[Top][All Lists]

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

Re: [Gcl-devel] More Code Cleanup...

From: Camm Maguire
Subject: Re: [Gcl-devel] More Code Cleanup...
Date: 11 Jan 2002 13:11:52 -0500

Greetings, and thanks for this work here.  I agree with you that we
should work with this gmp stuff as is in the short term, and your
contributions are much appreciated in this regard!  Over the medium
term, it would be much to our advantage to get gcl working with gmp
linked in as a dynamic shared library.  Would this work on vms?  I.e.,
have you tried building the full gmp lib package?  As far as I can
tell, there is just one garbage-collecting-related change WFS made to
the sources, and if/when we get the Boehm gc working, that may no
longer be needed, not to mention the possibility that the gmp
maintainers might be persuaded to adopt the change WFS submitted to
them.  We should follow up on this.  Anyone attracted to this project
is encouraged to step forward!

Take care,

Robert Alan Byer <address@hidden> writes:

> >
> >> Also, while compiling gcl/gmp/random.c under OpenVMS I got the following
> >> warning.
> >>
> >>      return mrand48 () | (mrand48 () << 32);
> >>    .......................^
> >>    %CC-W-SHIFTCOUNT, In this statement, the shift count "32" is negative 
> >> or is
> >>    greater than or equal to the promoted size of the operand 
> >> "mrand48(...)".
> >>    at line number 38 in file DISK$WORK:[TEST.GCL.GMP]URANDOM.H;1
> >>
> >> and also got the same above error when compiling gcl/gmp/mpz/random.c
> >>
> >
> >This looks a little alarming.  It would appear that there is some
> >fixed size misconfiguration going on.  My 32 bit machine does not
> >select this version of random.c, so I don't have this code in my
> >build.  Perhaps you can see what type mrand48 is declared, and why the
> >code thinks it can shift it be 32 bits.  (It would seem that a 48bit
> >object would be a likely candidate.  But apparently the compiler does
> >not agree.)
> >
> I've tracked this down and have fixed it.  The reason was because on
> Tru64 Alpha the "random()" function returns a double and it was shifting
> it down to 32-bits.
> I just changed the #ifdef to check for OpenVMS and if so igore that part of 
> the code and have it use another definition, which it does (just like the VAX
> version does) and I don't get the warning.
> To MAKE SURE that I didn't break anything in the process I recompiled
> everything on OpenVMS VAX and OpenVMS Alpha and ran the tests and they
> all passed and no compiler warnings.
> Here's a summary...
> MPN Test:     OpenVMS Alpha:          OpenVMS VAX:
> ADDMUL_1      PASS                    PASS
> ADD_N         PASS                    PASS
> COPY          PASS                    PASS
> DIVMOD_1      PASS                    PASS
> LSHIFT                PASS                    PASS
> MUL_1         PASS                    PASS
> RSHIFT                PASS                    PASS
> SUBMUL_1      PASS                    PASS
> SUB_N         PASS                    PASS
> TST-ADDSUB    PASS                    PASS
> (Note: The MPN tests arn't built on Tru64 so I don't know if they work)
> MPZ Test:     Tru64 Alpha:    OpenVMS Alpha:          OpenVMS VAX:
> T-MUL         PASS            PASS                    PASS
> T-TDIV                PASS            PASS                    PASS
> T-TDIV_UI     PASS            PASS                    PASS
> T-FDIV                PASS            PASS                    PASS
> T-FDIV_UI     PASS            PASS                    PASS
> T-GCD         FAIL            FAIL                    FAIL
> DIVE          PASS            PASS                    PASS
> T-SQRTREM     FAIL            FAIL                    FAIL
> CONVERT               PASS            PASS                    PASS
> IO            PASS            PASS                    PASS
> LOGIC         PASS            PASS                    PASS
> BIT           PASS            PASS                    PASS
> T-POW         FAIL            FAIL                    FAIL
> T-POWM                FAIL            FAIL                    FAIL
> T-POWM_UI     FAIL            FAIL                    FAIL
> T-2EXP                PASS            PASS                    PASS
> REUSE         FAIL            FAIL                    FAIL
> T-ROOT                PASS            PASS                    PASS
> T-JAC         FAIL            FAIL                    FAIL
> T-BIN         FAIL            FAIL                    FAIL
> T-MISC                PASS            PASS                    PASS
> So the OpenVMS Alpha and OpenVMS VAX versions both PASS all the
> tests that the Tru64 Alpha GMP tests pass (as well as others have reported 
> to pass).
> It would be good to determine which of the above tests are good indications
> of a working GMP library.
> So this is good news as it now looks like we'll have the VAX platform for
> OpenVMS as well as Alpha for GCL.
>  +------------------+--------------------------+---------------+
>  | Robert Alan Byer | address@hidden | ICQ #65926579 |
>  +------------------+--------------------------+---------------+
>  | Send an E-mail request to obtain a copy of my PGP key.      |
>  +-------------------------------------------------------------+
>  | "It is by caffeine alone I set my mind in motion.  It is by |
>  |  cans of cola the thoughts aquire speed, the hands acquire  |
>  |  shakes, the shakes become a warning.  It is by caffeine    |
>  |  alone I set my mind in motion."                            |
>  +-------------------------------------------------------------+
> Version: 2.6.3ia
> Charset: noconv
> iQEVAwUBPCUrJKVSqzlBVJbBAQH3pggAxMkxinuMza85/cgkJ9P3c1hwi/u2Ztlr
> GKW9ovTzLpNYUrZ5IPK6DZkJ7Qr1MhuoBYJef0TSlFyWhBPX6lc4GbFleqbh2UVb
> hwVnsVoVDiXmZOfw0vYut9XNNQW5R51r2AKtOs2NS6XZ0dk3bSlH0T6y9pr9j6Ja
> 61z7dHAqAf7m7a0cG6moEUgrcgrhdgzyTLJlDB8FGb8KoRyxDeMrvS5bSHgHmMZp
> jSC031r5mZ9gUfgL8+VtVnJReNjX+ZQsGG2GzrkFjwJKDMT0C+bk6A==
> =N0/X

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]