> On Tue, Oct 28, 2008 at 8:03 AM, Myles Watson <[EMAIL PROTECTED]> wrote: > > Here's a list of the #defines that I think we should look at. I would > like > > to detect as many as possible, and add the rest to the dts or Kconfig. > > > ./coherent_ht.c:#ifdef CONFIG_K8_HT_FREQ_1G_SUPPORT > > we only support f2 or later. This should die.
All right. > > ./coherent_ht.c:#if CONFIG_CROSS_BAR_47_56 > > don't know how to autodetect this ... it is a wiring on the mainboard > thing. Naively I would say that you could detect it after you had 4 & 5 initialized. If they were connected to each other than it was ladder, otherwise crossbar. I realize it may not be that simple. Should we put it in the dts then? > > > ./coherent_ht.c:#if CONFIG_USE_PRINTK_IN_CAR > > Is this in v3? I thought I had killed all these. I just killed the last one. > > ./coherent_ht.c:#if CONFIG_LOGICAL_CPUS==1 > > Ditto. That should be dead. I'll take it out. > > ./pci.c:#if SB_HT_CHAIN_ON_BUS0 > 0 > > This is detectable and stefan is going to tell us how. Good. There's a place in the code where it does a test to find the southbridge link by checking the value of one of the k8 registers. > > ./pci.c:#if SB_HT_CHAIN_UNITID_OFFSET_ONLY == 1 This one ? > > ./pci.c:#ifdef CONFIG_MULTIPLE_VGA_INIT > > config variable. Don't know how to do this dynamically. > > > ./incoherent_ht.c: #if (CONFIG_K8_SCAN_PCI_BUS == 1) > > ./incoherent_ht.c:#if CONFIG_K8_ALLOCATE_IO_RANGE == 1 > > ./raminit.c:#if defined(CONFIG_LB_MEM_TOPK) & ((CONFIG_LB_MEM_TOPK -1) > != 0) > > ./raminit.c:#ifdef QRANK_DIMM_SUPPORT > > > ./raminit.c:#if CPU_SOCKET_TYPE == SOCKET_L1 > > ./raminit.c:#if CPU_SOCKET_TYPE == SOCKET_AM2 > > yes, these are trivial to detect at runtime. > > > ./raminit.c:#ifdef DRAM_MIN_CYCLE_TIME > > um. > > > ./raminit.c:#if DIMM_SUPPORT==0x0104 /* DDR2 and REG */ > > This ought to be determined somehow but not sure how. > > > ./raminit.c:#if MEM_TRAIN_SEQ == 0 > > ./dqs.c: #if CONFIG_AP_CODE_IN_CAR == 0 > > These are hard, but we can try. > > . > > ./domain.c:#ifdef CONFIG_PCI_64BIT_PREF_MEM > > agree. But let's do the easy ones first. Agreed. > Starting with > FREQ_1G_support: all F2 are ok with 1G, right? We don't need that > test? Isn't this just for processors that don't report their supported frequency correctly? You can find F2 Semprons that don't support 1G. Thanks, Myles -- coreboot mailing list: [email protected] http://www.coreboot.org/mailman/listinfo/coreboot
