RE: [sv-ec] Eratta: Semaphore put method should be a function, not a task

I don't believe preventing starvation was the reason why put is a task.
I think that the main reason was for symmetry with the put/get methods
of mailboxes.
A mailbox's put method is a task since a bounded mailbox may cause put
to block.

	Arturo

-----Original Message-----
From: owner-sv-ec@eda.org [mailto:owner-sv-ec@eda.org] On Behalf Of
Steven Sharp
Sent: Monday, October 03, 2005 1:53 PM
To: sv-ec@eda.org; jamiel@Model.com
Subject: RE: [sv-ec] Eratta: Semaphore put method should be a function,
not a task


>From: "LaFlamme, Jamie" <jamiel@model.com>

>Are you sure that put() should never block?  The LRM isn't specific,
but
>one reasonable implementation would be for put() to execute threads
>blocked on a get() before put() returns.  That would reduce the risk of
>thread starvation.

Having the scheduler suspend a thread is not the same thing as the
thread blocking.  A Verilog scheduler is allowed to suspend a thread
pretty much any time it wants to, including within a function call,
even though a function call is not allowed to block.  Since the
scheduler can suspend a thread in a function call at any statement,
it can certainly suspend it at a put() if it likes.

Steven Sharp
sharp@cadence.com