Musings

So I decided to run a memory test:

% sudo memtester 2048 1
memtester version 4.0.8 (32-bit)
Copyright (C) 2007 Charles Cazabon.
Licensed under the GNU General Public License version 2 (only).

pagesize is 4096
pagesizemask is 0xfffffffffffff000
want 2048MB (2147483648 bytes)
got  2048MB (2147483648 bytes), trying mlock ...locked.
Loop 1/1:
 Stuck Address       : testing   1FAILURE: possible bad address line at offset 0x10ec6907.
Skipping to next test...
 Random Value        : ok
 Compare XOR         : ok
 Compare SUB         : ok
 Compare MUL         : ok
 Compare DIV         : ok
 Compare OR          : ok
 Compare AND         : ok
 Sequential Increment: ok
 Solid Bits          : testing   2FAILURE: 0x00000000 != 0x01000000 at offset 0x00ec6907.
 Block Sequential    : testing  14FAILURE: 0x0e0e0e0e != 0x0f0e0e0e at offset 0x00ec6907.
 Checkerboard        : testing   4FAILURE: 0xaaaaaaaa != 0xabaaaaaa at offset 0x00ec6907.
 Bit Spread          : testing  22FAILURE: 0xfebfffff != 0xffbfffff at offset 0x00ec6907.
 Bit Flip            : testing   4FAILURE: 0x00000001 != 0x01000001 at offset 0x00ec6907.
 Walking Ones        : testing  39FAILURE: 0xfeffffff != 0xffffffff at offset 0x00ec6907.
 Walking Zeroes      : testing  26FAILURE: 0x04000000 != 0x05000000 at offset 0x00ec6907.

Done.

and discovered that there’s a stuck bit on one of the memory modules.

Things should be better now (no more random crashes) with new RAM installed. One thing I discovered along the way was that the version of memtester that comes with Fink is incapable, at least on PPC hardware, of testing more than 2GB of RAM. (Use a number larger than 2048 in the above, and you get a slew of malloc errors.)

The thing to do is download the current source, apply this patch and type

% make
% sudo make install

This will build a Universal binary, which is 64bit-enabled on G5 (and Intel Core 2/Xeon) hardware.

That’s good, because — when all the dust settles — golem will have 6.5GB of RAM. Now, if only my own memory were similarly upgradable…