Difference between revisions of "GstCUDA - Performance Profiling"

From RidgeRun Developer Connection
Jump to: navigation, search
(Created page with "{{GstCUDA Page | Example 3: cudadebayer| Home| This page shows GstCUDA performance profiling. __TOC__ ==Description== Ri...")
 
Line 7: Line 7:
 
__TOC__
 
__TOC__
  
−
==Description==
+
= Glass to glass latency =  
−
RidgeRun offers ad-ons for GstCUDA. Those consists in full complete and ready to use elements that executes a specific CUDA algorithm that is integrated into the element code. Those ad-ons elements are based on the GstCUDA framework, and clearly shows the potential of this framework being used to generate a final product.
+
This wiki contains the glass to glass latency measurements results of GstCUDA simple capture and display pipelines on a TX2. It contains the results for all the possible GstCUDA (cudafilter and cudamux) configurations and uses cases.
  
 +
All the measurements were taken using the TX2 on the high-performance mode by running the following commands:
 +
<pre>
 +
sudo nvpmodel -m 0 #Reboot after running it, so changes can take effect.
 +
reboot
 +
sudo ~/jetson_clocks
 +
</pre>
  
−
Each add-on is sold separately. They are not included in the GstCUDA framework, but require it to work. So, to acquire an add-on you must have purchased GstCUDA at least once.
+
= Jetpack 3.3 - IMX274 camera 4K@60fps glass to glass latency =
 +
=Simple Capture to Display pipeline (without GstCUDA)=
 +
This measurement should be used as a reference to compare the glass to glass latency of the below pipelines with GstCUDA.
 +
* '''''Glass to Glass latency = 112.2042693 ms'''''
 +
Test pipeline:
 +
<pre>
 +
gst-launch-1.0 -v  nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
 +
</pre>
  
  
−
Over time, more ad-ons will be added. If you are interested in a new different ad-on, that isn't provided yet, please don't hesitate in [http://www.ridgerun.com/contact/ contact us].
+
== Cudafilter ==
 +
=== NVMM Direct Handling ===
 +
==== In-place:True ====
 +
* '''''Glass to Glass latency = 178.9331237 ms'''''
 +
Test pipeline:
 +
<pre>
 +
gst-launch-1.0 nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=NV12,framerate=60/1" ! nvvidconv ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! cudafilter in-place=true location=/home/nvidia/gst-cuda/tests/examples/cudafilter_algorithms/gray-scale-filter/gray-scale-filter.so ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
 +
</pre>
 +
==== In-place:False ====
 +
* '''''Glass to Glass latency = 230.3850304 ms'''''
 +
Test pipeline:
 +
<pre>
 +
gst-launch-1.0 nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=NV12,framerate=60/1" ! nvvidconv ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! cudafilter in-place=false location=/home/nvidia/gst-cuda/tests/examples/cudafilter_algorithms/gray-scale-filter/gray-scale-filter.so ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
 +
</pre>
 +
=== Unified Memory Allocator ===
 +
==== In-place:True ====
 +
* '''''Glass to Glass latency = 188.1192285 ms'''''
 +
Test pipeline:
 +
<pre>
 +
gst-launch-1.0 nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! nvvidconv ! "video/x-raw,width=3840,height=2160,format=I420,framerate=60/1" ! cudafilter in-place=true location=/home/nvidia/gst-cuda/tests/examples/cudafilter_algorithms/gray-scale-filter/gray-scale-filter.so ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
 +
</pre>
 +
==== In-place:False ====
 +
* '''''Glass to Glass latency = 306.2578894 ms'''''
 +
Test pipeline:
 +
<pre>
 +
gst-launch-1.0 nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! nvvidconv ! "video/x-raw,width=3840,height=2160,format=I420,framerate=60/1" ! cudafilter in-place=false location=/home/nvidia/gst-cuda/tests/examples/cudafilter_algorithms/gray-scale-filter/gray-scale-filter.so ! nvvidconv ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
 +
</pre>
  
  
−
 
+
== Cudamux ==
−
== Ad-ons Index ==
+
=== NVMM Direct Handling ===
−
The following index gives a detailed description of the GstCUDA ad-ons.
+
==== In-place:True ====
−
 
+
* '''''Glass to Glass latency = 145.5713375 ms'''''
−
<html>
+
Test pipeline:
−
  <div class="toc" style="font-size:80%;">
+
<pre>
−
    <ol>
+
gst-launch-1.0 -v cudamux name=cuda in-place=true location=/home/nvidia/gst-cuda/tests/examples/cudamux_algorithms/mixer/mixer.so nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=NV12,framerate=60/1" ! nvvidconv ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! queue max-size-buffers=3 leaky=2 ! cuda.sink_0 nvcamerasrc queue-size=10 sensor-id=2 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=NV12,framerate=60/1" ! nvvidconv ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! queue max-size-buffers=3 leaky=2 ! cuda.sink_1 cuda. ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
−
      <li> <a href=https://developer.ridgerun.com/wiki/index.php?title=GstCUDA_-_cudadebayer>cudadebayer</a></li>
+
</pre>
−
    </ol>
+
==== In-place:False ====
−
  </div>
+
* '''''Glass to Glass latency = 332.9231919 ms'''''
−
</html>
+
Test pipeline:
 +
<pre>
 +
gst-launch-1.0 -v cudamux name=cuda in-place=false location=/home/nvidia/gst-cuda/tests/examples/cudamux_algorithms/mixer/mixer.so nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=NV12,framerate=60/1" ! nvvidconv ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! queue max-size-buffers=3 leaky=2 ! cuda.sink_0 nvcamerasrc queue-size=10 sensor-id=2 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=NV12,framerate=60/1" ! nvvidconv ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! queue max-size-buffers=3 leaky=2 ! cuda.sink_1 cuda. ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
 +
</pre>
 +
=== Unified Memory Allocator ===
 +
==== In-place:True ====
 +
* '''''Glass to Glass latency = 136.4211149 ms'''''
 +
Test pipeline:
 +
<pre>
 +
gst-launch-1.0 -v cudamux name=cuda in-place=true location=/home/nvidia/gst-cuda/tests/examples/cudamux_algorithms/mixer/mixer.so nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! nvvidconv ! "video/x-raw,width=3840,height=2160,format=I420,framerate=60/1" ! queue max-size-buffers=3 leaky=2 ! cuda.sink_0 nvcamerasrc queue-size=10 sensor-id=2 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! nvvidconv ! "video/x-raw,width=3840,height=2160,format=I420,framerate=60/1" ! queue max-size-buffers=3 leaky=2 ! cuda.sink_1 cuda. ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
 +
</pre>
 +
==== In-place:False ====
 +
* '''''Glass to Glass latency = 197.1957698 ms'''''
 +
Test pipeline:
 +
<pre>
 +
gst-launch-1.0 -v cudamux name=cuda in-place=false location=/home/nvidia/gst-cuda/tests/examples/cudamux_algorithms/mixer/mixer.so nvcamerasrc queue-size=10 sensor-id=1 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! nvvidconv ! "video/x-raw,width=3840,height=2160,format=I420,framerate=60/1" ! queue max-size-buffers=3 leaky=2 ! cuda.sink_0 nvcamerasrc queue-size=10 sensor-id=2 fpsRange='60 60' ! "video/x-raw(memory:NVMM),width=3840,height=2160,format=I420,framerate=60/1" ! nvvidconv ! "video/x-raw,width=3840,height=2160,format=I420,framerate=60/1" ! queue max-size-buffers=3 leaky=2 ! cuda.sink_1 cuda. ! perf print-arm-load=true ! nvoverlaysink enable-last-sample=false
 +
</pre>
  
  
 
|keywords=GstCUDA add-ons,GstCUDA framework}}
 
|keywords=GstCUDA add-ons,GstCUDA framework}}

Revision as of 19:18, 15 July 2019

←

Example 3: cudadebayer

⌂
Home 		→

Home

{{{3}}}

←

Example 3: cudadebayer

⌂
Home 		→

Home

  1. Features and Limitations
    1. Supported Platforms
  2. Evaluating GstCUDA
  3. Building and Installation Guide
  4. Project Structure
  5. GstCUDA Framework
    1. libGstCUDA API
    2. GstCUDABaseFilter
    3. GstCUDABaseMISO
    4. GstCUDABaseSIMO
    5. GstCUDABaseMIMO
  6. GstCUDA Quick Prototyping Elements
    1. cudafilter
    2. cudamux
    3. cudademux
    4. cudamimo
  7. GstCUDA Algorithm Developer’s Guide
  8. GstCUDA Add-Ons
    1. cudadebayer
  9. Examples
    1. cudafilter
      1. cudafilter: NVMM direct mapping
      2. cudafilter: Unified memory allocator
    2. cudamux
      1. cudamux: NVMM direct mapping
      2. cudamux: Unified memory allocator
    3. cudadebayer
    4. opencv
      1. opencvwarp
      2. opencvfilter
  10. Performance Profiling
  11. Contact Us


Retrieved from "http://developer.ridgerun.com/wiki/index.php?title=GstCUDA_-_Performance_Profiling&oldid=24474"