Board index Linux

DIFFERENT RESULTS: I am confused!

DIFFERENT RESULTS: I am confused!

Post by I. Ahma » Fri, 28 Mar 1997 04:00:00



Hello folks:

Correct if I am making wrong assumption. As far as I know, c and c++ are
portable and well they seem to be portable in my case but giving me different
results in Linux and SunOS.

I am running my program on two different platform, Linux slackware 3.0, version
1.3.18 using gcc 2.7.1 and sunos 4.1, gcc 2.6 (?). For the exact same input, I
am getting entirely different results. My programs involve some mathematical
calulations using real numbers. The exact same program compiles/runs fine
without any problem but produces different results. At this point I have not
spend time to figure out which of the two results are better and don't know
why.

Is there a difference in precision in linux and SUNOS? I am confused and don't
know where exactly to look for the problem. Any pointers? I cannot attribute
the difference in results on different versions of gcc.

I will appreciate any response, preferrable by email as I cannot read this

-------------------------------------------------------------------------------
         ___              Imran Ahmad                          ___
       ( . . )            Dept. of Computer Science          ( . . )
       _\ ` /_            Wayne State University             _\ ` /_
      (_  .  _)           Detroit, MI 48202                 (_  .  _)
       / ___ \            Phone (off): (313)577-6771         / ___ \

        -----      Home page: http://www.cs.wayne.edu/~ima    -----

 
 
 
Top

DIFFERENT RESULTS: I am confused!

Post by Chung-Yen Cha » Thu, 03 Apr 1997 04:00:00


Hi,
        I had similar problems of the same C++ program generating
different results. However, in my case, it was two different compilers
running on the same HP station in school. I believe it must be something
with the compiler rather than OS. I could be wrong though. In my case I
was lucky because one set of results generated by the "CC" compiler is
totally unreasonable so I knew I should stick with "gcc". Try to find a
reliable system and compare the results so you know which one has
problem.

: Hello folks:

: Correct if I am making wrong assumption. As far as I know, c and c++ are
: portable and well they seem to be portable in my case but giving me different
: results in Linux and SunOS.

: I am running my program on two different platform, Linux slackware 3.0, version
: 1.3.18 using gcc 2.7.1 and sunos 4.1, gcc 2.6 (?). For the exact same input, I
: am getting entirely different results. My programs involve some mathematical
: calulations using real numbers. The exact same program compiles/runs fine
: without any problem but produces different results. At this point I have not
: spend time to figure out which of the two results are better and don't know
: why.

: Is there a difference in precision in linux and SUNOS? I am confused and don't
: know where exactly to look for the problem. Any pointers? I cannot attribute
: the difference in results on different versions of gcc.

: I will appreciate any response, preferrable by email as I cannot read this

: -------------------------------------------------------------------------------
:          ___              Imran Ahmad                          ___
:        ( . . )            Dept. of Computer Science          ( . . )
:        _\ ` /_            Wayne State University             _\ ` /_
:       (_  .  _)           Detroit, MI 48202                 (_  .  _)
:        / ___ \            Phone (off): (313)577-6771         / ___ \

:         -----      Home page: http://www.cs.wayne.edu/~ima    -----

--
Chung-yen Chang


 
 
 
Top

1. diff -r return different results on different runs using DMA

I experience some errors on DMA transferts (write but also read) that
do not happen when I switch DMA off (hdparm -d 0)

I copy a whole CDROM on a partition of my disk /dev/hdb3 using tar

cd /mnt/cdrom;tar cf - . | (cd /mnt/hdb3;tar xvf -)
then I copy again the content of /dev/hdb3 on /dev/hdg3
cd /mnt/hdb3;tar cf - . | (cd /mnt/hdg3;tar xvf -)

then I run a diff -r /mnt/hdb3 /mnt/mdg3
and I find many differences. The differences are real, because od -x
confirm them. Generally it's just one byte in each file concerned
(different)

The same diff -r run many times gives differents results. In fact the
number of differences tends to increase every time.

This is also true in a diff -r between /mnt/cdrom and /mnt/hdb3

My mother board is a Abit KG7 Raid with an AMD 761/VIA 686 Chipset.
My PCU is a duron 1200 MHz running on a 100MHz front bus clock.
I use both the primary IDE controllers and the HPT 37x ones (1 disk on
each)
I have 2 128MB DDR RAMs I tested successfully with memtest one night
long.
So both my L2 cache and memory seems to be ok.
My disks also because everything works fine without DMA.

I tried the test with kernels 2.6.0, 2.4.23 and the 2.4.20-8 comming
with the redhat 9 and a redhat 9 fresh install. I also try it with a
debian-woody and a 2.2.x kernel, same result. Same thing with a
knoopix 3.3
Same thing with both ext2 and ext3 file system. I tried the test on
every disk (I have 4)... same same.

Is anybody would have experienced this kind of problem ?
IS there any tools that can provide a revelant test of ultra DMA ?

any advice is welcome, I'm ready to change my board ;-)

2. Floppy overrun/underrun SLS???

3. Different winmanagers - different results (motif)

4. ethernet via ppp0 gateway flakey

5. Same command, same os, different machines, different results

6. switching window managers

7. different browser produces different result

8. Device driver fo SCSI medium changer under Solaris 2.

9. Confused by 'free' command results.

10. i am confused: DIP, SLIP, PPP, TCP ....

11. I am confused about Sendmail.

12. I was cracked, and am somewhat confused...

13. Disk Recovery and lde: Am I horribly confused?


All times are UTC
Board index