On Monday 29 June 2009, jassi brar wrote: > Hello, > I see drivers/spi/spi_s3c24xx.c uses bitbang i/f eventhough 24xx has a > dedicated SPI controller. Wasn't bitbang is supposed to be for SOCs which > don't have a SPI controller.
Yes and no. The framework is a bit more general than that; have a look at the spi_bitbang.c code. The lowest level has various options for per-word I/O, with bitbanging as the default; but you can use hardware shift registers too, if you like. And some folk *do* like. It turns out to be handy to get a driver working quickly, when "have it working yesterday" is more important than just "have it working really fast". > I was wondering why cudn't we make it work without the Bitbang abstraction? > Any potential issues that I am failing to fathom? > And, s3c24xx.c doesn't make use of DMA. Any particular reason against it? I don't know those s3c parts, but I'd presume the main reasons boil down to nobody spending time to create a speedy DMA-centric driver. - Dave ------------------------------------------------------------------------------ _______________________________________________ spi-devel-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/spi-devel-general
