Sandbox::https://github.com/kwccoin/thinlisp/blob/master/docs/tl-manual

Repository navigation

History

<HTML>

<HEAD>

<TITLE>ThinLisp Rant and Notes - Footnotes</TITLE>

</HEAD>

<BODY>

<H1>ThinLisp Manual</H1>

<H2>Version 1.0.1</H2>

<H2>12 October 2001</H2>

<ADDRESS>The ThinLisp Group:</ADDRESS>

<ADDRESS>Jim Allard</ADDRESS>

<P>

<P>Of course this is ignoring

other, more broadly applicable developments, such as multi-tasking operating

systems.

<P>All apologies in advance

to my father-in-law Ed (a scientist) and my brother-in-law Holger (a

mathematician) from me (an engineer).

<P>In the rest of this discussion, the scientist has gone for coffee,

uninterested because there is nothing pre-existing to study, only new things

being spun out of the air.

<P><CITE>The C Programming

Language, Second Edition</CITE>, by Kernigan, Brian W. and Ritchie, Dennis M.,

P T R, Englewood Cliffs, NewJersey 07632

<P>For consing operations that can be called from either context, you can

declare a consing area of <CODE>either</CODE>, but in practice that is fairly rare.

<P>There is an argument that they should be aligned on 8 byte

addresses. There are two advantages to 8 byte alignment. One is that it would

enable us to reliably store immediate double floats and pointers into the same

array. The second is that it would can give us an immediate type tag for conses

that is itself the offset to the cdr of a cons. The argument for 4 byte

alignment is that all current C compilers we use align structures on 4 byte

addresses, but some won't align them on 8 byte addresses, even if they contain

doubles. The other argument for 4 byte alignment is that we would not need to

occasionally skip forward 4 bytes when allocating from heap in order to find the

next 8 byte aligned address. For now, the 4 byte alignment wins.

<P>Kim Barrett recently made a convincing argument for an immediate type

tag of zero for fixnums. With this approach there is no bit fiddling needed for

fixnum numeric operations. I had originally dismissed this approach since it

would require pointer arithmatic to be done to fetch the type tag of an object.

However, since most processors have a single instruction that can fetch from an

address plus offset, this becomes a moot point. For now, the existing scheme

carries the day due to inertia. -jallard, 10/31/99

<P>The Alpha OSF has since been renamed DEC UNIX.

<P>The following comment applies only to 8 byte alignment: For

example, we could put immediate floats into odd-indexed simple vector locations,

and those floats would consume 2 elements. For structures and frames, this

could provide significant savings. -jra 8/30/95

<P>This note about classes is being written in anticipation of their

implementation. -jallard 10/29/99

<P>This latter idea of having a

"reclaimed flag" has never been implemented. -jallard 10/12/99

<P>Rest

arguments have since been added. -jallard 10/9/99.

This document was generated on 12 October 2001 using

<A HREF="http://wwwinfo.cern.ch/dis/texi2html/">texi2html</A> 1.56k.

</BODY>

</HTML>

Navigation Menu

Repository navigation

tl-manual_foot.html

tl-manual_foot.html

FilesExpand file tree

tl-manual_foot.html

Latest commit

History

tl-manual_foot.html

File metadata and controls