Where should the type information be 178

....................

I cannot do it easily with the present buttemblers. I believe I could do it with buttemblers with notation designed more for users than for the convenience of designers, without resorting to lengthy expressions.

I wonder how many CS students now do anything with buttembler language except get a nodding acquaintance. Also, how many see the difficulties in using languages with the semantic complexity of Lisp.

No; I am investigating the feasibility of efficient algorithms to produce non-uniform random numbers from uniform. If these do not do as expected, this would be a test of the random number generators.

The particular code segment I posted was for the purpose of producing an "infinite precision" exponential random number. This means one for which there is no computer roundoff, etc. For most distributions, the computational complexity of generating non-uniform random numbers from uniform bits is finite random, even if the corresponding operations are not.

Frankly, I question whether this will pay in this case, but it will in others.

However, other properties I have looked at are not of this type, and again, whether they will pay may well be hardware dependent. Try making efficient use of acceptance-rejection methods on SIMD machines; I believe I can get rid of most of the problems at low cost by what looks like a slightly inefficient approach. I have even found that to often be an improvement on the usual type of machine.

I am quite capable of utilizing hardware instructions, which could only be done in the HLLs very inefficiently. If the information was given in concise mathematical terms, so would many mathematicians. -- 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