-- another long message ..

Thanks to Greg and Patrick for their recent discussion (30 Jan 2001)
relating to the SARITH (Safe ARITHmetic) project:

http://groups.yahoo.com/group/smalleiffel/message/2992 - Greg
http://groups.yahoo.com/group/smalleiffel/message/2993 - Patrick

in which Patrick wrote in response to Greg:

--8<--
Greg C wrote:

> I've been thinking about the integer representation problem. I also
> took a look at the SARITH spec that was floating around last year. The
> problem I have with SARITH is that it starts to get complicated and
> ugly when Eiffel tries to mirror all of the permutations of integer
> types available in C.

Personally, I think SARITH handles integer types better than C. They have
really thought about all the consequences of combining different kinds of
integers, and they always do the right (intuitive) thing. AFAIK, this is
a first! :-)

Plus, if you want efficient unchecked versions, as provided with C,
they're still there.
-->8--

For those interested on the details of the SARITH Project, take a look
starting here:

http://www.elj.com/open-source/sarith/sarith.html
http://www.elj.com/open-source/sarith/history.txt

The project is currently stalled as the key player in the project (dd)
had to find a new job after the dotcom stock market meltdown in early
2000. Guus also got busy in his job (70 hour weeks) and I did not have
enough skills (still don't) to do the necessary work.

The status of the project as of April 2000 is described in detail here:

http://groups.yahoo.com/group/sarith/message/211

In summary, Guus had the following classes fundamentally integrated
into the SE distribution of the time (-0.77beta4):

INTEGER_08, INTEGER_16, INTEGER_32 and
UNSIGNED_INTEGER_08, UNSIGNED_INTEGER_16, UNSIGNED_INTEGER_32.

See messages:

http://groups.yahoo.com/group/sarith/message/161
http://groups.yahoo.com/group/sarith/message/160

However, there were some unresolved issues relating to the use of these
new classes as generic parameters:

http://groups.yahoo.com/group/sarith/message/180
http://groups.yahoo.com/group/sarith/message/203

Daniel had also done a lot coding on the C/assembler side of things:

http://www.elj.com/open-source/sarith/download/
-- sarith_upd_20000127.zip 30-Jan-2000 17:46 23k

This was the first step and there was much more to be done:

http://www.elj.com/open-source/sarith/sarith.html#work-to-be-done

--8<--
Modifications to the SmallEiffel source in order to:

[* Create 8-, 16-, 32-bit signed and unsigned types ]
[ -- done as described above ]

* Stick them in class structs, mindful of alignment issues

* Implement new promotional chart for the balancing rules

* Add keywords "signed", "unsigned", "saturable", "critical", etc. that
simply boil down to selecting an appropriate class, so as to not have
to manually instantiate cryptically named types.

* Testing

Low-level coding:

* Porting vc5 code to lcc-win32 (and other compilers?)
-- dd seemed quite excited by Jacob's note on intrinsics:
-- http://groups.yahoo.com/group/sarith/message/209
-- I also summarised some recent posts from the comp.compilers.lcc
-- newsgroups in relation lcc-win32 and inline assembler:
-- http://groups.yahoo.com/group/elj-win32-dev/message/449

* porting assembly code to other CPU's (DSP's?)

* Multiplication, division, modulo, left- and right-shift
saturable floating-point types

* 2, 4 and 8-pack types

* SUID probing once on boot function for class ANY or APPLICATION
MMX, SSI/3DNow! interfaces
-->8--

The ideas behind SARITH went beyond classic Eiffel and it would be
difficult to get enough support for it to be incorporated into the
SE distribution. Therefore it was always going to be hard to support
and keep uptodate with the current official SE distribution.

Anyway, hope this help give some background to the SARITH project.
It was a great project to be involved in and it even pushed (I think)
Bertrand Meyer to quantify some ideas that might find its way into the
next version of the language standard:

http://groups.yahoo.com/group/sarith/message/181

Maybe one day someone will take up the ball again .. Guus ?

Geoff

-- geoff@...
-- http://www.elj.com/elj-daily.cgi

========================================================================

A copy of the message announcing SARITH to the SmallEiffel mailing list:

--8<--
http://groups.yahoo.com/group/smalleiffel/message/1345

From: Geoff Eldridge <geldridg@p...>
Date: Fri Feb 4, 2000 2:39am
Subject: [Ann/Update] SARITH: Safe/Saturating ARITHmetic

SARITH (Safe/Saturating ARITHmetic) is an open-source project to provide
new handling in Eiffel for 8-, 16-, 32-bit integer types, both signed and
unsigned. Combinations of modulo- and saturation-arithmetic, as well as
execution-halting behaviour on overflow/out-of-boundary assignment via
require clauses, and exception-raising behaviours on both -debug and
-boost compilation modes will be provided via class-types.

The project is open-source'd (Eiffel Forum Freeware License, version 1),
and although it is presently intended for an upcoming elj-win32
test-release, use of the sources by other Eiffel compiler vendors, or
ports to other languages, are hereby formally granted and welcome.

Saturation ARITHmetic is applicable in any area of the problem domain
where the size of numerical representation is critical, the limit of
numerical representation is conceivably within reach, and where a closest
representable value in the result is preferable to a "wrap-around" result.
It is certainly not new: Ada is an example of a high-level language that
implements saturation; and a good example of saturation support in
hardware comes with MMX.

SARITH, however, goes far beyond Ada or MMX, in that, probably for the
first time in CS history, arbitrary mixtures of signed and unsigned types
in both inputs and outputs are allowed, defined, and working correctly.
Adding a signed and an unsigned, and saturating the result to an unsigned
is permitted, for example. Mixed signed and unsigned types are allowed in
all operations: addition and subtraction, assignments, relational
comparators, etc.

If, say, for an 8-bit example, p is signed 0x99 (a negative hex value) and
q is unsigned 0x99 (a larger, positive hex value), p = q returns False,
and p < q returns True.

Safe ARITHmetic is the natural complement to saturation arithmetic. It is
applicable most typically (though not always) to situations of the problem
domain where "an overflow IS a bug"; --i.e.: where we want to *detect*
rather than "adjust for" overflows.

A future goal of SARITH is to bring MMX support to the Eiffel language.
Unlike previous (inelegant) attempts to bring SIMD support to high level
languages, SARITH aims to first support saturation and other behaviours
with non-SIMD types, so as to "pave a road" to MMX in particular, and to
SIMD more generally.

The current status of SARITH is:

* 8, 16, 32-bit signed and unsigned base types have been implemented
at compiler level (these have been integrated into the -0.77beta6
release).

Not yet integrated in the compiler mods:

* Inter-type casting/saturating assignments, mixed types and sizes
(C macros).

* Relational comparisions, mixed types allowed (C macros).

* Addition/subtraction, mixed types, saturating and modulo (inline
assembly, x86).

Everything so far is 100% tested (for all possible input value
combinations).

Please note that SARITH is optimized for speed, by inlining, and by
minimizing of register usage and branch prediction cache pollution;

--code size concerns were last in the list.

We are looking for support for the SARITH project and would welcome
people with expertise in areas to help progress SARITH.

SARITH related links:

* Homepage:
http://www.elj.com/open-source/sarith/

* Mailing List:
http://www.egroups.com/group/sarith/

* Latest News:
http://www.elj.com/open-source/sarith/news/
http://www.elj.com/open-source/sarith/news/20000201.html

* History:
http://www.elj.com/open-source/sarith/history.txt

* Downloads:
http://www.elj.com/open-source/sarith/download/

* ETL3: Conformance and Conversion
http://www.egroups.com/group/sarith/181.html

* SmallEiffel upgrade patch instructions and results
http://www.egroups.com/group/sarith/161.html
http://www.egroups.com/group/sarith/160.html

Enjoy, from the Sarith Team.

Daniel Watkins - Guus Leeuw - Geoff Eldridge
dd@a... - guus.leeuw@t... - geoff@e...
-->8--
========================================================================