Where should the type information be

It is not necessary that the entire opcode be decoded before the instruction is issued. One of the oddest I have seen was GEORGE, in which there was a 16 bit opcode, four of the bits used by the instruction decoder, another four by the unit to which the code was issued, and the other 8 almost bit by bit by the central processor, with only a small connection to the rest of the opcode.

As someone who has made good use of strings of bits in more than one way, I must disagree with the idea of the type being in the "numbers". However, it should be that way in buttEMBLER code, with type override.

For example, most computers have a subtract operator, which I would like to write as x = y - z, with the buttembler putting in the appropriate instruction, which is usually now something like SUBPQR u v w, where PQR denotes the type, and u v w are the arguments and the result in SOME order. I have not seen x in the middle, but I have seen three of the four other orders.

Now if one want to have the machine treat the words as of another type, it becomes x =(type) y - z; this does not convert the strings of bits, as C does, but uses the bit strings as stated.

-- This address is for information only. I do not claim that these views are those of the Statistics Department or of Purdue University. Herman Rubin, Department of Statistics, Purdue University