Hello,
I have a question regarding CPU load when using an SDIO-based WiFi module on the dm365 EVM @ 297MHz. I see a dramatic difference in CPU load when using the original 2.6.18 kernel (with the SDIO driver patch) vs. a newer git tree kernel (2.6.32-rc2-davinci1) using the native Linux SDIO driver. In both cases I am using the same WiFi driver from the chip vendor (there are compile options to build for the different kernel SDIO interfaces).
Problem: Using the new git davinci 2.6.32 kernel I see about 4x MORE CPU load for the same WiFi traffic test (vs. the old 2.6.18 kernel).
For example, a bi-directional UDP traffic test with 160 byte packets spaced 20 ms apart (to simulate encoded voice traffic), the CPU usage using the 2.6.18 kernel is ~20%, where using the 2.6.32 kernel the usage is ~90%.
For such an order of magnitude difference, it makes me think there is some kernel configuration or something else that is causing the high CPU load using the new kernel. I have tried disabling netfilter and other networking options that may cause more CPU overhead, but nothing seems to change the CPU load.
Question 1: Does anyone else have CPU usage stats for SDIO based WiFi traffic using 297MHz dm365?
Question 2: Can anyone suggest a configuration or something else to look for to see why this is happening?
Thanks.
-Craig