Date: Mon, 14 Nov 1994 21:41:37 -0500 (CDT)
From: Aaron Ucko <UCKO AT VAX1 DOT ROCKHURST DOT EDU>
Subject: Re: Paranoia
To: BILLM AT vaxc DOT cc DOT monash DOT edu DOT au
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu
Organization: Rockhurst College; Kansas City, MO

>> DJGPP (optimization):  Gets _very_ confused.  First thinks base is 4096 (!),
>> then 0.  Dies because 1 - 1/2 - (base) >= 1/2.
>
>'paranoia' should never be compiled with optimizations.

Really?  It doesn't warn against them, and optimized TC++ got a perfect score. 
But I guess TC++ doesn't exactly "optimize." :-)
>
>> DJGPP(no opt., single prec):  Crashes on computing Z^i.
>
>Whoops!  It's a long time since I began running 'paranoia'.  My memory
>is hazy, but I think that I may have had to use a replacement for the
>djgpp pow() function.  What does the current djgpp pow() function do
>if called as pow(-1.0,-2.0) ?

Haven't tested it--but my point was that DJGPP should have configured the FPU
to trap it, not trigger an exception.
>
>[stuff deleted]
>> 
>> I have not yet tested with QC or QP, as I don't have them installed right
>> now.  The general conclusion, though, seems to be that the 8087/287/387/486+
>> does _not_ round correctly in double-precision mode.
>
>As I said, 'paranoia' incorrectly reports one FLAW.  The 80x87 FPUs do
>correctly round/truncate (unless you are unfortunate enough to have a
>faulty chip ;-).  You will probably get this incorrect FLAW reported
>with any modern 'C' compiler (i.e. one which supports long doubles)
>and an 80x86 machine (where the normal convention with 80x87 FPUs is
>to use the 53 bit precision type as a double, and the 64 bit precision
>type as a long double).

Really?  And why would that be?  It comments on there being the 11 extra
digits of precision, but wouldn't that, if anything, improve accuracy??  It
certainly doesn't have a problem with the 40 extra bits in single-precision
mode.

--- Aaron Ucko (ucko AT vax1 DOT rockhurst DOT edu; finger for PGP public key) -=- httyp!
-=*=-Just because you're paranoid doesn't mean they aren't out to get you.-=*=-
Geek code 2.1 [finger hayden AT vax1 DOT mankato DOT msus DOT edu for explanation]: 
 GCS/M/S d(-) H s g+ p? !au a-- w+ v+ C++(+++)>++++ U-(S+)>++++ P+ L>++ 3(-) 
 E-(----) !N>++ K- W(--) M-(--) V(--) po-(--) Y+(++) t(+) !5 j R G tv--(-) 
 b+++ !D(--) B--(---) e>++++(*) u++(@) h!() f(+) r-(--)>+++ n+(-) y?