Re: [clisp-list] *-float-*-epsilon in clisp

[Sam S. <sd...@gn...>]

> * In message <3C6...@us...>
> * On the subject of "[clisp-list] *-float-*-epsilon in clisp"
> * Sent on Sun, 10 Feb 2002 15:14:42 +0100
> * Honorable Stefan Kain <sm...@us...> writes:
>
> I get discrepancies for the double-float-*-epsilon values
> on my machine.

this is specific to linux on i386 (the fpu uses 80 bit floats).

apply the appended patch and rebuild _both_ lisp.run and lispinit.mem

please report your experiences

-- 
Sam Steingold (http://www.podval.org/~sds) running RedHat7.2 GNU/Linux
Keep Jerusalem united! <http://www.onejerusalem.org/Petition.asp>
Read, think and remember! <http://www.iris.org.il> <http://www.memri.org/>
The early worm gets caught by the bird.


--- lisparit.d.~1.30.~	Mon Nov 19 18:49:08 2001
+++ lisparit.d	Mon Feb 11 15:07:07 2002
@@ -2253,6 +2253,8 @@
         var object obj;
         #ifdef intQsize
         encode_DF(0,-DF_mant_len,bit(DF_mant_len)+1, obj=);
+        #elif (defined(unix) && defined(linux) && defined(i386))
+        encode2_DF(0,-DF_mant_len,bit(DF_mant_len-32)|bit(DF_mant_len-32-DF_exp_len),1, obj=);
         #else
         encode2_DF(0,-DF_mant_len,bit(DF_mant_len-32),1, obj=);
         #endif
@@ -2262,6 +2264,8 @@
         var object obj;
         #ifdef intQsize
         encode_DF(0,-DF_mant_len-1,bit(DF_mant_len)+1, obj=);
+        #elif (defined(unix) && defined(linux) && defined(i386))
+        encode2_DF(0,-DF_mant_len-1,bit(DF_mant_len-32)|bit(DF_mant_len-32-DF_exp_len),1, obj=);
         #else
         encode2_DF(0,-DF_mant_len-1,bit(DF_mant_len-32),1, obj=);
         #endif