[Bug c/28325] New: -fno-delayed-branch does not seem to work with the MIPS branch instructions

[Bug c/28325] New: -fno-delayed-branch does not seem to work with the MIPS branch instructions

[ska at bth dot se]

I'm working on a binary translator that translates MIPS binaries into Java
bytecode (http://spel.bth.se/index.php/Cibyl) with the goal of "recompiling" C
programs to run on J2ME.  To simplify things, I use compile programs for a
subset of the MIPS1 instruction set, and ideally I would like to get rid of
delayed instructions.

So, I used the -fno-delayed-branch instruction when compiling. The GCC version
is

mips-linux-gcc (GCC) 3.4.4 20050314 (prerelease) (Debian 3.4.3-13)

(an emdebian.org binary), and the command line i used is

mips-linux-gcc -Wall -mips1 -fno-delayed-branch -O2 -fno-pic -mno-abicalls
-I/home/ska/projects/private/games/mophun/cibyl/trunk//include  -O2 -c main.c
-o main.o

For the code below,

#define _syscall1(type,name,atype,a) \
type name(atype a) \
{ \
        register unsigned long __a0 asm("$4") = (unsigned long) a; \
        register unsigned long __v0 asm("$2"); \
        \
        __asm__ volatile ( \
        ".set\tnoreorder\n\t" \
        "li\t$2, %2\t\t\t# " #name "\n\t" \
        "syscall\n\t" \
        ".set\treorder" \
        : "=&r" (__v0) \
        : "r" (__a0), "i" (__NR_##name) \
        ); \
        \
        return (type) __v0; \
}
#define __NR_exit 0
static inline _syscall1(void,exit , int, code );

int tst(int *argv)
{
  int x;

  for (x = 0; x < 24; x++)
    exit(argv[x]);

  return 0;
}

I get instructions in the delay slots of bgez and jr:

00000000 <tst>:
   0:   00801821        move    v1,a0
   4:   24050017        li      a1,23
   8:   8c640000        lw      a0,0(v1)
   c:   24020000        li      v0,0
  10:   0000000c        syscall
  14:   24a5ffff        addiu   a1,a1,-1
  18:   04a1fffb        bgez    a1,8 <tst+0x8>
  1c:   24630004        addiu   v1,v1,4
  20:   03e00008        jr      ra
  24:   00001021        move    v0,zero
        ...

Which is not what I had expected. If i replace the "syscall" exit with a
function call instead, I get the expected behavior:

00000000 <exit3>:
   0:   08000000        j       0 <exit3>
   4:   00000000        nop

00000008 <tst>:
   8:   27bdffe0        addiu   sp,sp,-32
   c:   afb10014        sw      s1,20(sp)
  10:   afb00010        sw      s0,16(sp)
  14:   afbf0018        sw      ra,24(sp)
  18:   00808021        move    s0,a0
  1c:   24110017        li      s1,23
  20:   8e040000        lw      a0,0(s0)
  24:   2631ffff        addiu   s1,s1,-1
  28:   0c000000        jal     0 <exit3>
  2c:   00000000        nop
  30:   26100004        addiu   s0,s0,4
  34:   0621fffa        bgez    s1,20 <tst+0x18>
  38:   00000000        nop
  3c:   8fbf0018        lw      ra,24(sp)
  40:   8fb10014        lw      s1,20(sp)
  44:   8fb00010        lw      s0,16(sp)
  48:   00001021        move    v0,zero
  4c:   27bd0020        addiu   sp,sp,32
  50:   03e00008        jr      ra
  54:   00000000        nop
        ...


-- 
           Summary: -fno-delayed-branch does not seem to work with the MIPS
                    branch instructions
           Product: gcc
           Version: 3.4.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ska at bth dot se


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28325

[Bug middle-end/28325] -fno-delayed-branch does not seem to work with the MIPS branch instructions

[ian at airs dot com]

------- Comment #1 from ian at airs dot com  2006-07-10 16:27 -------
This is not a bug in the compiler.  The ".set reorder" directive tells the
assembler that it should reorder instructions into branch delay slots when
possible.  The compiler just copies the ".set reorder" directly from the asm
statement.  Both the compiler and the assembler are acting as expected.

You should rewrite your asm statement to not use .set reorder.  Do this
instead:
    .set push
    .set noreorder
    ...
    .set pop


-- 

ian at airs dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28325