gcl-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gcl-devel] RE: GCL bugs (decode-float, rationalize, expt)


From: Camm Maguire
Subject: Re: [Gcl-devel] RE: GCL bugs (decode-float, rationalize, expt)
Date: 12 Jan 2003 21:48:43 -0500

Greetings!  Just to followup on this, the only time one can get a GMP
illegal instruction crash is if

1) They don't configure with --enable-dynsysgmp
2) they compile on a box with one set of ISA extensions (e.g. 3dnow,
   SSE, ...)
3) they run on a box which doesn't support the ISA extensions
supported by the compiling box.

Both the solaris and x86 Linux packages I'm making under the CVS
directory either use dynsysgmp, or configure gmp3 explicitly for the
lowest denominator, e.g. by preceding the main configure and build
with 'cd gmp3 && ./configure --build=i486-unknown-linux-gnu && cd ..'
I would encourage the mingw developers of binary GCL snapshots to do
the same if compiling in GMP.

You can tell if this is the problem with a little objdump.  E.g.

objdump -d saved_gcl |grep xmm     (tests for SSE intructions)
objdump -d saved_gcl |grep pfadd,
objdump -d saved_gcl |grep pfmul   (tests for 3dNow/Athlon intructions)

Take care,


"Mike Thomas" <address@hidden> writes:

> Hi Stavros, Camm.
> 
> > Mike, can you confirm/comment on the mingw crashes?
> 
> I get no crashes and as the comparison below shows, my results echo Camm's
> exactly whether incorrect or not.  I assume Camm's were obtained from x86
> Linux.
> 
> I suppose that at the bottom of all these problems is GMP - the crashes
> probably depend on the CPU for which it was built as compared to the CPU on
> which it is running or else the version of GMP used.
> 
> The current GCL uses a lowest common denominator CPU on x86 platforms and (I
> believe) a newer version of GMP compared to GCL 2.3.6 which is probably why
> it is not crashing.
> 
> The questions which come to my mind are:
> 
> Is your GCL self built?
>   Me: yes - from CVS.  I don't have 2.3.6 at hand.
> 
> If yes, which gcc compiler are you using?
>   Me: MinGW32 gcc version 3.2.1
> 
> If no, what is the date of your binary build?
> 
> What CPU have you got?
>   Me: P4
> 
> What OS?
>   Me: XP
> 
> I suggest you try the new binary build of Windows GCL and see what happens.
> 
> Cheers
> 
> Mike Thomas.
> 
> -------------------------------------------------------------
> REPORT
> > (rationalize 1.0e-300) => Illegal Instruction (CRASH)
> 
> MIKE
> GCL (GNU Common Lisp)  Version(2.5.0) Mon Dec 30 13:56:10 EAST 2002
> Licensed under GNU Library General Public License
> Contains Enhancements by W. Schelter
> 
> Use (help) to get some basic information on how to use GCL.
> 
> >(rationalize 1.0e-300)
> 
> 6032057205060441/60320572050604408488421245431577356770502522517485057817966
> 1506
> 4961622344493727293370973578138265743708225425014400837164813540499979063179
> 1059
> 1959776695102219335509170789603485068403905907918039678834910609558429008744
> 6076
> 4137714689404772415506707531455176029312243924240295474299938241298892351581
> 4561
> 4364972941312
> 
> CAMM
> intech19:/fix/t2/camm/b/gcl$ gcl
> GCL (GNU Common Lisp)  Version(2.5.0) Thu Dec  5 08:07:35 EST 2002
> Licensed under GNU Library General Public License
> Contains Enhancements by W. Schelter
> 
> >(rationalize 1.0e-300)
> 
> 6032057205060441/60320572050604408488421245431577356770502522517485057817966
> 1506496162234449372729337097357813826574370822542501440083716481354049997906
> 3179105919597766951022193355091707896034850684039059079180396788349106095584
> 2900874460764137714689404772415506707531455176029312243924240295474299938241
> 29889235158145614364972941312
> 
> -------------------------------------------------------------
> REPORT
> >   (float (rationalize 1.0e-320)) =>
> >      1.1125369292541007E-308
> 
> MIKE
> >(float (rationalize 1.0e-320))
> 
> 0.0
> 
> CAMM
> >(float (rationalize 1.0e-320))
> 
> 0.0
> 
> -------------------------------------------------------------
> REPORT
> >  (expt 10.0 -310) => 0.0 on both 2.3.6 and 2.5.0
> >     1.0e-310 is a perfectly good number....
> 
> MIKE
> >(expt 10.0 -310)
> 
> 0.0
> 
> CAMM
> >(expt 10.0 -310)
> 
> 0.0
> 
> -------------------------------------------------------------
> REPORT
> >  (expt 10 -300) => Illegal Instruction (CRASH)
> 
> MIKE
> >(expt 10 -300)
> 
> 1/10000000000000000000000000000000000000000000000000000000000000000000000000
> 0000
> 0000000000000000000000000000000000000000000000000000000000000000000000000000
> 0000
> 0000000000000000000000000000000000000000000000000000000000000000000000000000
> 0000
> 000000000000000000000000000000000000000000000000000000000000000
> 
> CAMM
> >(expt 10 -300)
> 
> 1/10000000000000000000000000000000000000000000000000000000000000000000000000
> 0000000000000000000000000000000000000000000000000000000000000000000000000000
> 0000000000000000000000000000000000000000000000000000000000000000000000000000
> 000000000000000000000000000000000000000000000000000000000000000000000000000
> 
> -------------------------------------------------------------
> REPORT
>   (decode-float 1.0e-300) => 0.72, -1022, 1.0
> 
> MIKE
> >(decode-float 1.0e-300)
> 
> 0.66969287949141709
> -996
> 1.0
> 
> CAMM
> >(decode-float 1.0e-300)
> 
> 0.66969287949141709
> -996
> 1.0
> 
> -------------------------------------------------------------
> REPORT
>   (decode-float 1.0e-320) => 0.50, -1022, 1.0
> 
> MIKE
> >(decode-float 1.0e-320)
> 
> 0.50000000000022471
> -1022
> 1.0
> 
> CAMM
> >(decode-float 1.0e-320)
> 
> 0.50000000000022471
> -1022
> 1.0
> 
> -------------------------------------------------------------
> REPORT
> >  (integer-decode-float 1.0) =>
>      Illegal Instruction (CRASH)
> 
> MIKE
> >(integer-decode-float 1.0)
> 
> 4503599627370496
> -52
> 1
> 
> CAMM
> >(integer-decode-float 1.0)
> 
> 4503599627370496
> -52
> 1
> 
> 
> 
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gcl-devel
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah




reply via email to

[Prev in Thread] Current Thread [Next in Thread]