Yeah, I do and that would make sense. Afterall, most things aren't constrained like that. How would I know? Ask and ye shall receive. Anyway, it's not that serious, relax. You guys don't exactly make all of this right up at the top of the front page to be obvious. So this is how we all figure it out.
Btw, I changed the repo name so that page is now here: https://albazzaztariq.github.io/Firmware-Param-Catalog/ . But I may delete it if it's not really of much use. Cheers On Wed, Mar 4, 2026 at 3:51 PM Angel Pons <[email protected]> wrote: > Hi, > > On Wed, Mar 4, 2026 at 8:54 PM Tariq A <[email protected]> wrote: > >> Hi all, I got claude to make this page for the Intel FPS >> "documentation"/comments contained in the code. >> https://albazzaztariq.github.io/Intel-FSP-Param-Catalog/ I'm trying to >> build a tool that'll log a series of sessions using different hardware >> configurations, then optimize to find the best presets for the given >> workload. Basically an auto optimizer for coreboot settings. Could anyone >> help me in figuring out what would be a suitable range of configurations to >> try to capture a reasonable space representing the possibilities? >> > > Honestly, do not bother. Unless you understand the hardware enough to > reasonably understand what FSP settings do (which requires using the brain > quite a bit), this is a waste of time and computing resources. Have you > considered the possibility that hardware damage could occur with certain > combinations of FSP settings? > > There is no such thing as "optimising" FSP parameters. There are a some > which are mapped to user-configurable options (e.g. VT-d support). The rest > depend on hardware-specific details and have to be set to a specific value > in order for things to work. For example, DQ/DQS mapping in boards that use > LPDDRx memory: these settings reflect how the CPU and the RAM are > connected; if one of these values is incorrect, it simply will not work. To > make matters worse: sometimes the comments in the FSP headers aren't > accurate, so you cannot rely on those as the sole source of truth. There's > so many variables that are NOT controlled by FSP parameters, but that the > values of FSP parameters indirectly depend on. There's a reason why we have > wrappers for FSP settings in coreboot. > > Do you really expect a glorified text predictor to be able to "understand" > all of this, and more? > > _______________________________________________ >> coreboot mailing list -- [email protected] >> To unsubscribe send an email to [email protected] >> > > Best regards, > Angel >
_______________________________________________ coreboot mailing list -- [email protected] To unsubscribe send an email to [email protected]
