Python 1300 Sensor Video Processing in HW with EdkDSP 8xSIMD Accelerator for TE0720-03-2IF SoM on TE0701-05 Carrier
Key features
This application note describes use of an evaluation package with these demos:
- 3 edge detection video processing designs (sh01, sh02, sh03) with separate HW accelerated data paths.
- These demos document the possibility to define different HW paths by different source C/C++ functions. This is important for covering of the borders lines of the parallel processed parts of the frame.
- HW accelerators can be programmed for the number of processed micro-lines.
- These demos enable effective, synchronised parallel execution with ARM C user code.
- 1 motion detection video processing.
- This demonstrated the pipelined parallel execution of HW video processing accelerators.
- HW accelerators work with fixed number of processed micro-lines (1024 micro-lines).
All demos work in parallel with the 8xSIMD EdkDSP run-time reprogrammable floating point accelerator.
- C programs can be compiled for the MicroBlaze and for the EdkDSP accelerator and used in the accelerator, without need to re-compile the design in Vivado 2015.4.
- C programs for the MicroBlaze processor and for the EdkDSP accelerator can be edited in the same SDK 2015.4 environment.
All demos are designed for the Trenz TE0701-05 platform with industrial grade Zynq XC7Z020-2I device on System on Module TE0720-03-2I.
All demonstrated video processing algorithms have been developed, debugged and tested in Xilinx SDSoC 2015.4 environment.
SW algorithms have been compiled by Xilinx SDSoC 2015.4 system level compiler (based on the Xilinx HLS compiler) to Vivado 2015.4 HW projects, and compiled by Vivado 2015.4 to the bitstreams for Zynq XC7Z020-2I device.
Created SW access functions controlling the HW accelerators have been exported from the Xilinx SDSoC 2015.4 projects to the Xilinx SDK 2015.4 SW C projects as static .a libraries for standalone ARM Cortex A9 processor.
Main objectives
- To demonstrate how to install, compile, modify and use the enclosed SW projects in the SDK 2015.4.
- To demonstrate the HW accelerated video processing algorithms and the energy per pixel reduction in comparison to the original SW versions.
- To demonstrate parallel execution of predefined video processing HW paths with C user code on ARM.
- To demonstrate HW accelerated video processing working in parallel with the 8xSIMD EdkDSP run-time re-programmable floating point accelerator.
See Sections 5 of this application note for specification of deliverables and license details.
Package Summary
| Title | Python 1300 Sensor Video Processing in HW with EdkDSP 8xSIMD Accelerator for TE0720-03-2IF SoM on TE0701-05 Carrier |
| Filename | t20i2pm2_V54_IMPORT.zip t20i2pm2_2015_4_te0720.pdf |
| License | t20i2pm2_2015_4_te0720.pdf for licensing conditions. |
| Package content | ZIP archive with precompiled Vivado 2015.4 projects demonstrating Python 1300 Sensor Video Processing in HW with EdkDSP 8xSIMD Accelerator for TE0720-03-2IF SoM on TE0701-05 Carrier. |
| Size | ZIP file: 25068215 Bytes PDF file: 3085814 Bytes |
| Required tools & platform |
Xilinx SDK 2015.4 |
| Installation notes | See application note |
Result Category
| Project number | Year | RIV category | Comment |
| 7H14004 | 2016 | Gfunk | Functional sample (demo) |