I wonder if the Intel compiler is still good on speeding the compiled code. I finally got Visual Studio 2017 Professional and decided to give it a go. I was using Express 2013 for years. First tests, CPU usage on Wusik EVE V5, on a test preset. 70%. Compiled with Visual Studio using all options I could find. Still around 70%. Installed the Intel 2019 compiler free version (seems to be the same as the paid version), still 70%. So I wonder if I should bother. OR, if I’m doing something wrong (most likely). I did turn the compiler on, I see it is using the intel compiler as some messages are different, and I did turn all optimizations I could find and added extra SSE optimizations too. No changes… Maybe I’m just tired and don’t see the problem, as I had to install everything TWICE in order to make it work… aghh…
Anyway, just wondering about this. Would appreciate any tips and hints.
Best Regards, WilliamK
I tried using Express quite a few years ago, and was horrified at how slow the compiled code was, in RELEASE! The latest ‘community’ version appears to be just as good as the pro version from what I gather from other people. I don’t know anything about a separate Intel compiler, sorry.
BTW - Setting ‘Relax IEEE Compliance’ in Projucer enables /fp:fast in the compiler. ie. SSE enabled.
Thanks Dave. I will check this out right now.
Wait a minute… I just found out that it is not compiling with the Intel compiler. I have selected it, it shows as been used, but the resulting file is the same exact size as the previous one. What gives? Anyone managed to use it yet? :-o
What’s your experience with Juce and Intel Compiler?
We are thinking about trying it
I had some problems and had to gave up for now. But in the past the Intel Compiler always got me faster code.
We gave the Intel Compiler a try a few month ago but the performance improvements where neglectable and probably within margin of error of the measurement. MSVC and Clang got a lot better in optimizing code over the last years.
We were able to improve performance significantly only by rewriting a handful of the most time-consuming functions using SSE intrinsics.
Oh that’s good news. Around 3 years ago, on a codebase doing heavy number crunching whilst doing atomistic simulations of materials, the Intel compiler could consistently take 5% or so off the run time. The downsides were that it was relatively expensive and you’d hit compiler bugs every now and then, but as these simulations could take days to complete it was worth it. If you were compiling Fortran you’d consistently get a 10% improvement.
It was in the 1% range for us but I guess it depends a bit on the overall structure of the application.
But for anything less than 5% I would not go through the hassle of keeping the project running with the Intel Compiler.