This patch contains a new CPU model named `Minor'. Minor models a four
stage in-order execution pipeline (fetch lines, decompose into
macroops, decompose macroops into microops, execute).
The model was developed to support the ARM ISA but should be fixable
to support all the remaining gem5 ISAs. It currently also works for
Alpha, and regressions are included for ARM and Alpha (including Linux
boot).
Documentation for the model can be found in src/doc/inside-minor.doxygen and
its internal operations can be visualised using the Minorview tool
utils/minorview.py.
Minor was designed to be fairly simple and not to engage in a lot of
instruction annotation. As such, it currently has very few gathered
stats and may lack other gem5 features.
Minor is faster than the o3 model. Sample results:
Benchmark | Stat host_seconds (s)
---------------+--------v--------v--------
(on ARM, opt) | simple | o3 | minor
| timing | timing | timing
---------------+--------+--------+--------
10.linux-boot | 169 | 1883 | 1075
10.mcf | 117 | 967 | 491
20.parser | 668 | 6315 | 3146
30.eon | 542 | 3413 | 2414
40.perlbmk | 2339 | 20905 | 11532
50.vortex | 122 | 1094 | 588
60.bzip2 | 2045 | 18061 | 9662
70.twolf | 207 | 2736 | 1036
Only printing one rather than two args for the ignored syscall
warning means the count of register accesses has changed on
a few runs. Oddly only Alpha Tru64 seems to have any ignored
syscalls in the regression tests.
Stop setting the use_default_range flag in PioBus in order to
have random bad addresses result in a BadAddress response and
not a gem5 fatal error. This is necessary in Ruby as Ruby is
connected directly to PioBus, so misspeculated addresses will
be sent there directly. For the classic memory system, this
change has no effect, as bad addresses are caught by the
memory bus before being sent to the PioBus.
This work was done while Binh was an intern at AMD Research.
The System object has a static MemoryModeStrings array
that's (1) unused and (2) redundant, since there's an
auto-generated version in the Enums namespace. No
point in leaving it in.
When we switched getSyscallArg() from explicit arg indices to
the implicit method, some DPRINTF arguments were left as calls
to getSyscallArg(), even though C/C++ doesn't guarantee
anything about the order of invocation of these calls. As a
result, the args could be printed out in arbitrary orders.
Interestingly, this bug has been around since 2009:
http://repo.gem5.org/gem5/rev/4842482e1bd1
this operator uses memcmp() to detect if two EthAddr object have the same
address, however memcmp() will return 0 if all bytes are equal. operator==
returns the return value of memcmp() to indicate whether or not two
address are equal. this is incorrect as it will always give the opposite of
the intended behavior. this patch fixes that problem.
per the IEEE 802 spec:
1) fixed broadcast() to ensure that all bytes are equal to 0xff.
2) fixed unicast() to ensure that bit 0 of the first byte is equal to 0
3) fixed multicast() to ensure that bit 0 of the first byte is equal to 1, and
that it is not a broadcast.
also the constructors in EthAddr are fixed so that all bytes of data are
initialized.
Adds DVFS capabilities to gem5, by allowing users to specify lists for
frequencies and voltages in SrcClockDomains and VoltageDomains respectively.
A separate component, DVFSHandler, provides a small interface to change
operating points of the associated domains.
Clock domains will be linked to voltage domains and thus allow separate clock,
but shared voltage lines.
Currently all the valid performance-level updates are performed with a fixed
transition latency as specified for the domain.
Config file example:
...
vd = VoltageDomain(voltage = ['1V','0.95V','0.90V','0.85V'])
tsys.cluster1.clk_domain.clock = ['1GHz','700MHz','400MHz','230MHz']
tsys.cluster2.clk_domain.clock = ['1GHz','700MHz','400MHz','230MHz']
tsys.cluster1.clk_domain.domain_id = 0
tsys.cluster2.clk_domain.domain_id = 1
tsys.cluster1.clk_domain.voltage_domain = vd
tsys.cluster2.clk_domain.voltage_domain = vd
tsys.dvfs_handler.domains = [tsys.cluster1.clk_domain,
tsys.cluster2.clk_domain]
tsys.dvfs_handler.enable = True
This patch adds a first version of a script that processes the debug
output and generates a command trace for DRAMPower. This is work in
progress and is intended as a snapshot of ongoing work at this point.
The longer term plan is to link in DRAMPower as a library and have one
instance of the model per rank, and instantiate it based on a struct
passed from gem5. Each command will then be a call to the model and no
parsing of traces will be necessary.
This patch adds a DRAMPower flag to enable off-line DRAM power
analysis using the DRAMPower tool. A new DRAMPower flag is added
and a follow-on patch adds a Python script to post-process the output
and order it based on time stamps.
The long-term goal is to link DRAMPower as a library and provide the
commands through function calls to the model rather than first
printing and then parsing the commands. At the moment it is also up to
the user to ensure that the same DRAM configuration is used by the
gem5 controller model and DRAMPower.
This patch adds the index of the bank and rank as a field so that we can
determine the identity of a given bank (reference or pointer) for the
power tracing. We also grab the opportunity of cleaning up the
arguments used for identifying the bank when activating.
Mostly small differences in total ticks, but O3 stall causes
shifted significantly.
30.eon does speed up by ~6% on Alpha and ARM, and 50.vortex
by 4.5% on ARM. At the other extreme, X86 70.twolf is 0.8%
slower.
In a cycle, we could see a R and W requests corresponding to the same
page walk being sent to the memory. During the cycle that assertion
happens, we have 2 responses corresponding to the R and W above. We
also have a 'read' variable to keep track of the inflight Read
request, this gets reset to NULL right after we send out any R
request; and gets set to the next R in the page walk when a response
comes back.
The issue we are seeing here is when we get a response for W request,
assert(!read) fires because we got a response for R request right
before this, hence we set 'read' to NOT NULL value, pointing to the
next R request in the pagewalk!
This work was done while Binh was an intern at AMD Research.
Check for free entries in Load Queue and Store Queue separately to
avoid cases when load cannot be renamed due to full Store Queue and
vice versa.
This work was done while Binh was an intern at AMD Research.
This patch bumps the supported version of gcc from 4.4 to 4.6, and
clang from 2.9 to 3.0. This enables, amongst other things, range-based
for loops, lambda expressions, etc. The STL implementation shipping
with 4.6 also has a full functional implementation of unique_ptr and
shared_ptr.
The style checker used to traverse symlinks if they pointed to files, which can
result in style checker failure if the pointed-to file doesn't exist. This
style check is actually unnecessary, since symlinks either point to other files
that are already style checked, or files outside gem5, which shouldn't be
checked. Skip symlinks.
The language describing the clockEdge and nextCycle functions were ambiguous,
and so were prone to misinterpretation/misuse. Clear up the comments to more
rigorously describe their functionality.
Add a regression tester to McPAT. Joel Hestness wrote these tests and Yasuko
Eckert modified them to reflect the new McPAT interface and other changes
the previous patch made.
This patch includes software engineering changes and some generic bug fixes
Joel Hestness and Yasuko Eckert made to McPAT 0.8. There are still known
issues/concernts we did not have a chance to address in this patch.
High-level changes in this patch include:
1) Making XML parsing modular and hierarchical:
- Shift parsing responsibility into the components
- Read XML in a (mostly) context-free recursive manner so that McPAT input
files can contain arbitrary component hierarchies
2) Making power, energy, and area calculations a hierarchical and recursive
process
- Components track their subcomponents and recursively call compute
functions in stages
- Make C++ object hierarchy reflect inheritance of classes of components
with similar structures
- Simplify computeArea() and computeEnergy() functions to eliminate
successive calls to calculate separate TDP vs. runtime energy
- Remove Processor component (now unnecessary) and introduce a more abstract
System component
3) Standardizing McPAT output across all components
- Use a single, common data structure for storing and printing McPAT output
- Recursively call print functions through component hierarchy
4) For caches, allow splitting data array and tag array reads and writes for
better accuracy
5) Improving the usability of CACTI by printing more helpful warning and error
messages
6) Minor: Impose more rigorous code style for clarity (more work still to be
done)
Overall, these changes greatly reduce the amount of replicated code, and they
improve McPAT runtime and decrease memory footprint.
Using '== true' in a boolean expression is totally redundant,
and using '== false' is pretty verbose (and arguably less
readable in most cases) compared to '!'.
It's somewhat of a pet peeve, perhaps, but I had some time
waiting for some tests to run and decided to clean these up.
Unfortunately, SLICC appears not to have the '!' operator,
so I had to leave the '== false' tests in the SLICC code.
This patch removes the stat totalCommittedInsts. This variable was used for
recording the total number of instructions committed across all the threads
of a core. The instructions committed by each thread are recorded invidually.
The total would now be generated by summing these individual counts.
in makeDualRoot() the etherlink interfaces are set using the tsunami interface
however, they are set again a few lines later based on whether or not the system
is a realview or tsunami system; the original assignment is always overwritten
or there will be a fatal. this seems like an artifact from when tsunami was the
only type of system capable of running with the dual option.
This patch bumps the bus clock speed such that the interconnect does
not become a bottleneck with a DDR4-2400-x64 DRAM delivering 19.2
GByte/s theoretical max.
This patch reflects the recent name change in the DRAM TrafficGen
tests and also tidies up the test directory.
--HG--
rename : tests/configs/tgen-simple-dram.py => tests/configs/tgen-dram-ctrl.py
rename : tests/quick/se/70.tgen/ref/null/none/tgen-simple-dram/config.ini => tests/quick/se/70.tgen/ref/null/none/tgen-dram-ctrl/config.ini
rename : tests/quick/se/70.tgen/ref/null/none/tgen-simple-dram/simerr => tests/quick/se/70.tgen/ref/null/none/tgen-dram-ctrl/simerr
rename : tests/quick/se/70.tgen/ref/null/none/tgen-simple-dram/simout => tests/quick/se/70.tgen/ref/null/none/tgen-dram-ctrl/simout
rename : tests/quick/se/70.tgen/ref/null/none/tgen-simple-dram/stats.txt => tests/quick/se/70.tgen/ref/null/none/tgen-dram-ctrl/stats.txt
rename : tests/quick/se/70.tgen/tgen-simple-dram.cfg => tests/quick/se/70.tgen/tgen-dram-ctrl.cfg
This patch makes a more firm connection between the DDR3-1600
configuration and the corresponding datasheet, and also adds a
DDR3-2133 and a DDR4-2400 configuration. At the moment there is also
an ongoing effort to align the choice of datasheets to what is
available in DRAMPower.
This patch extends the current timing parameters with the DRAM cycle
time. This is needed as the DRAMPower tool expects timestamps in DRAM
cycles. At the moment we could get away with doing this in a
post-processing step as the DRAMPower execution is separate from the
simulation run. However, in the long run we want the tool to be called
during the simulation, and then the cycle time is needed.
This patch adds the basic ingredients for a precharge all operation,
to be used in conjunction with DRAM power modelling.
Currently we do not try and apply any cleverness when precharging all
banks, thus even if only a single bank is open we use PREA as opposed
to PRE. At the moment we only have a single tRP (tRPpb), and do not
model the slightly longer all-bank precharge constraint (tRPab).
This patch adds the tRTP timing constraint, governing the minimum time
between a read command and a precharge. Default values are provided
for the existing DRAM types.
This patch merges the two control paths used to estimate the latency
and update the bank state. As a result of this merging the computation
is now in one place only, and should be easier to follow as it is all
done in absolute (rather than relative) time.
As part of this change, the scheduling is also refined to ensure that
we look at a sensible estimate of the bank ready time in choosing the
next request. The bank latency stat is removed as it ends up being
misleading when the DRAM access code gets evaluated ahead of time (due
to the eagerness of waking the model up for scheduling the next
request).
This patch adds the write recovery time to the DRAM timing
constraints, and changes the current tRASDoneAt to a more generic
preAllowedAt, capturing when a precharge is allowed to take place.
The part of the DRAM access code that accounts for the precharge and
activate constraints is updated accordingly.
