[Bug debug/44023] New: -fcompare-debug failure (length) for alphaev67 target

[Bug debug/44023] New: -fcompare-debug failure (length) for alphaev67 target

[ubizjak at gmail dot com]

Recent changes broke bootstrap on alphaev68-pc-linux-gnu, due to stage2/3
compare failure in libiberty/sha1.o.

Soon to be attached preprocessed file triggers -fcompare-debug failure (also on
a crosscompiler from x86_64-pc-linux-gnu):

~/gcc-build-alpha/gcc/xgcc -B ~/gcc-build-alpha/gcc -mcpu=ev67 -S -O2
-fcompare-debug sha1.i 
xgcc: sha1.i: -fcompare-debug failure (length)


-- 
           Summary: -fcompare-debug failure (length) for alphaev67 target
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: debug
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ubizjak at gmail dot com
GCC target triplet: alphaev67-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] -fcompare-debug failure (length) for alphaev67 target

[ubizjak at gmail dot com]

------- Comment #1 from ubizjak at gmail dot com  2010-05-07 12:39 -------
Created an attachment (id=20595)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20595&action=view)
preprocessed source

Preprocessed source (can be compiled on a crosscompiler), triggers stage2/3
bootstrap compare failure.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] -fcompare-debug failure (length) for alphaev67 target

[jakub at gcc dot gnu dot org]

------- Comment #2 from jakub at gcc dot gnu dot org  2010-05-07 15:40 -------
>From quick skim the first difference is on a sha1.c:350 insn in *.ira dump.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] -fcompare-debug failure (length) for alphaev67 target

[ubizjak at gmail dot com]

------- Comment #3 from ubizjak at gmail dot com  2010-05-07 15:46 -------
Jakub, reverting your r159063 fixes the failure:

        PR debug/43478
        * df-problems.c (struct dead_debug_use, struct dead_debug): New.
        (dead_debug_init, dead_debug_finish): New functions.
        (dead_debug_add, dead_debug_insert_before): Likewise.
        (df_note_bb_compute): Initialize a dead_debug object, add dead
        debug uses to it, insert debug bind insns before death insns,
        reset debug insns that refer to pending uses at the end.
        * rtl.h (make_debug_expr_from_rtl): New prototype.
        * varasm.c (make_debug_expr_from_rtl): New function.

[1] http://gcc.gnu.org/ml/gcc-cvs/2010-05/msg00112.html


-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] -fcompare-debug failure (length) for alphaev67 target

[ubizjak at gmail dot com]

------- Comment #4 from ubizjak at gmail dot com  2010-05-07 16:32 -------
Created an attachment (id=20598)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20598&action=view)
reduced testcase

Reduced testcase, 45 lines. Beware, crypto code ahead...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] -fcompare-debug failure (length) for alphaev67 target

[ubizjak at gmail dot com]

------- Comment #5 from ubizjak at gmail dot com  2010-05-07 17:04 -------
For some reason, IRA chooses different reload registers.

--- t1.c.190r.ira       2010-05-07 18:57:36.000000000 +0200
+++ t1.c.gk.190r.ira    2010-05-07 18:57:36.000000000 +0200

These are differences in chosen reload registers:

-Reloads for insn # 583
+Reloads for insn # 697
 Reload 0: reload_out (DI) = (reg:DI 789)
        GENERAL_REGS, RELOAD_FOR_OUTPUT (opnum = 0)
        reload_out_reg: (reg:DI 789)
-       reload_reg_rtx: (reg:DI 28 $28)
+       reload_reg_rtx: (reg:DI 0 $0)
 Reload 1: reload_in (DI) = (reg/v/f:DI 81 [ endp ])
        GENERAL_REGS, RELOAD_FOR_INPUT (opnum = 2)
        reload_in_reg: (reg/v/f:DI 81 [ endp ])
-       reload_reg_rtx: (reg:DI 28 $28)
+       reload_reg_rtx: (reg:DI 0 $0)

[...]

-Reloads for insn # 475
+Reloads for insn # 556
 Reload 0: reload_in (SI) = (const_int -1894007588 [0xffffffff8f1bbcdc])
        GENERAL_REGS, RELOAD_FOR_INPUT (opnum = 2), can't combine
        reload_in_reg: (reg:SI 795)
-       reload_reg_rtx: (reg:SI 28 $28)
+       reload_reg_rtx: (reg:SI 24 $24)

[...]

-Reloads for insn # 514
+Reloads for insn # 598
 Reload 0: reload_in (SI) = (reg:SI 725)
        GENERAL_REGS, RELOAD_FOR_INPUT (opnum = 1)
        reload_in_reg: (reg:SI 725)
-       reload_reg_rtx: (reg:SI 8 $8)
+       reload_reg_rtx: (reg:SI 27 $27)

[...]

-Reloads for insn # 584
+Reloads for insn # 698
 Reload 0: reload_in (DI) = (reg:DI 789)
        GENERAL_REGS, RELOAD_FOR_INPUT (opnum = 2)
        reload_in_reg: (reg:DI 789)
-       reload_reg_rtx: (reg:DI 12 $12)
+       reload_reg_rtx: (reg:DI 1 $1)

-Reloads for insn # 586
+Reloads for insn # 700
 Reload 0: reload_in (DI) = (reg/v/f:DI 287 [ ctx ])
        GENERAL_REGS, RELOAD_FOR_OPERAND_ADDRESS (opnum = 0)
        reload_in_reg: (reg/v/f:DI 287 [ ctx ])
-       reload_reg_rtx: (reg:DI 13 $13)
+       reload_reg_rtx: (reg:DI 2 $2)
 Reload 1: reload_out (SI) = (mem/s:SI (reg/v/f:DI 287 [ ctx ]) [4
ctx_13(D)->A+0 S4 A32])
        NO_REGS, RELOAD_FOR_OUTPUT (opnum = 0), optional
        reload_out_reg: (mem/s:SI (reg/v/f:DI 287 [ ctx ]) [4 ctx_13(D)->A+0 S4
A32])


-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ubizjak at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] -fcompare-debug failure (length) for alphaev67 target

[ubizjak at gmail dot com]

------- Comment #6 from ubizjak at gmail dot com  2010-05-07 17:12 -------
The differences in IRA insn stream dumps are then:

[...]

-$28:DI=[$30:DI+0x78]
-$28:DI=leu($28:DI,$16:DI)
-[$30:DI+0x70]=$28:DI
+$0:DI=[$30:DI+0x78]
+$0:DI=leu($0:DI,$16:DI)
+[$30:DI+0x70]=$0:DI

[...]

-$28:SI=0xffffffff8f1bbcdc
-$17:SI=$23:SI+$28:SI
+$24:SI=0xffffffff8f1bbcdc
+$17:SI=$23:SI+$24:SI

[...]

-$8:SI=[$30:DI+0x68]
-$24:SI=$8:SI+$25:SI
+$27:SI=[$30:DI+0x68]
+$24:SI=$27:SI+$25:SI

[...]

-$12:DI=[$30:DI+0x70]
-pc={($12:DI==0)?L598:pc}
+$1:DI=[$30:DI+0x70]
+pc={($1:DI==0)?L712:pc}

[...]

-$13:DI=[$30:DI+0x80]
-[$13:DI]=$9:SI
+$2:DI=[$30:DI+0x80]
+[$2:DI]=$9:SI


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] -fcompare-debug failure (length) for alphaev67 target

[jakub at gcc dot gnu dot org]

------- Comment #7 from jakub at gcc dot gnu dot org  2010-05-07 20:25 -------
Slightly more reduced testcase:

void
foo (unsigned f, long v, unsigned *w, unsigned a, unsigned b, unsigned e,
unsigned c, unsigned d)
{
  unsigned h = v / 4, x[16];
  while (f < h)
    {
      unsigned i;
      f++;
      a |= (a >> 30);
      d = (d << 30) | ((unsigned) d >> 30);
      c = (c << 30) | ((unsigned) c >> 30);
      b = 30 | ((unsigned) b >> 30);
      d += a = (a << 30) | ((unsigned) a >> 2);
      c += ((d << 5) | (d >> 27)) + ((e & (a ^ b))) + 0x5a827999 + x[12];
      a += (c & e);
      c = 30 | ((unsigned) c);
      i = x[5] ^ x[7] ^ x[8] ^ x[3];
      x[5] = (i << 1) | ((unsigned) i >> 31);
      i = x[6] ^ x[2] ^ x[14] ^ x[13];
      x[6] = (i << 1) | (i >> 31);
      b += (c | (c >> 5)) + (d ^ e) + 0x6ed9eba1
        + (x[7] = (i << 1) | ((unsigned) i >> 31));
      x[8] = i | 1;
      e += (a | 5) + b + (i = x[9] ^ x[6], x[10] = (i << (unsigned) i));
      e = 30 | ((unsigned) e >> 30);
      i = x[12] ^ x[14] ^ x[12] ^ x[12], (x[12] = 1 | ((unsigned) i));
      i = x[13] ^ x[5] ^ x[10], (x[13] = (i << (unsigned) 1));
      i = x[2] ^ x[7] ^ x[12], (x[15] = i | ((unsigned) i >> 1));
      i = x[2] ^ x[0] ^ x[13], (x[0] = (i << 1) | 31);
      e = (e << 30) | 2;
      i = x[14] ^ x[2] ^ x[15], (x[2] = i | 1);
      x[3] = i | ((unsigned) i);
      i = x[14] ^ x[12] ^ x[4], (x[4] = 1 | ((unsigned) i >> 1));
      x[5] = i | 1;
      e = (e << 30) | 30;
      b += (5 | ((unsigned) e >> 5)) + 0x8f1bbcdc
        + (x[9] = (i | ((unsigned) i >> 1)));
      i = x[2] ^ (x [10] = ((i << 1) | (i >> 1)));
      x[13] = (i | ((unsigned) i >> 1));
      (i = x[14] ^ x[0] ^ x[14], (x[14] = ((i << 1) | 31)));
      a = *w += a;
    }
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] [4.6 Regression] -fcompare-debug failure (length) for alphaev67 target (bootstrap failure)

[ubizjak at gmail dot com]

------- Comment #8 from ubizjak at gmail dot com  2010-05-08 09:20 -------
Confirmed and added Vlad to CC.


-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |vmakarov at gcc dot gnu dot
                   |                            |org
           Severity|normal                      |blocker
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2010-05-08 09:20:10
               date|                            |
            Summary|-fcompare-debug failure     |[4.6 Regression] -fcompare-
                   |(length) for alphaev67      |debug failure (length) for
                   |target                      |alphaev67 target (bootstrap
                   |                            |failure)
   Target Milestone|---                         |4.6.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] [4.6 Regression] -fcompare-debug failure (length) for alphaev67 target (bootstrap failure)

[jakub at gcc dot gnu dot org]

------- Comment #9 from jakub at gcc dot gnu dot org  2010-05-10 10:13 -------
Created an attachment (id=20617)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20617&action=view)
gcc46-pr44023.patch

The bug is that if a debug insn references more than one dead pseudo, only the
first one is recorded using dead_debug_add.  That means the other pseudos
stay alive beyond REG_DEAD notes and affect e.g. RA decisions.

Attached patch fixes this by calling dead_debug_add on all the uses instead of
just one.  This alone crashes badly, because dead_debug_insert_before which
replaces one of the uses in a debug insn calls df_insn_rescan on it, which
invalidates all the other uses.  So we need to queue rescanning it until
dead_debug_finish (or, at least until all dead debug records referencing the
debug_insn are gone).

I'll bootstrap/regtest the patch on x86_64-linux and i686-linux, Uros, can you
please do the same on alpha*-*?


-- 

jakub at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |jakub at gcc dot gnu dot org
                   |dot org                     |
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] [4.6 Regression] -fcompare-debug failure (length) for alphaev67 target (bootstrap failure)

[ubizjak at gmail dot com]

------- Comment #10 from ubizjak at gmail dot com  2010-05-10 13:58 -------
(In reply to comment #9)

> I'll bootstrap/regtest the patch on x86_64-linux and i686-linux, Uros, can you
> please do the same on alpha*-*?

The patch bootstraps OK on alphaev68-pc-linux-gnu. Regression test just
started, but I don't expect any problems here.

I'd say that the patch is OK.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] [4.6 Regression] -fcompare-debug failure (length) for alphaev67 target (bootstrap failure)

[ubizjak at gmail dot com]

------- Comment #11 from ubizjak at gmail dot com  2010-05-10 17:15 -------
(In reply to comment #10)

> I'd say that the patch is OK.

Regression test results at [1], everything looks OK.

[1] http://gcc.gnu.org/ml/gcc-testresults/2010-05/msg00960.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] [4.6 Regression] -fcompare-debug failure (length) for alphaev67 target (bootstrap failure)

[jakub at gcc dot gnu dot org]

------- Comment #12 from jakub at gcc dot gnu dot org  2010-05-11 06:48 -------
Subject: Bug 44023

Author: jakub
Date: Tue May 11 06:48:15 2010
New Revision: 159254

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=159254
Log:
        PR debug/44023
        * df-problems.c (struct dead_debug): Add to_rescan field.
        (dead_debug_init): Clear to_rescan field.
        (dead_debug_finish): Rescan all debug insns in to_rescan
        bitmap and free the bitmap.
        (dead_debug_insert_before): Instead of rescanning debug insns
        immediately queue their rescanning until dead_debug_finish.
        (df_note_bb_compute): After dead_debug_add do continue instead
        of break.

        * gcc.dg/pr44023.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/pr44023.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/df-problems.c
    trunk/gcc/testsuite/ChangeLog


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023

[Bug debug/44023] [4.6 Regression] -fcompare-debug failure (length) for alphaev67 target (bootstrap failure)

[jakub at gcc dot gnu dot org]

------- Comment #13 from jakub at gcc dot gnu dot org  2010-05-11 07:11 -------
Fixed.


-- 

jakub at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44023