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

Re: [MiNT] Daily freemint kernel builds



On 1/16/10 8:41 AM, Paul Wratt wrote:
2010/1/16 Mark Duckworth<mduckworth@atari-source.org>:
On 1/16/10 7:34 AM, Lars Schmidbauer wrote:
Hm, $fptest shows (no matter what CPU speed):

rounding is double, result = 0x640006, expected result = 0x640005
rounding is float, result = 0x640006, expected result = 0x640005

????

BENCH.TOS 66 Mhz: 0.810 seconds
           90 MHz: 0.620 seconds
          100 MHz: 0.565 seconds

FPUTEST.TOS pass until ~100000 without problems at 66 and 90 Mhz, then i
stopped it manually. At 100 MHz, errors are shown right from the start.
How long shall i run FPUTEST.TOS?

One more: The first block seems always to be compressed correct (at
           90 and 100 MHz):

bzip2 -vv /h/ozk-15012010.tar
   /h/ozk-15012010.tar:
     block 1: crc = 0x820d2ecd, combined CRC = 0x820d2ecd, size = 899981

bzip2/libbzip2: internal error number 1007.
...


Again, thank you all. Perhaps i'll test some other SDRAM. But now i look
how the builds are working on my Falcon.

Lars

That rounding issue happens on mine at any speed.  The fact that you get
errors on the lengthy test at 100MHz immediately suggests that your system
is not healthy at 100MHz.  Sorry but you have to crank it down.  It's
interesting that certain parts of your system are working and certain parts
aren't.  I suspect only certain opcodes are giving a bad result so you can
boot and run the system but certain math functions die.

Thanks,
Mark

Out of interest, is there a way to compile for 060 that does not use FPU?

BTW, the others who have problems at higher speeds, can it be isolated
to the FPU?

Paul


No, because most programs in easymint do not make use of fpu. Like bzip2. It's unlikely that his was compiled to even use the fpu. I don't even know if bzip2 would use fpu functions. What I have seen though is that if there is a problem somewhere else, there is a problem with the fpu too ;)

Thanks,
Mark