the primary identifier for a hardware context should be contextId(). The
concept of threads within a CPU remains, in the form of threadId() because
sometimes you need to know which context within a cpu to manipulate.
SE. Process still keeps track of the tc's it owns, but registration occurs
with the System, this eases the way for system-wide context Ids based on
registration.
across the subclasses. generally make it so that member data is _cpuId and
accessor functions are cpuId(). The ID val comes from the python (default -1 if
none provided), and if it is -1, the index of cpuList will be given. this has
passed util/regress quick and se.py -n4 and fs.py -n4 as well as standard
switch.
The constructor no-longer schedules an event at construction and the implict conversion between int and bool was allowing the old code to compile without warning.
Signed-off By: Ali Saidi
the instruction after the hwrei to be fetched before the ITB/DTB_CM register is updated in a call pal
call sys and thus the translation fails because the user is attempting to access a super page address.
Minimally, it seems as though some sort of fetch stall or refetch after a hwrei is required. I think
this works currently because the hwrei uses the exec context interface, and the o3 stalls when that occurs.
Additionally, these changes don't update the LOCK register and probably break ll/sc. Both o3 changes were
removed since a great deal of manual patching would be required to only remove the hwrei change.
Since I never implemented a proper solution, put it back to something that
at least works for now. Once I add more event queues, I'll have to really
fix this though
The IntDev class is a base for anything that supports IntPins. IntPins allow
devices to generically trigger interrupts on a particular pin of an IntDev
device without having to know what the device is or what pin they're attached
to.
The major thrust of this change is to limit the amount of code
duplication surrounding the code for these functions. This code also
adds two new message types called info and hack. Info is meant to be
less harsh than warn so people don't get confused and start thinking
that the simulator is broken. Hack is a way for people to add runtime
messages indicating that the simulator just executed a code "hack"
that should probably be fixed. The benefit of knowing about these
code hacks is that it will let people know what sorts of inaccuracies
or potential bugs might be entering their experiments. Finally, I've
added some flags to turn on and off these message types so command
line options can change them.
If the same dictionary option is seen in several options, those
dictionaries are composed. If you define the same dictionary key in
multiple options, the system flags an error.
Also, clean up the jobfile code so that it is more debuggable.
Make them easier to express by only having the cxx_type parameter which
has the full namespace name, and drop the cxx_namespace thing.
Add support for multiple levels of namespace.
Since the early days of M5, an event needed to know which event queue
it was on, and that data was required at the time of construction of
the event object. In the future parallelized M5, this sort of
requirement does not work well since the proper event queue will not
always be known at the time of construction of an event. Now, events
are created, and the EventQueue itself has the schedule function,
e.g. eventq->schedule(event, when). To simplify the syntax, I created
a class called EventManager which holds a pointer to an EventQueue and
provides the schedule interface that is a proxy for the EventQueue.
The intent is that objects that frequently schedule events can be
derived from EventManager and then they have the schedule interface.
SimObject and Port are examples of objects that will become
EventManagers. The end result is that any SimObject can just call
schedule(event, when) and it will just call that SimObject's
eventq->schedule function. Of course, some objects may have more than
one EventQueue, so this interface might not be perfect for those, but
they should be relatively few.
I've done a few things here. First, I invoke the script a little bit
differently so that pdb doesn't get confused. Second, I've stored the
actual filename in the module's __file__ so that pdb can find the
source file on your machine.
Targets look like libm5_debug.so. This target can be dynamically
linked into another C++ program and provide just about all of the M5
features. Additionally, this library is a standalone module that can
be imported into python with an "import libm5_debug" type command
line.
Even though we're not incorrect about operator precedence, let's add
some parens in some particularly confusing places to placate GCC 4.3
so that we don't have to turn the warning off. Agreed that this is a
bit of a pain for those users who get the order of operations correct,
but it is likely to prevent bugs in certain cases.
We should always refer to the specific ISA in that arch directory.
This is especially necessary if we're ever going to make it to the
point where we actually have heterogeneous systems.
Fix the logic in the LSQ that determines if there are any stores to
write back. In the commit stage, check for thread specific writebacks
instead of just any writeback.
These functions keep trying to read and write until all data has been
transferred, or an error occurrs. In the case where an end of file
hasn't been reached, but all of the bytes have not been read/written,
try again. On EINTR, try again.
We haven't used the preprocessor feature of the inifile stuff in a
very long time, so let's get rid of it since it would otherwise take
effort to maintain.
python type of a latency. In addition, the multiple definitions of profile in the different cpu models caused
problems for intialization of the interval value. If a child class's profile value was defined, the parent
BaseCPU::ProfileEvent interval field would be initialized with a garbage value. The fix was to remove the
multiple redifitions of profile in the child CPU classes.
A whole bunch of stuff has been converted to use the new params stuff, but
the CPU wasn't one of them. While we're at it, make some things a bit
more stylish. Most of the work was done by Gabe, I just cleaned stuff up
a bit more at the end.
When invoking several copies of m5 on the same machine at the same
time, there can be a race for TCP ports for the terminal connections
or remote gdb. Expose a function to disable those ports, and have the
regression scripts disable them. There are some SimObjects that have
no other function than to be used with ports (NativeTrace and
EtherTap), so they will panic if the ports are disabled.
This should allow m5 to be more easily embedded into other simulators.
The m5 binary adds a simple main function which then calls into the m5
libarary to start the simulation. In order to make this work
correctly, it was necessary embed python code directly into the
library instead of the zipfile hack. This is because you can't just
append the zipfile to the end of a library the way you can a binary.
As a result, Python files that are part of the m5 simulator are now
compile, marshalled, compressed, and then inserted into the library's
data section with a certain symbol name. Additionally, a new Importer
was needed to allow python to get at the embedded python code.
Small additional changes include:
- Get rid of the PYTHONHOME stuff since I don't think anyone ever used
it, and it just confuses things. Easy enough to add back if I'm wrong.
- Create a few new functions that are key to initializing and running
the simulator: initSignals, initM5Python, m5Main.
The original code for creating libm5 was inspired by a patch Michael
Adler, though the code here was done by me.
I decided that separating some of the scons code into generate.py was
just a bad idea because it caused the dependency system to get all
messed up. If separation is the right way to go in the future, we
should probably use the sconscript mechanism, not the mechanism that I
just removed.
- Add the option of redirecting stderr to a file. With the old
behaviour, stderr would follow stdout if stdout was to a file, but
stderr went to the host stderr if stdout went to the host stdout. The
new default maintains stdout and stderr going to the host. Now the
two can specify different files, but they will share a file descriptor
if the name of the files is the same.
- Add --output and --errout options to se.py to go with --input.
- insert warnings for deprecated m5ops
- reserve opcodes for Ali's stuff
- remove code for stuff that has been deprecated forever
- simplify m5op_alpha
The status quo is preferred since it is less likely that people will
rely on LIFO than FIFO, and when we move to a parallelized M5, no
ordering between events of the same time/priority will be guaranteed.
linked list sorted by time and priority. For things of the same time
and priority, a second, circularly linked list maintains the data
structure. Events of the same time and priority are now inserted in
FIFO order instead of LIFO order. This dramatically improves the
performance of systems that schedule multiple events at the same time.
The FIFO order version is not preferred to LIFO (because it may cause
people to rely on it), but I'm going to commit it anyway and
immediately commit the preferred LIFO version on top.
This appears to work, but I don't want to commit it until it gets tested a lot more.
I haven't deleted the functionality in this patch that will come later, but one question
is how to enforce encourage objects that call getVirtPort() to not cache the virtual port
since if the CPU changes out from under them it will be worse than useless. Perhaps a null
function like delVirtPort() is still useful in that case.
The notIdleFraction statistic isn't updated when the statistics reset, probably because the cpu Status information
was pulled into the atomic and timing cpus. This changeset pulls Status back into the BaseSimpleCPU object. Anyone
care to comment on the odd naming of the Status instance? It shouldn't just be status because that is confusing
with Port::Status, but _status seems a bit strage too.
they're all in the same place. This also involves having just one
jobfile.py and moving it into the utils directory to avoid
duplication. Lots of improvements to the utility as well.
--HG--
rename : src/python/m5/attrdict.py => src/python/m5/util/attrdict.py
rename : util/pbs/jobfile.py => src/python/m5/util/jobfile.py
rename : src/python/m5/util.py => src/python/m5/util/misc.py
rename : src/python/m5/multidict.py => src/python/m5/util/multidict.py
rename : util/stats/orderdict.py => src/python/m5/util/orderdict.py
It runs out that if a MemObject turns around and does a send in its
receive callback, and there are other sends already scheduled, then
it could observe a state where it's not at the head of the list but
the bus's sendEvent is not scheduled (because we're still in the
middle of processing the prior sendEvent).
I was asserting that the only reason you would defer targets is if
a write came in while you had an outstanding read miss, but there's
another case where you could get a read access after you've snooped
an invalidation and buffered it because it applies to a prior
outstanding miss.