Currently the Switch Allocator takes up most of the simulation
wall clock time. This function checks for all VCs to see if it
should wakeup next. The input units which are simulated before
the switch allocator could have scheduled it already. This patch
adds a check for it.
Change-Id: I8609d4e7f925aa5e97198f6cd07466530f6fcf4c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32600
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
This upgrades the garnet model to support HeteroGarnet
1) Static and dynamic multi-freq domains in network
2) Support for CDC
3) Separate links for each message class
4) Separate linkwidth for each message class
5) Support for SerDes
Change-Id: I6d00e3b5cb3745e849d221066cb46b2138c47871
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32597
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Garnet utilizes round robin policy to select a VC for
transmission ar Network Interface and Routers. The current logic
for round robin is only fair if all the virtual networks are active
at a given router. If the router or network interface is not
receiving traffic in from any vnet then the priority is always taken
up by the next vnet in numerically (or loops back to 0).
This fix changes the way we perform round robin arbitration. When
a VC is selected in a cycle, the round robin pointer is set to the VC
next to it and is iterated from there on. If any VC does not have a
flit in a given cycle, it will lose its turn until the next round.
At maximum traffic this will model round robin correctly even if
a certain VNET is not active at that unit.
Change-Id: I9bf805221054f9f25bee14b57ff521f4ce4ca980
Reviewed-on: https://gem5-review.googlesource.com/c/16688
Reviewed-by: Jieming Yin <Jieming.Yin@amd.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
This patch fix some statistics that in presence of a resetStats
instruction were not reseted. This bug makes impossible to obtain
reliable network statistics when the simulation doesn't start from tick
zero.
Change-Id: Ibec45f08d95bf0a533d94b70ec960719206ae945
Maintainer: Tushar Krishna <tushar@ece.gatech.edu>
Reviewed-on: https://gem5-review.googlesource.com/3700
Reviewed-by: Jieming Yin <bjm419@gmail.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
The rr arbiter pointer in garnet was getting updated on every request,
even if there is no grant. This was leading to a huge variance in wait
time at a router at high injection rates.
This patch corrects it to update upon a grant.
