Hi,
I have a DaVinci DM8168 EVM with our DM8168 board connected via PCIe.
In RC side we will allocate a buffer for data using kmalloc() and an ioremap() address point to EP buffer,
Then we measure the throughput as following,
using memcpy:
EP write to RC buffer around 50MB/s
EP read from RC buffer only 5MB/s
using EDMA in EP side:
EP write to RC buffer 23MB/s
EP read from RC buffer 22MB/s
Questions:
1. The throughput is reasonable?
2. Our application is EP decode video stream, our design is right? I means using the RAM buffer in RC instead of EP.
Regards,
Vic.