[coreboot] [PATCH] 440BX registered SDRAM support

Keith Hui buurin at gmail.com
Thu Mar 31 05:38:24 CEST 2011


On Tue, Mar 29, 2011 at 5:34 PM, Stefan Reinauer
<stefan.reinauer at coreboot.org> wrote:
> * Keith Hui <buurin at gmail.com> [110329 06:11]:
>> >> +    if ((edosd & 0x84) == 0x84) {
>> >> +        edosd = 0x10; // Registered SDRAM
>> >> +    } else {
>> >> +        // Clear [4:3] in case it's EDO.
>> >> +        edosd &= 0x07;
>> >> +//    } else if (edosd & 0x02) {
>> > Besides being commented out, this piece of code would never be executed,
>> > as there already is an else case.
>> > Also, modifying edosd in place is semi nice.
>>
>> So is this good, not so good, or bad?
>>
>> I want to know if I should split up edosd.
>
> Please do.
>

And so I did. Signoff in the patch.

edosd was a romcc-inspired trick because variables were a scarce resource.

The nbxecc simplification in this patch completed one full pass of
memtest86+ each with a regular and registered ECC DIMM installed.

With this patch the 440BX romstage is 60 bytes smaller, freeing up an
extra 64 bytes in the image.

Thanks
Keith
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 440bx.rdimm.patch
Type: application/octet-stream
Size: 4357 bytes
Desc: not available
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20110330/235005d8/attachment.obj>


More information about the coreboot mailing list