Now that I think of it, the _multiplier_ (as I declared it here) cannot be
an integer, so this is all wrong.

I don't know how I got this working... I have a 50 Mhz oscillator with 66
Mhz system clock.

Why not use the oscillator define (CFG_8XX_XIN) for MPC866_et_al?

David


owner-linuxppc-embedded@lists.linuxppc.org wrote on 05/27/2004 03:19:26 PM:

quoted

The last if condition is wrong in this function.   The change is made
against 0.4.8.  AFAIK, 1.0.2 still has the same problem.

And I wonder why it has to reverse engineer the oscclk when I pass the
value as a define to come up with the gd->cpu_clk for the MPC866

family...

quoted

It is a bit useless I think.

Sorry, the last patch breaks on the old 8xx.

This is a bit better.


/*
 * Get timebase clock frequency (like cpu_clk in Hz)
 *
 * See table 15-5 pp. 15-16, and SCCR[RTSEL] pp. 15-27.
 */
unsigned long get_tbclk (void)
{
        DECLARE_GLOBAL_DATA_PTR;

        volatile immap_t *immr = (volatile immap_t *) CFG_IMMR;
        ulong oscclk, factor, multiplier;

        if (immr->im_clkrst.car_sccr & SCCR_TBS) {
                return (gd->cpu_clk / 16);
        }
#define PLPRCR_val(a) (((CFG_PLPRCR) & PLPRCR_ ## a ## _MSK) >> PLPRCR_

##

a ## _SHIFT)
#ifdef CONFIG_MPC866_et_al
        /*                   MFN
                     MFI + -------
                           MFD + 1
          factor =  -----------------
                     (PDF + 1) * 2^S
         */

        factor = (PLPRCR_val(MFI) + PLPRCR_val(MFN)/(PLPRCR_val(MFD)+1));

        multiplier = (PLPRCR_val(MFI) +
PLPRCR_val(MFN)/(PLPRCR_val(MFD)+1))/
                     (PLPRCR_val(PDF)+1) / (1<<PLPRCR_val(S));
#else
        multiplier = factor = PLPRCR_val(MF)+1;
#endif

        oscclk = gd->cpu_clk / multiplier;

        if ((immr->im_clkrst.car_sccr & SCCR_RTSEL) == 0 || factor > 2) {
                return (oscclk / 4);
        }
        return (oscclk / 16);
}

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/