previous post
http://www.garlic.com/~lynn/2007c.html#47 SVCs

... and old email on dispatching "disabled" from long ago and far away

To: wheeler
Date: 21 January 1986, 07:11:46 CST
Subject: Dispatcher change for VM/XA

I am planning on changing the XA dispatcher to execute the SIE
instruction with I/O interrupts disabled (external interrupts will
still be enabled).  The SIE instruction is a very expensive
instruction and I want to give the virtual machine a chance to do some
productive work before taking an interrupt.  With I/O interrupts
disabled, the virtual machine will get to run until it relinquishes
control to CP or hits the end of the dispatcher timeslice.  The I/O
supervisor already uses the TPI instruction to process all pending
interrupts before returning control to the dispatcher.

Do you have any thoughts on the matter?  I have read CPDESIGN FORUM on
the IBMVM disk, so I know what has been discussed there.  Do you have
a ballpark figure for the maximum allowable dispatcher timeslice which
would allow satisfactory I/O throughput?  I am thinking about
disabling I/O interrupts for a maximum of 10 milliseconds.  Another
alternative would be to have DMKSTP set/reset the interrupt mask based
upon the observed I/O interrupt rate.

... snip ...

i.e. some amount of the stuff in the original vm370 resource manager
got modified in unusual ways over extended period of time (i.e from
decade earlier that above email)

misc. collected posts mentioning resource manager, dynamic adaptive
resource control, fair share, etc
http://www.garlic.com/~lynn/subtopic.html#fairshare

previous post
http://www.garlic.com/~lynn/2006j.html#27 virtual memory

with old email discussing SIE implementation differences between 3081
and 3090 in some detail
http://www.garlic.com/~lynn/2006j.html#email810630

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html

Reply via email to