🚀 go-pugleaf

In article <1992Sep20.141802.4612@gdr.bath.ac.uk> mascdb@gdr.bath.ac.uk (C D Burdorf) writes:
>In article <9209181825.AA08510@thelonius.mitre.org> john@linus.mitre.org writes:
>>
>>True, but many (most?) of the "built-in" functions in a Lisp
>>implementation are neither implemented in assembly language, nor are
>>they usually open-coded into calling code. 
>
>I'm suprised people are arguing with me on this point.  I C coded
>the exponential distribution function and did a foreign function
>interface to CMU Lisp and it was many times faster than what the
>CMU common lisp compiler could produce with declares and speed at 3
>and safety at 0.  

1. The question was whether functions you write yourself can be as
fast as (equivalent) built-in functions.  And the answer is very
often "yes".

2. It happens to be the case that most built-in functions are
written in Lisp or C, not assembler (at least in most Lisps where I
know something about the implementation).

3. Neither 1 nor 2 is in any way an argument that compiled Lisp is
always as fast as hand-coded assembler.  They're not even an argument
that compiled Lisp is as fast as C.

4. Nonetheless, there are cases where Lisp is as fast, or faster 
than, C.  There are also times when compiled code is faster than
hand-coded assembler.  (Indeed, not all assembly code is designed
for max speed; sometimes readability considerations intervene.)

>I've seen over and over again cases where people can write
>assembly code that is faster than what Lisp compilers produce.
>
>HCL is a very fast (on sun3's) lisp and part of the reason is
>that much of it is written in assembly language.

Is it faster than other Lisps?  How much faster?  Is it faster
than C?

-- jd