forums.ps2dev.org

This was the one of the only forums not seeing a post from this month
so I thought I would liven things up a bit.

This is an academic exercise thats been on my mind for a couple of
years. I know, why don't I just write code, play around, and try it
out ?

Can one still kick via path1 (vumem to gif) during a reversed path2
setting where transfers occur from gs2eehost ? Can the data kicked
to the GS via path1 automatically be rerouted back through path2
to the EE ?

Enquiring minds want to know...
Gorim

a pure academic response would be no, the GIF fifo is reversed, you cannot access general purpose registers, but I wonder if it does stall ( or if the xgkick reaches GIF control logic at all ), or it is one of those "undefined behaviour".

I guess time will tell.

What actually happens is indeed interesting.

If a transfer is actually in progress, I believe one
of the paths is supposed to stall in deference to the
other.

But what happens when the fifo is reversed, but
no transfer is in progress on path 2 when a kick
occurs to path 1 ? If the data from the kick is routed
to the front of the reversed fifo, it might seem there
is no where to go.

During this time, the VIF1 fifo is also reversed. If
the data enters the bus lines connecting VIF1 to GIF,
theoretically it could come back out the VIF1, but only
if there is an interconnect between path 1 and 2, instead
of physically separate lines.

But if it did, one thinks we would have heard more on
this topic. I remember back when the kit was released
in the US, on the sony playstation2-linux forums, this
topic receive a huge amount of discussion, and the
consensus was that one pretty much had to kick data
to GS memory store and then do gs2eehost transfer
to get it back out.

Well, if I have time on this vacation I might play around.

Gorim :)