A recent change got rid of the strict Master => Slave port relationship
which used to be checked in python and instead left the checking up to
C++. One major downside to this approach is that it was no longer
obvious in the configuration what was supposed to be connected to what,
and it still left the arbitrary and misleading MasterPort and SlavePort
types in the Ethernet devices which could now connect with each other
symmetrically but couldn't actually connect to an arbitrary
MasterPort/SlavePort.
This change exposes the base Port and VectorPort types, and makes them
accept a "role" parameter in __init__ which used to be set directly by
their subclasses. This role can be any string, and will be used later
to check for compatiblity and to give a hint as to what can be
connected to what in the SimObject definitions.
To make the checks work with arbitrary compatible pairs, the base Port
type now has a class method called compat() which accepts a pair of
roles which will become mutually compatible, ie any port with the first
role will be allowed to connect to any port with the second role, and
vice versa. To be self compatible, the same role should be passed in
for both parameters.
To maintain compatibility, the MasterPort and SlavePort types are
retained, but now they're nothing special and could have been set up
in any arbitrary SimObject .py file. The same is true for
MasterVectorPort and SlaveVectorPort.
Also, since we can no longer assume that all edges in the dot graph of
the config should start with a port with the MASTER role and end with
a port with the SLAVE role, Ports now track an is_source property which
says whether the arrow head should be surpressed at that end of the
edge representing the connection.
Change-Id: Ifcc6faab05e437ad87cd21f0ba613b09cf21c321
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18168
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
This changeset enables clone to work with X86KvmCPU model, which
will allow running multi-threaded applications at near hardware
speeds. Even though the application is multi-threaded, the KvmCPU
model uses one event queue, therefore, only one hardware thread
will be used, through KVM, to simulate multiple application threads.
Change-Id: I2b2a7b1edb1c56eeb9c4fa0553cd236029cd53f8
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18268
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Commit 7976b561de tried fixing
replacement update when a single location can be associated to
multiple blocks.
Although the comment of the correct action was added, the proper
validation check was forgotten. This change adds that check and
moves doing the eviction to when there is a valid block.
Change-Id: I31d8bb914ccfd1849e9d97464d70a58a62f59533
Signed-off-by: Daniel <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18210
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Replacements should be increased when there is any evicted
block, which does not necessarily have to be the victim.
For example, assume a superblock contains 4 blocks, and both
A and C are stored compressed (belonging to SB_1). Then F,
from SB2 needs to make room by replacing SB1. If F map to
location 2, the number of replacements should be increased,
even though 2 had no valid blocks:
Tag Data Tag Data
|SB_1|--|A|X|C|X| --> |SB_2| |X|F|X|X|
1 2 3 4 1 2 3 4
Change-Id: I7b3735d28a35faa8d8fa613a1555bb258da65859
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18208
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Having the caller decide the matching logic is error-prone, and
frequently ends up with the secure bit being forgotten. This
change adds matching functions to the QueueEntry to avoid this
problem.
As a side effect the signature of findPending has been changed.
Change-Id: I6e494a821c1e6e841ab103ec69632c0e1b269a08
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17530
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
The backing store pointer is added to the back door when it's set,
assuming that the range isn't interleaved. If it is interleaved, then
there isn't a way to get a flat pointer to the backing store.
Depending on how the backing store is set up, it may be possible to
return a larger backdoor which applies to all interleaved memories at
the same time and to avoid problems with interleaving. I'm leaving this
as a todo.
Change-Id: I0e531c22835ec10954ab39f761b3d87666b59220
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17668
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
The RRPV values for RRIP and NRU replacment policies.
Long re-rereference interval was used instead of
distant re-rereference interval and vice-versa.
The btp value permit to choose beetwen distant and
long insertion ratio. A btp value of 0 force the
policy to always insert at a distant re-reference
interval and a btp value of 100 force the policy to
always insert at a long (intermediate) re-rereference
interval.
Change-Id: I516098f73942b769dcc31fe0edfe07c3e9c3effd
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17851
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
These both perform atomic accesses like their non-backdoor equivalents,
and also request a backdoor corresponding to the access.
The default implementation for recvAtomicBackdoor prints a warning
(once per port instance), calls recvAtomic to do the actual access,
and leaves the backdoor pointer as nullptr. That way if an object
doesn't know how to handle or transfer requests for a back door, it
automatically replies in a safe way that ignores the back door request.
Change-Id: Ia9fbbe9996eb4b71ea62214d203aa039a05f1618
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17590
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Gabe Black <gabeblack@google.com>
Previously satisfied clean requests would not snoop in-service
MSHRs. This is a problem when a clean request is also invalidating, in
which case we have to post-invalidate or post-downgrade outstanding
requests. This changes fixes this bug.
Change-Id: I31e42aa94dd3637b2818e00fbaae68c810145eaf
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17728
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Use SimpleMemory instead of DDR3 so we can use the timing results in
tests. By using SimpleMemory, even if the DRAM timing changes the timing
of this test won't change. I expect the timing of SimpleMemory to never
change.
Change-Id: I4c75981d7b8bfc4dcca59e628e89f5a6ea4c0e36
Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17871
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
garnet_synth_traffic.py imports common.Options on line 40, so exec'ing
the Options.py file again seems redundant.
It also runs Options.py as a script rather than a module, which throws
ValueError: Attempted relative import in non-package due to the recent
change to python3 imports.
Change-Id: Id729a8dfa776af0d14312e765168aff6900eb727
Signed-off-by: Ryan Gambord <gambordr@oregonstate.edu>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17888
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
These are similar to the structures TLM's DMI mechanism uses. Instead
of having an invalidation broadcast which propogates backwards up the
port hierarchy, this mechanism tracks a set of callbacks which are
triggered when a back door is invalidated to let other holders clean
up their bookkeeping.
Change-Id: If24489258dcaee14d7b6e5b996dfb1c2636f26ab
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17589
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Gabe Black <gabeblack@google.com>
For atomic RMW instructions that go directly to memory, we want to put
them on the write queue instead of the read queue. Swap the if/else
condition to accomplish this.
Note: This is ignoring the read latency of the RMW, but these
instructions should usually be handled in caches anyway.
Change-Id: I62dbfff3a16ac470f1ebdb489abe878962b20bb6
Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17828
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
