On Sat, 2008-01-12 at 12:24, Jerry Van Baren wrote:
Hi Kim,
The isdram command is controlled separately from the rest of the i2c commands (CONFIG_CMD_I2C + CONFIG_CMD_SDRAM). I did this originally because the isdram command was arguably a memory consuming luxury. Luxuries tend to become necessities over time...
Larry Johnson submitted a patch yesterday to extend the decoding and printout to include DDR2 fields, making this even more useful (and jdl is claiming a share of the credit ;-).
gvb
I'm not claiming any credit for it, but if I do the following command:
$ git show d9b94f28a442b001 -- cpu/mpc85xx/spd_sdram.c
Someone who impersonated me quite well appears to have checked in whole bunch of DDR2 decode stuff way back in July of 2005.
For example:
+/* + * Convert a two-nibble BCD value into a cycle time. + * While the spec calls for nano-seconds, picos are returned. + * + * This implements the tables for bytes 9, 23 and 25 for both + * DDR I and II. No allowance for distinguishing the invalid + * fields absent for DDR I yet present in DDR II is made. + * (That is, cycle times of .25, .33, .66 and .75 ns are + * allowed for both DDR II and I.) + */ + unsigned int -banksize(unsigned char row_dens) +convert_bcd_tenths_to_cycle_time_ps(unsigned int spd_val) { - return ((row_dens >> 2) | ((row_dens & 3) << 6)) << 24; + /* + * Table look up the lower nibble, allow DDR I & II. + */ + unsigned int tenths_ps[16] = { + 0, + 100, + 200, + 300, + 400, + 500, + 600, + 700, + 800, + 900, + 250, + 330, /* FIXME: Is 333 better/valid? */ + 660, /* FIXME: Is 667 better/valid? */ + 750, + 0, /* undefined */ + 0 /* undefined */ + }; +
Weird, huh?
jdl