Why is the FPGA compiler server so slow?

[sparkymark567]

FPGA Cloud compiler is faster because NI servers are high performance, right???? Well, my Windows 7 PC is high performance, but no luck there! Total CPU usage is never more than 25% and total memory usage (including other applications) is less than 16%. Why does the compile server not use all available resources? has it been restricted? How can I leverage more power from my PC? If NI Servers can compile code quicker, there must be plausible explanation / reason? Please advise, how to make the compile server run faster locally?

[dan_u]

The Xilinx compiler is single core. See here.

To get the fastest FPGA compile time you need a high clock rate CPU, not a multicore CPU.

[sparkymark567]

OK, thanks. My PC is (3.33Ghz i5) and the compiler takes approximately 1.5 hours. If the Xlinx compiler is truely limited to 1 core, the NICloud compiler will not be much faster than my own PC, is that correct?

[dan_u · Accepted solution]

I don't know what machines the cloud compile service runs on, but I wouldn't think so.

It seems the compilation is much faster on a Linux machine than on a Windows machine (I think I heard up to 30% faster). The Linux FPGA compile server is available for LV2012. Set up a Linux machine with a high clock speed CPU and sufficient RAM and you will get about the fastest compilation possible with the current tools.

[sparkymark567]

OK, thanks for info.

[bwv1043]

Try this if you haven't yet:

http://forums.ni.com/t5/LabVIEW-FPGA-Idea-Exchange/Turbo-Boost-the-speed-of-FPGA-compiles-under-Wind...

In my experience, taking advantage of Intel's Turbo Boost (my PC has i5) reduced compile time from 85 minutes to 45.

I'm currently testing Cloud Compiler and it took 110 minutes.  How can Cloud Compile Server be the slowest compile?

[sparkymark567]

Thanks for this, will try it next time.