Currently, in the OSS flow all memory operations are assumed to granted/responded to as soon as the protocol allows (immediately for req_o/gnt_i and after one cycle for gnt_i/rvalid_i).
For Jasper we allow anything between the minimum and a 10 cycle delay. Ideally we would do this for the OSS flow too.
In principle this should not be difficult, but it would be reasonable to expect a fairly significant slowdown on some properties, though hopefully not too many.
This may be as simple as just running the flow with relaxed constraints and seeing what happens, or it may require SST work.
Currently, in the OSS flow all memory operations are assumed to granted/responded to as soon as the protocol allows (immediately for
req_o/gnt_iand after one cycle forgnt_i/rvalid_i).For Jasper we allow anything between the minimum and a 10 cycle delay. Ideally we would do this for the OSS flow too.
In principle this should not be difficult, but it would be reasonable to expect a fairly significant slowdown on some properties, though hopefully not too many.
This may be as simple as just running the flow with relaxed constraints and seeing what happens, or it may require SST work.