1. I can't remember learning this specifically but for some reason I think you are not supposed to use blocking in sequential blocks? If so why is this? 2. I was taught that if blocking is used, but would still be functionally the same as non blocking, it synthesizes the same as non blocking, i.e a single = statement within an always block. I answered this way in an interview once and was told I was wrong, that they do synthesize differently. So, which is it? How does synthesis handle blocking vs non-blocking?
Trevor H. wrote: > 1. This rule applies to variables driven and read by different processes. The core of the issue is in Verilog's scheduling semantics (cf. IEEE1800-2012, sec. 4.4 “Stratified event scheduler”). Blocking assignments take place in the Active region, that is as soon as the corresponding statements are executed. Non-blocking assignments will be scheduled into the NBA region, that is after all statements have executed.¹ IOW, the non-blocking assignment ensures that all processes that are looking at a variable will see the value it had before any changes in the current time step. Much like a VHDL signal (but not quite the same). > 2. As far as a single assignment statement is concerned, I'd say this is true. But if there is an assignment before reading back the variable, blocking vs non-blocking does synthesize differently. Footnotes: ¹ Simplified. There are loopholes. Verilog gives you enough rope to hang yourself.
Reply
Please log in before posting.