Dale, thanks for the sample IOCP.
I guess my question still stands. It seems that you don't have to
have a one to one correspondence between 'real' control units and
CNTLUNIT macros. Based on your sample, you define as many CNTLUNIT
macros as you need to satisfy your number of devices where each
control unit can have 256 devices.
The way these statements look on paper do not appear as good from a
performance point of view. This may not matter but it seems that
each set of 256 devices are accessible only thru one control unit.
For example, something like this:
CU#800 CNTLUNIT CUNUMBR=800,
CUADD=0,
PATH=(80,E9,88,E5),
LINK=(F8,FB,F8,FB),
UNIT=2105,
UNITADD=((00,256))
DA#5000 IODEVICE ADDRESS=(5000,256),
CUNUMBR=(800),
STADET=Y,
UNIT=3390
As compared to something like this for the old 3990/3390 control units:
CNTLUNIT CUNUMBR=A00,PATH=(07,13),UNIT=3990,
UNITADD=((00,256))
CNTLUNIT CUNUMBR=A01,PATH=(08,14),UNIT=3990,
UNITADD=((00,256))
CNTLUNIT CUNUMBR=A02,PATH=(15,09),UNIT=3990,
UNITADD=((00,256))
CNTLUNIT CUNUMBR=A03,PATH=(16,0A),UNIT=3990,
UNITADD=((00,256))
IODEVICE ADDRESS=(A00,256),CUNUMBR=(A00,A01,A02,A03),
UNIT=3390,UNITADD=00
In this case, the devices have four control units backing them rather
than just one.
Can someone shed some light on this?
Thanks.
