We were getting a spurious warning in the regressions that turned
out to be due to having the wrong value for TGT_MAP_ANONYMOUS for
Power Linux, but in the process of tracking it down I ended up
doing some cleanup of the mmap handling in general.
A significant contributor to the need for adoptOrphanParams()
is the practice of appending to SimObjectVectors which have
already been assigned as children. This practice sidesteps the
assignment operation for those appended SimObjects, which is
where parent/child relationships are typically established.
This patch reworks the config scripts that use append() on
SimObjectVectors, which all happen to be in the x86 system
configuration. At some point in the future, I hope to make
SimObjectVectors immutable (by deriving from tuple rather than
list), at which time this patch will be necessary for correct
operation. For now, it just avoids some of the warning
messages that get printed in adoptOrphanParams().
Re-enabling implicit parenting (see previous patch) causes current
Ruby config scripts to create some strange hierarchies and generate
several warnings. This patch makes three general changes to address
these issues.
1. The order of object creation in the ruby config files makes the L1
caches children of the sequencer rather than the controller; these
config ciles are rewritten to assign the L1 caches to the
controller first.
2. The assignment of the sequencer list to system.ruby.cpu_ruby_ports
causes the sequencers to be children of system.ruby, generating
warnings because they are already parented to their respective
controllers. Changing this attribute to _cpu_ruby_ports fixes this
because the leading underscore means this is now treated as a plain
Python attribute rather than a child assignment. As a result, the
configuration hierarchy changes such that, e.g.,
system.ruby.cpu_ruby_ports0 becomes system.l1_cntrl0.sequencer.
3. In the topology classes, the routers become children of some random
internal link node rather than direct children of the topology.
The topology classes are rewritten to assign the routers to the
topology object first.
Last summer's big rewrite of the initialization code (in
particular cset 6efc3672733b) got rid of the implicit parenting
that used to occur when an unparented SimObject was assigned as
a parameter value to another SimObject. The idea was that the
new adoptOrphanParams() step would catch these anyway so it was
unnecessary.
Unfortunately it turns out that adoptOrphanParams() has some
inherent instability in that the parent that does the adoption
depends on the config tree traversal order. Even making this
order deterministic (e.g., by traversing children in
alphabetical order) can introduce unwanted and unexpected
hierarchy changes between similar configs (e.g., when adding a
switch_cpu in place of a cpu), causing problems when trying to
restore checkpoints across similar configs. The hierarchy
created by implicit parenting is more stable and more
controllable, so this patch turns that behavior back on.
This patch also cleans up some long-standing holes regarding
parenting of SimObjects that are created in class definitions
(either in the body of the class, or as default parameters).
To avoid breaking some existing config files, this necessitated
changing the error on reparenting children to a warning. This
change fixes another bug where attempting to print the prior
error message would fail on reparenting SimObjectVectors
because they lack a _parent attribute. Some further issues
with SimObjectVectors were cleaned up by getting rid of the
get_parent() call (which could cause errors with some
SimObjectVectors where there was no single parent to return)
with has_parent() (since all the uses of get_parent() were just
boolean tests anyway).
Finally, since the adoptOrphanParam() step turned out to be so
problematic, we now issue a warning when it actually has to do
an adoption. Future cleanup of config files will get rid of
current warnings.
Calculation of offset to copy from storeQueue[idx].data structure for load to
store forwarding fixed to be difference in bytes between store and load virtual
addresses. Previous method would induce bug where a load would index into
buffer at the wrong location.
If a split load fails on a blocked cache wbOutstanding can be decremented
twice if the first part of the split load succeeds and the second part fails.
Condition the decrementing on not having completed the first part of the load.
This patch fixes two problems with the O3 cpu model. The first is an issue
with an instruction fetch causing a fault on the next address while the
current macro-op is being issued. This happens when the micro-ops exceed
the fetch bandwdith and then on the next cycle the fetch stage attempts
to issue a request to the next line while it still has micro-ops to issue
if the next line faults a fault is attached to a micro-op in the currently
executing macro-op rather than a "nop" from the next instruction block.
This leads to an instruction incorrectly faulting when on fetch when
it had no reason to fault.
A similar problem occurs with interrupts. When an interrupt occurs the
fetch stage nominally stops issuing instructions immediately. This is incorrect
in the case of a macro-op as the current location might not be interruptable.
The virtual channels within "response" vnets are made buffers_per_data_vc
deep (default=4), while virtual channels within other vnets are made
buffers_per_ctrl_vc deep (default = 1). This is for accurate power estimates.
Identifying response vnets versus other vnets will allow garnet to
determine which vnets will carry data packets, and which will carry
ctrl packets, and use appropriate buffer sizes (since data packets are larger
than ctrl packets). This in turn allows the orion power model to accurately
estimate buffer power.
Renamed (message) class to vnet for consistency with rest of ruby.
Moved some parameters specific to fixed/flexible garnet networks into their
corresponding py files.
This change further eliminates cases where condition codes were being read
just so they could be written without change because the instruction in
question was supposed to preserve them. This is done by creating the condition
code code based on the input rather than just doing a simple substitution.
If one of the condition codes isn't being used in the execution we should only
read it if the instruction might be dependent on it. With the preeceding changes
there are several more cases where we should dynamically pick instead of assuming
as we did before.
Break up the condition code bits into NZ, C, V registers. These are individually
written and this removes some incorrect dependencies between instructions.
Move the saturating bit (which is also saturating) from the renamed register
that holds the flags to the CPSR miscreg and adds a allows setting it in a
similar way to the FP saturating registers. This removes a dependency in
instructions that don't write, but need to preserve the Q bit.
This change splits out the condcodes from being one monolithic register
into three blocks that are updated independently. This allows CPUs
to not have to do RMW operations on the flags registers for instructions
that don't write all flags.
Debug flags are ExecUser, ExecKernel, and ExecAsid. ExecUser and
ExecKernel are set by default when Exec is specified. Use minus
sign with ExecUser or ExecKernel to remove user or kernel tracing
respectively.
Add registers and components to better support the VersatileEB board.
Made the MIDR and SYS_ID register parameters to ArmSystem and RealviewCtrl
respectively.
Instructions that load an address and are control instructions can
execute down the wrong path if they were predicted correctly and then
instructions following them are squashed. If an instruction is a
memory and control op use the predicted address for the next PC instead
of just advancing the PC. Without this change NPC is used for the next
instruction, but predPC is used to verify that the branch was successful
so the wrong path is silently executed.
The network tester terminates after injecting for sim_cycles
(default=1000), instead of having to explicitly pass --maxticks from the
command line as before. If fixed_pkts is enabled, the tester only
injects maxpackets number of packets, else it keeps injecting till sim_cycles.
The tester also works with zero command line arguments now.
The RubyMemory flag wasnt used in the code, creating large gaps in trace output. Replace cprintfs w/dprintfs
using RubyMemory in memory controller. DPRINTF also deprecate the usage of the setDebug() pure virtual
function in the AbstractMemoryOrCache Class as well the m_debug/cprintf functions in MemoryControl.hh/cc
The simple network's endpoint bandwidth value is used to adjust the overall
bandwidth of the network. Specifically, the ration between endpoint bandwidth
and the MESSAGE_SIZE_MULTIPLIER determines the increase. By setting the value
to 1000, that means the bandwdith factor specified in the links translates to
the link bandwidth in bytes. Previously, it was increasing that value by 10.
This patch will likely require a reset of the ruby regression tester stats.
Moved the buffer_size, endpoint_bandwidth, and adaptive_routing params out of
the top-level parent network object and to only those networks that actually
use those parameters.
This patch ensures that both Garnet and the simple networks use the bw value
specified in the topology. To do so, the patch generalizes the specification
of bw for basic links. This value is then translated to the specific value
used by the simple and Garnet networks. Since Garent does not support
non-uniformed link bandwidth, the patch also adds a check to ensure all bws are
equal.
--HG--
rename : src/mem/ruby/network/BasicLink.cc => src/mem/ruby/network/simple/SimpleLink.cc
rename : src/mem/ruby/network/BasicLink.hh => src/mem/ruby/network/simple/SimpleLink.hh
rename : src/mem/ruby/network/BasicLink.py => src/mem/ruby/network/simple/SimpleLink.py
This patch converts links and switches from second class simobjects that were
virtually ignored by the networks (both simple and Garnet) to first class
simobjects that directly correspond to c++ ojbects manipulated by the
topology and network classes. This is especially true for Garnet, where the
links and switches directly correspond to specific C++ objects.
By making this change, many aspects of the Topology class were simplified.
--HG--
rename : src/mem/ruby/network/Network.cc => src/mem/ruby/network/BasicLink.cc
rename : src/mem/ruby/network/Network.hh => src/mem/ruby/network/BasicLink.hh
rename : src/mem/ruby/network/Network.cc => src/mem/ruby/network/garnet/fixed-pipeline/GarnetLink_d.cc
rename : src/mem/ruby/network/Network.hh => src/mem/ruby/network/garnet/fixed-pipeline/GarnetLink_d.hh
rename : src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.py => src/mem/ruby/network/garnet/fixed-pipeline/GarnetLink_d.py
rename : src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.py => src/mem/ruby/network/garnet/fixed-pipeline/GarnetRouter_d.py
rename : src/mem/ruby/network/Network.cc => src/mem/ruby/network/garnet/flexible-pipeline/GarnetLink.cc
rename : src/mem/ruby/network/Network.hh => src/mem/ruby/network/garnet/flexible-pipeline/GarnetLink.hh
rename : src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.py => src/mem/ruby/network/garnet/flexible-pipeline/GarnetLink.py
rename : src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.py => src/mem/ruby/network/garnet/flexible-pipeline/GarnetRouter.py
Moved the Topology class to the top network directory because it is shared by
both the simple and Garnet networks.
--HG--
rename : src/mem/ruby/network/simple/Topology.cc => src/mem/ruby/network/Topology.cc
rename : src/mem/ruby/network/simple/Topology.hh => src/mem/ruby/network/Topology.hh
Due to certain changes made via changeset 8229, the compilation was failing
in certain cases. The compiler pointed to base/stats/mysql.hh for not naming
a certain types like uint64_t. To rectify this, base/types.hh is being
included in base/stats/mysql.hh.
This change makes the decoder figure out if an instruction that only supports
memory is using a register encoding and decodes directly to "Unknown" which will
behave appropriately. This prevents other parts of the instruction creation
process from seeing the mismatch and asserting.
This is similar to guards on mercurial queues and they're used for selecting
which files are compiled into some given object. We already do something
similar, but it's mostly hard coded for the m5 binary and the m5 library
and I'd like to make it more flexible to better support the unittests
At the same time, rename the trace flags to debug flags since they
have broader usage than simply tracing. This means that
--trace-flags is now --debug-flags and --trace-help is now --debug-help
This is basically like the range_map stuff in src/base (range already
exists in Python). This code is like a set of ranges. I'm using it
to keep track of changed lines in source code, but it could be use to
keep track of memory ranges and holes in memory regions. It could
also be used in memory allocation type stuff. (Though it's not at all
optimized.)
Frame buffer and boot linux:
./build/ARM_FS/m5.opt configs/example/fs.py --benchmark=ArmLinuxFrameBuf --kernel=vmlinux.touchkit
Linux from a CF card:
./build/ARM_FS/m5.opt configs/example/fs.py --benchmark=ArmLinuxCflash --kernel=vmlinux.touchkit
Run Android
./build/ARM_FS/m5.opt configs/example/fs.py --benchmark=ArmAndroid --kernel=vmlinux.android
Run MP
./build/ARM_FS/m5.opt configs/example/fs.py --benchmark=ArmLinuxCflash --kernel=vmlinux.mp-2.6.38
This change fixes a small bug in the arm copyRegs() code where some registers
wouldn't be copied if the processor was in a mode other than MODE_USER.
Additionally, this change simplifies the way the O3 switchCpu code works by
utilizing TheISA::copyRegs() to copy the required context information
rather than the adhoc copying that goes on in the CPU model. The current code
makes assumptions about the visibility of int and float registers that aren't
true for all architectures in FS mode.
The comment in the code suggests that the checking granularity should be 16
bytes, however in reality the shift by 8 is 256 bytes which seems much
larger than required.
Fixed an error reguarding DMA for uninprocessor systems. Basically removed an
overly agressive optimization that lead to inconsistent state between the
cache and the directory.
This function duplicates the functionality of allocate() exactly, except that it does not return
a return value. In protocols where you just want to allocate a block
but do not want that block to be your implicitly passed cache_entry, use this function.
Otherwise, SLICC will complain if you do not consume the pointer returned by allocate(),
and if you do a dummy assignment Entry foo := cache.allocate(address), the C++
compiler will complain of an unused variable. This is kind of a hack to get around
those issues, but suggestions welcome.
Before this changeset, all local variables of type Entry and TBE were considered
to be pointers, but an immediate use of said variables would not be automatically
deferenced in SLICC-generated code. Instead, deferences occurred when such
variables were passed to functions, and were automatically dereferenced in
the bodies of the functions (e.g. the implicitly passed cache_entry).
This is a more general way to do it, which leaves in place the
assumption that parameters to functions and local variables of type AbstractCacheEntry
and TBE are always pointers, but instead of dereferencing to access member variables
on a contextual basis, the dereferencing automatically occurs on a type basis at the
moment a member is being accessed. So, now, things you can do that you couldn't before
include:
Entry foo := getCacheEntry(address);
cache_entry.DataBlk := foo.DataBlk;
or
cache_entry.DataBlk := getCacheEntry(address).DataBlk;
or even
cache_entry.DataBlk := static_cast(Entry, pointer, cache.lookup(address)).DataBlk;
This is a substitute for MessageBuffers between controllers where you don't
want messages to actually go through the Network, because requests/responses can
always get reordered wrt to one another (even if you turn off Randomization and turn on Ordered)
because you are, after all, going through a network with contention. For systems where you model
multiple controllers that are very tightly coupled and do not actually go through a network,
it is a pain to have to write a coherence protocol to account for mixed up request/response orderings
despite the fact that it's completely unrealistic. This is *not* meant as a substitute for real
MessageBuffers when messages do in fact go over a network.
It is useful for Ruby to understand from whence request packets came.
This has all request packets going into Ruby pass the contextId value, if
it exists. This supplants the old libruby proc_id value passed around in
all the Messages, so I've also removed the unused unsigned proc_id; member
generated by SLICC for all Message types.
***
(1): get rid of expandForMT function
MIPS is the only ISA that cares about having a piece of ISA state integrate
multiple threads so add constants for MIPS and relieve the other ISAs from having
to define this. Also, InOrder was the only core that was actively calling
this function
* * *
(2): get rid of corespecific type
The CoreSpecific type was used as a proxy to pass in HW specific params to
a MIPS CPU, but since MIPS FS hasnt been touched for awhile, it makes sense
to not force every other ISA to use CoreSpecific as well use a special
reset function to set it. That probably should go in a PowerOn reset fault
anyway.
The goal of the patch is to do away with the CacheMsg class currently in use
in coherence protocols. In place of CacheMsg, the RubyRequest class will used.
This class is already present in slicc_interface/RubyRequest.hh. In fact,
objects of class CacheMsg are generated by copying values from a RubyRequest
object.
The tester code is in testers/networktest.
The tester can be invoked by configs/example/ruby_network_test.py.
A dummy coherence protocol called Network_test is also addded for network-only simulations and testing. The protocol takes in messages from the tester and just pushes them into the network in the appropriate vnet, without storing any state.
I had recently committed a patch that removed the WakeUp*.py files from the
slicc/ast directory. I had forgotten to remove the import calls for these
files from slicc/ast/__init__.py. This resulted in error while running
regressions on zizzer. This patch does the needful.
This patch fixes the problem where Ruby would fail to call sendRetry on ports
after it nacked the port. This patch is particularly helpful for bursty dma
requests which often include several packets.
In SLICC, in order to define a type a data type for which it should not
generate any code, the keyword external_type is used. For those data types for
which code should be generated, the keyword structure is used. This patch
eliminates the use of keyword external_type for defining structures. structure
key word can now have an optional attribute external, which would be used for
figuring out whether or not to generate the code for this structure. Also, now
structures can have functions as well data members in them.
In order to add stall and wait facility for protocols, a keyword
wake_up_dependents was introduced. This patch removes the keyword,
instead this functionality is now implemented as function call.
In order to add stall and wait facility for protocols, a keyword
wake_up_all_dependents was introduced. This patch removes the keyword,
instead this functionality is now implemented as function call.
Thanks to swig this was interfering with the standard Python
random module. The only function in that module was seed(),
which erroneously called srand48(). Moved the function to
m5.internal.core, renamed it seedRandom(), and made it call
random_mt.init() instead.
FastAlloc's reuse policies can mask allocation bugs, so
we typically want it disabled when debugging. Set
FORCE_FAST_ALLOC to enable even when debugging, and set
NO_FAST_ALLOC to disable even in non-debug builds.
The ISAR registers describe which features the processor supports.
Transcribe the values listed in section B5.2.5 of the ARM ARM
into the registers as read-only values
This change speeds up booting, especially in MP cases, by not executing
udelay() on the core but instead skipping ahead tha amount of time that is being
delayed.
This patch prevents not executed conditional instructions marked as
IsQuiesce from stalling the pipeline indefinitely. If the instruction
is not executed the quiesceSkip psuedoinst is called which schedules a
wakes up call to the fetch stage.
This changes the RFE macroop into 3 microops:
URa = [sp]; URb = [sp+4]; // load CPSR,PC values from stack
sp = sp + offset; // optionally auto-increment
PC = URa; CPSR = URb; // write to the PC and CPSR.
Importantly:
- writing to PC is handled in the last micro-op.
- loading occurs prior to state changes.
This change fixes the problem for all the cases we actively use. If you want to try
more creative I/O device attachments (E.g. sharing an L2), this won't work. You
would need another level of caching between the I/O device and the cache
(which you actually need anyway with our current code to make sure writes
propagate). This is required so that you can mark the cache in between as
top level and it won't try to send ownership of a block to the I/O device.
Asserts have been added that should catch any issues.
Without this change the a store can be issued to the cache multiple times.
If this case occurs when the l1 cache is out of mshrs (and thus blocked)
the processor will never make forward progress because each cycle it will
send a single request using the recently freed mshr and not completing the
multipart store. This will continue forever.
This causes a lot of rebuilds that could have otherwise possibly been
avoided, and, more annoyingly, a lot of unnecessary rerunning of the
regressions. The benefits of having the revision in the output haven't
materialized, so this change removes it.
None of the code in the ruby tester directory is compiled or referred to
outside of that directory. This change eliminates it. If it's needed in the
future, it can be revived from the history. In the mean time, this removes
clutter and the only use of the GEMS_ROOT scons variable.
The internet says this instruction was created by accident when an Intel CPU
failed to decode x87 instructions properly. It's been documented on a few rare
occasions and has generally worked to ensure backwards compatability. One
source claims that the gcc toolchain is basically the only thing that emits
it, and that emulators/binary translators like qemu and bochs implement it.
We won't actually implement it here since we're hardly implementing any other
x87 instructions either. If we were to implement it, it would behave the same
as ffree but then also pop the register stack.
http://www.pagetable.com/?p=16
There may not be a formally correct spelling for the past tense of mmap, but
mmapped is the spelling Google doesn't try to autocorrect. This makes sense
because it mirrors the past tense of map->mapped and not the past tense of
cape->caped.
--HG--
rename : src/arch/alpha/mmaped_ipr.hh => src/arch/alpha/mmapped_ipr.hh
rename : src/arch/arm/mmaped_ipr.hh => src/arch/arm/mmapped_ipr.hh
rename : src/arch/mips/mmaped_ipr.hh => src/arch/mips/mmapped_ipr.hh
rename : src/arch/power/mmaped_ipr.hh => src/arch/power/mmapped_ipr.hh
rename : src/arch/sparc/mmaped_ipr.hh => src/arch/sparc/mmapped_ipr.hh
rename : src/arch/x86/mmaped_ipr.hh => src/arch/x86/mmapped_ipr.hh
At a couple of places in PerfectSwitch.cc and MessageBuffer.cc, DPRINTF()
has not been provided with correct number of arguments. The patch fixes these
bugs.
This patch removes the store buffer from Ruby. It is not in use currently.
Since libruby is being and store buffer makes calls to libruby, it is not
possible to maintain it until substantial changes are made.
This patch changes Address.hh so that it is not dependent on RubySystem.
This dependence seems unecessary. All those functions that depend on
RubySystem have been moved to Address.cc file.
This patch changes DataBlock.hh so that it is not dependent on RubySystem.
This dependence seems unecessary. All those functions that depende on
RubySystem have been moved to DataBlock.cc file.
This patch integrates permissions with cache and memory states, and then
automates the setting of permissions within the generated code. No longer
does one need to manually set the permissions within the setState funciton.
This patch will faciliate easier functional access support by always correctly
setting permissions for both cache and memory states.
--HG--
rename : src/mem/slicc/ast/EnumDeclAST.py => src/mem/slicc/ast/StateDeclAST.py
rename : src/mem/slicc/ast/TypeFieldEnumAST.py => src/mem/slicc/ast/TypeFieldStateAST.py
Because int and not InstSeqNum was used in a couple of places, you can
overflow the int type and thus get wierd bugs when the sequence number
is negative (or some wierd value)
remove constructors that werent being used (it just gets confusing)
use initialization list for all the variables instead of relying on initVars()
function
-use a pointer to CacheReqPacket instead of PacketPtr so correct destructors
get called on packet deletion
- make sure to delete the packet if the cache blocks the sendTiming request
or for some reason we dont use the packet
- dont overwrite memory requests since in the worst case an instruction will
be replaying a request so no need to keep allocating a new request
- we dont use retryPkt so delete it
- fetch code was split out already, so just assert that this is a memory
reference inst. and that the staticInst is available
If there is an outstanding table walk and no other activity in the CPU
it can go to sleep and never wake up. This change makes the instruction
queue always active if the CPU is waiting for a store to translate.
If Gabe changes the way this code works then the below should be removed
as indicated by the todo.
We only support EABI binaries, so there is no reason to support OABI syscalls.
The loader detects OABI calls and fatal() so there is no reason to even check
here.
The ARM performance counters are not currently supported by the model.
This patch interprets a 'reset performance counters' command to mean 'reset
the simulator statistics' instead.
"executing" isnt a very descriptive debug message and in going through the
output you get multiple messages that say "executing" but nothing to help
you parse through the code/execution.
So instead, at least print out the name of the action that is taking
place in these functions.
Overall, continue to progress Ruby debug messages to more of the normal M5
debug message style
- add a name() to the Ruby Throttle & PerfectSwitch objects so that the debug output
isn't littered w/"global:" everywhere.
- clean up messages that print over multiple lines when possible
- clean up duplicate prints in the message buffer
In certain actions of the L1 cache controller, while creating an outgoing
message, the machine type was not being set. This results in a
segmentation fault when trace is collected. Joseph Pusudesris provided
his patch for fixing this issue.
keep track of when an instruction needs the execution
behind it to be serialized. Without this, in SE Mode
instructions can execute behind a system call exit().
resources don't need to call getLatency because the latency is already a member
in the class. If there is some type of special case where different instructions
impose a different latency inside a resource then we can revisit this and
add getLatency() back in
each resource has a certain # of requests it can take per cycle. update the #s here
to be more realistic based off of the pipeline width and if the resource needs to
be accessed on multiple cycles
---
need to delete the cache request's data on clearRequest() now that we are recycling
requests
---
fetch unit needs to deallocate the fetch buffer blocks when they are replaced or
squashed.
formerly, to free up bandwidth in a resource, we could just change the pointer in that resource
but at the same time the pipeline stages had visibility to see what happened to a resource request.
Now that we are recycling these requests (to avoid too much dynamic allocation), we can't throw
away the request too early or the pipeline stage gets bad information. Instead, mark when a request
is done with the resource all together and then let the pipeline stage call back to the resource
that it's time to free up the bandwidth for more instructions
*** inteface notes ***
- When an instruction completes and is done in a resource for that cycle, call done()
- When an instruction fails and is done with a resource for that cycle, call done(false)
- When an instruction completes, but isnt finished with a resource, call completed()
- When an instruction fails, but isnt finished with a resource, call completed(false)
* * *
inorder: tlbmiss wakeup bug fix
take away all instances of reqMap in the code and make all references use the built-in
request vectors inside of each resource. The request map was dynamically allocating
a request per instruction. The request vector just allocates N number of requests
during instantiation and then the surrounding code is fixed up to reuse those N requests
***
setRequest() and clearRequest() are the new accessors needed to define a new
request in a resource
we are going to be getting away from creating new resource requests for every
instruction so no more need to keep track of a reqRemoveList and clean it up
every tick
first change in an optimization that will stop InOrder from allocating new memory for every instruction's
request to a resource. This gets expensive since every instruction needs to access ~10 requests before
graduation. Instead, the plan is to allocate just enough resource request objects to satisfy each resource's
bandwidth (e.g. the execution unit would need to allocate 3 resource request objects for a 1-issue pipeline
since on any given cycle it could have 2 read requests and 1 write request) and then let the instructions
contend and reuse those allocated requests. The end result is a smaller memory footprint for the InOrder model
and increased simulation performance
Currently the wakeup function for the PerfectSwitch contains three loops -
loop on number of virtual networks
loop on number of incoming links
loop till all messages for this (link, network) have been routed
With an 8 processor mesh network and Hammer protocol, about 11-12% of the
was observed to have been spent in this function, which is the highest
amongst all the functions. It was found that the innermost loop is executed
about 45 times per invocation of the wakeup function, when each invocation
of the wakeup function processes just about one message.
The patch tries to do away with the redundant executions of the innermost
loop. Counters have been added for each virtual network that record the
number of messages that need to be routed for that virtual network. The
inner loops are only executed when the number of messages for that particular
virtual network > 0. This does away with almost 80% of the executions of the
innermost loop. The function now consumes about 5-6% of the total execution
time.
In x86, 32 and 64 bit writes to registers in which registers appear to be 32 or
64 bits wide overwrite all bits of the destination register. This change
removes false dependencies in these cases where the previous value of a
register doesn't need to be read to write a new value. New versions of most
microops are created that have a "Big" suffix which simply overwrite their
destination, and the right version to use is selected during microop
allocation based on the selected data size.
This does not change the performance of the O3 CPU model significantly, I
assume because there are other false dependencies from the condition code bits
in the flags register.
These faults can panic/warn/warn_once, etc., instead of instructions doing
that themselves directly. That way, instructions can be speculatively
executed, and only if they're actually going to commit will their fault be
invoked and the panic, etc., happen.
When redirecting fetch to handle branches, the npc of the current pc state
needs to be left alone. This change makes the pc state record whether or not
the npc already reflects a real value by making it keep track of the current
instruction size, or if no size has been set.
The patch changes the order in which L1 dcache and icache are looked up when
a request comes in. Earlier, if a request came in for instruction fetch, the
dcache was looked up before the icache, to correctly handle self-modifying
code. But, in the common case, dcache is going to report a miss and the
subsequent icache lookup is going to report a hit. Given the invariant -
caches under the same controller keep track of disjoint sets of cache blocks,
we can move the icache lookup before the dcache lookup. In case of a hit in
the icache, using our invariant, we know that the dcache would have reported
a miss. In case of a miss in the icache, we know that icache would have
missed even if the dcache was looked up before looking up the icache.
Effectively, we are doing the same thing as before, though in the common case,
we expect reduction in the number of lookups. This was empirically confirmed
for MOESI hammer. The ratio lookups to access requests is now about 1.1 to 1.
resource skeds are divided into two parts: front end (all insts) and back end (inst. specific)
each of those are implemented as separate lists, so this iterator wraps around
the traditional list iterator so that an instruction can walk it's schedule but seamlessly
transfer from front end to back end when necessary
add a stage scheduler class to replace InstStage in pipeline_traits.cc
use that class to define a default front-end, resource schedule that all
instructions will follow. This will also replace the back end schedule in
pipeline_traits.cc. The reason for adding this is so that we can cache
instruction schedules in the future instead of calling the same function
over/over again as well as constantly dynamically alllocating memory on
every instruction to try to figure out it's schedule
When a table walk is initiated by the fetch stage, the CPU can
potentially move to the idle state and never wake up.
The fetch stage must call cpu->wakeCPU() when a translation completes
(in finishTranslation()).
Uncacheable requests were set as such only in atomic mode.
currState->delayed is checked in place of currState->timing for resetting
currState in atomic mode.
This change fixes an issue where a DTLB fault occurs and redirects fetch to
handle the fault and the ITLB requires a walk which delays translation. In this
case the status of the cpu isn't updated appropriately, and an additional
instruction fetch occurs. Eventually this hits an assert as multiple instruction
fetches are occuring in the system and when the second one returns the
processor is in the wrong state.
Some asserts below are removed because it was always true (typo) and the state
after the initiateAcc() the processor could be in any valid state when a
d-side fault occurs.
