I have just started reading the gcc code a bit, and while looking at the avr port I noticed that in the function avr_rtx_costs in avr.c there is some code to calculate the cost of a multiplication:

    case MULT:
      switch (mode)
        {
        case QImode:
          if (AVR_HAVE_MUL)
            *total = COSTS_N_INSNS (optimize_size ? 3 : 4);
          else if (optimize_size)
            *total = COSTS_N_INSNS (AVR_MEGA ? 2 : 1);
          else
            return false;

        case HImode:
          if (AVR_HAVE_MUL)
            *total = COSTS_N_INSNS (optimize_size ? 7 : 10);
          else if (optimize_size)
            *total = COSTS_N_INSNS (AVR_MEGA ? 2 : 1);
          else
            return false;

        default:
          return false;
        }
      *total += avr_operand_rtx_cost (XEXP (x, 0), mode, code);
      *total += avr_operand_rtx_cost (XEXP (x, 1), mode, code);
      return true;

if AVR_HAVE_MUL is true, the above code will be equal to the following:

    case MULT:
      switch (mode)
        {
        case QImode:
          *total = COSTS_N_INSNS (optimize_size ? 3 : 4);

        case HImode:
          *total = COSTS_N_INSNS (optimize_size ? 7 : 10);

        default:
          return false;
        }
      *total += avr_operand_rtx_cost (XEXP (x, 0), mode, code);
      *total += avr_operand_rtx_cost (XEXP (x, 1), mode, code);
      return true;

which I belive is incorrect since it means that the assignments to *total in the switch statement are useless.  It probably doesn't break anything serious but might give slightly worse optimizations.