Gtuner CV Display Capture using RTX 2080 Super GPU

GCV scripting for Gtuner IV and Titan Two. Configuration, examples, questions.

Gtuner CV Display Capture using RTX 2080 Super GPU

Postby Squidward » Sun Sep 13, 2020 3:42 pm

I play Apex on pc using an Xbox One controller. I use Gtuner on the same pc to capture my display for CV. I do not have a second pc to run Gtuner CV capture. I only have one monitor. I play at 1440p@165fps.

GPU: RTX 2080 Super
CPU: Intel Core i7 9700k

When playing Apex without CV, my average framerate is ~150-165fps and fairly stable. While using CV, my average framerate is ~100-125fps and WAY less stable. If I'm understanding this thread correctly, should I be able to configure Gtuner CV to capture my display using my gpu rather than my cpu? I ask because the only reason my framerate is suffering while using CV for Apex is because my cpu is bottlenecking like crazy. My gpu is never >55% load when playing Apex but my cpu reaches 100% while running CV.
User avatar
Squidward
Lieutenant
Lieutenant
 
Posts: 397
Joined: Tue Dec 23, 2014 4:39 pm

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby J2Kbr » Mon Sep 14, 2020 10:19 pm

Squidward wrote:should I be able to configure Gtuner CV to capture my display using my gpu rather than my cpu?

You are correct. However, the script should be re-written to make use of the CUDA cores (Nvidia GPUs), instead of the CPU. You can find more information about this here:

https://opencv.org/platforms/cuda/
ConsoleTuner Support Team
User avatar
J2Kbr
General of the Army
General of the Army
 
Posts: 20323
Joined: Tue Mar 18, 2014 1:39 pm

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby Squidward » Tue Sep 15, 2020 4:49 pm

J2Kbr wrote:
Squidward wrote:should I be able to configure Gtuner CV to capture my display using my gpu rather than my cpu?

You are correct. However, the script should be re-written to make use of the CUDA cores (Nvidia GPUs), instead of the CPU. You can find more information about this here:

https://opencv.org/platforms/cuda/


Hmmm. I just read that but I am unsure where to start to reconfigure CV so Gtuner can capture with my gpu. Is it just the apex_antirecoil.py that needs to be modified to get this to work? Do you know of any guides and such?
User avatar
Squidward
Lieutenant
Lieutenant
 
Posts: 397
Joined: Tue Dec 23, 2014 4:39 pm

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby J2Kbr » Tue Sep 15, 2020 9:42 pm

After more researching it seems using CUDA is more complicated than I was initially under impression. You will need first rebuild the OpenCV library with CUDA support enabled. If you are really inclined for that, this article explains how to proceed:

https://jamesbowley.co.uk/accelerate-op ... -bindings/
ConsoleTuner Support Team
User avatar
J2Kbr
General of the Army
General of the Army
 
Posts: 20323
Joined: Tue Mar 18, 2014 1:39 pm

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby Squidward » Thu Sep 17, 2020 8:57 pm

J2Kbr wrote:After more researching it seems using CUDA is more complicated than I was initially under impression. You will need first rebuild the OpenCV library with CUDA support enabled. If you are really inclined for that, this article explains how to proceed:

https://jamesbowley.co.uk/accelerate-op ... -bindings/


Well, I guess we will see how this build goes. It's been running for almost 2 hours now. Fingers crossed that it works. Though I'm not sure what to do once it finishes building.

Code: Select all
CUSTOMBUILD : nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). [C:\Users\ICS\Documents\CV\opencv\build\modules\world\opencv_world.vcxproj]
  multiband_blend.cu
  Processing OpenCL kernels (tracking)
  Building NVCC (Device) object modules/world/CMakeFiles/cuda_compile_1.dir/__/__/__/opencv_contrib/modules/cudaoptflow
  /src/cuda/Release/cuda_compile_1_generated_farneback.cu.obj
  farneback.cu
CUSTOMBUILD : nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). [C:\Users\ICS\Documents\CV\opencv\build\modules\world\opencv_world.vcxproj]
CUSTOMBUILD : nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). [C:\Users\ICS\Documents\CV\opencv\build\modules\world\opencv_world.vcxproj]
  pyrlk.cu
  farneback.cu
  Building NVCC (Device) object modules/world/CMakeFiles/cuda_compile_1.dir/__/__/__/opencv_contrib/modules/cudaoptflow
  /src/cuda/Release/cuda_compile_1_generated_pyrlk.cu.obj
CUSTOMBUILD : nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). [C:\Users\ICS\Documents\CV\opencv\build\modules\world\opencv_world.vcxproj]
User avatar
Squidward
Lieutenant
Lieutenant
 
Posts: 397
Joined: Tue Dec 23, 2014 4:39 pm

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby jaj » Sat Oct 03, 2020 6:35 pm

Squidward wrote:
J2Kbr wrote:After more researching it seems using CUDA is more complicated than I was initially under impression. You will need first rebuild the OpenCV library with CUDA support enabled. If you are really inclined for that, this article explains how to proceed:

https://jamesbowley.co.uk/accelerate-op ... -bindings/


Well, I guess we will see how this build goes. It's been running for almost 2 hours now. Fingers crossed that it works. Though I'm not sure what to do once it finishes building.

Code: Select all
CUSTOMBUILD : nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). [C:\Users\ICS\Documents\CV\opencv\build\modules\world\opencv_world.vcxproj]
  multiband_blend.cu
  Processing OpenCL kernels (tracking)
  Building NVCC (Device) object modules/world/CMakeFiles/cuda_compile_1.dir/__/__/__/opencv_contrib/modules/cudaoptflow
  /src/cuda/Release/cuda_compile_1_generated_farneback.cu.obj
  farneback.cu
CUSTOMBUILD : nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). [C:\Users\ICS\Documents\CV\opencv\build\modules\world\opencv_world.vcxproj]
CUSTOMBUILD : nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). [C:\Users\ICS\Documents\CV\opencv\build\modules\world\opencv_world.vcxproj]
  pyrlk.cu
  farneback.cu
  Building NVCC (Device) object modules/world/CMakeFiles/cuda_compile_1.dir/__/__/__/opencv_contrib/modules/cudaoptflow
  /src/cuda/Release/cuda_compile_1_generated_pyrlk.cu.obj
CUSTOMBUILD : nvcc warning : The 'compute_35', 'compute_37', 'compute_50', 'sm_35', 'sm_37' and 'sm_50' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). [C:\Users\ICS\Documents\CV\opencv\build\modules\world\opencv_world.vcxproj]


Can you update if you got it to work? This is something I've been working on for days. The incompatibilities keep popping up. There are downloadable opencv w/CUDA Python 3.7 bindings but Gtuner needs 3.8. Only option seems to be building the 3.8 bindings
User avatar
jaj
Staff Sergeant
Staff Sergeant
 
Posts: 13
Joined: Tue Mar 31, 2020 10:50 am

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby Squidward » Fri Oct 09, 2020 7:32 pm

J2Kbr wrote:After more researching it seems using CUDA is more complicated than I was initially under impression. You will need first rebuild the OpenCV library with CUDA support enabled. If you are really inclined for that, this article explains how to proceed:

https://jamesbowley.co.uk/accelerate-op ... -bindings/


Hi J2Kbr, I just finished following and testing jaj's tutorial (here: viewtopic.php?f=15&t=16399 ) to get the Cuda Python bindings working with OpenCV. At this point, we believe that all that's left is for the Apex CV script to be modified to utilize Cuda OpenCV.

Is a Cuda-modified version of the Apex GCV script something you think you might be able to do at some point? I recognize that it's complex and will likely be a large, time-consuming task. I would love to help in any way I can. Like I mentioned before, I have an RTX 2080 Super, so I can easily test anything you need me to.
User avatar
Squidward
Lieutenant
Lieutenant
 
Posts: 397
Joined: Tue Dec 23, 2014 4:39 pm

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby J2Kbr » Mon Oct 19, 2020 7:26 pm

Yes, definitively I can take a look on this. :) -- Please be a little patient, with the new console just around the corner, I have finish some important other tasks urgently. thanks.
ConsoleTuner Support Team
User avatar
J2Kbr
General of the Army
General of the Army
 
Posts: 20323
Joined: Tue Mar 18, 2014 1:39 pm

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby Squidward » Mon Oct 19, 2020 7:33 pm

J2Kbr wrote:Yes, definitively I can take a look on this. :) -- Please be a little patient, with the new console just around the corner, I have finish some important other tasks urgently. thanks.


Alright, thank you. Any idea when you might be able to get to it? I'm just curious. :)
User avatar
Squidward
Lieutenant
Lieutenant
 
Posts: 397
Joined: Tue Dec 23, 2014 4:39 pm

Re: Gtuner CV Display Capture using RTX 2080 Super GPU

Postby J2Kbr » Mon Oct 19, 2020 7:40 pm

I will take a quick took shortly, but probably the final implementation will take more time as I will need learn how to work with CUDA (never did this before). thanks.
ConsoleTuner Support Team
User avatar
J2Kbr
General of the Army
General of the Army
 
Posts: 20323
Joined: Tue Mar 18, 2014 1:39 pm

Next

Return to Gtuner Computer Vision

Who is online

Users browsing this forum: No registered users and 50 guests