Allow fetches to bypass prefetches to L1

[micprog]

No description provided.

[paulsc96]

LGTM. I would have instantiated another stream_arbiter with .ARBITER ("prio") for the final selection between prefetch and fetch though; would make it a bit more readable.

[micprog]

LGTM. I would have instantiated another stream_arbiter with .ARBITER ("prio") for the final selection between prefetch and fetch though; would make it a bit more readable.

I don't think this will work, as LockIn is required (to ensure stable signalling when a low priority fetch is stalled and a high priority fetch arrives) but not supported in the rr_arb_tree when configured to use ExtPrio. I agree on the sentiment, though. Should the rr_arb_tree be modified to support this? In another thought, should this stream_arbiter with priority levels be collected into a common_cell? @niwis