Am Samstag, den 17.07.2010, 00:59 +0200 schrieb Carl-Daniel Hailfinger:
Why so slow? SPI was designed to be fast (if you want a slow system, use I2C instead ;) ). It might be quite sensible on programmers at the parallel port having some cable length and no driver chip at the flash piece (maybe that applies to SPIPGM), but that is specific to this programmer. nVidia on-board stuff should run slow enough without any programmer_delay calls.
Not sure. I figured that 1 usec would be a sane default, and could be reduced to zero by individual programmers depending on their knowledge about timing.
My idea was that everything that is not "broken" in hardware doesn't need a delay, so 0 is a sane default and programmers can set this to a bigger value in their code if they think that this programmer is so slow, but I don't really care, 1 as default is OK too.
Signed-off-by: Carl-Daniel Hailfinger c-d.hailfinger.devel.2006@gmx.net
I'd rather kill the whole buffer stuff (see attached patch) before applying your patch (and of course remove the buffer stuff from your patch, too). No other reason not to ack.
Regards, Michael Karcher