🚀 go-pugleaf

In article <1992Nov18.154754.1848@udel.edu> carroll@dori.cis.udel.edu (Mark C. Carroll) writes:
>In article <Bxr9vx.KBD@mentor.cc.purdue.edu> hrubin@mentor.cc.purdue.edu (Herman Rubin) writes:
>]In article <1992Nov15.032459.3069@udel.edu> carroll@thorin.cis.udel.edu (Mark C. Carroll) writes:
>]]In article <BxMCxA.3nM@mentor.cc.purdue.edu> hrubin@pop.stat.purdue.edu (Herman Rubin) writes:

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

>Any optimization, particularly on modern architectures, has got to
>consider its affect on pipeline fills, branch predictions, register
>allocations, memory and cache uses, and dozens of other factors.  Do
>you think that you're qualified to judge what affect your two-cycle
>savings will have on the rest of your program? If so, you're in the
>wrong field - you should be designing compilers.

I am quite aware of all of these.  In most of the cases I have discussed
in any detail, I can see ways that the language design prohibits many of
these considerations.  

As to designing compilers, this is a massive operation, not to be done by
one or two people.  And when it comes to optimization, these compiler
designers should ask for things to be considered which they do not now
consider.  This applies to language design, as well.  The people who 
seem to dominate the field assume that their considerations are ALL
that matter.

>]The semantic model of the language can, however, greatly inhibit the
>]programmer.  Try communicating floating binary from one machine to
>]another; the only semi-portable way I have seen are to convert to
>]higher precision floating decimal, transmit that, and convert back.

>What does that have to do with the semantic model of the language?
>That's a very hardware related issue. If I design a language that uses
>floating point numbers, I'm restricted to using whatever internal
>float format the hardware designer chose for the architecture I'm
>learning. That's got nothing to do with the semantic model of the
>language.

Did I say anything about matching the internal representations of
the numbers?  We can communicate binary integers across machines
and implementations; the much used notation 0ddddd for octal and
0xdddddddd for hexadecimal, bad as they are, do this explicitly.
Algol had a different way to do this, and it could handle other
bases explicitly; I believe it could also do this for non-integers,
but the syntax did not carry over well.  Most languages and 
compilers allow for explicit writing of fixed-point decimals 
and floating-point decimals.  However, there are many situations
in which people may wish to communicate non-integers in hex.  I
have found cases in which decimal is inconvenient and even leads
to confusion.

I have seen the problem addressed by coming up with a characterization
of the number of decimal digits which need to be communicated so that
a given number of bits are correctly communicated by conversion.  This
I do not find to be satisfactory, and I see no reason why this insistence
on decimal as the only means of communication is made.
-- 
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
Phone: (317)494-6054
hrubin@snap.stat.purdue.edu (Internet, bitnet)  
{purdue,pur-ee}!snap.stat!hrubin(UUCP)