HQPlayer Linux Desktop and HQplayer embedded

[asdf1000]

15 minutes ago, Miska said:

 

Then there is likely a problem with the matrix and the filter is not working for some reason. Check what HQPlayer log says about it. To create the plot, HQPlayer runs test signal through the pipeline processor.

 

 

Ok i have to try and connect a keyboard/monitor to UpBoard to get log file.

But here is the mono wav file.  What output do you get? 

 

Elegia Harman PEQ - stock MONO.wav

 

 

 

[asdf1000]

 PCM96k playing well now. No pops

 

PCM192k plays at half the speed - no pops but in slow motion 😃

 

Some progress though, with getting the clocking correct and 100ms buffer.

[asdf1000]

1 hour ago, Miska said:

 

Now it looks correct, this sample rate zone is clock slaved to the Anubis.

 

 

If running HQPlayer Desktop (CoreAudio on macOS) would you still keep Anubis on the crown/master like this:

 

[Miska]

28 minutes ago, asdf1000 said:

 

If running HQPlayer Desktop (CoreAudio on macOS) would you still keep Anubis on the crown/master like this:

 

 

Yes, that's correct...

 

[Miska]

59 minutes ago, asdf1000 said:

 PCM96k playing well now. No pops

 

PCM192k plays at half the speed - no pops but in slow motion 😃

 

Some progress though, with getting the clocking correct and 100ms buffer.

 

Likely the device sample rate doesn't change, so 192k data plays at half speed when actual rate is 96k. Just set the rate from Anubis and make HQPlayer stick to the constant rate and it should be fine.

 

[Miska]

18 hours ago, Miska said:

Tomorrow I'll check this build more on Intel CPUs too.

 

After some quick testing, the normal build performs better on Intel CPU and the new build just suffers performance-wise on Intel. So it really seems to be AMD specific version, but I was surprised how much difference it made.

 

[asdf1000]

16 hours ago, Miska said:

 

Then there is likely a problem with the matrix and the filter is not working for some reason. Check what HQPlayer log says about it. To create the plot, HQPlayer runs test signal through the pipeline processor.

 

 

See log attached. There's nothing in the log showing "plot".

 

The wav file is definitely in the pipeline because the -9dB gain is easily audible.

 

The wav file attached also

 

Elegia Harman PEQ - stock MONO.wav

 

plot error log.pdf

[Miska]

2 hours ago, asdf1000 said:

See log attached. There's nothing in the log showing "plot".

 

The wav file is definitely in the pipeline because the -9dB gain is easily audible.

 

The wav file attached also

 

And you have selected only one or two of the pipelines for plotting? You can have max two pipelines selected for the plot.

 

[maya]

16 hours ago, Miska said:

 

After some quick testing, the normal build performs better on Intel CPU and the new build just suffers performance-wise on Intel. So it really seems to be AMD specific version, but I was surprised how much difference it made.

 

You mean ver 4.22 for HQP Desktop?

How about HQPe 4.21.1.60 is it an AMD specific version?

[Miska]

Just now, maya said:

You mean ver 4.22 for HQP Desktop?

How about HQPe 4.21.1.60 is it an AMD specific version?

 

Version 4.10 of the Desktop. I will try to make a similar build of Embedded for AMD later today.

 

[maya]

13 minutes ago, Miska said:

 

Version 4.10 of the Desktop. I will try to make a similar build of Embedded for AMD later today.

 

i7 8700 doing DSD128 7ECext2 with HOPe 4.19 and 4.21.

But under 4.21 DSD256 its SQ seems ro be worse than DSD128.  

i7 8700 cannot do DSD256 but can do under 4.21.

There was big SQ enhancements in SQ with 4.19 compared with 4.14 etc.

[maya]

8 minutes ago, maya said:

i7 8700 doing DSD128 7ECext2 with HOPe 4.19 and 4.21.

But under 4.21 DSD256 its SQ seems ro be worse than DSD128.  

i7 8700 cannot do DSD256 but can do under 4.21.

There was big SQ enhancements in SQ with 4.19 compared with 4.14 etc.

 

[hifi25nl]

Where can I find version HOPe 4.19? I see only the buster version on https://www.signalyst.eu/bins/hqplayerd/

[asdf1000]

1 hour ago, Miska said:

 

And you have selected only one or two of the pipelines for plotting? You can have max two pipelines selected for the plot.

 

 

Correct, there's only 2 pipelines in total.

 

Initially tried both of them and then tried just one.

 

I am ticking "plot" checkbox for channel 1 and then below the big matrix, I press "plot"

 

HQP Desktop handles the same wav file fine and generates correct plot.

 

 

[maya]

3 hours ago, hifi25nl said:

Where can I find version HOPe 4.19? I see only the buster version on https://www.signalyst.eu/bins/hqplayerd/

Please ask Jessi, not  available on his website now.

 

[maya]

6 minutes ago, maya said:

Please ask Jessi, not  available on his website now.

 

What HOPe version now in Audiolinux ?

 

3 hours ago, hifi25nl said:

Where can I find version HOPe 4.19? I see only the buster version on https://www.signalyst.eu/bins/hqplayerd/

What version of ? HQPe in Audiolinux now ?

[hifi25nl]

Last available version is 4.22.0 based on Fedora package

On request user can install another version based on Focal package with CUDA and AMD GPU acceleration.

[maya]

34 minutes ago, hifi25nl said:

Last available version is 4.22.0 based on Fedora package

On request user can install another version based on Focal package with CUDA and AMD GPU acceleration.

I see thanks

[Miska]

6 hours ago, Miska said:

Version 4.10 of the Desktop. I will try to make a similar build of Embedded for AMD later today.

 

Now there's also AMD optimized build of Embedded available. Performance on Intel Core CPUs is equal to the regular build. Note that this version doesn't include GPU support. AVX2 support is required on the CPU.

 

Regular build requires just SSE4.2 at minimum, but supports also AVX, AVX2 and AVX-512.

 

[Miska]

One notable preliminary result is that I managed to get poly-sinc-ext2 + ASDM7EC working to DSD256 on my passive cooled i5-7600T CPU with the AMD optimized build... ;)

 

[maya]

3 hours ago, Miska said:

One notable preliminary result is that I managed to get poly-sinc-ext2 + ASDM7EC working to DSD256 on my passive cooled i5-7600T CPU with the AMD optimized build... ;)

 

After install focal 4.22.0-62 ?

Great ! 

 

[maya]

4 hours ago, Miska said:

One notable preliminary result is that I managed to get poly-sinc-ext2 + ASDM7EC working to DSD256 on my passive cooled i5-7600T CPU with the AMD optimized build... ;)

 

i5 7600T Max 3.7GHz can do DSD256 7EC-ext2 !

Great !!

[Miska]

1 hour ago, maya said:

i5 7600T Max 3.7GHz can do DSD256 7EC-ext2 !

Great !!

 

Yes, I changed multicore to "1" from "auto" and played for couple of hours without glitches.

 

Kind of black magic when something works and doesn't. It is not very straightforward or systematic. Because the very same code performs in different ways on different CPU models. And I'm not nearly as secretive as CPU manufacturers about what is going on inside!

 

Good result for a 35W TDP CPU.

 

[maya]

7 hours ago, Miska said:

 

Yes, I changed multicore to "1" from "auto" and played for couple of hours without glitches.

 

Kind of black magic when something works and doesn't. It is not very straightforward or systematic. Because the very same code performs in different ways on different CPU models. And I'm not nearly as secretive as CPU manufacturers about what is going on inside!

 

Good result for a 35W TDP CPU.

 

Great !

Passive cooling can do a better

SQ !

 

I have successfully installed 4.22.0-62 in my system under Debian 11 too.

There is further 5 % decrease in CPU usage !

 

In 4.21 it was not changed to 1?

 

If you change multicore back from auto again then No more ROCm optimization?

 

Can the user try to change this ?

If can where to do?

 

Many thanks !

 

[maya]

9 hours ago, Miska said:

 

Yes, I changed multicore to "1" from "auto" and played for couple of hours without glitches.

 

Kind of black magic when something works and doesn't. It is not very straightforward or systematic. Because the very same code performs in different ways on different CPU models. And I'm not nearly as secretive as CPU manufacturers about what is going on inside!

 

Good result for a 35W TDP CPU.

 

This one no more warning after reboot