On Wednesday 24 March 2010, Øyvind Harboe wrote:
> Target scripts without a adapter_khz should be fixed to have one,
> otherwise the khz is undefined.

Make that "board config scripts", as a rule.

It's routine for different boards to use different crystals, even if
they have the same chip/target ... and thus for speed settings in
target config scripts to be wrong.  

This isn't straightforward.  The right speed is often also a function
of how firmware has initialized things and where it is in the init
sequence, as well as board design variables.

(The User's GUide *does* have information about this.  IN several
places.  It's a bit more scattered than I'd prefer.)

Consider the Atmel boards reset to a 32 KiHz "slow clock" and
then later turn on the main oscillator (tens of MHz) and then a PLL
(to run at maybe 200 MHz).  Event handlers have to update the JTAG
clock.  (Not at all restricted to Atmel chips of course; those are
notable mostly because the initial clock is unusually slow.  And...
because that initial clock is a chip characteristic, not specific
to the board.)

... but it is virtually NEVER correct to have an adapter driver choose
a clock rate.


> If a card *could* support RCLK, then use jtag_rclk instead. It will
> default to rclk if the adapter supports it, otherwise use the khz
> specified by the script.

... only do that if the target/chip supports RCLK though.  If it does,
then using adaptive clocking probably means you won't need to have
event handlers updating the JTAG clock rate.

- Dave
_______________________________________________
Openocd-development mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/openocd-development

Reply via email to