On 11/23/22 22:20, Jon Hunter wrote: > > On 22/11/2022 16:07, Greg KH wrote: >> On Fri, Nov 18, 2022 at 03:40:05PM +0000, Jon Hunter wrote: >>> From: Sing-Han Chen <singhanc@xxxxxxxxxx> >>> >>> This change adds Tegra234 XUSB host mode controller support. >>> >>> In Tegra234, some of the registers have moved to bar2 space. >>> The new soc variable has_bar2 indicates the chip with bar2 >>> area. This patch adds new reg helper to let the driver reuse >>> the same code for those chips with bar2 support. >>> >>> Signed-off-by: Sing-Han Chen <singhanc@xxxxxxxxxx> >>> Co-developed-by: Wayne Chang <waynec@xxxxxxxxxx> >>> Signed-off-by: Wayne Chang <waynec@xxxxxxxxxx> >>> Signed-off-by: Jon Hunter <jonathanh@xxxxxxxxxx> >> >> This is should be much slower with the additional redirection. Is it >> noticable on this hardware platform with, and without this change? Or >> is the hardware slow enough that it doesn't even show up as a speed >> decrease? > Wayne, do we have any inputs on this? > > I know that we have been using this implementation now for some time on > the kernels we ship and that would be tested on Tegra210, Tegar186, > Tegra194 and Tegra234. So I assume that the performance there is good, > but not sure about Tegra124. > Hi Jonathan and Greg, Thanks for the review. The implementation is out of the difference in hardware implementation between old chips and tegra234. Most of the read/write ops are one-time calls during initialization. Even the mbox msg are seldom used and thus the redirection should not cause any noticeable performance drop. thanks, Wayne. > Jon >
