It's still the case that the default is to use a fixed time,
which was the main point of the last changeset.
--HG--
extra : convert_revision : be74d808112fd4d0324475de7739e58b51a23c81
we want to keep revision control of which system
we're referring to
*NOTE* this means that this file must manually be changed if
we move to a new system!
--HG--
extra : convert_revision : cf42b07a7e4bcab86ecc96a06ac0a9c8da8895a3
checking in places to make sure stuff is behaving properly.
base/statistics.cc:
separate the per stat check from the general stats check.
We always want the general stuff to happen
base/statistics.hh:
- separate the per stat check from the general stats check.
we always want the general stuff to happen
- make every stat check that its bin is at least initialized
- set the vector2d x and y coordinates in init to prevent an
uninitialized memory access
test/Makefile:
don't need sim_time.o to test stats
test/stattest.cc:
don't make x and y the same on the 2d test so that we make
sure that the two dimensions are correct
--HG--
extra : convert_revision : 81320325056ac1c09f6842474fb6ee3bcc030a8e
base/cprintf.cc:
Get rid of unnecessary stream state check.
Bug is fixed now, and I'm not sure this did much anyway.
cpu/exetrace.cc:
Get rid of unnecessary flush() call.
--HG--
extra : convert_revision : 107dcdc578b2b0ad1652ee52ea0a33b0f3cd4e39
arch/alpha/alpha_tru64_process.cc:
So, I don't know why linux uses an off_t here.
I'm also not sure why linux defines an off_t to be a long
Let's just use long here since it works for linux, and that's
what bsd does
base/inifile.cc:
correct #include for OpenBSD
dev/disk_image.cc:
the correct type for this is streampos
--HG--
extra : convert_revision : f3ac3a3b8515d66e07ffb9780d8a9e387297b6a0
Most notable is that the rundiff trace-output lossage is fixed!
base/pollevent.cc:
Use SA_RESTART to keep blocked write() calls from failing
with EINTR when signals happen.
base/trace.hh:
Make TRACING_ON depend on NDEBUG not being set (instead of
DEBUG being set) so m5.opt/m5.fast do what we want.
util/rundiff:
Print header lines showing file/cmd names.
--HG--
extra : convert_revision : 4966814ad65aedb6d62a5a62a6ad3e3f326d61b8
makefile, such that decoder.cc was not getting rebuilt.
Also add -fno-strict-aliasing to fix all the bizarre problems
I've been having with g++ 3.3.x.
arch/alpha/isa_desc:
Fix compilation problems. AlphaISA is a class now, not a namespace.
--HG--
extra : convert_revision : 1583cebc1258c57cbd286c1955d11648150fa1f4
arch/alpha/alpha_memory.cc:
Rename md_mode_type to mode_type.
arch/alpha/ev5.cc:
simPalCheck() only gets called on correct path now, so
there's no need to test misspeculating().
arch/alpha/isa_desc:
Get privileged call_pall detection right this time (I hope).
ExecContext::simPalCheck() and Annotate::Callpal() are now
called only on non-speculative executions... this should fix
the bogus pal-call stats we've been seeing (since these are
incremented in simPalCheck()).
Also check for invalid call_pall function codes.
--HG--
extra : convert_revision : 465d6724884007d3fa066d14cd5e6db0cd3954e1
util/rundiff:
Dump buffered lines on loss of sync so we see where things went wrong.
Also return proper exit code from checkmatch() function.
--HG--
extra : convert_revision : cd88064ee784a50b675bf3fededbb3901c3c266c
base/statistics.hh:
Make a couple of typedefs public rather than protected.
g++ 3.3 complains otherwise.
--HG--
extra : convert_revision : 6a8e4edb0c39474c0e0c7bc474ad0f22b9fb8505
disable FastAlloc. Gcc 3.3.x appears to have a bug that
breaks FastAlloc when compiled with optimization.
base/fast_alloc.cc:
base/fast_alloc.hh:
#define NO_FAST_ALLOC makes FastAlloc a no-op.
--HG--
extra : convert_revision : bf4bb8e963331c6782f2cfd475a811aa520e31b5
base/str.hh:
this should really be inline
base/time.cc:
base/time.hh:
clean up how the time class works. Export only one, and let
people calculate their own elapsed times, etc.
sim/main.cc:
sim/sim_time.X -> base/time.X
--HG--
rename : sim/sim_time.cc => base/time.cc
rename : sim/sim_time.hh => base/time.hh
extra : convert_revision : f3888fe3a1fdd1022084c282b58407c631a6d9a0
- make stuff compile with NDEBUG
- remove the TRACE setup option since it isn't useful anymore
base/statistics.cc:
Compile with -DNDEBUG
--HG--
extra : convert_revision : a75b3881be1513abb5915b6463c6cdc1f23a62e1
cpu/memtest/memtest.hh:
Oops... make memtest compile with new event priorities too.
--HG--
extra : convert_revision : 8143f2e078bb785991d351803fc3a6a50aa31ae5
This lets us centralize priorities so we can see what's going on.
- Shift serialize & cpu-switch events to happen before CPU ticks (to be
consistent with starting new CPU on same cycle instead of next cycle).
- Get rid of unnecessary bus stats reset callback.
cpu/simple_cpu/simple_cpu.cc:
sim/debug.cc:
sim/eventq.hh:
sim/serialize.cc:
sim/sim_events.cc:
sim/sim_events.hh:
Switch events to use a priority enum instead of integers.
This lets us centralize priorities so we can see what's going on.
--HG--
extra : convert_revision : 510d79b43c0a1c97a10eb65916f7335b1de8b956
a checkpoint now gives identical results to running from scratch
and doing at switchover at the same cycle!
- CPUs start at cycle 0 again, not cycle 1.
- curTick is now serialized & unserialized.
- Stats get reset in main (before event loop). Since this is done
after curTick is unserialized, simTicks gets set correctly for
running from a checkpoint.
- Simplify serialization to happen in a single pass.
- s/Serializeable/Serializable/
arch/alpha/isa_traits.hh:
dev/etherlink.hh:
sim/eventq.cc:
sim/eventq.hh:
s/Serializeable/Serializable/
kern/tru64/tru64_system.cc:
sim/process.cc:
Make initial CPU activation on cycle 0 again (not 1).
sim/main.cc:
Reset stats before getting started.
Make error message on falling out of event loop
more meaningful.
sim/serialize.cc:
sim/serialize.hh:
Get rid of now-useless initial pass; serialization is
done in a single pass now.
Serialize & unserialize curTick.
Wrap curTick and mainEventQueue in a "globals" Serializable object.
s/Serializeable/Serializable/
sim/sim_object.cc:
Add static function to serialize all SimObjects.
sim/sim_object.hh:
Add static function to serialize all SimObjects.
s/Serializeable/Serializable/
--HG--
extra : convert_revision : 9dcc411d0009b54b8eb61c3a509680b81b9f6f68
and SimpleCPU::setStatus() into separate functions. For example,
setStatus(Active) is now activate().
--HG--
extra : convert_revision : 4392e07caf6c918db0b535f613175109681686fe
on test5 because of a wrong exit code; fixed that.
sim/main.cc:
No need to ignore SIGPIPE, as far as I know.
sim/sim_events.cc:
A CountedExitEvent is a normal termination, so
should have an exit code of 0.
--HG--
extra : convert_revision : 8b5072aca54f3ca08b2815b73ac01c00e4da49b6
base/hashmap.hh:
Not all machines need these extra hash functions
also make Counter int64_t since that's really what is needed
--HG--
extra : convert_revision : b5b5d65db157d01d76adbf4de4882b1516c9f2e7
restarting from a checkpoint at tick N.
cpu/simple_cpu/simple_cpu.cc:
On a CPU switchover, schedule the new CPU's first tick event
for curTick+1 instead of curTick.
--HG--
extra : convert_revision : f0757d6f028214d36c1cff992db688fd6e8a6fdc
base/statistics.cc:
- Merge the m5 display and simplescalar compatible display functions.
- Use functors for the stats printing instead of functions. This
gets rid of the very long argument lists and improves clarity.
- Fix a bug in binning that caused an invalid allocation to occur.
base/statistics.hh:
- Instead of using a compile time variable to choose whether
to print in ss compat mode, we use a runtime variable.
This is how we'll choose python output.
- There are no more virtual functions in the statistics themselves.
All virtual functions have been moved into a secondary helper class.
Every stat has an associated helper class that knows how to access
certain variables in that stat.
There is a hash_map from the stat to it's helper class data. This was
done because the helper data is only used during setup, stats reset,
and printing. All of which happen rarely, and you want to avoid
any performance hit that you can.
- To provide the name(), desc(), etc functions to classes so that the
user can set various parameters to the stats class, a single class
containing all of those functions was created. An odd trick was
done to actually make this class derive from the stat class because
a base class with no data actually does end up taking up space.
- The detail namespace was removed for now. I'll put it back when the
package is not in so much flux.
- Standard deviation, and mean were added to all distribution stats.
- There are several bugfixes and changes that aren't mentioned
test/Makefile:
Don't test sim_stats foo.
test/stattest.cc:
Don't test sim_stats foo.
Fix bin usage so that it corresponds to the current usage.
--HG--
extra : convert_revision : ce8d9a8d485a84d55799f253d851e83650684170
idle, keep track of the fraction of time we're not idle. This
works better because the default processor state is idle, and
the default stat value is 0.
Keep the stat as idleFraction which is a formula that is equal
to 1 - notIdleFraction
--HG--
extra : convert_revision : 331c2e46f45ae0abda46988567ac2c4f7c42ccad
I noticed that checkpoints dumped from m5.debug and m5.opt had
differences, which is no longer the case (other than addresses
encoded in names).
cpu/full_cpu/smt.hh:
Get rid of unused SMT_MAX_CPUS.
sim/debug.cc:
Use AutoDelete for DebugBreakEvent.
sim/serialize.cc:
sim/serialize.hh:
Move constructor & destructor to header so they can be inlined.
--HG--
extra : convert_revision : bc68c2c9b053b1de0d655ed555734419fafd0b83
