Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: foo_dsp_srcresample (Read 18226 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

foo_dsp_srcresample

This was done as a response to this completely retarded rant.

Binary
Source code

Code: [Select]
Resampling 44100 Hz WAV (total length: 1:44:23.923) to 11025 Hz

PPHS (Normal): 658.870x realtime
PPHS (Ultra): 155.373x realtime
SoX (Normal): 604.723x realtime
SoX (Best): 489.187x realtime
Secret Rabbit Code (Best): 17.187x realtime
Secret Rabbit Code (Medium): 53.297x realtime
Secret Rabbit Code (Fastest): 111.648x realtime
Secret Rabbit Code (Zero Order Hold): 1028.030x realtime
Secret Rabbit Code (Linear): 943.277x realtime


People complaining about lost funds when people will do it for *free*? And complaining about "hours and hours of work"?
Sheesh.

foo_dsp_srcresample

Reply #1 –
Developing outside your primary platform? Unheard of. That would ordinarily deserve a large cash payment, befitting of an experienced software engineer. I can't imagine developing for, much less even using, software outside of my comfort zone. Which is why I've brought both Linux and OS X into my comfort zone, and have dabbled in developing for both.

There may be problems with this, even with source released, since I believe the author of the library has shown in the past that he is strict with enforcing his GPL or commercial licensing, since he does want to make money for his work.

Such a departure from the typical FOSS attitude of suggesting that software developers should not be entitled to make any money off their work, and instead should work for McDonald's or Wal-mart or something. Just ask Richard Stallman.

foo_dsp_srcresample

Reply #2 –
Quote from: kode54 on
There may be problems with this, even with source released, since I believe the author of the library has shown in the past that he is strict with enforcing his GPL or commercial licensing, since he does want to make money for his work.


What can he do, sue me? In my current situation, that would be like trying to get blood from a stone...
Or I can be a douche and just make the source only available, leaving people to compile the code, absolving me of any responsibility. Which if that happens, I am willing to, but the source can stay up, as FB2K's SDK is BSD.

I also don't find a issue in coding OSS in spare time. Dual licensing and then basing your entire revenue on software seems kinda insane to me. Then again, the author's attitudes are completely and utterly offensive, so payback seems reasonable to me, to be honest.

foo_dsp_srcresample

Reply #3 –
Hah this is just epic and made my day, thanks mudlord.
"I hear it when I see it."

foo_dsp_srcresample

Reply #4 –
Yeah, I am honestly quite shocked at how bad it performs. And to think that was once touted as the best thing since sliced bread....
Very sad indeed. xD

foo_dsp_srcresample

Reply #5 –
Quote
this completely retarded rant.

I do hope that some day you will be able to overcome your ghetto form of expression, mudlord.
This is HA. Not the Jerry Springer Show.

foo_dsp_srcresample

Reply #6 –
What do you expect me to say? The exposition is wonderful?
Now who is being naive. I just simply cannot stand when FOSS zealots try to justify calling Windows a "legacy" operating system.

Anywayz, the component is done. Point being, use SoX or the standard FB2K resampler. Both kick this resampler's ass, even my libspeex one does too.

foo_dsp_srcresample

Reply #7 –
I know it is irrelevant in this context, but i think of reason to want money today for such an algorithm in contrast with its current free competition rather than think if this plugin violates something or not, what isn´t probably correct of course. There could be some hidden advantages (i doubt, but it isn´t important), that are significant for people, who pay for it and know what they are paying for, i suppose..
Anyway, thank for this, btw. did you use the last libsamplerate version (0.1.8)?

foo_dsp_srcresample

Reply #8 –
libsamplerate 0.1.8 is indeed used.

and really, SoX is better than libsamplerate, you are wasting your money.

foo_dsp_srcresample

Reply #9 –
Quote from: mudlord on
and really, SoX is better than libsamplerate, you are wasting your money.

Did you make some quality comparision?

Why is accent on processing power so important when talking about resamplers? - I guess simplest one could beat 'em all

Then to use it after your obvious aversion is same as like making something just be buried. - Crazy!

foo_dsp_srcresample

Reply #10 –
now you are just being a dick.

SoX has been proven many times on this forum to be more efficient than SRC....
so there.

foo_dsp_srcresample

Reply #11 –
Quote from: mudlord on
now you are just being a dick.

No, it's just you being fool again. If you do have some skills, use it to make something better if you can, instead misleading with your mockups

Quote from: mudlord on
SoX has been proven many times on this forum to be more efficient than SRC....
so there.

I guess you are talking about images at http://src.infinitewave.ca/

foo_dsp_srcresample

Reply #12 –
Quote from: klonuo on
Quote from: mudlord on
now you are just being a dick.

No, it's just you being fool again. If you do have some skills, use it to make something better if you can, instead misleading with your mockups

Quote from: mudlord on
SoX has been proven many times on this forum to be more efficient than SRC....
so there.

I guess you are talking about images at http://src.infinitewave.ca/


Not to mention the benchmark results. Which don't lie.

Quote
If you do have some skills, use it to make something better if you can, instead misleading with your mockups


How is a actual implementation of the resampler misleading? Huh? Well?
See? I can be just as obtuse, too. 

I wouldn't be surprised if you are a sockpuppet of Fidel Castro Loco.

foo_dsp_srcresample

Reply #13 –
I've no doubts the hype around SRC algorithm could be justified around a time of its origin or main improvements, when there wasn't enough higher quality free competition. However, as with everything, a progress is obvious and suddenly there are equal or better quality algorithms in terms of efficiency. I thing with licence modification, SRC could be perceived much better these days, because it is still very good but not yet miles ahead from competition.
Maybe mudlord's expression is more straight forward, personally if i'd be an author of SRC, i would be thankfull to him for testing and comparing my code with competition, unfortunately SRC isn't my effort...

foo_dsp_srcresample

Reply #14 –
From zita-resampler homepage:

Quote
Libsamplerate uses a giant fixed filter coefficient table which is interpolated at run time to find the actual filter coefficients. This allows it to perform variable ratio resampling [...] The interpolation has no significant impact on quality, but of course it takes CPU cycles.


No wonder that there are more efficient algorithms if variable-ratio resampling is not required.

foo_dsp_srcresample

Reply #15 –
I have done a simple test comparing the graphs (impulse response, frequency response, spectrogram) and speed of five foobar2000 resamplers with highest settings and ASIO4ALL resampler also. Maybe someone will find it interesting.
Link (russian, but logs are on english)
🇺🇦 Glory to Ukraine!

foo_dsp_srcresample

Reply #16 –
clickable thumbs - 180db range (the wider screen the better): http://db.tt/07oYoQ6
SoX v14.3.2 (simple -v switch)
sndfile-resample - libsamplerate-0.1.7 (-c 0)
zita-resampler 0.4.0

Quote from: mudlord on
I wouldn't be surprised if you are a sockpuppet of Fidel Castro Loco.

Shoot blanks, but everyone that used foobar SRC component before, knows about performance issue. With Mega I'm not related in any way, but I'm grateful for his work. I would make guess, based on your many times demonstrated GPL disgust, that your aggressive creation is just an attack, sourced by author's strict licensing policy, which somehow offends you. Heh, nothing new here, you said that in OP

To put it simple - you could do better in civilized way. No need for malicious approach. Take as example zita resampler rant

foo_dsp_srcresample

Reply #17 –
Quote from: Steve Forte Rio on
I have done a simple test comparing the graphs (impulse response, frequency response, spectrogram) and speed of five foobar2000 resamplers

Steve, I'd recommend to show the impulse response amplitude in dB rather than linear as this (with suitable axes annotation) gives a better idea as to the length of the filter and the potential audibility of pre- and post-echo on transients when the target nyquist freq is in the audible range—a subject of occasional discussion here at HA.

foo_dsp_srcresample

Reply #18 –
Imho adding "-w Kaiser" should give better pics when dynamic range is 180db.

foo_dsp_srcresample

Reply #19 –
Done + additionally compressed images
I'll add upsampled images later today on same link

foo_dsp_srcresample

Reply #20 –
Quote from: klonuo on
Done + additionally compressed images

What is your conclusion from the graphs?  BTW, I suspect that you've got 24-bit quantization in some graphs but not (i.e. perhaps 32-bit or double-precision) in others.

foo_dsp_srcresample

Reply #21 –
Sure professor!
Those are just images as popularly produced at http://src.infinitewave.ca/ at 180dB range. Samples were also taken from there and batch processed. From those spectrograms bare mortals can't conclude much, as tested resamplers shows great, which just proves itself if you see how all perform at 150dB range. None audible by-products
Train pulse and stereo 1KHz tone spectrograms are provided as bonus

So my conclusion is that for testing state of the art resamplers you'll need something more then sine chirp spectrograms

foo_dsp_srcresample

Reply #22 –
Quote from: klonuo on
which somehow offends you.


Somehow? Somehow?

How is one not to be offended when a GPL zealot calls my primary OS "legacy"?
Last time I checked, Windows 7 is not deprecated, thnx.

foo_dsp_srcresample

Reply #23 –
Quote from: klonuo on
So my conclusion is that for testing state of the art resamplers you'll need something more then sine chirp spectrograms

You can check the phase response too but neither SoX nor SRC has any issue there.

After that, folk tend to look for low CPU-usage and low cost, as mentioned above.

foo_dsp_srcresample

Reply #24 –
Quote from: bandpass on
After that, folk tend to look for low CPU-usage and low cost, as mentioned above.


Which applies to SoX. Efficient and good quality. I don't see SRC as efficient at least...