Have I found my problem in synchronization for pci-6111?

Hallo,David Schleef ,

I think I have found the problem in synchronization for my pci-6111 and now
begin to change my linux Kernel.

My problem :  When synchronizing AO (through INTTRIG)and AI(EXTTRIG),  they
delayed  always so much that I can not use this new function. For example ,the
  Simultaneous start sample has followed results(no changings of the sample
program, only different freq):
____________________
when freq=1000, write times 4 > read times 3.
....................write 4096
....................write 4096
....................write 4096
....................write 4096
read 4096
read 60
read 4040
....................write 4096
....................write 4096
....................write 4096
....................write 4096
(continue)......
_____________________
when freq=10000, write times >> read times.
..write 4096
write 8
..write 4096
write 8
..write 4096
write 8
..write 4096
write 8
..write 4096
write 8
..write 4096
write 8
..write 4092
..write 4096
write 16
..write 4076
..write 4092
..write 4096
write 28
..write 4096
write 4
..write 4084
..write 4092
..write 4036
write 128
..write 4056
..write 4092
..write 4080
..write 4096
write 4
..write 4092
..write 4096
write 4
..write 4096
write 4
..write 4080
..write 4096
write 8
..write 4092
..write 4096
write 20
..write 4084
..write 4096
..write 4096
write 8
..write 4076
..write 4036
write 64
..write 4096
write 12
..write 4092
..read 4096
read 116
read 3984
write 4076
..write 4096
write 12
(continue)

....
....
And I have compared with Dr. Rolf Mueller's results( he uses pci-6110card),
see that: This example program has different performance ,the times of write
and read are always same  by his card under different freq . But my card'
results here is not so.

Then I compared my Comedi Install information with his (using comedi_test
program), and find that comedi detect the AI FIFO, but not AO FIFO by my card,
the results are only:
         ......
         I: testing cmd_write_fast_1chan...
         comedi_command: Resource temporarily unavailable
         I: testing cmd_fifo_depth_check...
         not applicable
         ....
Comedi detect much more by his card(especially AO FIFO):
          .......

           I: testing cmd_write_fast_1chan...
            can't do timed?!?
           not supported
           64,0
            can't do timed?!?
           not supported
           128,0
            can't do timed?!?
           not supported
           256,0
            can't do timed?!?
           not supported
           512,0
            can't do timed?!?
           not supported
             1025,0
            can't do timed?!?
           not supported
            2048,0
            can't do timed?!?
           not supported
             4096,0
            can't do timed?!?
           not supported
           8192,0
            can't do timed?!?
           not supported
            16384,0
            can't do timed?!?
           not supported
            32768,0
            can't do timed?!?
           not supported
             ....

I know FIFO take control of  this eyample's sychronization in background. My
pci-6111 has same FIFO Buffer as  his pci-6110(AO 2048, AI 8196). I think :
the diffrent results come from same hardware(one manul,pci-611x fast same),
same comedi ,same comedi_test program, same example program. Only Linux kernel
different, I have kernel from Redhatlinux7.3 package CD, he has it direct from
ftp.de.kernel.org.  I mean this difference of Kerne source  leads to my
problem.

Is my analyse  right?

Wang Maosen