[Bug c/38869] New: valgrind find problem with -O3 -march=native

[Bug c/38869] New: valgrind find problem with -O3 -march=native

[dcb314 at hotmail dot com]

I just tried to compile this file with the GNU C compiler
version 4.4 snapshot 20090109 using valgrind and the flags
-O3 march=native on a amd64 box.

The debug output was

==20395== Invalid read of size 2
==20395==    at 0xAFDA6E: T.911 (ira-costs.c:590)
==20395==    by 0xAFE0F0: T.910 (ira-costs.c:987)
==20395==    by 0xAFE27A: process_bb_node_for_costs (ira-costs.c:1034)
==20395==    by 0xAF66DA: ira_traverse_loop_tree (ira-build.c:1390)
==20395==    by 0xAFC207: ira_costs (ira-costs.c:1170)
==20395==    by 0xAFAA17: ira_build (ira-build.c:2775)
==20395==    by 0xAF4F19: rest_of_handle_ira (ira.c:1926)
==20395==    by 0x664FFC: execute_one_pass (passes.c:1279)
==20395==    by 0x665244: execute_pass_list (passes.c:1328)
==20395==    by 0x66525C: execute_pass_list (passes.c:1329)
==20395==    by 0x75BA9B: tree_rest_of_compilation (tree-optimize.c:419)
==20395==    by 0x8DB993: cgraph_expand_function (cgraphunit.c:1047)
==20395==  Address 0x7216754 is 804 bytes inside a block of size 1,352 free'd
==20395==    at 0x4C243AF: free (in
/usr/lib64/valgrind/amd64-linux/vgpreload_memcheck.so)
==20395==    by 0x68CEF1: init_regs (regclass.c:687)
==20395==    by 0xAAE762: gimple_expand_cfg (cfgexpand.c:2324)
==20395==    by 0x664FFC: execute_one_pass (passes.c:1279)
==20395==    by 0x665244: execute_pass_list (passes.c:1328)
==20395==    by 0x75BA9B: tree_rest_of_compilation (tree-optimize.c:419)
==20395==    by 0x8DB993: cgraph_expand_function (cgraphunit.c:1047)
==20395==    by 0x8DD6F4: cgraph_optimize (cgraphunit.c:1106)
==20395==    by 0x416C62: c_write_global_declarations (c-decl.c:8074)
==20395==    by 0x70C36C: toplev_main (toplev.c:981)

I suspect that valgrind may have found a problem.


-- 
           Summary: valgrind find problem with -O3 -march=native
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dcb314 at hotmail dot com
  GCC host triplet: x86_64-suse-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] valgrind find problem with -O3 -march=native

[hjl dot tools at gmail dot com]

------- Comment #1 from hjl dot tools at gmail dot com  2009-01-16 14:18 -------
Where is the input file?


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hjl dot tools at gmail dot
                   |                            |com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] valgrind find problem with -O3 -march=native

[jakub at gcc dot gnu dot org]

------- Comment #2 from jakub at gcc dot gnu dot org  2009-01-16 14:27 -------
Also for -march=native you should really post what -fverbose-asm tells you has
been selected for -march and other option (and you should retry with those
options instead of -march=native).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] valgrind find problem with -O3 -march=native

[jv244 at cam dot ac dot uk]

------- Comment #3 from jv244 at cam dot ac dot uk  2009-01-16 14:38 -------
(In reply to comment #2)
> Also for -march=native you should really post what -fverbose-asm tells you has
> been selected for -march and other option (and you should retry with those
> options instead of -march=native).

just compiling with -v should be enough, no?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] valgrind find problem with -O3 -march=native

[hjl dot tools at gmail dot com]

------- Comment #4 from hjl dot tools at gmail dot com  2009-01-16 14:59 -------
(In reply to comment #3)
> (In reply to comment #2)
> > Also for -march=native you should really post what -fverbose-asm tells you has
> > been selected for -march and other option (and you should retry with those
> > options instead of -march=native).
> 
> just compiling with -v should be enough, no?
> 

Yes, if there is an input file on unmodified gcc source.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] valgrind find problem with -O3 -march=native

[dcb314 at hotmail dot com]

------- Comment #5 from dcb314 at hotmail dot com  2009-01-16 15:54 -------
(In reply to comment #1)
> Where is the input file?

gcc/testsuite/gcc.dg/callabi/func-1.c 

My apologies for not getting the cut'n'paste correct.



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] valgrind find problem with -O3 -march=native

[hjl dot tools at gmail dot com]

------- Comment #6 from hjl dot tools at gmail dot com  2009-01-17 01:41 -------
I can reproduce it on Linux/x86-64:

[hjl@gnu-6 gcc]$ valgrind --tool=memcheck  ./cc1 -fpreprocessed x.i -quiet
-dumpbase x.i -mtune=generic -auxbase x  -O -version -o x.s
==30493== Memcheck, a memory error detector.
==30493== Copyright (C) 2002-2007, and GNU GPL'd, by Julian Seward et al.
==30493== Using LibVEX rev 1804, a library for dynamic binary translation.
==30493== Copyright (C) 2004-2007, and GNU GPL'd, by OpenWorks LLP.
==30493== Using valgrind-3.3.0, a dynamic binary instrumentation framework.
==30493== Copyright (C) 2000-2007, and GNU GPL'd, by Julian Seward et al.
==30493== For more details, rerun with: -v
==30493== 
GNU C (GCC) version 4.4.0 20090114 (experimental) [trunk revision 143401]
(x86_64-unknown-linux-gnu)
        compiled by GNU C version 4.3.2 20081105 (Red Hat 4.3.2-7), GMP version
4.2.2, MPFR version 2.3.2.
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
Compiler executable checksum: c755410401678804359bd441117db6be
==30493== Invalid read of size 2
==30493==    at 0xE8FB49: process_bb_node_for_hard_reg_moves (ira-costs.c:1428)
==30493==    by 0xE8772C: ira_traverse_loop_tree (ira-build.c:1390)
==30493==    by 0xE8FF69: setup_allocno_cover_class_and_costs
(ira-costs.c:1496)
==30493==    by 0xE90254: ira_costs (ira-costs.c:1596)
==30493==    by 0xE8B9F7: ira_build (ira-build.c:2775)
==30493==    by 0xE830DB: ira (ira.c:1926)
==30493==    by 0xE837A1: rest_of_handle_ira (ira.c:2095)
==30493==    by 0x7DCCE1: execute_one_pass (passes.c:1279)
==30493==    by 0x7DCF2A: execute_pass_list (passes.c:1328)
==30493==    by 0x7DCF48: execute_pass_list (passes.c:1329)
==30493==    by 0x95210C: tree_rest_of_compilation (tree-optimize.c:419)
==30493==    by 0xB9A0F5: cgraph_expand_function (cgraphunit.c:1047)
==30493==  Address 0x9f24506 is 806 bytes inside a block of size 1,352 free'd
==30493==    at 0x4A0609F: free (vg_replace_malloc.c:323)
==30493==    by 0x814995: init_reg_sets_1 (regclass.c:687)
==30493==    by 0x814C95: init_regs (regclass.c:738)
==30493==    by 0xB503E8: ix86_maybe_switch_abi (i386.c:4627)
==30493==    by 0xE1B437: gimple_expand_cfg (cfgexpand.c:2313)
==30493==    by 0x7DCCE1: execute_one_pass (passes.c:1279)
==30493==    by 0x7DCF2A: execute_pass_list (passes.c:1328)
==30493==    by 0x95210C: tree_rest_of_compilation (tree-optimize.c:419)
==30493==    by 0xB9A0F5: cgraph_expand_function (cgraphunit.c:1047)
==30493==    by 0xB9A2A9: cgraph_expand_all_functions (cgraphunit.c:1106)
==30493==    by 0xB9A847: cgraph_optimize (cgraphunit.c:1311)
==30493==    by 0x42EBA0: c_write_global_declarations (c-decl.c:8074)
==30493== 
==30493== Invalid read of size 2
==30493==    at 0xE8FB10: process_bb_node_for_hard_reg_moves (ira-costs.c:1428)
==30493==    by 0xE8772C: ira_traverse_loop_tree (ira-build.c:1390)
==30493==    by 0xE8FF69: setup_allocno_cover_class_and_costs
(ira-costs.c:1496)
==30493==    by 0xE90254: ira_costs (ira-costs.c:1596)
==30493==    by 0xE8B9F7: ira_build (ira-build.c:2775)
==30493==    by 0xE830DB: ira (ira.c:1926)
==30493==    by 0xE837A1: rest_of_handle_ira (ira.c:2095)
==30493==    by 0x7DCCE1: execute_one_pass (passes.c:1279)
==30493==    by 0x7DCF2A: execute_pass_list (passes.c:1328)
==30493==    by 0x7DCF48: execute_pass_list (passes.c:1329)
==30493==    by 0x95210C: tree_rest_of_compilation (tree-optimize.c:419)
==30493==    by 0xB9A0F5: cgraph_expand_function (cgraphunit.c:1047)
==30493==  Address 0x9f244a2 is 706 bytes inside a block of size 1,352 free'd
==30493==    at 0x4A0609F: free (vg_replace_malloc.c:323)
==30493==    by 0x814995: init_reg_sets_1 (regclass.c:687)
==30493==    by 0x814C95: init_regs (regclass.c:738)
==30493==    by 0xB503E8: ix86_maybe_switch_abi (i386.c:4627)
==30493==    by 0xE1B437: gimple_expand_cfg (cfgexpand.c:2313)
==30493==    by 0x7DCCE1: execute_one_pass (passes.c:1279)
==30493==    by 0x7DCF2A: execute_pass_list (passes.c:1328)
==30493==    by 0x95210C: tree_rest_of_compilation (tree-optimize.c:419)
==30493==    by 0xB9A0F5: cgraph_expand_function (cgraphunit.c:1047)
==30493==    by 0xB9A2A9: cgraph_expand_all_functions (cgraphunit.c:1106)
==30493==    by 0xB9A847: cgraph_optimize (cgraphunit.c:1311)
==30493==    by 0x42EBA0: c_write_global_declarations (c-decl.c:8074)
==30493== 
==30493== Invalid read of size 2
==30493==    at 0xE93708: update_copy_costs (ira-color.c:287)
==30493==    by 0xE947AB: assign_hard_reg (ira-color.c:655)
==30493==    by 0xE96D15: pop_allocnos_from_stack (ira-color.c:1336)
==30493==    by 0xE98264: color_allocnos (ira-color.c:1836)
==30493==    by 0xE988E0: color_pass (ira-color.c:1936)
==30493==    by 0xE876FB: ira_traverse_loop_tree (ira-build.c:1381)
==30493==    by 0xE99096: do_coloring (ira-color.c:2080)
==30493==    by 0xE9C13A: color (ira-color.c:3232)
==30493==    by 0xE9C592: ira_color (ira-color.c:3335)
==30493==    by 0xE83124: ira (ira.c:1938)
==30493==    by 0xE837A1: rest_of_handle_ira (ira.c:2095)
==30493==    by 0x7DCCE1: execute_one_pass (passes.c:1279)
==30493==  Address 0x9f25fac is 628 bytes inside a block of size 1,352 free'd
==30493==    at 0x4A0609F: free (vg_replace_malloc.c:323)
==30493==    by 0x814995: init_reg_sets_1 (regclass.c:687)
==30493==    by 0x814C95: init_regs (regclass.c:738)
==30493==    by 0xB503E8: ix86_maybe_switch_abi (i386.c:4627)
==30493==    by 0xE1B437: gimple_expand_cfg (cfgexpand.c:2313)
==30493==    by 0x7DCCE1: execute_one_pass (passes.c:1279)
==30493==    by 0x7DCF2A: execute_pass_list (passes.c:1328)
==30493==    by 0x95210C: tree_rest_of_compilation (tree-optimize.c:419)
==30493==    by 0xB9A0F5: cgraph_expand_function (cgraphunit.c:1047)
==30493==    by 0xB9A2A9: cgraph_expand_all_functions (cgraphunit.c:1106)
==30493==    by 0xB9A847: cgraph_optimize (cgraphunit.c:1311)
==30493==    by 0x42EBA0: c_write_global_declarations (c-decl.c:8074)
==30493== 
==30493== Invalid read of size 2
==30493==    at 0xE936CF: update_copy_costs (ira-color.c:287)
==30493==    by 0xE947AB: assign_hard_reg (ira-color.c:655)
==30493==    by 0xE96D15: pop_allocnos_from_stack (ira-color.c:1336)
==30493==    by 0xE98264: color_allocnos (ira-color.c:1836)
==30493==    by 0xE988E0: color_pass (ira-color.c:1936)
==30493==    by 0xE876FB: ira_traverse_loop_tree (ira-build.c:1381)
==30493==    by 0xE99096: do_coloring (ira-color.c:2080)
==30493==    by 0xE9C13A: color (ira-color.c:3232)
==30493==    by 0xE9C592: ira_color (ira-color.c:3335)
==30493==    by 0xE83124: ira (ira.c:1938)
==30493==    by 0xE837A1: rest_of_handle_ira (ira.c:2095)
==30493==    by 0x7DCCE1: execute_one_pass (passes.c:1279)
==30493==  Address 0x9f25db8 is 128 bytes inside a block of size 1,352 free'd
==30493==    at 0x4A0609F: free (vg_replace_malloc.c:323)
==30493==    by 0x814995: init_reg_sets_1 (regclass.c:687)
==30493==    by 0x814C95: init_regs (regclass.c:738)
==30493==    by 0xB503E8: ix86_maybe_switch_abi (i386.c:4627)
==30493==    by 0xE1B437: gimple_expand_cfg (cfgexpand.c:2313)
==30493==    by 0x7DCCE1: execute_one_pass (passes.c:1279)
==30493==    by 0x7DCF2A: execute_pass_list (passes.c:1328)
==30493==    by 0x95210C: tree_rest_of_compilation (tree-optimize.c:419)
==30493==    by 0xB9A0F5: cgraph_expand_function (cgraphunit.c:1047)
==30493==    by 0xB9A2A9: cgraph_expand_all_functions (cgraphunit.c:1106)
==30493==    by 0xB9A847: cgraph_optimize (cgraphunit.c:1311)
==30493==    by 0x42EBA0: c_write_global_declarations (c-decl.c:8074)
==30493== 
==30493== ERROR SUMMARY: 19 errors from 4 contexts (suppressed: 4 from 1)
==30493== malloc/free: in use at exit: 414,777 bytes in 1,526 blocks.
==30493== malloc/free: 12,711 allocs, 11,185 frees, 8,719,282 bytes allocated.
==30493== For counts of detected errors, rerun with: -v
==30493== searching for pointers to 1,526 not-freed blocks.
==30493== checked 3,778,992 bytes.
==30493== 
==30493== LEAK SUMMARY:
==30493==    definitely lost: 16 bytes in 1 blocks.
==30493==      possibly lost: 0 bytes in 0 blocks.
==30493==    still reachable: 414,761 bytes in 1,525 blocks.
==30493==         suppressed: 0 bytes in 0 blocks.
==30493== Rerun with --leak-check=full to see details of leaked memory.
[hjl@gnu-6 gcc]$ 


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |vmakarov at redhat dot com
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-01-17 01:41:24
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug rtl-optimization/38869] [4.4 Regression] valgrind find problem with -O3 -mtune=generic

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|valgrind find problem with -|[4.4 Regression] valgrind
                   |O3 -mtune=generic           |find problem with -O3 -
                   |                            |mtune=generic
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] [4.4 Regression] valgrind find problem with -O3 -mtune=generic

[hjl dot tools at gmail dot com]

------- Comment #7 from hjl dot tools at gmail dot com  2009-01-17 01:59 -------
init_regs may be called more than once when switching ABIs. It
may introduce memory leaks.


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jh at suse dot cz, ktietz at
                   |                            |onevision dot com
          Component|rtl-optimization            |c
   Target Milestone|4.4.0                       |---


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] [4.4 Regression] valgrind find problem with -O -mtune=generic

[vmakarov at redhat dot com]

------- Comment #8 from vmakarov at redhat dot com  2009-01-19 00:37 -------
H.J. is right.  The problem is that IRA does not reinitialize some of its data
after changing ABI.  config/i386/i386.c::ix86_maybe_switch_abi calls only
init_regs (this is the single port which does it).  Function ira_init should be
called too in this case.

I'll send a patch soon.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] [4.4 Regression] valgrind find problem with -O -mtune=generic

[vmakarov at gcc dot gnu dot org]

------- Comment #9 from vmakarov at gcc dot gnu dot org  2009-01-19 17:17 -------
Subject: Bug 38869

Author: vmakarov
Date: Mon Jan 19 17:17:14 2009
New Revision: 143498

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=143498
Log:
2009-01-19  Vladimir Makarov  <vmakarov@redhat.com>

        PR c/38869
        * rtl.h (reinit_regs): New prototype.
        * regclass.c: Include ira.h.
        (reinit_regs): New.
        * Makefile.in (regclass.o): Add ira.h.
        * config/i386/i386.c (ix86_maybe_switch_abi): Use reinit_regs.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/Makefile.in
    trunk/gcc/config/i386/i386.c
    trunk/gcc/regclass.c
    trunk/gcc/rtl.h


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869

[Bug c/38869] [4.4 Regression] valgrind find problem with -O -mtune=generic

[rguenth at gcc dot gnu dot org]

------- Comment #10 from rguenth at gcc dot gnu dot org  2009-01-23 16:18 -------
Fixed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=38869