This improves CPI for things like
lbu t0, (t0)
j foo
addi t0, t0, 1
where the addi would stall, causing j's misprediction check to fail,
flushing the pipeline.
Until now, the number of L1 client channels was set statically in the
configuration. This static configuration also assumed the same number of
cached and uncached channels per tile. As we plan to move towards
heterogenous multicore systems, this restriction should be removed.
This commit changes the generator so that number of channels per tile
can be independently set (using cde.Parameters.alterPartial).
The OuterMemorySystem will dynamically compute the number of cached and
uncached channels by summing the number of each kind of channel per core.
They now share common sub-transactions within traits, and use a common
set of state transitions and scoreboarding logic. Tracker allocation
logic has also been updated. No changes to external IOs or the TileLink protocol.
A new bufferless Broadcast hub is also included, but does not yet pass fuzzing checks.
If you have multi-target rules that don't have %s in them, make
interprets that as "run this recipe multiple times, once to produce each
target". If you have %s in the rules, then make interprets it as "run
this recipe once to produce all targets". We want the second one.
I'm trying to get someone to attach their stuff to Rocket Chip for the
upcoming tapout. TileLink sounded too complicated, but Smi went over
well. Since the mmioNetwork in Rocket Chip is based on TileLink, it
seemed like the easiest thing to do was to write a TileLink to Smi
converter so people could use it.
It turns out there was already one inside the groundtest unit tests, so
I just moved that into uncore (it was inlined into a test case so you
couldn't actually use it before). Internally the converter uses Nasti,
but I figured that's good enough for now.
