Jump to content

v1.9.0.737 and Intel HD 400 memory leak


Recommended Posts

Hi,

On my humble Celeron laptop, AP v1.9.0.734 works like a champ. But when I enable OpenCL it becomes unusable.

Some weird behaviour I experienced:

1. The benchmark stalls when it arrives to the third test 'Raster (Multi CPU)'. There is no activity. CPU and GPU usage drops to almost 0 and I can't close the benchmark dialog.

2. On an edited photo, I can't disable the levels layer. I can uncheck and check the checkbox but there is no change to the image. Then if I try to quit and AP hangs. I have to kill the task.

3. A couple of tools crashed on me. For example the undo brush. When I alt click, AP quits unexpectedly.

Maybe there is an issue with the OpenGL lib?

I have the latest Intel drivers from august 2020. OpenCL version in the system32 folder is 2.0.2. CPU is Celeron N3060 and GPU is HD 400.

Congratulations for this great piece of software.

Keep up the good work!

Link to comment
Share on other sites

8 hours ago, hoch said:

 

1. The benchmark stalls when it arrives to the third test 'Raster (Multi CPU)'. There is no activity. CPU and GPU usage drops to almost 0 and I can't close the benchmark dialog.

 

exactly the same if I use older drivers for HD 630, latest one from Intel solves that (but only that)

Link to comment
Share on other sites

  • Staff

Hey, we have a newer intenal build that should work around the benchmark freeze. My AMD desktop was doing the same and with the new built, it works fine. We should maybe wait until you get that to proceed as we might be flogging something that ultimately gets fixed.

Also saying that, I'm not sure what, if any gains you will get by enabling OpenCL without a discrete card.

Link to comment
Share on other sites

Hi Mark and Chris,

Thank you for your fast answer.

Chris, you gain a lot, even with this tiny GPU. 

I tried 1.9.0.767 and in my humble but silent, ultra low power and low emissions system... :-)... it's simply awesome... It's almost as having a new computer! 

The benchmarks ran ok and show a speed up of 8x when OpenCL is enabled, very impressive!

I did some test with a 20MP image and the few things I tried are definitely much faster.

For example, switching on and off the levels layer, shows the changes almost instantly. Without OpenGL you can see AP drawing tiles for a second or so, you can almost count them... :-)

But it seems that there is a memory leak. This same action, switching on and off a level layers, consumes a lot of memory. Everytime you switch on or off, the memory usage grows.

You can see it in the appended screenshots of Process Explorer.

The first one shows the memory usage when you open the image. The next one when you switch 30 times the levels layout, on and off. The second 50 times. If you keep going, AP crashes. 

When OpenGL is switched off, the same actions do not consume memory, almost, only a few Kb.

Hope you can fix this. This new OpenCL acceleration is amazing, at least on my system.

Thank you and keep up the good work, guys!

OpenCL_Open.png

OpenCL_30.png

OpenCL_50.png

Link to comment
Share on other sites

12 minutes ago, hoch said:

This same action, switching on and off a level layers, consumes a lot of memory.

I've just tried this, a 20MP image with a Levels Adjustment, enabling and disabling the Levels Adjustment doesn't change the memory usage. Could you possibly provide a video, or some other steps to reproduce?

Link to comment
Share on other sites

Hi Mark,

I append a video where you can see how memory usage grows as I switch on and off the levels layer.

This is what I did:

  1. Open a 20MP jpeg,
  2. Add a levels layer. Make some adjustments.
  3. Swith it on and off.

That's it. This does not happend with OpenCL switched off. Clearly there is something wrong with OpenCL.

  • CPU is Celeron N3060 and GPU is HD 400,
  • Intel drivers are the latest, from august 2020,
  • OpenCL version in the system32 folder, is 2.0.2,
  • Windows 10 is the latest before 2004.

If you want, I can send you the Chrome GPU report. Chrome has a few issues and some workarounds for this GPU.

 

Link to comment
Share on other sites

×
×
  • Create New...

Important Information

Terms of Use | Privacy Policy | Guidelines | We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.