GitLab

make lint should pass again now

Stops enqueuing messages but keeps synchronizing with the network

Includes setting a terminated flag and a call to detect this.

This will be used for orderly termination of simulators at different times
without blocking others.

Empty structs are not standard C and sizeof for them in gcc is zero but for c++
is 1.

SimBricksBaseIfEstablish handles parallel connections and handshakes on
multiple simbricks connections.

This can help break ties consistently when multiple events have the same
timestamp.

Also splits out arg parsing from running so code can use the mac address in
initializationg before calling run.

Determines how much shm a particular connection will need based on the
parameters..

libsimbricks combines the base, pcie, and network libraries in one
archive.

High-level idea is to separate a out a base-protocol and library that
the PCIe and network protocols layer over. The base-protocol includes
synchronization and the basics of setting up queues, and transferring
messages. So this will enable us to build protocol-agnostic pieces like
proxies, loggers etc. A secondary goal down the line is to remove the
fixed roles of  "nics listen on pcie & network" while hosts and networks
connect, instead making this dynamic.

Unfortunately this is a big change touching just about every piece
interfacing with simbricks. So this is also consolidating and
restructuring our simbricks libraries and adapters, reducing copying of
code etc. After this commit all simulators, ours and external ones, are
broken till the respective commits fixing them.

set is broken if multiple events have the same timestamp (even though
the actual set members are pointers to the event :-/)

This allows multiple behavioral nic models to run on one OS thread
using boost fibers (user-level threads)

Enables overriding for multi NIC runner in next step

All devices will somehow need this, might as well stick it in the base
class.

Reduces the risk of getting the same mac in multi-host clusters

A bit too short for a fast qemu timing.

This breaks large mtu experiments, but when proxied makes a huge
difference. Should make this a runtime parameter.

Only show one message when we start to wait, and one when we succeeded.