[Bug target/65260] New: [SH] Convert add add sub -> sub sub sub

["olegendo at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65260

            Bug ID: 65260
           Summary: [SH] Convert add add sub -> sub sub sub
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: olegendo at gcc dot gnu.org
            Target: sh*-*-*

I've spotted the following sequence in CSiBE bzip2/blocksort:

        dmulu.l r10,r8
        mov.l   r8,@(8,r15)
        sts     mach,r0
        shlr    r0
        mov     r0,r1
        add     r1,r1
        add     r1,r0
        sub     r0,r8
        tst     r8,r8
        bt/s    .L253
        mov     r8,r0

A reduced test case is:

unsigned int test (unsigned int a, unsigned int b)
{
  unsigned int x = (a * b) / 2;
  unsigned int y = a - (x * 3);
  return y;
}

compiled with -m4 -O2:

        mul.l   r5,r4
        mov     r4,r0
        sts     macl,r1
        shlr    r1
        mov     r1,r2
        add     r2,r2
        add     r2,r1
        rts
        sub     r1,r0   // r0 = r0 - (r1 + r2 + r2)
            //    = r0 - r1 - r2 - r2

which can be converted to the following, saving one register:

        mul.l   r5,r4
        mov     r4,r0
        sts     macl,r1
        shlr    r1
        sub     r1,r0
        sub     r1,r0
        rts
        sub     r1,r0