gem5.git
13 years agoARM: Use a stl queue for the table walker state
Dam Sunwoo [Mon, 23 Aug 2010 16:18:39 +0000 (11:18 -0500)]
ARM: Use a stl queue for the table walker state

13 years agoCPU: Set a default value when readBytes faults.
Ali Saidi [Mon, 23 Aug 2010 16:18:39 +0000 (11:18 -0500)]
CPU: Set a default value when readBytes faults.

This was being done in read(), but if readBytes was called directly it
wouldn't happen. Also, instead of setting the memory blob being read to -1
which would (I believe) require using memset with -1 as a parameter, this now
uses bzero. It's hoped that it's more specialized behavior will make it
slightly faster.

13 years agoCompiler: Fixes for GCC 4.5.
Ali Saidi [Mon, 23 Aug 2010 16:18:39 +0000 (11:18 -0500)]
Compiler: Fixes for GCC 4.5.

13 years agoBASE: Fix genrand to generate both 0s and 1s when max equals one.
Ali Saidi [Mon, 23 Aug 2010 16:18:39 +0000 (11:18 -0500)]
BASE: Fix genrand to generate both 0s and 1s when max equals one.
previously was only generating 0s.

13 years agostats: Fix off-by-one error in distributions.
Ali Saidi [Mon, 23 Aug 2010 16:18:39 +0000 (11:18 -0500)]
stats: Fix off-by-one error in distributions.

bkt size isn't evenly divisible by max-min and it would round down,
it's possible to sample a distribution and have no place to put the sample.
When this case occured the simulator would assert.

13 years agoX86: Get rid of unused file arguments.hh.
Gabe Black [Mon, 23 Aug 2010 01:42:23 +0000 (18:42 -0700)]
X86: Get rid of unused file arguments.hh.

13 years agoSPARC: Fix some style issues in utility.hh.
Gabe Black [Mon, 23 Aug 2010 01:39:39 +0000 (18:39 -0700)]
SPARC: Fix some style issues in utility.hh.

13 years agoX86: Get rid of the unused getAllocator on the python base microop class.
Gabe Black [Mon, 23 Aug 2010 01:24:09 +0000 (18:24 -0700)]
X86: Get rid of the unused getAllocator on the python base microop class.

This function is always overridden, and doesn't actually have the right
signature.

13 years agoregress: Regression tester updates
Brad Beckmann [Sat, 21 Aug 2010 00:44:26 +0000 (17:44 -0700)]
regress: Regression tester updates

Regression tester updates required by the following patches:

brad/moved_python_protocol_files: config: moved python protocol config files
brad/ruby_options_movement: config: reorganized how ruby specifies command-line options
brad/config_token_bcast: ruby: added token broadcast config params to cmd options
brad/topology_name: config: Added the topology description to m5 config.ini
brad/ruby_system_names: config: Improve ruby simobject names
brad/consolidated_protocol_stats: slicc: Consolidated the protocol stats printing
brad/ruby_request_type_ostream_fix: ruby: Added ruby_request_type ostream def to libruby.hh
brad/memtest_dma_extension: memtest: Memtester support for DMA
brad/token_dma_lockdown_fix: MOESI_CMP_token: Fixed dma persistent lockdown bugs
brad/profile_generic_mach_type: ruby: Reincarnated the responding machine profiling
brad/network_msg_consolidated_stats: ruby: Added consolidated network msg stats
brad/bcast_msg_profiling: ruby: Added bcast msg profiling to hammer and token
brad/l2cache_profiling_fix: ruby: Fixed L2 cache miss profiling
brad/llsc_ruby_m5_fix: ruby: fix ruby llsc support to sync sc outcomes
brad/ruby_latency_fixes: ruby: Reduced ruby latencies
brad/hammer_l2_cache_latency: ruby: Updated MOESI_hammer L2 latency behavior
brad/deterministic_resurrection: ruby: Resurrected Ruby's deterministic tests
brad/token_dma_fixes: ruby: MOESI_CMP_token dma fixes
brad/ruby_cmd_options: config: added cmd options to control ruby debug
brad/token_owner_fixes: ruby: fixed token bugs associated with owner token counts
brad/ruby_remove_try_except: ruby: Improved try except blocks in ruby creation
brad/ruby_port_callback_fix: ruby: Fixed RubyPort sendTiming callbacks
brad/interrupt_drain_fix: devices: Fixed periodic interrupts to work with draining
brad/llsc_trace_profile: ruby: Added SC fail indication to trace profiling
brad/no_migrate_atomic: ruby: Disable migratory sharing for token and hammer
brad/ruby_start_time_fix: ruby: Reset ruby stats in RubySystem unserialize
brad/numa_bit_select_fix: ruby: fixed DirectoryMemory's numa_high_bit configuration
brad/hammer_probe_filter: ruby: added probe filter support to hammer
brad/miss_latency_detail_profile: MOESI_hammer: break down miss latency stalled cycles
brad/recycle_latency_fix: ruby: Recycle latency fix for hammer
brad/stall_and_wait: ruby: Stall and wait input messages instead of recycling
brad/rubytest_request_flag_fix: ruby: Fixed minor bug in ruby test for setting the request type
brad/hammer_merge_gets: ruby: Added merge GETS optimization to hammer
brad/regress_updates: regress: Regression tester updates

13 years agoruby: Added merge GETS optimization to hammer
Brad Beckmann [Fri, 20 Aug 2010 18:46:14 +0000 (11:46 -0700)]
ruby: Added merge GETS optimization to hammer

Added an optimization that merges multiple pending GETS requests into a
single request to the owner node.

13 years agoruby: Fixed minor bug in ruby test for setting the request type
Brad Beckmann [Fri, 20 Aug 2010 18:46:14 +0000 (11:46 -0700)]
ruby: Fixed minor bug in ruby test for setting the request type

13 years agoruby: Stall and wait input messages instead of recycling
Brad Beckmann [Fri, 20 Aug 2010 18:46:14 +0000 (11:46 -0700)]
ruby: Stall and wait input messages instead of recycling

This patch allows messages to be stalled in their input buffers and wait
until a corresponding address changes state.  In order to make this work,
all in_ports must be ranked in order of dependence and those in_ports that
may unblock an address, must wake up the stalled messages.  Alot of this
complexity is handled in slicc and the specification files simply
annotate the in_ports.

--HG--
rename : src/mem/slicc/ast/CheckAllocateStatementAST.py => src/mem/slicc/ast/StallAndWaitStatementAST.py
rename : src/mem/slicc/ast/CheckAllocateStatementAST.py => src/mem/slicc/ast/WakeUpDependentsStatementAST.py

13 years agoruby: Recycle latency fix for hammer
Brad Beckmann [Fri, 20 Aug 2010 18:46:14 +0000 (11:46 -0700)]
ruby: Recycle latency fix for hammer

Patch allows each individual message buffer to have different recycle latencies
and allows the overall recycle latency to be specified at the cmd line. The
patch also adds profiling info to make sure no one processor's requests are
recycled too much.

13 years agoMOESI_hammer: break down miss latency stalled cycles
Brad Beckmann [Fri, 20 Aug 2010 18:46:14 +0000 (11:46 -0700)]
MOESI_hammer: break down miss latency stalled cycles

This patch tracks the number of cycles a transaction is delayed at different
points of the request-forward-response loop.

13 years agoruby: added probe filter support to hammer
Brad Beckmann [Fri, 20 Aug 2010 18:46:14 +0000 (11:46 -0700)]
ruby: added probe filter support to hammer

13 years agoruby: fixed DirectoryMemory's numa_high_bit configuration
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: fixed DirectoryMemory's numa_high_bit configuration

This fix includes the off-by-one bit selection bug for numa mapping.

13 years agoruby: Reset ruby stats in RubySystem unserialize
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: Reset ruby stats in RubySystem unserialize

The main purpose for clearing stats in the unserialize process is so
that the profiler can correctly set its start time to the unserialized
value of curTick.

13 years agoruby: Disable migratory sharing for token and hammer
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: Disable migratory sharing for token and hammer

This patch allows one to disable migratory sharing for those cache blocks that
are accessed by atomic requests.  While the implementations are different
between the token and hammer protocols, the motivation is the same.  For
Alpha, LLSC semantics expect that normal loads do not unlock cache blocks that
have been locked by LL accesses.  Therefore, locked blocks should not transfer
write permissions when responding to these load requests.  Instead, only they
only transfer read permissions so that the subsequent SC access can possibly
succeed.

13 years agoruby: Added SC fail indication to trace profiling
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: Added SC fail indication to trace profiling

13 years agodevices: Fixed periodic interrupts to work with draining
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
devices: Fixed periodic interrupts to work with draining

Added drain functions to the RTC and 8254 timer so that periodic interrupts
stop when the system is draining.  This patch is needed to checkpoint in
timing mode.  Otherwise under certain situations, the event queue will never
be completely empty.

13 years agoruby: Fixed RubyPort sendTiming callbacks
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: Fixed RubyPort sendTiming callbacks

Fixed RubyPort schedSendTiming calls to match ruby frequency.

13 years agoruby: Improved try except blocks in ruby creation
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: Improved try except blocks in ruby creation

Replaced the sys.exit in the try-except blocks with raise so that the python
call stack will be printed

13 years agoruby: fixed token bugs associated with owner token counts
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: fixed token bugs associated with owner token counts

This patch fixes several bugs related to previous inconsistent assumptions on
how many tokens the Owner had.  Mike Marty should have fixes these bugs years
ago.  :)

13 years agoconfig: added cmd options to control ruby debug
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
config: added cmd options to control ruby debug

13 years agoruby: MOESI_CMP_token dma fixes
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: MOESI_CMP_token dma fixes

This patch fixes various protocol bugs regarding races between dma requests
and persistent requests.

13 years agoruby: Resurrected Ruby's deterministic tests
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: Resurrected Ruby's deterministic tests

Added the request series and invalidate deterministic tests as new cpu models
and removed the no longer needed ruby tests

--HG--
rename : configs/example/rubytest.py => configs/example/determ_test.py
rename : src/mem/ruby/tester/DetermGETXGenerator.cc => src/cpu/directedtest/DirectedGenerator.cc
rename : src/mem/ruby/tester/DetermGETXGenerator.hh => src/cpu/directedtest/DirectedGenerator.hh
rename : src/mem/ruby/tester/DetermGETXGenerator.cc => src/cpu/directedtest/InvalidateGenerator.cc
rename : src/mem/ruby/tester/DetermGETXGenerator.hh => src/cpu/directedtest/InvalidateGenerator.hh
rename : src/cpu/rubytest/RubyTester.cc => src/cpu/directedtest/RubyDirectedTester.cc
rename : src/cpu/rubytest/RubyTester.hh => src/cpu/directedtest/RubyDirectedTester.hh
rename : src/mem/ruby/tester/DetermGETXGenerator.cc => src/cpu/directedtest/SeriesRequestGenerator.cc
rename : src/mem/ruby/tester/DetermGETXGenerator.hh => src/cpu/directedtest/SeriesRequestGenerator.hh

13 years agoruby: Updated MOESI_hammer L2 latency behavior
Brad Beckmann [Fri, 20 Aug 2010 18:46:13 +0000 (11:46 -0700)]
ruby: Updated MOESI_hammer L2 latency behavior

Previously, the MOESI_hammer protocol calculated the same latency for L1 and
L2 hits.  This was because the protocol was written using the old ruby
assumption that L1 hits used the sequencer fast path.  Since ruby no longer
uses the fast-path, the protocol delays L2 hits by placing them on the
trigger queue.

13 years agoruby: Reduced ruby latencies
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
ruby: Reduced ruby latencies

The previous slower ruby latencies created a mismatch between the faster M5
cpu models and the much slower ruby memory system.  Specifically smp
interrupts were much slower and infrequent, as well as cpus moving in and out
of spin locks.  The result was many cpus were idle for large periods of time.

These changes fix the latency mismatch.

13 years agoruby: fix ruby llsc support to sync sc outcomes
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
ruby: fix ruby llsc support to sync sc outcomes

Added support so that ruby can determine the outcome of store conditional
operations and reflect that outcome to M5 physical memory and cpus.

13 years agoruby: Fixed L2 cache miss profiling
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
ruby: Fixed L2 cache miss profiling

Fixed L2 cache miss profiling for the MOESI_CMP_token protocol

13 years agoruby: Added bcast msg profiling to hammer and token
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
ruby: Added bcast msg profiling to hammer and token

13 years agoruby: Added consolidated network msg stats
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
ruby: Added consolidated network msg stats

13 years agoruby: Reincarnated the responding machine profiling
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
ruby: Reincarnated the responding machine profiling

This patch adds back to ruby the capability to understand the response time
for messages that hit in different levels of the cache heirarchy.
Specifically add support for the MI_example, MOESI_hammer, and MOESI_CMP_token
protocols.

13 years agoMOESI_CMP_token: Fixed dma persistent lockdown bugs
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
MOESI_CMP_token: Fixed dma persistent lockdown bugs

13 years agomemtest: Memtester support for DMA
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
memtest: Memtester support for DMA

This patch adds DMA testing to the Memtester and is inherits many changes from
Polina's old tester_dma_extension patch.  Since Ruby does not work in atomic
mode, the atomic mode options are removed.

13 years agoruby: Added ruby_request_type ostream def to libruby.hh
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
ruby: Added ruby_request_type ostream def to libruby.hh

13 years agoslicc: Consolidated the protocol stats printing
Brad Beckmann [Fri, 20 Aug 2010 18:46:12 +0000 (11:46 -0700)]
slicc: Consolidated the protocol stats printing

Created a separate ProfileDumper that consolidates the generated stats for
each controller of a certain type.

13 years agoconfig: Improve ruby simobject names
Brad Beckmann [Fri, 20 Aug 2010 18:46:11 +0000 (11:46 -0700)]
config: Improve ruby simobject names

This patch attaches ruby objects to the system before the topology is
created so that their simobject names read their meaningful variable
names instead of their topology name.

13 years agoconfig: Added the topology description to m5 config.ini
Brad Beckmann [Fri, 20 Aug 2010 18:46:11 +0000 (11:46 -0700)]
config: Added the topology description to m5 config.ini

13 years agoruby: added token broadcast config params to cmd options
Brad Beckmann [Fri, 20 Aug 2010 18:46:11 +0000 (11:46 -0700)]
ruby: added token broadcast config params to cmd options

13 years agoconfig: reorganized how ruby specifies command-line options
Brad Beckmann [Fri, 20 Aug 2010 18:44:09 +0000 (11:44 -0700)]
config: reorganized how ruby specifies command-line options

13 years agoruby: Fixed printout when Sequencer detects a deadlock
Brad Beckmann [Fri, 20 Aug 2010 18:41:35 +0000 (11:41 -0700)]
ruby: Fixed printout when Sequencer detects a deadlock

13 years agoMESI_CMP_directory: bug fix for old PUTX requests
Brad Beckmann [Fri, 20 Aug 2010 18:41:35 +0000 (11:41 -0700)]
MESI_CMP_directory: bug fix for old PUTX requests

13 years agoconfig: moved python protocol config files
Brad Beckmann [Fri, 20 Aug 2010 18:41:35 +0000 (11:41 -0700)]
config: moved python protocol config files

Moved the python protocol config files back to their original location to avoid
addToPath calls.

--HG--
rename : configs/ruby/protocols/MESI_CMP_directory.py => configs/ruby/MESI_CMP_directory.py
rename : configs/ruby/protocols/MI_example.py => configs/ruby/MI_example.py
rename : configs/ruby/protocols/MOESI_CMP_directory.py => configs/ruby/MOESI_CMP_directory.py
rename : configs/ruby/protocols/MOESI_CMP_token.py => configs/ruby/MOESI_CMP_token.py
rename : configs/ruby/protocols/MOESI_hammer.py => configs/ruby/MOESI_hammer.py

13 years agomisc: add some AMD copyright notices
Steve Reinhardt [Tue, 17 Aug 2010 12:49:05 +0000 (05:49 -0700)]
misc: add some AMD copyright notices
Meant to add these with the previous batch of csets.

13 years agox86: minor checkpointing bug fixes
Steve Reinhardt [Tue, 17 Aug 2010 12:20:39 +0000 (05:20 -0700)]
x86: minor checkpointing bug fixes

13 years agosim: revamp unserialization procedure
Steve Reinhardt [Tue, 17 Aug 2010 12:17:06 +0000 (05:17 -0700)]
sim: revamp unserialization procedure

Replace direct call to unserialize() on each SimObject with a pair of
calls for better control over initialization in both ckpt and non-ckpt
cases.

If restoring from a checkpoint, loadState(ckpt) is called on each
SimObject.  The default implementation simply calls unserialize() if
there is a corresponding checkpoint section, so we get backward
compatibility for existing objects.  However, objects can override
loadState() to get other behaviors, e.g., doing other programmed
initializations after unserialize(), or complaining if no checkpoint
section is found.  (Note that the default warning for a missing
checkpoint section is now gone.)

If not restoring from a checkpoint, we call the new initState() method
on each SimObject instead.  This provides a hook for state
initializations that are only required when *not* restoring from a
checkpoint.

Given this new framework, do some cleanup of LiveProcess subclasses
and X86System, which were (in some cases) emulating initState()
behavior in startup via a local flag or (in other cases) erroneously
doing initializations in startup() that clobbered state loaded earlier
by unserialize().

13 years agosim: fold checkpoint restore code into instantiate()
Steve Reinhardt [Tue, 17 Aug 2010 12:17:06 +0000 (05:17 -0700)]
sim: fold checkpoint restore code into instantiate()
The separate restoreCheckpoint() call is gone; just pass
the checkpoint dir as an optional arg to instantiate().
This change is a precursor to some more extensive
reworking of the startup code.

13 years agoconfigs: clean up checkpoint code in Simulation.py
Steve Reinhardt [Tue, 17 Aug 2010 12:17:06 +0000 (05:17 -0700)]
configs: clean up checkpoint code in Simulation.py
Small change to clean up some redundant code.
Should not have any functional impact.

13 years agotest: Update stats for python object iteration.
Steve Reinhardt [Tue, 17 Aug 2010 12:14:03 +0000 (05:14 -0700)]
test: Update stats for python object iteration.
Small changes in tests with data races due to new object creation
order.

13 years agosim: clean up child handling
Steve Reinhardt [Tue, 17 Aug 2010 12:11:00 +0000 (05:11 -0700)]
sim: clean up child handling
The old code for handling SimObject children was kind of messy,
with children stored both in _values and _children, and
inconsistent and potentially buggy handling of SimObject
vectors.  Now children are always stored in _children, and
SimObject vectors are consistently handled using the
SimObjectVector class.

Also, by deferring the parenting of SimObject-valued parameters
until the end (instead of doing it at assignment), we eliminate
the hole where one could assign a vector of SimObjects to a
parameter then append to that vector, with the appended objects
never getting parented properly.

This patch induces small stats changes in tests with data races
due to changes in the object creation & initialization order.
The new code does object vectors in order and so should be more
stable.

13 years agosim: move iterating over SimObjects into Python.
Steve Reinhardt [Tue, 17 Aug 2010 12:08:50 +0000 (05:08 -0700)]
sim: move iterating over SimObjects into Python.

13 years agosim: fail on implicit creation of orphans via ports
Steve Reinhardt [Tue, 17 Aug 2010 12:06:22 +0000 (05:06 -0700)]
sim: fail on implicit creation of orphans via ports
Orphan SimObjects (not in the config hierarchy) could get
created implicitly if they have a port connection to a SimObject
that is in the hierarchy.  This means that there are objects on
the C++ SimObject list (created via the C++ SimObject
constructor call) that are unknown to Python and will get
skipped if we walk the hierarchy from the Python side (as we are
about to do).  This patch detects this situation and prints an
error message.

Also fix the rubytester config script which happened to rely on
this behavior.

13 years agosim: make Python Root object a singleton
Steve Reinhardt [Tue, 17 Aug 2010 12:06:22 +0000 (05:06 -0700)]
sim: make Python Root object a singleton
Enforce that the Python Root SimObject is instantiated only
once.  The C++ Root object already panics if more than one is
created.  This change avoids the need to track what the root
object is, since it's available from Root.getInstance() (if it
exists).  It's now redundant to have the user pass the root
object to functions like instantiate(), checkpoint(), and
restoreCheckpoint(), so that arg is gone.  Users who use
configs/common/Simulate.py should not notice.

13 years agotests: update reference config.ini files for previous cset
Steve Reinhardt [Tue, 17 Aug 2010 12:06:22 +0000 (05:06 -0700)]
tests: update reference config.ini files for previous cset
Rename 'responder_set' to 'use_default_range'.

13 years agobus: clean up default responder code.
Steve Reinhardt [Tue, 17 Aug 2010 12:06:21 +0000 (05:06 -0700)]
bus: clean up default responder code.
Clean up some minor things left over from the default responder
change in rev 9af6fb59752f.  Mostly renaming the 'responder_set'
param to 'use_default_range' to actually reflect what it does...
old name wasn't that descriptive in the first place, but now
it really doesn't make sense at all.

Also got rid of the bogus obsolete assignment to 'bus.responder'
which used to be a parameter but now is interpreted as an
implicit child assignment, and which was giving me problems in
the config restructuring to come.  (A good argument for not
allowing implicit child assignments, IMO, but that's water under
the bridge, I'm afraid.)

Also moved the Bus constructor to the .cc file since that's
where it should have been all along.

13 years agoInorder: Fix compilation of m5.fast.
Gabe Black [Sat, 14 Aug 2010 08:00:45 +0000 (01:00 -0700)]
Inorder: Fix compilation of m5.fast.

printMemData is only used in DPRINTFs. If those are removed by compiling
m5.fast, that function is unused, gcc generates a warning, that gets turned
into an error, and the build fails. This change surrounds the function
definition with #if TRACING_ON so it only gets compiled in if the DPRINTFs do
to.

13 years agoMerge with head.
Gabe Black [Fri, 13 Aug 2010 13:16:30 +0000 (06:16 -0700)]
Merge with head.

13 years agoCPU: Add readBytes and writeBytes functions to the exec contexts.
Gabe Black [Fri, 13 Aug 2010 13:16:02 +0000 (06:16 -0700)]
CPU: Add readBytes and writeBytes functions to the exec contexts.

13 years agoInOrder: Clean up some DPRINTFs that print data sent to/from the cache.
Gabe Black [Fri, 13 Aug 2010 13:16:00 +0000 (06:16 -0700)]
InOrder: Clean up some DPRINTFs that print data sent to/from the cache.

13 years agoCPU: Tidy up endianness handling for mmapped "IPR"s.
Gabe Black [Fri, 13 Aug 2010 13:10:45 +0000 (06:10 -0700)]
CPU: Tidy up endianness handling for mmapped "IPR"s.

13 years agoutil/m5/m5.c: ensure readfile() buffer pages are in page table
Joel Hestness [Fri, 13 Aug 2010 00:16:04 +0000 (17:16 -0700)]
util/m5/m5.c: ensure readfile() buffer pages are in page table
(and marked dirty, in case that matters) by touching them beforehand

13 years agoTimingSimpleCPU: fix NO_ACCESS memory op handling
Joel Hestness [Fri, 13 Aug 2010 00:16:02 +0000 (17:16 -0700)]
TimingSimpleCPU: fix NO_ACCESS memory op handling

When a request is NO_ACCESS (x86 CDA microinstruction), the memory op
doesn't go to the cache, so TimingSimpleCPU::completeDataAccess needs
to handle the case where the current status of the CPU is Running
and not DcacheWaitResponse or DTBWaitResponse

13 years agoNone, not none
Nathan Binkert [Mon, 9 Aug 2010 05:57:16 +0000 (22:57 -0700)]
None, not none

13 years ago.hgignore: added src/doxygen
Steve Reinhardt [Wed, 28 Jul 2010 03:00:38 +0000 (20:00 -0700)]
.hgignore: added src/doxygen

13 years agoARM: Add regression tests
Ali Saidi [Tue, 27 Jul 2010 05:03:44 +0000 (01:03 -0400)]
ARM: Add regression tests

13 years agoPower: The condition register should be set or cleared upon a system call
Timothy M. Jones [Thu, 22 Jul 2010 17:54:37 +0000 (18:54 +0100)]
Power: The condition register should be set or cleared upon a system call
return to indicate success or failure.

13 years agoLSQ Unit: After deleting part of a split request, set it to NULL so that it
Timothy M. Jones [Thu, 22 Jul 2010 17:54:37 +0000 (18:54 +0100)]
LSQ Unit: After deleting part of a split request, set it to NULL so that it
isn't accidentally deleted again later (causing a segmentation fault).

13 years agoPort: Only indicate that a SimpleTimingPort is drained if its send event is
Timothy M. Jones [Thu, 22 Jul 2010 17:54:37 +0000 (18:54 +0100)]
Port: Only indicate that a SimpleTimingPort is drained if its send event is
not scheduled, as well as the transmit list being empty.

13 years agoO3CPU: Fix a bug where stores in the cpu where never marked as split.
Timothy M. Jones [Thu, 22 Jul 2010 17:52:02 +0000 (18:52 +0100)]
O3CPU: Fix a bug where stores in the cpu where never marked as split.

13 years agoSyscall: Don't close the simulator's standard file descriptors.
Timothy M. Jones [Thu, 22 Jul 2010 17:47:52 +0000 (18:47 +0100)]
Syscall: Don't close the simulator's standard file descriptors.

13 years agoO3CPU: O3's tick event gets squashed when it is switched out. When repeatedly
Timothy M. Jones [Thu, 22 Jul 2010 17:47:43 +0000 (18:47 +0100)]
O3CPU: O3's tick event gets squashed when it is switched out.  When repeatedly
switching between O3 and another CPU, O3's tick event might still be scheduled
in the event queue (as squashed).  Therefore, check for a squashed tick event
as well as a non-scheduled event when taking over from another CPU and deal
with it accordingly.

13 years agoPower: Provide a utility function to copy registers from one thread context
Timothy M. Jones [Thu, 22 Jul 2010 17:47:03 +0000 (18:47 +0100)]
Power: Provide a utility function to copy registers from one thread context
to another in the Power ISA.

13 years agostats: unify the two stats distribution type better
Nathan Binkert [Thu, 22 Jul 2010 01:54:53 +0000 (18:54 -0700)]
stats: unify the two stats distribution type better

13 years agostats: cleanup a few small problems in stats
Nathan Binkert [Wed, 21 Jul 2010 22:53:53 +0000 (15:53 -0700)]
stats: cleanup a few small problems in stats

13 years agopython: add a sorted dictionary class
Nathan Binkert [Wed, 21 Jul 2010 22:53:53 +0000 (15:53 -0700)]
python: add a sorted dictionary class
It would be nice if python had a tree class that would do this for real,
but since we don't, we'll just keep a sorted list of keys and update
it on demand.

13 years agopython: Add mechanism to override code compiled into the exectuable
Nathan Binkert [Wed, 21 Jul 2010 22:53:52 +0000 (15:53 -0700)]
python: Add mechanism to override code compiled into the exectuable
If the user sets the environment variable M5_OVERRIDE_PY_SOURCE to
True, then imports that would normally find python code compiled into
the executable will instead first check in the absolute location where
the code was found during the build of the executable.  This only
works for files in the src (or extras) directories, not automatically
generated files.

This is a developer feature!

13 years agoFix x86 XCHG macro-op to use locked micro-ops for all memory accesses
Tushar Krishna [Wed, 21 Jul 2010 16:55:57 +0000 (09:55 -0700)]
Fix x86 XCHG macro-op to use locked micro-ops for all memory accesses

13 years agoSimObject: transparently forward Python attribute refs to C++.
Steve Reinhardt [Sat, 17 Jul 2010 15:56:49 +0000 (08:56 -0700)]
SimObject: transparently forward Python attribute refs to C++.
This tidbit was pulled from a larger patch for Tim's sake, so
the comment reflects functions that haven't been exported yet.
I hope to commit them soon so it didn't seem worth cleaning up.

13 years agoARM: Make an SRS instruction with a bad mode cause an undefined instruction fault.
Gabe Black [Thu, 15 Jul 2010 09:11:56 +0000 (02:11 -0700)]
ARM: Make an SRS instruction with a bad mode cause an undefined instruction fault.

13 years agoARM: Adjust the FP_Base_DepTag to be larger than the largest int reg index.
Gabe Black [Wed, 14 Jul 2010 05:41:47 +0000 (22:41 -0700)]
ARM: Adjust the FP_Base_DepTag to be larger than the largest int reg index.

13 years agocache: fix bug in SC upgrade handling
Steve Reinhardt [Fri, 9 Jul 2010 00:56:13 +0000 (17:56 -0700)]
cache: fix bug in SC upgrade handling
This bug was introduced with the recent rework of SC
failure handling in cset f97b62be544f.

13 years agogarnet: Added topology print function to Garnet printStats
Brad Beckmann [Thu, 8 Jul 2010 23:18:20 +0000 (16:18 -0700)]
garnet: Added topology print function to Garnet printStats

13 years agoNetworkMessage copy constructor fix
Tushar Krishna [Thu, 8 Jul 2010 23:18:20 +0000 (16:18 -0700)]
NetworkMessage copy constructor fix

13 years agocheckpointing: another small overload fix
Steve Reinhardt [Tue, 6 Jul 2010 05:57:23 +0000 (22:57 -0700)]
checkpointing: another small overload fix
On Nate's advice, overload 'char' as well as 'signed char'
and 'unsigned char'.

13 years agosim: allow SimObject subclasses to define classmethods
Steve Reinhardt [Tue, 6 Jul 2010 04:39:38 +0000 (21:39 -0700)]
sim: allow SimObject subclasses to define classmethods
(without requiring a leading underscore)
Also a little cleanup on type names in SimObject.py.

13 years agosim: fold StartupCallback into SimObject
Steve Reinhardt [Tue, 6 Jul 2010 04:39:38 +0000 (21:39 -0700)]
sim: fold StartupCallback into SimObject
There used to be a reason to have StartupCallback
be a separate object, but not any more.  Now
it's just confusing.

13 years agocheckpointing: minor cleanup.
Steve Reinhardt [Tue, 6 Jul 2010 04:39:38 +0000 (21:39 -0700)]
checkpointing: minor cleanup.
Move some static checkpoint stuff into the
Checkpoint object namespace.

13 years agocheckpointing: fix minor bug
Steve Reinhardt [Tue, 6 Jul 2010 04:39:38 +0000 (21:39 -0700)]
checkpointing: fix minor bug
Somehow we now need to explicitly specialize on
'signed char' and not just 'char' to catch cases
like int8_t

13 years agoutil: add a script for testing checkpointing
Steve Reinhardt [Tue, 6 Jul 2010 04:39:38 +0000 (21:39 -0700)]
util: add a script for testing checkpointing
See comments in util/checkpoint-tester.py for details.

13 years agoprocess: get rid of some unused code & vars
Steve Reinhardt [Tue, 6 Jul 2010 04:39:38 +0000 (21:39 -0700)]
process: get rid of some unused code & vars

13 years agoprocess: minor format/style cleanup
Steve Reinhardt [Tue, 6 Jul 2010 04:39:38 +0000 (21:39 -0700)]
process: minor format/style cleanup

14 years agostyle: updated garnet to match M5 coding style
Tushar Krishna [Tue, 22 Jun 2010 22:36:07 +0000 (15:36 -0700)]
style: updated garnet to match M5 coding style

13 years agoinorder: remove another debug stat
Korey Sewell [Mon, 28 Jun 2010 11:33:33 +0000 (07:33 -0400)]
inorder: remove another debug stat

13 years agoinorder: remove debugging stat
Korey Sewell [Sat, 26 Jun 2010 13:41:39 +0000 (09:41 -0400)]
inorder: remove debugging stat
m5 doesnt do stats specific to binary and this resource request stat is probably only
useful for people who really know the ins/outs of the model anyway

13 years agoinorder: update regressions from RAS fix
Korey Sewell [Fri, 25 Jun 2010 21:42:55 +0000 (17:42 -0400)]
inorder: update regressions from RAS fix

13 years agoinorder: Return Address Stack bug
Korey Sewell [Fri, 25 Jun 2010 21:42:35 +0000 (17:42 -0400)]
inorder: Return Address Stack bug
the nextPC was getting sent to the branch predictor not the current PC, so
the RAS was returning the wrong PC and mispredicting everything.

13 years agoinorder: resource scheduling backend
Korey Sewell [Fri, 25 Jun 2010 21:42:34 +0000 (17:42 -0400)]
inorder: resource scheduling backend
replace priority queue with vector of lists(1 list per stage) and place inside a class
so that we have more control of when an instruction uses a particular schedule entry
...
also, this is the 1st step toward making the InOrderCPU fully parameterizable. See the
wiki for details on this process

13 years agoX86: Fix div2 flag calculation.
Gabe Black [Fri, 25 Jun 2010 07:21:48 +0000 (00:21 -0700)]
X86: Fix div2 flag calculation.

14 years agoinorder: update regressions
Korey Sewell [Thu, 24 Jun 2010 19:34:21 +0000 (15:34 -0400)]
inorder: update regressions