Although they provide the exact same behavior, the params
created in the tests did not have the type expected by the
internal safe cast.
The following error was triggered:
storage.test.debug: build/NULL/base/cast.hh:47: T safe_cast(U)
[with T = const Stats::SampleStor::Params*;
U = const Stats::StorageParams*]:
Assertion `ret' failed.
Change-Id: I4f2ba51f3ccdb44589e61f235997245e7d9bf3c9
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40555
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
These files were originally used to provide a more gtest like mechanism
for the UnitTest executables, many of which didn't actually test
anything. With the definitions in those files, the tests could check
whether their expectations were met, and either pass or fail without a
human having to inspect the output and knowing what output to expect.
Change-Id: Ie0601391b994859eb544b37201333838fa3ba02a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40618
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
This "UnitTest" was really not a unit test, it was a timing utility for
measuring the performance of gem5's cprintf implementation. The name was
misleading, but more than that, it was linked against all of gem5 which
created a approximately 1.5 gigabyte binary for what is a very small
program.
Instead, the new version of cprintftime, which has the same
functionality as the old version, weighs in at a svelte 500k with debug
information.
This also trims down the number of misleading "UnitTest" entries to 3,
getting us closer to the point where we can eliminate that type of
entity entirely.
Change-Id: Id30d094f2844e948fe67e820c89412f8667aaa52
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40617
Maintainer: Gabe Black <gabe.black@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
That tag was intended to mark an Executable subclass as abstract, aka
only suitable for using as bases for other Executable subclasses and not
for direct instantiation. The only place it was used was the base
Executable class however, and that class is actually directly useful
when setting up a generic executable from other gem5 sources.
Change-Id: I70204b63c03bb45bf21b8c312a7b8581be5e0cab
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40616
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
This DPRINTF accesses the ExtMachInst typed machInst member of the
StaticInst class, and so is ISA dependent. Move the DPRINTF to where the
instructions are actually decoded where that type doesn't have to be
disambiguated.
Also, this change makes this DPRINTF more accurate, since microops are
not really "decoded" when they are extracted from a macroop. The process
of unpacking them to feed into the rest of the CPU should be fairly
trivial, so really they're just being retrieved. With the DPRINTF in
this new position, it will only trigger when an instruction is actually
decoded from memory.
Change-Id: I14145165b93bb004057a729fa7909cd2d3d34d29
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40099
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The only way to allocate fixed sized arrays which will definitely be big
enough for all source/destination registers for a given instruction is
to track the maximum number of each at compile time, and then size the
arrays appropriately. That creates a point of centralization which
prevents breaking up decoder and instruction definitions into more
modular pieces, and if multiple ISAs are ever built at once, would
require coordination between all ISAs, and wasting memory for most of
them.
The dynamic allocation overhead is minimized by allocating the storage
for all variable arrays in one chunk, and then placing the arrays there
using placement new. There is still some overhead, although less than it
might be otherwise.
Change-Id: Id2c42869cba944deb97da01ca9e0e70186e22532
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/38384
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Per the AMD64 Architecture Programming Manual:
The size of the count register (CX, ECX, or RCX) depends on the
address-size attribute of the JrCXZ instruction. Therefore, JRCXZ can
only be executed in 64-bit mode
and
In 64-bit mode, the operand size defaults to 64 bits. The processor
sign-extends the 8-bit displacement value to 64 bits before adding it
to the RIP.
This patch also renames the instruction from JRCX to JRCXZ to match the
language in the programming manual.
Change-Id: Id55147d0602ff41ad6aaef483bef722ff56cae62
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40195
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Add an "All" compound debug flag, which encapsulates all
debug flags.
Since this is the broadest compound flag, allowing users
to include it would imply in extremely generic includes.
Moreover, it is highly unlikely that any correct C++ code
would ever use all debug flags. Therefore, a header file
for this flag is not generated to force users to directly
include only the debug flags they need.
Change-Id: If40f2f708be1495fa2b2380266164d5d44d7cffa
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/39077
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nathanael Premillieu <nathanael.premillieu@huawei.com>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
This takes the place of direct access to the machInst field as used in
the MinorCPU model which makes the incorrect assumption that it can
arbitrarily treat the ExtMachInst as an integer, and that masking in a
certain way can meaningfully classify what the instruction will do.
Because that assumption is not correct in general, that had been
ifdef-ed out in most ISAs except ARM, and for the other ISAs the value
was simply set to zero.
Change-Id: I8ac05e65475edc3ccc044afdff09490e2c05ba07
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40098
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
Create unit tests for the stats storage types. As a side effect
storage-related classes have been moved to separate files.
HistStor's grow_up, grow_out, and grow_convert have been made
private and renamed to comply with gem5's naming convention
and make grow_convert match its grow_up counterpart (growDown)
which is more suitable for its expected behavior.
The params declarations have been moved to be close to their
storage class' constructor.
HistStor has a explicit condition stating that there must be
at least 2 buckets.
Added documentation!
Fixed grow_convert so that it yields consistent histograms.
Previously buckets could not fully intersect, so doubling their
bucket size would make them steal contents innaproprietly. For
example, the neighbors [-6,-2[, [-2,2[, [2,6[, when doubled,
become [-12,-4[, [-4,4[, [4,12[; however, since the individual
values are not stored, it is impossible to know how to populate
the middle bucket with its neighbor's partial contents.
This fix forces the middle bucket of a storage to have its lower
bound at 0, solving the partial intersection issue.
Change-Id: Idb063e3dbda3cce3a8969e347660143162146eb9
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/25425
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Debug flags are flags that aid with debugging by printing
relevant information when enabled. Debug-formatting flags
define how the debug flags will print the information.
Although a viability, this patch does not support declaring
compound format flags.
As a side effect, now debug flags and debug-formatting flags
are printed in different lists, when using --debug-help.
Change-Id: Ieae68745276218cf4e9c1d37d7bf3bd1f19709ae
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/39076
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The current tests included don't require boost header to work. Remove
the dependency. This also gets rid of the warning message generated by
the latest boost headers.
Tested by running systemC tests:
src/systemc/tests/verify.py --update-json \
--filter-file=src/systemc/tests/working.filt -j 56 build/ARM/
Change-Id: I9d3bfe145597335abdf24f2de85ed3c0708aea27
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40315
Reviewed-by: Gabe Black <gabe.black@gmail.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This should help reduce warning spew when building with newer compilers.
The pybind11::module type has been renamed pybind11::module_ to avoid
conflicts with c++20 modules, according to the pybind11 changelog, so
this CL also updates gem5 source to use the new type. There is
supposedly an alias pybind11::module which is for compatibility, but we
still get linker errors without changing to pybind11::module_.
Change-Id: I0acb36215b33e3a713866baec43f5af630c356ee
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40255
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>
Previously we use attribute and event for communication between gem5
SimObject to systemC fastmodel sc_module. Creating a base class allows us
to perform casting once and get all the interface required. Also,
instead of warning on attribute not found, we should make simulator
panic if the sc_module does not provide the interface we need.
Change-Id: I91e1036cb792d556dfc4010e7a0f138b1519b079
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40277
Reviewed-by: Gabe Black <gabe.black@gmail.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The register cntfrq should be set to system counter frequency.
However, the current fastmodel implementation accidentally set it to
core frequency. This CL removes the wrong implementation, and real
cntfrq setting is performed in the initState.
Change-Id: I6c62822a4fbbcc0c499f79f6003dabb0c133f997
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/40276
Reviewed-by: Gabe Black <gabe.black@gmail.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
In Gem5ToTlmBridge::pec, the function expects blockingRequest should be
set no matter the tlm peer returns TLM_UPDATE or TLM_ACCEPTED.
However, current implementation only sets blockingRequest when the tlm
peer returns TLM_ACCEPTED. We should also set blockingRequest when the
tlm peer returns TLM_UPDATE.
Change-Id: I87bba3201cd68d52ded93c9c200f4fa4a40bdf5b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/39815
Reviewed-by: Earl Ou <shunhsingou@google.com>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Event creation and management support from emulated drivers is required
to support interruptible signals in HSA and this support was not
available. This changeset adds the event creation and management support
in the emulated driver. With this patch, each interruptible signal
created by the HSA runtime is associated with a signal event. The HSA
runtime can then put a thread waiting on a signal condition to sleep
asking the driver to monitor the event associated with that signal. If
the signal is modified by the GPU, the dispatcher notifies the driver
about signal value change. If the modifier is a CPU thread, the thread
will have to make HSA API calls to modify the signal and these API calls
will notify the driver about signal value change. Once the driver is
notified about a change in the signal value, the driver checks to see if
any thread is sleeping on that signal and wake up the sleeping thread
associated with that event. The driver has also implemented the time_out
wakeup that can wake up the thread after a certain time period has
expired. This is also true for barrier packets.
Each signal has an event address in a kernel managed and allocated
event page that can be used as a mailbox pointer to notify an event.
However, this feature used by non-CPU agents to communicate with the
driver is not implemented by this changeset because the non-CPU HSA
agents in our model can directly communicate with driver in our
implementation. Having said that, adding that feature should be trivial
because the event address and event pages are correctly setup by this
changeset and just adding the event page's virtual address to our PIO
doorbell interface in the page tables and registering that pio address
to the driver should be sufficient. Managing mailbox pointer for an
event is based on event ID and using this event ID as an index into
event page, this changeset already provides a unique mailbox pointer for
each event.
Change-Id: Ic62794076ddd47526b1f952fdb4c1bad632bdd2e
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/38335
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Some CSR register is physically shared between different privilige
level. Current implementation of CSR setting only considers to verify
the bits visable in current privilige level, and directly writes the
masked bits back to register. This leads to other bits invisable
to current mode is overwritten and wrong behavior across the modes.
Thus, CSR updating should always keep the bits value for other modes.
e.g. disabling interrupt in S mode with setting
SSTATUS SIE bit will lead to clear MIE bit as well (the interrupt
is disabled unintentionally).
All CSR register sharing same physical register in different mode
may have similar issue. I only fixed some important ones.
The fix is verified in FS.
Jira Issue: https://gem5.atlassian.net/browse/GEM5-860
Change-Id: I34d4766a4b483b5add2c3bbefd28b21b9abf37f6
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/39036
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ayaz Akram <yazakram@ucdavis.edu>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
The base/refcnt.hh header was not used in base/types.hh at all, and
enum/ByteOrder.hh was there just so other files could find it. Instead,
this change moves enum/Byteorder.hh to sim/byteswap.hh where it's fits
with the purpose of the header.
This change also fixes some style problems with the code in
base/types.hh itself.
Change-Id: I471ae5cb2cca9169ba8616fb8411b40108a3ffb2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/39855
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Gabe Black <gabe.black@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
