Subject: Re: Performance issues
From: Ajayharsh Varikat (ajay@cadence.com)
Date: Thu Apr 10 2003 - 00:08:12 PDT
>> What other performance issues are there then? It seems to me that simulation
>> is the most important "performance" issue and if that just has to do with
>> simulation vendors doing a better job then what are we talking about?
>>
>> Steve
>>
One of the main reasons why VITAL has not been accelerated more is
related to customer demand.
At the time that the VITAL standard was introduced, mixed language simulations
were not as effective or widespread as it is now. Many VHDL designers wanted
a pure VHDL gate level simulation mechanism that simulated reasonably fast -
and this was demonstrated in the significant amount of user participation
in the standard development.
However, one of the major issues that came up was the lack of library
and tool support for VITAL. It took time for library vendors to
come out with VITAL equivalents of their Verilog libraries. Maintaining
a parallel VITAL library costs in terms of development and testing effort,
so library vendors would want to avoid doing this unless there is strong
push from users. And there may have been some sort of a catch-22 situation
here - were library and simulation vendors waited for customer demand,
while users waited for wider support and better performance.
At the same time, there was an established Verilog gate level simulation
market (which was significantly larger than the VHDL market), and tool
vendors continued investing in Verilog performance. Also, mixed language
simulation became more widely available. These reasons made Verilog
more attractive for doing gate level simulations when performance was
important, even for traditional VHDL designers. To add to that, the
effective price gap between VITAL licenses and high performance Verilog
licenses has been narrowing.
As I mentioned earlier, there is a significant optimization effort
involved in accelerating VITAL. Which means improved VITAL performance
would likely cost more per simulation seat. Would there be sufficient demand
for these higher cost VITAL simulators in the current situation? I don't
see users asking for it.
I believe these are some of the main reasons why VITAL performance is
not better than it is now. And I don't think tweaking with the VITAL
standard or developing a new one is going to help - these basic issues
would remain the same.
-ajay
This archive was generated by hypermail 2b28 : Thu Apr 10 2003 - 00:08:32 PDT