Jump to content

Holoscan Framework/Holoscan Sensor Bridge/Performance: Difference between revisions

Line 153: Line 153:
On the other hand, an important consideration is that CUDA ISP does not offer RGBA64, needed for the Holoviz. This integrates the necessary conversions for RGBA64 to RGBA32 back and forth.
On the other hand, an important consideration is that CUDA ISP does not offer RGBA64, needed for the Holoviz. This integrates the necessary conversions for RGBA64 to RGBA32 back and forth.


The following table highlights the results by using the Holoscan Sensor Bridge, the NVIDIA Jetson AGX Orin and the IMX274 imager:
The following tables highlight the configurations and results by using the Holoscan Sensor Bridge, the NVIDIA Jetson AGX Orin and the IMX274 imager:


<center>
{| class="wikitable"  
{| class="wikitable"  
|-
|-
! Pipeline
! Configurations
! Image Dimensions (px)
!  
! CPU Usage (%)
! RAM Usage (MiB)
! GPU Usage (%)
! Glass-to-Glass Latency (ms)
|-
|-
| Baseline
| DP
| 3840x2160 (4K)
| Display Port
| 2.56 (total) 30.6 (core)
| 347
| 14
| 96
|-
|-
| CUDA ISP (RGB24)
| HDMI
| 3840x2160 (4K)
| HDMI Port
| 2.96 (total) 35.6 (core)
| 401
| 35
| 58
|-
|-
| CUDA ISP (RGBA64)
| JC
| 3840x2160 (4K)
| Jetson Clocks
| 3.2 (total) 39.9 (core)
|-
| 448
| ED
| 32
| Exclusive Display
| 73
|-
| NED
| Non Exclusive Display
|-
| MAXN
| Maximum Power
|-
| CISP-GW
| CudaISP GrayWorld
|-
| CISP-HS
| CudaISP HistogramStretch
|-
| DBOnly
| Only Debayer without ISP
|}
|}
</center>
<center>
{| class="wikitable" style="text-align:center; vertical-align:bottom;"
|- style="font-weight:bold;"
! style="text-align:left;" | Configuration
! Mean Latency (ms)
! Uncertainty (ms)
|-
| style="text-align:left;" | Baseline
| 41.61
| 8.35
|-
| style="text-align:left;" | DP+ED+MAXN+JC+CISP-HS
| 37.93
| 8.35
|-
| style="text-align:left;" | DP+ED+MAXN+JC+CISP-GW
| 49.76
| 8.35
|-
| style="text-align:left;" | DP+ED+MAXN+JC+DBOnly
| 35.96
| 8.35
|}
</center>


This involves configuring the Jetson into a maximum performance mode. With the CUDA ISP in 24-bit colour depth. The latency lowered from 96 to 58 ms by optimizing the pipeline, leading to a 39.5% reduction. By preserving the 64-bit colour colour depth, the latency is still better, leading to 73 ms of glass-to-glass latency.
This involves configuring the Jetson into a maximum performance mode. With the CUDA ISP in 24-bit colour depth. The latency lowered from 96 to 58 ms by optimizing the pipeline, leading to a 39.5% reduction. By preserving the 64-bit colour colour depth, the latency is still better, leading to 73 ms of glass-to-glass latency.
2,070

edits

Cookies help us deliver our services. By using our services, you agree to our use of cookies.