This patch enables cache controllers to make response
messages in advance, store them in a per-address saved
map in an output message buffer and enqueue them altogether
in the future. This patch introduces new slicc statement
called defer_enqueueing. This patch would help simplify
the logic of state machines that deal with coalesing
multiple requests from different requestors.
Change-Id: I566d4004498b367764238bb251260483c5a1a5e5
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/28132
Reviewed-by: Tuan Ta <qtt2@cornell.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Doing arithmetic on a null pointer is undefined behavior in C/C++. Clang
compilers complain when this occurs. As this MACRO is used twice, and
does nothing important, it has been removed in favor of a more simple
solution. A comment has been added explaining the MACRO's removal.
Change-Id: I42d9356179ee0fa5cb20f827af34bb11780ad1a9
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29534
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Without this fix `error: call to function 'operator<<' that is neither
visible in the template definition nor found by argument-dependent
loopup` is thrown when compiling HSAIL_X86 using a clang compiler (at
`base/cprintf_formats.hhi:139`).
This error is due to a "<<" operator in a template declared prior to its
definition in the code. The operator is used in
`base/cprintf_formats.hh`, included in `base/cprintf.hh`, and defined in
`mem/ruby/common/BoolVec.hh`. Therefore, for clang to compile without
error, `mem/ruby/common/BoolVec.hh` must be included before
`base/cprintf.hh` when generating the
`mem/ruby/protocol/RegionBuffer_Controller.cc` in
`mem/slicc/symbols/StateMachine.py`.
Due to the gem5 style-checker, an overly-verbose solution was required
to permit this patch to be committed to the codebase.
Change-Id: Ie0ae4053e4adc8c4e918e4a714035637925ca104
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29532
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Add support for multiple networks per RubySystem. This is done by
introducing local IDs to each network and translating from a global ID
passed around through Ruby and SLICC code. The local IDs represents the
NodeID of a MachineType in the network and are ordered the same way
that NodeIDs are ordered using MachineType_base_number. If there are
not multiple networks in a RubySystem the local and global IDs are the
same value.
This is useful in cases where multiple isolated networks are needed to
support devices with Ruby caches which do not interact with other
networks. For example, a dGPU device will have a cache hierarchy that
will not interact with the CPU cache hierachy.
Change-Id: I33a917b3a394eec84b16fbf001c3c2c44c047f66
JIRA: https://gem5.atlassian.net/browse/GEM5-445
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/27927
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Bradford Beckmann <brad.beckmann@amd.com>
Maintainer: Bradford Beckmann <brad.beckmann@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>
These false positives break the build. The error is below, and is bogus
as best I can tell. The constructor for the sc_unsigned and sc_signed
types, defined with some macro goop in sc_nbcommon.inc, have a call to
vec_copy_and_zero to copy over some data and zero the data that isn't
copied. That only happens if the source is smaller than the destination.
Then in vec_copy_and_zero, it calls vec_zero to set the last elements to
zero. Because of the check back at the constructor, only values that
exist should ever be set.
Also, in gem5, SC_MAX_NBITS is not set, so the definition of the array
it's bounds checking is declared right near where it's used and is sized
based on the variable being passed into vec_copy_and_zero.
In file included from build/ARM/systemc/ext/dt/bit/../int/../fx/sc_fxdefs.hh:52,
from build/ARM/systemc/ext/dt/bit/../int/sc_length_param.hh:63,
from build/ARM/systemc/ext/dt/bit/sc_bv_base.hh:56,
from build/ARM/systemc/dt/int/sc_unsigned.cc:83:
In function 'void sc_dt::vec_zero(int, int, sc_dt::sc_digit*)',
inlined from 'void sc_dt::vec_copy_and_zero(int, sc_dt::sc_digit*, int, const sc_digit*)' at build/ARM/systemc/ext/dt/bit/../int/../fx/../int/sc_nbutils.hh:407:13,
inlined from 'sc_dt::sc_unsigned::sc_unsigned(sc_dt::small_type, int, int, sc_dt::sc_digit*, bool)' at build/ARM/systemc/dt/int/sc_nbcommon.inc:2285:26:
build/ARM/systemc/ext/dt/bit/../int/../fx/../int/sc_nbutils.hh:379:14: error: 'void* __builtin_memset(void*, int, long unsigned int)' offset [12, 15] is out of the bounds [0, 12] [-Werror=array-bounds]
379 | u[i] = 0;
|
Change-Id: Ica721178b24de56dbeabf4af7d3422dea6336a23
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29432
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This uses a "name()" method which is not defined by the Ticked class,
and isn't a global method. This was probably originally supposed to be
the name() method of the Serializable class that Ticked inherits from,
but a while ago that was removed. It's not clear how this has been
compiling.
Instead, use the name() method of the ClockedObject which is the first
constructor argument.
Change-Id: Icfb71732c58ea9984ef7343bbaa46097a25abf28
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29406
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This constant isn't in normalized units, ie doesn't scale when the time
value of a Tick changes, is global, has an extremely generic name even
though it's only used by a few ethernet devices, and has an arbitrary
value.
Get rid of it, and replace it with 1ns, what it would typically be
equivalent to when using the default 1ps time scale.
Change-Id: I31d9dad438f854b4152cd53c9a7042a25d13e0a6
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29398
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This commit changes m5's tick rounding mechanism from python's round()
to python's ceil() function.
Currently, non intergral ticks are rounded by round() function in python.
In python2, this function rounds values >= 0.5 to 1. However, in python3,
0.5 is rounded to 0. This causes the function to return 0 ticks for
non-zero second values, which doesn't make sense, and also causes
several tests to fail.
ceil() function is now used to round up the tick values. This makes more
sense as non-zero second values won't be rounded to zero in any cases.
Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Change-Id: I14c43e38e8c678f77baf13407f7eeff4b86f1014
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29372
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
By default, DOT configs are always generated when pydot is present.
This change allows a user to pass an empty --dot-config='' to disable
generating the DOT configuration. This can be useful to save space, or
to reduce Gem5 startup time when running many small regression tests.
This brings the behavior in-line with providing an empty
--dump_config='' and/or --json_config='' which similarly disables
generation of those output files.
Change-Id: I5bf39fda0409b948a8d14f3afa95db8fc78de6ee
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29232
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
These tables are based on passing the symbols in the current table
through some sort of operator function which can chose to add those
symbols, modified versions of those symbols, or nothing at all into a
new symbol table.
The new table is returned as a shared_ptr so its memory will be
managed automatically.
Change-Id: I8809336e2fc2fda63b16a0400536116ca852ca13
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/24786
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This singleton object is used thruoughout the simulator. There is
really no reason not to have it statically allocated, except that
whether it was allocated seems to sometimes be used as a signal that
something already put symbols in it, specifically in SE mode.
To keep that functionality for the moment, this change adds an "empty"
method to the SymbolTable class to make it easy to check if the symbol
table is empty, or if someone already populated it.
Change-Id: Ia93510082d3f9809fc504bc5803254d8c308d572
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/24785
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The SymbolTable class had been tracking symbols as two independent
pieces, a name and an address, and acted as a way to translate between
them. Symbols can be more complex than that, and so this change
encapsulates the information associated with a symbol in a new class.
As a step towards simplifying the API for reading symbols from a
binary, this change also adds a "binding" field to that class so that
global, local and weak symbols can all go in the same table and be
differentiated later as needed. That should unify the current API
which has a method for each symbol type.
While the innards of SymbolTable were being reworked, this change
also makes that class more STL like by adding iterators, and begin
and end methods. These iterate over a new vector which holds all the
symbols. The address and name keyed maps now hold indexes into that
vector instead of the other half of the symbol.
Change-Id: I8084f86fd737f697ec041bac86a635a315fd1194
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/24784
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
When gem5.fast is compiled, an error on a variable
used only for debug purposes is raised:
build/X86/cpu/o3/mem_dep_unit_impl.hh:262:19: error: unused variable 'producing_store' [-Werror=unused-variable]
for (auto producing_store : producing_stores)
This patch remove the variable when *.fast is used.
Change-Id: Ib77c26073db39644e3525bc16edcb7d3bc871d76
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29252
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
The physical address has already been set (it's read earlier in the
function), and so doesn't need to be set again. Reading the virtual
address can cause an assert if the virtual address had never been set in
the first place, for example when an access comes from KVM which might
give you an access to complete which is based on a physical address
only.
Change-Id: Ic46a40b1a94235538b5bd53065e5019273b3d3f3
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/29172
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Pouya Fotouhi <pfotouhi@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
AbstractController sends requests using a QueuedMasterPort which has an
implicit buffer which is unbounded. Remove this by changing the port to
a MasterPort and implement a retry mechanism for AbstractController.
Although the request remains in the MessageBuffer if a retry is needed,
the additional retry logic optimizes serviceMemoryQueue slightly and
prevents the DRAMCtrl retry stats from being incorrect due to multiple
calls to sendTimingReq.
Change-Id: I8c592af92a1a499a418f34cfee16dd69d84803ad
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/28387
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Bradford Beckmann <brad.beckmann@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>
In the base Result and Argument templates, there were private static
functions which weren't meant to be used, but which would act as
documentation for what those functions should look like. They were
marked as private to prevent them from being accidentally used and
causing confusing, hard to debug errors.
Unfortunately, that also meant that those functions exist, and
apparently cause inconsistent problems with SFINAE. I assume if the
functions don't exist at all, then SFINAE will work properly. When
they're private, that seems to cause a substitution failure which
actually is an error which makes the build fail.
Change-Id: I326e9e1d05eafe1b00732ae10264354b07426e74
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/28308
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
