First of all, I've compiled BOINC from CVS for a variety of Intel architectures... i686, Pentium3, Pentium4, Athlon, Athlon-XP, etc. with GCC optimizations (Currently running GCC 3.3.4 on most of my Linux boxes). They all work great and show greatly improved and much more accurate benchmark results.

I have several, admitedly old Pentium 200 and Pentium 233 machines that work ok with the stock, pre-compiled boinc_4.13_i686-pc-linux-gnu client but every version from CVS I've tried to compile on these i586 class machines has a problem.

When I compile the client from source using the CFLAGS="-march=pentium-mmx" gcc option or the CFLAGS="-march=i586" option, everything compiles fine and the benchmarks run fine and actually show much faster benchmarks than the stock precompiled version:

-----
stock boinc_4.13_i686-pc-linux-gnu:
89 double precision MIPS (Whetstone) per CPU
138 integer MIPS (Dhrystone) per CPU

Compiled from cvs:
boinc_4.13_i586-pc-linux-gnu
181 double precision MIPS (Whetstone) per CPU
157 integer MIPS (Dhrystone) per CPU
-----

A considerable improvement.

But, when I go to --update prefs, the client succeeds in contacting the server but gets this message:
[SETI@home] Message from server: platform 'i586-pc-linux-gnu' not found

-----
2004-11-03 14:37:55 [---] Starting BOINC client version 4.13 for i586-pc-linux-gnu
2004-11-03 14:37:55 [SETI@home] Project prefs: using your defaults
2004-11-03 14:37:55 [SETI@home] Host ID is 139227
2004-11-03 14:37:55 [---] General prefs: from SETI@home (last modified 2004-08-29 14:35:25)
2004-11-03 14:37:55 [---] General prefs: using your defaults
2004-11-03 14:37:55 [SETI@home] Resuming computation for result 05mr04ab.23787.31554.254826.221_1 using setiathome version 4.02
2004-11-03 14:37:55 [SETI@home] Sending request to scheduler: http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi
2004-11-03 14:37:55 [SETI@home] Scheduler RPC to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded
2004-11-03 14:37:55 [SETI@home] Message from server: platform 'i586-pc-linux-gnu' not found
2004-11-03 14:37:55 [SETI@home] Project prefs: using your defaults
-------

Furthermore, my page does *not* get updated with the new benchmark results.

So my questions are, why, and what machine architecture is the stock client compiled on that works with i586 machines?

And if it is compiled on a i686 or higher, is GCC being given an option to compile the client so that it will run on i586 machines? If so, what is it?

The reason I ask is, if I specify CFLAGS="-march=i686", the resulting client, not suprisingly, segfaults if it is run on an i586 class machine.

If BOINC simply does not support the i586 (Pentium) architecture, I haven't been able to find that information stated anywhere in the docs. Have I overlooked it or is that the case?

How do I compile a client that will work on an i586 computer!!!???

Any help appreciated!


Thanks,

Chuck Bell

---

 We Listen and Compute!
            |/
         ( q p )
------oOOo-(_)-oOOo-------
http://www.dianetics.org

Have you tried using -mcpu=i586 instead of -march=i586 (ie just using -mcpu=i586)?

Here is the definitions of -mcpu and -march from the gcc online manual:

-mcpu=cpu-type
Tune to cpu-type everything applicable about the generated code, except for the ABI and the set of available instructions. The choices for cpu-type are i386, i486, i586, i686, pentium, pentium-mmx, pentiumpro, pentium2, pentium3, pentium4, k6, k6-2, k6-3, athlon, athlon-tbird, athlon-4, athlon-xp, athlon-mp, winchip-c6, winchip2 and c3.

While picking a specific cpu-type will schedule things appropriately for that particular chip, the compiler will not generate any code that does not run on the i386 without the -march=cpu-type option being used. i586 is equivalent to pentium and i686 is equivalent to pentiumpro. k6 and athlon are the AMD chips as opposed to the Intel ones.


-march=cpu-type
Generate instructions for the machine type cpu-type. The choices for cpu-type are the same as for -mcpu. Moreover, specifying -march=cpu-type implies -mcpu=cpu-type.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

> Have you tried using -mcpu=i586 instead of -march=i586 (ie just using
> -mcpu=i586)?
>
> Here is the definitions of -mcpu and -march from the gcc online manual:
>
> -mcpu=cpu-type
> Tune to cpu-type everything applicable about the generated code, except
> for the ABI and the set of available instructions. The choices for cpu-type
> are i386, i486, i586, i686, pentium, pentium-mmx, pentiumpro, pentium2,
> pentium3, pentium4, k6, k6-2, k6-3, athlon, athlon-tbird, athlon-4, athlon-xp,
> athlon-mp, winchip-c6, winchip2 and c3.
>
> While picking a specific cpu-type will schedule things appropriately for
> that particular chip, the compiler will not generate any code that does not
> run on the i386 without the -march=cpu-type option being used. i586 is
> equivalent to pentium and i686 is equivalent to pentiumpro. k6 and athlon are
> the AMD chips as opposed to the Intel ones.
>
>
> -march=cpu-type
> Generate instructions for the machine type cpu-type. The choices for
> cpu-type are the same as for -mcpu. Moreover, specifying -march=cpu-type
> implies -mcpu=cpu-type.
>
> Ned
>

Thanks Ned, I don't know why I didn't think of that! It might work, but meanwhile, I compiled the BOINC client on an actual i686 computer but just used -march=pentium to build it and then copied the client to my i586 computers.
The client calls itself boinc_4.13_i686-pc-linux-gnu but works fine now on these i586 computers and it updates my Computer Information page(s) properly with the new and faster benchmarks. So, nevermind! I'm happy.


Cheers -

Chuck

---

 We Listen and Compute!
            |/
         ( q p )
------oOOo-(_)-oOOo-------
http://www.dianetics.org

>
> Thanks Ned, I don't know why I didn't think of that! It might work, but
> meanwhile, I compiled the BOINC client on an actual i686 computer but just
> used -march=pentium to build it and then copied the client to my i586
> computers.
> The client calls itself boinc_4.13_i686-pc-linux-gnu but works fine now on
> these i586 computers and it updates my Computer Information page(s) properly
> with the new and faster benchmarks. So, nevermind! I'm happy.
>
>
> Cheers -
>
> Chuck
>

Hi Chuck,

So was the problem that you were originally compiling on an i586 computer?

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

> Hi Chuck,
>
> So was the problem that you were originally compiling on an i586 computer?
>
> Ned


Yep, so the original question actually remains unanswered as to wether -mcpu=i586 might work, but I figure it probably wouldn't since the message from the server seems to be that it rejects clients that call themselves 'i586-pc-linux-gnu'. At least that is how I interpret it.

Chuck

---

 We Listen and Compute!
            |/
         ( q p )
------oOOo-(_)-oOOo-------
http://www.dianetics.org