[Bug target/46127] New: Use 16bit add instead of 32bit in thumb2

[carrot at google dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46127

           Summary: Use 16bit add instead of 32bit in thumb2
           Product: gcc
           Version: 4.6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: carrot@google.com
                CC: carrot@google.com
              Host: i686-linux
            Target: arm-eabi
             Build: i686-linux


Compile the following code with options -march=armv7-a -mthumb -Os

unsigned long compressBound (unsigned long sourceLen)
{
      return sourceLen + (sourceLen >> 12) + (sourceLen >> 14) +
                     (sourceLen >> 25) + 13;
}

GCC 4.6 generates:

compressBound:
        add     r3, r0, #13           // A
        add     r3, r3, r0, lsr #12
        add     r3, r3, r0, lsr #14
        add     r0, r3, r0, lsr #25
        bx      lr

We can change the instruction order and register a little

compressBound:
        add     r3, r0, r0, lsr #12
        add     r3, r3, r0, lsr #14
        add     r0, r3, r0, lsr #25
        add     r0, r0, #13           // B
        bx      lr

Now instruction A becomes instruction B. Instruction A is 32 bit, instruction B
is 16 bit, so it becomes shorter.

Don't know how to handle it in compiler.