Computational strength between C2000 CLA and CPU
11 ビュー (過去 30 日間)
I'm using F28379D Ti LaunchPad for a system controller and computing Kalman Filter (which daw much computational cost). As you all know that TMSF28379D has 2 CPU and 2 CLA.
There are 2 options for me here. First is using CPU1 for controller and CPU2 for computing Kalman Filter. The second choise is using CPU1 for controller and CLA of CPU1 for computing Kalman Filter.
So, what is the better option here? In computational strength (specially in matrix computation), Which one is better, CLA or CPU?
回答 (1 件)
Venkatesh Chilapur 2020 年 11 月 2 日
As we know CLA is a 32-bit floating point math accelerator. So any part of the algorithm that involves Math based computation can be implemented on CLA. This would run in parallel with CPU and helps increase the overall bandwidth of the CPU.
Use CLA if the same computation when done on CPU would significantly impact your system performance as CPU in general would handle periodic tasks, events and interrupts. So the choice depends on how much bandwidth you need on the CPU including current and future enhancements.
If your processing on CPU is overrunning on your tasks, then use CLA to offload CPU.