Re: Reference to "First-Class Data Type"

In article 92-02-108 spot@CS.CMU.EDU (Scott Draves) writes:
| > On 22 Feb 92 22:06:53 GMT, eric@tfs.COM (Eric Smith) said:
| > Would [a runtime C compiler] satisfy the definition of first class?
|
| Not in a practical sense. Creating procedures in a higher order language
| should be cheap enough that programmers can use it with impunity. The
| above scheme is so expensive that it would never be used, except in
| desperate circumstances.


Maybe, maybe not. A library C compiler (especially a simple,
not-very-optimizing one) need not be all that expensive. You might want
to check out Henry Massalin's Synthesis operating system; it creates and
compiles customized pieces of itself on the fly, in order to improve
performance.


%A H. Massalin
%A C. Pu
%T Threads and Input/Output in the Synthesis Kernel
%P 191-201
%K sosp executable data structures
%J PROC of the Twelfth SOSP
%C Litchfield Park, AZ
%D 3-6 December 1989
%O In \f2\&OSR 23\fP:5


%A H. Massalin
%A C. Pu
%T Fine-Grain Scheduling
%K synthesis
%P 91-104
%J PROC of the First WEBDMS
%C Ft. Lauderdale, FL
%D 5-6 October, 1989
%O Also published as Technical Report No. CUCS-381-88,
DCS, Columbia UNIV, 28 October 1988,
revised November 1988


%A C. Pu
%A H. Massalin
%A J. Ioannidis
%A P. Metzger
%T The Synthesis System
%R Technical Report No. CUCS-259-87
%I DCS, Columbia UNIV
%D 1987
--
Michael L. Scott
Computer Science Dept. (716) 275-7745, 5671
University of Rochester FAX 461-2018
Rochester, NY 14627-0226 scott@cs.rochester.edu
--

Return to the comp.compilers page.
Search the comp.compilers archives again.