Well, well. I'm glad to see that there's at least one other person that's interested in this stuff besides Jan and me.

I think it was Quiz5 that also had a constant execution time, but it was nowhere as pretty or as fast as your table lookup method.

I sort of cringed at the hard-coded constants on the ADD instructions, which depend on the offsets between the MOVC instructions and the table. Talk about a maintenance nightmare! Better IMO to let the assembler calculate those offsets, something like the following.

But even this one isn't as nice as it ought to be because the "-1" term in the table offset calculations is so ugly. Can somebody suggest a good way to get rid of it?

-- Russ

00C9           Quiz14:
00C9 F8                mov     r0, a  
00CA 540F              anl     a, #0Fh
00CC 2409              add     a, #Table-Movc1-1
00CE 83        Movc1:  movc    a, @a+pc
00CF C8                xch     a, r0
00D0 C4                swap    a
00D1 540F              anl     a, #0Fh
00D3 2402              add     a, #Table-Movc2-1
00D5 83        Movc2:  movc    a, @a+pc
00D6 28                add     a, r0
00D7 22                ret

00D8 00        Table:  db      0       ; 0 - 0000
00D9 01                db      1       ; 1 - 0001
00DA 01                db      1       ; 2 - 0010
00DB 02                db      2       ; 3 - 0011
00DC 01                db      1       ; 4 - 0100
00DD 02                db      2       ; 5 - 0101
00DE 02                db      2       ; 6 - 0110
00DF 03                db      3       ; 7 - 0111
00E0 01                db      1       ; 8 - 1000
00E1 02                db      2       ; 9 - 1001
00E2 02                db      2       ; A - 1010
00E3 03                db      3       ; B - 1011
00E4 02                db      2       ; C - 1100
00E5 03                db      3       ; D - 1101
00E6 03                db      3       ; E - 1110
00E7 04                db      4       ; F - 1111