Juce FFT vs FFTW benchmarking?

General JUCE discussion
21

Well, all I can say is that ours is definitely the shortest, cleanest code! :slight_smile:

22

Does anybody have an idea of the market share between Intel vs other CPU vendors for Windows desktop/laptop devices? Iā€™m assuming that the IPP canā€™t be used on non-intel chips.

23

It should work but might be less optimized.

  1. https://software.intel.com/en-us/forums/intel-integrated-performance-primitives/topic/310239

quote from thread:

I see, I have to repeat it again. We do very thin optimization for Intel architecture, and we canā€™t guarantee any resultsfor this optimized codeon other architecture. We have ā€œgenericā€ code, which should work on 100% compatible architecture.

In terms of compatibility,
AMD ISA should be compatible upto some SIMD
(Advanced Vector Extensions - Wikipedia)

So I guess Intel and developers gets what they want -

  • Developer, got for the average use-case, optimized performance (since Intelā€™s market share and with Mac in-mind is larger) and broad x86 compatibility.
  • Intel gets better performance making a user prefer its CPU due to software optimized for it.

Market Share:
(keep in-mind on macOS itā€™s almost zero since even machines running macOS on non-Apple machines with AMD is very rare).

Seeing some graphs online the market share is between ~ 20%-40%

https://www.cpubenchmark.net/market_share.html

2 Likes
24

Ah, thatā€™s great to know! Wow thanks so much for the info.

25

Anyone have any idea how the FFT in Max/MSP stacks up against JUCE or any of these others?