Yes it's SDCC syntax.


The point is to tell the compiler that it's located at a 256 byte boundary. So it does not need a full 16 bit addition of array base address and index but instead can directly load dpl with the high byte of the address and dpl with the index.

So you the 8bit indexed xdata array access is only:

   0006 85*00 82            383         mov     dpl,_ms_tenth_countdown
   0009 75 83 00            384         mov     dph,#(_table_P2_over_time >> 8)
   000C E0                  385         movx    a,@dptr

 

instead of:

   000B E5*00               386         mov     a,_ms_tenth_countdown
   000D 24r00               387         add     a,#_table_P2_over_time
   000F F5 82               388         mov     dpl,a
   0011 E4                  389         clr     a
   0012 34s00               390         addc    a,#(_table_P2_over_time >> 8)
   0014 F5 83               391         mov     dph,a
   0016 E0                  392         movx    a,@dptr

 

AFAIK you'd tell Keil the absolute location of the variable at the linking stage (which is to late for code generation so the 16 bit addition has been generated). So out of the box Keil probably cannot do it.


For SDCC you can consider using using as well. But as SDCC doesn't use R0..R7 in the proposed code snippet there is no overhead to reduce:^)