On Thu, May 21, 2020 at 01:19:16PM -0700, Tim Harvey wrote: > Should I be submitting this patch with logic that only does > half-duplex if the spi controller doesn't support it (if > (spi->controller->flags & SPI_CONTROLLER_HALF_DUPLEX)) or is it > acceptable to simply make the driver half-duplex like this for all > cases? It seems likely that making the transfers explicitly half duplex will perform better, especially for PIO controllers since there's less FIFO stuffing to do but also just generally on longer messages. You will get some overhead setting up two transfers on write then read messages which might offset that but my best guess would be that it'll be negligable on most controllers. It's also just a more accurate representation of what the transfers are actually doing which seems nicer. If there *is* a performance win for doing full duplex messages on some controllers we should probably look at optimizing this in the SPI core since it'll affect a wide range of hardware and we already have some code for forcing full duplex anyway.
Attachment:
signature.asc
Description: PGP signature
