hi miod! really thanks for this reply, really answers most of my questions that can affect some of my future project with my SGIs
miod wrote:There is no choice but to use these chips. For example, in order to be able to use the SCSI controller, you need to be able to talk to the PCI Bridge it lies beneath. But in this particular example, the SGI XBow-PCI Bridge has specific features to do some sort of dynamic bandwidth shaping for the DMA operations, as well as the ability to use virtual adresses for DMA (in other words, scatter-gather for transfers spanning multiple pages), and these are not supported by BSD or Linux because the scarce information available about this chip does not cover these well enough.
yeah, also thought so, looking at the block diagrams from Ian for both O2 and Octane, there's no possible way without utilizing these chips, sorry for still asking tho, just want to confirm
hmm so that's why porters for BSD and Linux still saying that some chips are xx% in progress, and this is why these OS are slower than the IRIX.
miod wrote:The main source of information are the IRIX system header files. SGI also contributed some code to Linux to support the Altix systems, which acts as an unsorted source of tidbits and quirks about these chips (many hardware defects and their workarounds are mentioned in these contributions).
Another source of information is the patents SGI filed about some of its chips. They don't dive into the details, but give a good overview of what the chips are capable of.
Several comments in the Linux Octane-specific files hint that there was some disassembly involved, at least for the Impact support. I also believe that the GR2 ``texport'' IRIX routines were disassembled by NetBSD developrs, which led to the initial XZ support in NetBSD
ah i see, geeez i can imagine the porters, really need patience on gathering these infos.. but its worth especially after able to successfully boot these OS hehe i really salute to them. i just hope that maybe some old IRIX engineers still have some documents left and can share it here
miod wrote:Not really. You may be able to copy an FPGA, assuming there is a debug probe connector left on the board (assuming you recognize the FPGA flavour and have the proper tools for it), but ASIC are out of reach of mere mortals.
ah i see, geez all SGI chips use ASIC right? hahaha mere mortals, like this one