SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Advanced Micro Devices - Moderated (AMD)
AMD 248.09+6.2%10:34 AM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: muzosi who wrote (227336)3/3/2007 3:11:09 PM
From: Ali ChenRead Replies (2) of 275872
 
"i frequently find myself running ise & ncverilog simultaneously. that never happened to you? or running multiple ise sessions? or running a matlab simulation & ncverilog at the same time?"

There is a misconception. I also frequently, very frequently find myself with open Xilinx ISE, ModelSim, iMpact, PowerPCB, Internet Explorer, Acrobat with several documents, maybe some Word and/or Excel, and Outlook. But I rarely RUN them simultaneously.

I don't know about you, but I am an ordinary Human. Humans are sequential thinking machines. If I made some changes to RTL, I need to make sure that it compiles into netlist without errors. Then I need to make sure that the changes do what I wanted them to do, and that they do not break other functionality. So I have to run behavioral simulations. It takes time. Then I need to map and route the design, which takes time too, sometimes too much depending on design size, chip utilization level, and severity of timing constraints.

Then I have to run back-annotated simulations, to make sure that my timing didn't fall apart, and have to go back and change constraints if necessary, then repeat place and routing. Then, if hardware is already available, I need to load binaries and run physical experiments with the design. Then I find something wrong, or move on to implementation of another hardware feature, and the cycle repeats. Each step of this cycle depends on previous step. If I will try to make another change while the cycle is incomplete and start another cycle, I will quickly lose focus on which step verifies which variant, and the whole design would fall apart and likely never function properly. I am not a superscalar computer and cannot track all conditions for every step in a pipeline of inter-dependent changes, and make focused decisions that each step requires.

As you see, overall productivity depends on how quickly I can get an individual application to get the job done. Of course, multiple cores do help, but mainly in managing Windows [ir]responsiveness :-(. Sometimes two real apps can overlap in time, but usually they are different apps. Running several copies of the SAME application is not a characteristic workload for a workstation ever. That's why I declare once again: for bulk computing, SPECbase = YES!, SPECrate = NO. Servers are probably different, but I stressed many times that I speak about bulk computing and HPC, where CPU manufacturers get majority of their revenue AFAIK.

Cheers,

- Ali
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext