[Bug rtl-optimization/57439] New: [4.9 regression]

[Bug rtl-optimization/57439] New: [4.9 regression]

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

            Bug ID: 57439
           Summary: [4.9 regression]
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: schwab@linux-m68k.org
                CC: amylaar at gcc dot gnu.org
            Target: m68k-*-*

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
   Target Milestone|---                         |4.9.0
            Summary|[4.9 regression]            |[4.9 regression] FAIL:
                   |                            |gcc.c-torture/execute/92050
                   |                            |1-6.c execution,  -O1

--- Comment #1 from Andreas Schwab <schwab@linux-m68k.org> ---
Executing on host: /daten/aranym/gcc/gcc-20130528/Build/gcc/xgcc
-B/daten/aranym/gcc/gcc-20130528/Build/gcc/
/daten/aranym/gcc/gcc-20130528/gcc/testsuite/gcc.c-torture/execute/920501-6.c 
-fno-diagnostics-show-caret -fdiagnostics-color=never  -w  -O1   -lm   -o
/daten/aranym/gcc/gcc-20130528/Build/gcc/testsuite/gcc/920501-6.x1    (timeout
= 300)
spawn /daten/aranym/gcc/gcc-20130528/Build/gcc/xgcc
-B/daten/aranym/gcc/gcc-20130528/Build/gcc/
/daten/aranym/gcc/gcc-20130528/gcc/testsuite/gcc.c-torture/execute/920501-6.c
-fno-diagnostics-show-caret -fdiagnostics-color=never -w -O1 -lm -o
/daten/aranym/gcc/gcc-20130528/Build/gcc/testsuite/gcc/920501-6.x1
PASS: gcc.c-torture/execute/920501-6.c compilation,  -O1 
Executing on aranym:
LD_LIBRARY_PATH=:/daten/aranym/gcc/gcc-20130528/Build/gcc:/daten/aranym/gcc/gcc-20130528/Build/gcc
timeout 600 /daten/aranym/gcc/gcc-20130528/Build/gcc/testsuite/gcc/920501-6.x1 
  (timeout = 300)
Executed /daten/aranym/gcc/gcc-20130528/Build/gcc/testsuite/gcc/920501-6.x1,
status 1
Output: bash: line 1:  2912 Aborted                
LD_LIBRARY_PATH=:/daten/aranym/gcc/gcc-20130528/Build/gcc:/daten/aranym/gcc/gcc-20130528/Build/gcc
timeout 600 /daten/aranym/gcc/gcc-20130528/Build/gcc/testsuite/gcc/920501-6.x1
child process exited abnormally
FAIL: gcc.c-torture/execute/920501-6.c execution,  -O1 

--- /daten/aranym/gcc/gcc-20130527/Build/gcc/testsuite/gcc/920501-6.s   
2013-05-28 11:38:57.524381509 +0200
+++ /daten/aranym/gcc/gcc-20130528/Build/gcc/testsuite/gcc/920501-6.s   
2013-05-28 11:39:09.113314268 +0200
@@ -236,15 +236,13 @@ main:
     jne .L22
     move.l -72(%fp),%d0
     move.l -68(%fp),%d1
-    clr.l %d2
-    move.l #1234111121,%d3
+    move.b #-111,%d2
     sub.l %d3,%d1
     subx.l %d2,%d0
     jne .L22
     move.l -64(%fp),%d0
     move.l -60(%fp),%d1
-    clr.l %d2
-    move.l #1234111127,%d3
+    move.b #-105,%d2
     sub.l %d3,%d1
     subx.l %d2,%d0
     jne .L22

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #2 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
I can-t build the m68k-elf toolchain - it gets an error trying to build
libgloss.
Hence, I don't have stdio.h.

Could you attach a preprocessed testcase?

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #3 from Andreas Schwab <schwab@linux-m68k.org> ---
You don't need stdio.h.

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #5 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
(In reply to Jorn Wolfgang Rennecke from comment #4)
> Created attachment 30209 [details]
> experimental patch

bootstrap / regtest on i686-pc-linux-gnu and
build/regtest for i686-pc-linux-gnu X sh-elf
show no regressions.  Can't regtest for m68k, though.

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2013-05-28
                 CC|                            |ebotcazou at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #6 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> > Created attachment 30209 [details]
> > experimental patch
> 
> bootstrap / regtest on i686-pc-linux-gnu and
> build/regtest for i686-pc-linux-gnu X sh-elf
> show no regressions.  Can't regtest for m68k, though.

Is SH big-endian?  The patch looks good but we need to test it on a big-endian
platform.  I can give it a try on PowerPC if that can help.

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #7 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
(In reply to Eric Botcazou from comment #6)
> Is SH big-endian?

Yes, the default for sh-elf - which is what I have tested - is big endian.

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #8 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> Yes, the default for sh-elf - which is what I have tested - is big endian.

Thanks, the patch is OK then.

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #9 from Andreas Schwab <schwab@linux-m68k.org> ---
Not enough.

Executing on host: /daten/aranym/gcc/gcc-20130529/Build/gcc/xgcc
-B/daten/aranym/gcc/gcc-20130529/Build/gcc/
/daten/aranym/gcc/gcc-20130529/gcc/testsuite/gcc.c-torture/execute/pr42721.c 
-fno-diagnostics-show-caret -fdiagnostics-color=never  -w  -Og -g   -lm   -o
/daten/aranym/gcc/gcc-20130529/Build/gcc/testsuite/gcc/pr42721.x6    (timeout =
300)
spawn /daten/aranym/gcc/gcc-20130529/Build/gcc/xgcc
-B/daten/aranym/gcc/gcc-20130529/Build/gcc/
/daten/aranym/gcc/gcc-20130529/gcc/testsuite/gcc.c-torture/execute/pr42721.c
-fno-diagnostics-show-caret -fdiagnostics-color=never -w -Og -g -lm -o
/daten/aranym/gcc/gcc-20130529/Build/gcc/testsuite/gcc/pr42721.x6
PASS: gcc.c-torture/execute/pr42721.c compilation,  -Og -g 
Executing on aranym:
LD_LIBRARY_PATH=:/daten/aranym/gcc/gcc-20130529/Build/gcc:/daten/aranym/gcc/gcc-20130529/Build/gcc
timeout 600 /daten/aranym/gcc/gcc-20130529/Build/gcc/testsuite/gcc/pr42721.x6  
 (timeout = 300)
Executed /daten/aranym/gcc/gcc-20130529/Build/gcc/testsuite/gcc/pr42721.x6,
status 1
Output: bash: line 1: 32610 Aborted                
LD_LIBRARY_PATH=:/daten/aranym/gcc/gcc-20130529/Build/gcc:/daten/aranym/gcc/gcc-20130529/Build/gcc
timeout 600 /daten/aranym/gcc/gcc-20130529/Build/gcc/testsuite/gcc/pr42721.x6
child process exited abnormally
FAIL: gcc.c-torture/execute/pr42721.c execution,  -Og -g 

--- /daten/aranym/gcc/gcc-20130527/Build/gcc/testsuite/gcc/pr42721.s   
2013-05-29 10:05:31.865578609 +0200
+++ /daten/aranym/gcc/gcc-20130529/Build/gcc/testsuite/gcc/pr42721.s   
2013-05-29 10:05:40.025528853 +0200
@@ -42,7 +42,6 @@ main:
     move.l b,%d1
     eor.l %d1,%d0
     move.l %d0,b
-    moveq #1,%d2
     cmp.l %d0,%d2
     jeq .L4
     jsr abort

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #10 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
Created attachment 30212
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30212&action=edit
experimental patch for execute/pr42721.c failure

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #11 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
(In reply to Jorn Wolfgang Rennecke from comment #10)
> Created attachment 30212 [details]
> experimental patch for execute/pr42721.c failure

bootstrapped / regtested on i686-pc-linux-gnu
build / regtested for i686-pc-linux-gnu X sh-elf

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[schwab@linux-m68k.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #12 from Andreas Schwab <schwab@linux-m68k.org> ---
Successfully tested on m68k.

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

--- Comment #13 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
(In reply to Jorn Wolfgang Rennecke from comment #10)
> Created attachment 30212 [details]
> experimental patch for execute/pr42721.c failure

The patch is OK but let's beef up the comment a little bit:

+  if (mode != reg_mode[regno])
+    {
+      if (!MODES_OK_FOR_MOVE2ADD (mode, reg_mode[regno]))
+    return false;
+      /* For big endian, the starting regno for the subreg might be
+     different.  */

      /* The value loaded into regno in reg_mode[regno] is also valid in
         mode after truncation only if (REG:mode regno) is the lowpart of
         (REG:reg_mode[regno] regno).  Now, for big endian, the starting
         regno of the lowpart might be different.  */

+      int s_off = subreg_lowpart_offset (mode, reg_mode[regno]);
+      s_off = subreg_regno_offset (regno, reg_mode[regno], s_off, mode);
+      if (s_off != 0)
+    /* We could in principle adjust regno, check reg_mode[regno] to be
+       BLKmode, and return s_off to the caller (vs. -1 for failure),
+       but we currently have no callers that could make use of this
+       information.  */
+    return false;
+    }

[Bug rtl-optimization/57439] [4.9 regression] FAIL: gcc.c-torture/execute/920501-6.c execution, -O1

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57439

Richard Biener <rguenth at gcc dot gnu.org> changed:

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

--- Comment #14 from Richard Biener <rguenth at gcc dot gnu.org> ---
Patch was applied, thus fixed.