[Bug c/11965] New: On SPARC, gcc produces invalid assembler code for a shift << 32 operation

[Bug c/11965] New: On SPARC, gcc produces invalid assembler code for a shift << 32 operation

[jk at tools dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965

           Summary: On SPARC, gcc produces invalid assembler code for a
                    shift << 32 operation
           Product: gcc
           Version: 3.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jk at tools dot de
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: sparc-sun-solaris2.8
  GCC host triplet: sparc-sun-solaris2.8
GCC target triplet: sparc-sun-solaris2.8

GCC 3.3 on Solaris SPARC produces invalid assembler code, for shift operator
shifting 32 bits.

GCC 3.3 is compiled to use the system's assembler, /usr/ccs/bin/as


To reproduce the issue (admitted, this code has undefined behaviour):

% cat shift.c

#if __GNUC__
#define ATTR_ALWAYS_INLINE __attribute__ ((__always_inline__))
#else
#define ATTR_ALWAYS_INLINE
#endif


static ATTR_ALWAYS_INLINE unsigned
put_bits(int n, unsigned int value)
{
        return value << n;
}

main()
{
        unsigned val = 1;
        int i;

        for (i = 0; i < 4; i++)
                val = put_bits(32, val);

        printf("val=%x\n", val);
}

% gcc -O3 -mcpu=ultrasparc -o shift shift.c
/usr/ccs/bin/as: "/tmp/cc0VFu29.s", line 20: error: shift count negative or too
big: 32


Looking at the generated assembler source...

% gcc -O3 -mcpu=ultrasparc -S shift.c

... there's a shift instruction with a constant shift count of 32, where the
immediate shift count is restricted to the range 0 .. 31 on SPARC:

       sll     %o1, 32, %o1


The same code used to compile with gcc 3.2 and erlier, and it does compile with
gcc 3.3 at optimization level -O2 or when option -mcpu=ultrasparc is not used.

[Bug c/11965] On SPARC, gcc produces invalid assembler code for a shift << 32 operation

[falk at debian dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965


falk at debian dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1


------- Additional Comments From falk at debian dot org  2003-08-18 17:35 -------
I can reproduce this on Sparc Linux with gcc 3.3.1 20030728 (Debian prerelease).

[Bug c/11965] On SPARC, gcc produces invalid assembler code for a shift << 32 operation

[martin at netbsd dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965


martin at netbsd dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |martin at netbsd dot org


------- Additional Comments From martin at netbsd dot org  2003-08-19 20:04 -------
Forgot to mention: I used gcc version 3.3.1.
I initially left out the 32bit changes (since that cases currently seem to not
be triggered), but Ian L. Taylor told me you can not rely on that and pointed to
the mips backends way of handling it.

[Bug c/11965] On SPARC, gcc produces invalid assembler code for a shift << 32 operation

[dhazeghi at yahoo dot com]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965


dhazeghi at yahoo dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |wrong-code
   Target Milestone|3.4                         |---

[Bug target/11965] [3.3/3.4 Regression] On SPARC, gcc produces invalid assembler code for a shift << 32 operation

[pinskia at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c                           |target
            Summary|On SPARC, gcc produces      |[3.3/3.4 Regression] On
                   |invalid assembler code for a|SPARC, gcc produces invalid
                   |shift << 32 operation       |assembler code for a shift
                   |                            |<< 32 operation
   Target Milestone|---                         |3.3.2


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-08-23 14:24 -------
This is a regression from 3.2.3 and there is even a patch.


Martin can you send the patch to gcc-patches saying this fixes this PR?

[Bug target/11965] [3.3/3.4 Regression] invalid assembler code for a shift << 32 operation

[ebotcazou at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965


ebotcazou at gcc dot gnu dot org changed:

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


------- Additional Comments From ebotcazou at gcc dot gnu dot org  2003-09-09 14:51 -------
I'll bootstrap/regtest it for SPARC and SPARC64 on the 3.3 branch and install it.

[Bug target/11965] [3.3/3.4 Regression] invalid assembler code for a shift << 32 operation

[cvs-commit at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965



------- Additional Comments From cvs-commit at gcc dot gnu dot org  2003-09-10 12:59 -------
Subject: Bug 11965

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	ebotcazou@gcc.gnu.org	2003-09-10 12:59:36

Modified files:
	gcc            : ChangeLog 
	gcc/config/sparc: sparc.c sparc.md 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg: ultrasp10.c 

Log message:
	PR target/11965
	* config/sparc/sparc.c (sparc_v8plus_shift): Protect against
	constants greater than 63.
	* config/sparc/sparc.md (ashlsi3, ashrsi3, lshrsi3): Protect
	against constants greater than 31.
	(*ashldi3_sp64, *ashrdi3_sp64, *lshrdi3_sp64): Protect against
	constants greater than 63.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.1022&r2=2.1023
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/sparc/sparc.c.diff?cvsroot=gcc&r1=1.258&r2=1.259
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/sparc/sparc.md.diff?cvsroot=gcc&r1=1.190&r2=1.191
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.3047&r2=1.3048
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/ultrasp10.c.diff?cvsroot=gcc&r1=NONE&r2=1.1

[Bug target/11965] [3.3/3.4 Regression] invalid assembler code for a shift << 32 operation

[cvs-commit at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965



------- Additional Comments From cvs-commit at gcc dot gnu dot org  2003-09-10 13:04 -------
Subject: Bug 11965

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_3-branch
Changes by:	ebotcazou@gcc.gnu.org	2003-09-10 13:04:00

Modified files:
	gcc            : ChangeLog 
	gcc/config/sparc: sparc.c sparc.md 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg: ultrasp10.c 

Log message:
	PR target/11965
	* config/sparc/sparc.c (sparc_v8plus_shift): Protect against
	constants greater than 63.
	* config/sparc/sparc.md (ashlsi3, ashrsi3, lshrsi3): Protect
	against constants greater than 31.
	(*ashldi3_sp64, *ashrdi3_sp64, *lshrdi3_sp64): Protect against
	constants greater than 63.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.16114.2.734&r2=1.16114.2.735
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/sparc/sparc.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.233.4.5&r2=1.233.4.6
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/sparc/sparc.md.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.181.4.6&r2=1.181.4.7
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.2261.2.276&r2=1.2261.2.277
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/ultrasp10.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=NONE&r2=1.1.2.1

[Bug target/11965] [3.3/3.4 Regression] invalid assembler code for a shift << 32 operation

[ebotcazou at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=11965


ebotcazou at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|ebotcazou at gcc dot gnu dot|
                   |org                         |
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


------- Additional Comments From ebotcazou at gcc dot gnu dot org  2003-09-10 13:05 -------
Done.