Does software need a break-in period?

[Jud]

I've seen cases where over-zeolous optimization, especially of fortran, can result in binaries that give the wrong results for calculations.

 

Incorrect calculations could well change the sound of player software.

[cjf]

"Does software need a break-in period?"

 

Only if it wasn't installed on a computer running a SSD drive powered by an outboard LPS or Battery Pack and downloaded to a NAS drive using the same tweaks

[jabbr]

With dynamic optimizations schemes such as branch prediction, timing and hence jitter are affected.

Dynamic Branch Prediction

[esldude]

With dynamic optimizations schemes such as branch prediction, timing and hence jitter are affected.

Dynamic Branch Prediction

 

We talking jitter effects in the output? What is the size and type of effect and can you show me some measurements of it?

[miguelito]

I've seen cases where over-zeolous optimization, especially of fortran, can result in binaries that give the wrong results for calculations.

If the math is right the result should be correct - within all of the rounding framework of the implementation. I've seen different platforms give different rounding results (eg linux vs windows). The differences are in the least significant figures usually, but it is a little shocking when you see it the first time. I imagine it's possible different compilers on the same platform could show the same result. But you should not see different numbers much above the resolution threshold or the "optimization" was more like they fecked up the math.

[17629]

Does a freshly installed music player sound the same as after a some time of use? Is the quality of the power supply extra critical during such a break-in period? If music files can be permanently contaminated by electrical noise during processing, then presumably the same applies to software as well.

 

You can't break in software, but SQ can deteriorate over time. With Windows, for example, the registry keeps getting bigger with use. It could put stress on the processing power. Hard drives work slower when they approach max capacity. Linux sway file may not be big enough.....

[mansr]

If the math is right the result should be correct - within all of the rounding framework of the implementation. I've seen different platforms give different rounding results (eg linux vs windows). The differences are in the least significant figures usually, but it is a little shocking when you see it the first time. I imagine it's possible different compilers on the same platform could show the same result. But you should not see different numbers much above the resolution threshold or the "optimization" was more like they fecked up the math.

 

Careless coding can cause rounding errors to accumulate until they become very noticeable. Mixing floating-point numbers of vastly different magnitudes is a common way for this to happen. Of course, if the output can be verified somehow, there's no reason to worry, and this is usually the case.

[AnotherSpin]

This is the same listening experience when upgrading another audio component. There is a change, it may not be obvious right away but give a few days of listening, then judge. I find if there's no obvious preference for the new sound, then revert to the original and start again.

 

Sometimes the change is obnoxious, like many members posts, so that change is an easy choice to revert and quickly.

 

I also find out that if I compare two osx players which have most attention here at the forum one gives better immediate or short-term impression, the other better lasting or long-term impression, imho, of course and in my set/room/etc.

[barefoot]

These kinds of discussions are always fascinating. We learn so much, not about technology, but rather the human brain's propensity to allow superstition to creep into every little nook and cranny.

 

***That was the calm and considered response, after my initial burst of laughter from reading the thread title!

[semente]

Study: $90 wine tastes better than the same wine at $10 - CNET

[AudioDoctor]

I was wondering what would be the next head exploding post here...

[17629]

Study: $90 wine tastes better than the same wine at $10 - CNET

 

 

You can find a study that claims anything you can think of. A few years ago I read a study done from someone at a major university claiming dinosaur farts contained harmful greenhouse gasses. So much, in fact, the environment couldn't handle it and they ended up farting themselves into extension. So much for the meteor.

[miguelito]

Careless coding can cause rounding errors to accumulate until they become very noticeable. Mixing floating-point numbers of vastly different magnitudes is a common way for this to happen. Of course, if the output can be verified somehow, there's no reason to worry, and this is usually the case.

Trudat.

[jabbr]

We talking jitter effects in the output? What is the size and type of effect and can you show me some measurements of it?

 

I have great data which shows jitter settling to less than 1ps after several hours playing, but you need a reclocked browser to see it.

[Allan F]

Does a freshly installed music player sound the same as after a some time of use? Is the quality of the power supply extra critical during such a break-in period? If music files can be permanently contaminated by electrical noise during processing, then presumably the same applies to software as well.

I think a far more important issue is whether the size of the head of a pin changes when the number of angels that can dance on it dance.

[jabbr]

Careless coding can cause rounding errors to accumulate until they become very noticeable.

 

The problem being that if you accumulate sloppy coding practices while you are young, then the errors accumulate in subsequent code despite the use of better practices.

[Paul R]

The problem being that if you accumulate sloppy coding practices while you are young, then the errors accumulate in subsequent code despite the use of better practices.

 

Not to worry - they are writing programs to write the programs, so human error will be totally eliminated and business won't need software professionals anymore. They will simply talk to the computers, in English, and get what they want done.

 

Just ask those Corporate CEO's and CFO's that read the mags in the back of the plan seats. They can tell ya! :)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(Gads - hope nobody actually believes that - I am being sarcastic and (hopefully) a little bit humorous.... JIC!)

[jabbr]

Not to worry - they are writing programs to write the programs, so human error will be totally eliminated and business won't need software professionals anymore. They will simply talk to the computers, in English, and get what they want done.

 

Just ask those Corporate CEO's and CFO's that read the mags in the back of the plan seats. They can tell you

 

Well unless the computer that wrote the program itself was LPS, the jitter will accumulate and keep on accumulating the longer it runs.

[Boris75]

Why is it any more unlikely than audible differences in audio files with identical check-sums?

 

Very good question indeed.

[davide256]

Software doesn't require break in. Users like me who don't always read the startup instructions do :<)

[Distinctive]

they are writing programs to write the programs, so human error will be totally eliminated

This is in fact a very good point (and I am not kidding).

Let say you find yourselves at the foot of a mountain and your destination is the mountain peek. You start off in the track in front of you, but half way up the track is split into two tracks, both heading for the top. Let say you choose the right track and arrive at the peak in 30 mins. Then half way on your way down you try the left track and discover that if you had chosen this one the first time it would take you 25 mins, saving 5.

So now you know that if time matters you will select the left track the next time (learning by trying).

(I bet a lot of you wonder where this is going.....).

 

The point is that some modern (intelligent) compilers work just like this.

If the same code base is compiled over and over again, the compiler optimize its path by learning which is generally a good thing for computation of data. Maybe it doesn't matter if one subroutine etc. is executed before another one for what you are doing, asf.

But the thing is that this may play a significant (or even huge) role in audio where one 'might' benefit from predictable routing opposed to an unpredictable outcome.

So what we might end up with is different execution times and sequences even though the compiling process itself reports 100% success each and every time.

[Jud]

This is in fact a very good point (and I am not kidding).

Let say you find yourselves at the foot of a mountain and your destination is the mountain peek. You start off in the track in front of you, but half way up the track is split into two tracks, both heading for the top. Let say you choose the right track and arrive at the peak in 30 mins. Then half way on your way down you try the left track and discover that if you had chosen this one the first time it would take you 25 mins, saving 5.

So now you know that if time matters you will select the left track the next time (learning by trying).

(I bet a lot of you wonder where this is going.....).

 

The point is that some modern (intelligent) compilers work just like this.

If the same code base is compiled over and over again, the compiler optimize its path by learning which is generally a good thing for computation of data. Maybe it doesn't matter if one subroutine etc. is executed before another one for what you are doing, asf.

But the thing is that this may play a significant (or even huge) role in audio where one 'might' benefit from predictable routing opposed to an unpredictable outcome.

So what we might end up with is different execution times and sequences even though the compiling process itself reports 100% success each and every time.

 

Explanations like this that make technical concepts accessible to laypeople really delight me (as an ignorant layperson).

[Bystander]

Does a freshly installed music player sound the same as after a some time of use?

Not necessarily. But I think it's more likely to sound worse after some time of use. It's the same concept as Windows bogging down the PC after it's been used for a while. Best to take advantage of the short window where it still sounds good. If you have good hearing, that might mean minutes only, 1hr max. So, the opposite of a break-in period.

[pipis2010]

Yes. The problem is particularly acute with DSD software. With DSD we have a single stream of a single bit doing all the work. With PCM it is divided between multiple bits grouped in pulses of discrete samples. So DSD wears away the proper edges in the signal. With time it leads to a smoothed over sound that many actually like though it is really not highest fidelity. Think of it being similar to grooves on an LP being worn smooth with repeated playing. There's a reason many refer to DSD being more analog like than PCM.

 

Aha, so that's why I love DSD so much!! Thanks!!

[gstanley75]

My OS has been breaking for decades. Break in, not so much, but a grand thought.