Re: my action item for explaining semantics of notify callbacks in SCE-MI pipes

Shabtay,

Comments ...


Shabtay Matalon wrote:
>>>>Does your proposal allow the infrastructure to yield control from
> 
> the
> 
>>>>HDL side to the SW side under any other condition?
>>>
>>>
>>>How about when the buffer implementing the pipe is full?
>>
>>johnS:
>>Precisely. i.e. a blocking function will see a try function
>>fail due to a full (or empty) pipe and so will yield at that point.
>>Until it can try again. Again, refer to the example in the
>>spec for how this is done.
> 
> 
>  [Shabtay] 
> 
> John, I want to be very precise about the conditions. Fullness of
> emptiness of the buffer by your proposal does not cause either yield or
> callback to be called. Only the fact that #elements in the pipe have
> changed established the criteria for the callback to be called, and it
> is the blocking send on full buffer or blocking receive on empty buffer
> that causes the yield (among the other causes we discussed).
> 
> Please confirm.

johnS:
Not entirely true.

Detection of fullness (or emptiness) will ultimately cause
a blocking send (or receive) operation to yield.

In the process of trying to move a requested # elements,
a blocking call will do partial updates of the pipe between
yields. When this happens it is important to notify the other
end since this represents a change in state.


> 
> Taking this discussion to the next level of details,
> 
> Assume now that at simulation start an input pipe is empty. Assume the
> HW side used first try_receive() which obviously failed. Assume now that
> the HW side now calls the blocking receive() function.
> 
> According to your proposal, given that the "the pipe state (i.e.
> #elements in the pipe) hasn't changed", the ok_to_send callback will NOT
> be called.

johnS:
That's correct. And it does not need to be because the first
attempt to do a try a send on the pipe from the C side
will be successful since the pipe is empty.

> 
> Is this behavior in line with your proposal? 

johnS:
Yes.

> 
> Actually, even if the first call at simulation start is a blocking
> receive() on an empty pipe, by your proposal ok_to_send callback will
> NOT be called.
> 
> Is this behavior in line with your proposal or do you want to create an
> exception at simulation start?

johnS:
I don't believe any exceptions are needed.

-- johnS
<eom>

> 
> 
> 



-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.