On Sun, Mar 22, 2009 at 5:16 PM, Sascha Hauer s.hauer@pengutronix.de wrote:
On Sat, Mar 21, 2009 at 09:38:48AM -0400, Jon wrote:
struct mpc5xxx_i2c_tap { @@ -94,7 +94,7 @@ static int wait_for_bb(void) mpc_reg_out(®s->mcr, 0, 0); mpc_reg_out(®s->mcr, I2C_EN, 0); #endif
- udelay(1000);
- udelay(1);
status = mpc_reg_in(®s->msr); }
The actual condition we wait for takes only a few useconds to become true, but we wait for at least 1ms. As the I2C EEPROM environment routine reads nearly the complete EEPROM several times during startup the speed gain is enormous. I can't recall if it was intentional to reduce the overall timeout with this patch, but I can imagine that the i2c detect function will be faster aswell.
Even better, since this is entirely a busywait, the udelay(1) could be dropped entirely and the exit condition be based on the value of timebase. That would give the absolute maximum resolution to the busyloop. However, the incremental improvement probably isn't a great deal better than this patch.
g.