From: Uwe B. <bo...@el...> - 2009-04-29 18:26:30
|
>>>>> "Cary" == Cary R <cy...@ya...> writes: Cary> --- On Wed, 4/29/09, XINGZHI WEN <xin...@ho...> wrote: >> I started using Icarus from last week, but I noticed it is slower >> than Cadence ncverilog. >> >> I understand that it is unfair to compare Icarus Vs. Cadence >> ncverilog, considering the number of developers behind them, but I >> would like to know what is the average speed ratio. In my test it is >> about 50, 80 minutes in Icarus, but 90 seconds ncverilog. >> >> Is this a normal speed I should expect from Icarus? >> >> If this is a normal speed, do you have explanation, what makes >> slowing down Icarus? Is there any option/switch that can speed up >> Icarus? Cary> Yes Icarus is slow. It works well for RTL simulation, but for gate Cary> level simulations it can be too limiting. Add to this the fact Cary> that it does not fully support all SDF delays and that timing Cary> checks are missing means I would not recommend Icarus for gate Cary> level simulations. I personally get a short term license for a Cary> commercial tool to do final gate level simulations, but all my RTL Cary> design is done with Icarus. Cary> There is no switch to make Icarus faster. At some point in time we Cary> probably need to look at this again. Steve did some work last year Cary> that made Icarus noticeably faster, but to get the order of Cary> magnitude improvement we need for reasonable gate level Cary> simulations will likely require a significant restructuring of the Cary> run time/event processing. There is much learning that needs to Cary> happen before we can tackle this problem. Compiling with ICC migth also be an option to try... -- Uwe Bonnes bo...@el... Institut fuer Kernphysik Schlossgartenstrasse 9 64289 Darmstadt --------- Tel. 06151 162516 -------- Fax. 06151 164321 ---------- |