James Bowman]

> From: Joseph Myers
> ...
> > +@table @gcctabopt
> > +
> > +@item -mspace
> > +@opindex mspace
> > +Enable code-size optimizations.
> > +Some of these optimizations incur a minor performance penalty.
> 
> We already have -Os, so why is an architecture-specific option for this
> needed?
> 
> > +A 16-bit signed constant (-32768..32767)
> 
> Use @minus{} for a minus sign in Texinfo documentation, and @dots{}
> instead of literal ".." or "...".

I have updated the target options. Space-saving is now enabled by
-Os. There is also a new option -msim to enable building for the
simulator (the simulator is pending submission to gdb-binutils).

I have fixed the Texinfo formatting.



FT32 is a new high performance 32-bit RISC core developed by FTDI for embedded 
applications.

Support for FT32 has already been added to binutils. This patch adds FT32 
support to gcc.

Please can someone review it, and if appropriate commit it, as I do not have 
write access to the tree.

The FSF have acknowledged receipt of FTDI's copyright assignment papers.

Thanks very much. ChangeLog entry:

2014-02-16  James Bowman  

* configure.ac: FT32 target added
* libgcc/config.host: FT32 target added
* gcc/config/ft32/: FT32 target added
* libgcc/config/ft32/: FT32 target added
* gcc/doc/install.texi, invoke.texi, md.texi: FT32 details added
* gcc/doc/contrib.texi, MAINTAINERS: self added
* contrib/config-list.mk: FT32 target added
* configure: Regenerated

--
James Bowman
FTDI Open Source Liaison


gcc-ft32.txt.gz
Description: gcc-ft32.txt.gz

Re: [4.8 branch] PATCH: PR middle-end/53623: [4.7/4.8 Regression] sign extension is effectively split into two x86-64 instructions

[Jakub Jelinek]

On Sun, Feb 15, 2015 at 12:53:39PM -0800, H.J. Lu wrote:
> This is a backport of the patch for PR middle-end/53623 plus all bug
> fixes caused by it.  Tested on Linux/x86-32, Linux/x86-64 and x32.  OK
> for 4.8 branch?

What about PR64286 and PR63659, are you sure those aren't related?
I mean, they are on the 4.9 branch and I don't see why they couldn't affect
the 4.8 backport.

Jakub

Re: [4.8 branch] PATCH: PR middle-end/53623: [4.7/4.8 Regression] sign extension is effectively split into two x86-64 instructions

[Richard Biener]

On Sun, Feb 15, 2015 at 9:53 PM, H.J. Lu  wrote:
> Hi,
>
> This is a backport of the patch for PR middle-end/53623 plus all bug
> fixes caused by it.  Tested on Linux/x86-32, Linux/x86-64 and x32.  OK
> for 4.8 branch?

Ok if nobody objects within 24h (you changed the bug to wrong-code
which is why I am considering it - a missed-optimization fix this big
shouldn't be backported).

Richard.

> Thanks.
>
>
> H.J.
> ---
> diff --git a/gcc/ChangeLog b/gcc/ChangeLog
> index 469ee31..44bf322 100644
> --- a/gcc/ChangeLog
> +++ b/gcc/ChangeLog
> @@ -1,3 +1,82 @@
> +2015-02-15  H.J. Lu  
> +
> +   Backport from mainline:
> +   2014-06-13  Jeff Law  
> +
> +   PR rtl-optimization/61094
> +   PR rtl-optimization/61446
> +   * ree.c (combine_reaching_defs): Get the mode for the copy from
> +   the extension insn rather than the defining insn.
> +
> +   2014-06-02  Jeff Law  
> +
> +   PR rtl-optimization/61094
> +   * ree.c (combine_reaching_defs): Do not reextend an insn if it
> +   was marked as do_no_reextend.  If a copy is needed to eliminate
> +   an extension, then mark it as do_not_reextend.
> +
> +   2014-02-14  Jeff Law  
> +
> +   PR rtl-optimization/60131
> +   * ree.c (get_extended_src_reg): New function.
> +   (combine_reaching_defs): Use it rather than assuming location
> +   of REG.
> +   (find_and_remove_re): Verify first operand of extension is
> +   a REG before adding the insns to the copy list.
> +
> +   2014-01-17  Jeff Law  
> +
> +   * ree.c (combine_set_extension): Temporarily disable test for
> +   changing number of hard registers.
> +
> +   2014-01-15  Jeff Law  
> +
> +   PR tree-optimization/59747
> +   * ree.c (find_and_remove_re): Properly handle case where a second
> +   eliminated extension requires widening a copy created for elimination
> +   of a prior extension.
> +   (combine_set_extension): Ensure that the number of hard regs needed
> +   for a destination register does not change when we widen it.
> +
> +   2014-01-10  Jeff Law  
> +
> +   PR middle-end/59743
> +   * ree.c (combine_reaching_defs): Ensure the defining statement
> +   occurs before the extension when optimizing extensions with
> +   different source and destination hard registers.
> +
> +   2014-01-10  Jakub Jelinek  
> +
> +   PR rtl-optimization/59754
> +   * ree.c (combine_reaching_defs): Disallow !SCALAR_INT_MODE_P
> +   modes in the REGNO != REGNO case.
> +
> +   2014-01-08  Jeff Law  
> +
> +   * ree.c (get_sub_rtx): New function, extracted from...
> +   (merge_def_and_ext): Here.
> +   (combine_reaching_defs): Use get_sub_rtx.
> +
> +   2014-01-07  Jeff Law  
> +
> +   PR middle-end/53623
> +   * ree.c (combine_set_extension): Handle case where source
> +   and destination registers in an extension insn are different.
> +   (combine_reaching_defs): Allow source and destination
> +   registers in extension to be different under limited
> +   circumstances.
> +   (add_removable_extension): Remove restriction that the
> +   source and destination registers in the extension are the
> +   same.
> +   (find_and_remove_re): Emit a copy from the extension's
> +   destination to its source after the defining insn if
> +   the source and destination registers are different.
> +
> +   2013-12-12  Jeff Law  
> +
> +   * i386.md (simple LEA peephole2): Add missing mode to zero_extend
> +   for zero-extended MULT simple LEA pattern.
> +
>  2015-02-12  Jakub Jelinek  
>
> Backported from mainline
> diff --git a/gcc/config/i386/i386.md b/gcc/config/i386/i386.md
> index 372ae63..aabd6ec 100644
> --- a/gcc/config/i386/i386.md
> +++ b/gcc/config/i386/i386.md
> @@ -17265,7 +17265,7 @@
> && REGNO (operands[0]) == REGNO (operands[1])
> && peep2_regno_dead_p (0, FLAGS_REG)"
>[(parallel [(set (match_dup 0)
> -  (zero_extend (ashift:SI (match_dup 1) (match_dup 2
> +  (zero_extend:DI (ashift:SI (match_dup 1) (match_dup 2
>   (clobber (reg:CC FLAGS_REG))])]
>"operands[2] = GEN_INT (exact_log2 (INTVAL (operands[2])));")
>
> diff --git a/gcc/ree.c b/gcc/ree.c
> index c7e106f..bc566ad 100644
> --- a/gcc/ree.c
> +++ b/gcc/ree.c
> @@ -327,8 +327,30 @@ combine_set_extension (ext_cand *cand, rtx curr_insn, 
> rtx *orig_set)
>  {
>rtx orig_src = SET_SRC (*orig_set);
>enum machine_mode orig_mode = GET_MODE (SET_DEST (*orig_set));
> -  rtx new_reg = gen_rtx_REG (cand->mode, REGNO (SET_DEST (*orig_set)));
>rtx new_set;
> +  rtx cand_pat = PATTERN (cand->insn);
> +
> +  /* If the extension's source/destination registers are not the same
> + then we need to change the original load to reference the destination
> + of the extension.  Then we need to emit a copy from that destination
> + to the original destination of the load