mem-ruby: HTM mem implementation

This patch augments the MESI_Three_Level Ruby protocol with hardware
transactional memory support.

The HTM implementation relies on buffering of speculative memory updates.
The core notifies the L0 cache controller that a new transaction has
started and the controller in turn places itself in transactional state
(htmTransactionalState := true).

When operating in transactional state, the usual MESI protocol changes
slightly. Lines loaded or stored are marked as part of a transaction's
read and write set respectively. If there is an invalidation request to
cache line in the read/write set, the transaction is marked as failed.
Similarly, if there is a read request by another core to a speculatively
written cache line, i.e. in the write set, the transaction is marked as
failed. If failed, all subsequent loads and stores from the core are
made benign, i.e. made into NOPS at the cache controller, and responses
are marked to indicate that the transactional state has failed. When the
core receives these marked responses, it generates a HtmFailureFault
with the reason for the transaction failure. Servicing this fault does
two things--

(a) Restores the architectural checkpoint
(b) Sends an HTM abort signal to the cache controller

The restoration includes all registers in the checkpoint as well as the
program counter of the instruction before the transaction started.

The abort signal is sent to the L0 cache controller and resets the
failed transactional state. It resets the transactional read and write
sets and invalidates any speculatively written cache lines.  It also
exits the transactional state so that the MESI protocol operates as
usual.

Alternatively, if the instructions within a transaction complete without
triggering a HtmFailureFault, the transaction can be committed. The core
is responsible for notifying the cache controller that the transaction
is complete and the cache controller makes all speculative writes
visible to the rest of the system and exits the transactional state.

Notifting the cache controller is done through HtmCmd Requests which are
a subtype of Load Requests.

KUDOS:
The code is based on a previous pull request by Pradip Vallathol who
developed HTM and TSX support in Gem5 as part of his master’s thesis:

http://reviews.gem5.org/r/2308/index.html

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: Icc328df93363486e923b8bd54f4d77741d8f5650
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30319
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem: Add HTM fields to the Packet object

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: I39268825327f2387ca7e622093fdb42c24a6c82c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30318
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Base dyn inst HTM flags getter

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: Ie15d8849edcff34ee7d5c7dd5e6ee2e099f937fc
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30326
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

sim: Add HTM Generic Fault

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: Iedbf06d25330a92790123805cff50d57b613a7a5
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30325
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Add HTM ThreadContext API

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: I9d60f69592c8072e70cef18787b5a4f2fc737a9d
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30324
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Add HTM ExecContext API

* initiateHtmCmd(Request::Flags flags)
* getHtmTransactionUid()
* newHtmTransactionUid()
* inHtmTransactionalState()
* getHtmTransactionalDepth()

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: I438832a3c47fff1d12d0123425985cfa2150ab40
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30323
Tested-by: kokoro <noreply+kokoro@google.com>

mem: Re-remove the arch/isa_traits.hh include in the base prefetcher.

This was removed but then accidentally re-added by a following change,
probably from a slighly incorrect rebase.

Change-Id: Ia7e8c755f92343c8b5e82febea2c1db4683fa69a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34166
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem: Add HTM fields to Request

This starts the support of Hardware Transactional Memory on the mem side

* The following flags have been added:

HTM_START: The request starts a HTM transaction
HTM_COMMIT: The request commits a HTM transaction
HTM_CANCEL: The request cancels a HTM transaction
HTM_ABORT: The request aborts a HTM transaction

* The following fields have been added:

_instCount: The instruction count at the time this request is created
_htmAbortCause: The cause for HTM transaction abort

https://gem5.atlassian.net/browse/GEM5-587

Change-Id: Ic582a6566fdd23f30eb92723e629d0c4d4ca10e5
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30316
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Add HTM CPU API

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: Iff95eb97603b4cb9629c04382a824b02594ee5c7
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30322
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Add HTM Instruction Flags

IsHtmStart: Starts a HTM transaction
IsHtmStop: Stops (commits) a HTM transaction
IsHtmCancel: Explicitely aborts a HTM transaction

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: I33144f97a2009e28b0c64777f0313cd6eadb7ff9
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30321
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Add HtmCpu DebugFlag

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: Id4b86b8964bc64bce1d2e4af941217eb114f3cc4
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30320
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

python: Remove unused debug APIs

The following APIs are not exported from the _m5 namespace and not
used by any of the debug glue code:

 * m5.debug.findFlag
 * m5.debug.setDebugFlag
 * m5.debug.clearDebugFlag
 * m5.debug.dumpDebugFlags

All of them have a clean Python interface where flags are exported
using the m5.debug.flags dictionary. There is also an m5.debug.help
function that lists the available debug flags.

Remove the unused APIs to avoid confusion.

Change-Id: I74738451eb5874f83b135adaccd30a0c6b478996
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34120
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

python: Add the ability to check if a debug flag has been enabled

There is currently no Python API to check if a debug flag is
enabled. Add a new status property that can be read or set to control
the status of a flag. The stat of a flag can also be queried by
converting it to a bool.

For example:

  m5.debug.flags["XBar"].status = True

  if m5.debug.flags["XBar"]:
      print("XBar debugging is on")

Change-Id: I5a50c39ced182ab44e18c061c463d7d9c41ef186
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34119
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

base: Cleanup debug flags API

The debug flags API has a couple of quirks that should be cleaned
up. Specifically:

 * Only CompoundFlag should expose a list of children.
 * The global enable flag is just called "active", this isn't very
   descriptive.
 * Only SimpleFlag exposed a status member. This should be in the base
   class to make the API symmetric.
 * Flag::Sync() is an implementation detail and needs to be protected.

Change-Id: I4d7fd32c80891191aa04f0bd0c334c8cf8d372f5
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34118
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

base: Remove unused Debug::All flag

The Debug::All flag doesn't seem to be used. Remove it.

Change-Id: I3d6ad1b2f61a2a0a5c52cbc6d520112855946007
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34117
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

scons: Simplify arch enum generation

C++ allows a trailing comma after the last item in an enum, so there
is no need for a special case.

Change-Id: I6ead36b4a8562b4a7a5aec88e4f6390182eedf56
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34116
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

base: Cleanup Debug::CompoundFlag

Compound flags are currently constructed using a constructor with a
finite set of arguments that default to nullptr that refer to child
flags. C++11 introduces two cleaner ways to achieve the same thing,
variadic templates and initializer_list. Use an initializer list to
pass dependent flags.

Change-Id: Iadcd04986ab20efccfae9b92b26c079b9612262e
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34115
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

misc,scons,util: Drop support for GCC 4

Corresponding website update:
https://gem5-review.googlesource.com/c/public/gem5-website/+/33657

Issue-on: https://gem5.atlassian.net/browse/GEM5-218
Change-Id: Ia72edda6229214e2f9d548266a42a0affd49b340
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33659
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

learning-gem5: convert simple cache to new style stats

Change-Id: I6988c45c13955825fde974f390460f4473af017a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34135
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-garnet: Separable allocator in Garnet not fair enough.

Currently there are independent round robin arbiter at each
input port and output port. Every time a VC is selected for
output allocation round robin is incremented irrespective of
if it is selected by its output port or not. This leads to
unfair arbitration at input port and is well known[1]. This
patch fixes it to increment only if the output port also
selects it.

[1] D. U. Becker and W. J. Dally, "Allocator implementations
for network-on-chip routers," Proceedings of the Conference
on High Performance Computing Networking, Storage and
Analysis, Portland, OR, 2009, pp. 1-12

Change-Id: I65963fb8082c51c0e3c6e031a8b87b4f5c3626e1
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32601
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

mem-garnet: Add a check to see if router is already scheduled

Currently the Switch Allocator takes up most of the simulation
wall clock time. This function checks for all VCs to see if it
should wakeup next. The input units which are simulated before
the switch allocator could have scheduled it already. This patch
adds a check for it.

Change-Id: I8609d4e7f925aa5e97198f6cd07466530f6fcf4c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32600
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

mem-garnet: Flexible VCs per Vnet for each router

This change allows configuring each router with a certain number
of VCs for each VNET. This is beneficial when dealing with
heterogenous link widths in a system. Configuring VCs
for each router allows one to ensure equal throughput
within the network while avoiding head-of-line blocking.
Changing a router's VCs number can be done in topology files
using the vcs_per_vnet value argument of router.

Change-Id: Icf4f510248128429a1a11f19f9802ee96f340611
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32599
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-garnet: Initialize unused Credit members

The Credit class doesn't initialize a number of its unused base class
fields.  This leads to non-determanistic traces when printing flits that
are Credits.  This patch initializes all unused fields to 0.

Change-Id: Ib73c652c71a10be57b24c0d6e1ac22eafa421e11
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32598
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

mem-garnet: Integration of HeteroGarnet

This upgrades the garnet model to support HeteroGarnet
1) Static and dynamic multi-freq domains in network
2) Support for CDC
3) Separate links for each message class
4) Separate linkwidth for each message class
5) Support for SerDes

Change-Id: I6d00e3b5cb3745e849d221066cb46b2138c47871
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32597
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>

cpu: Failure to restore RAS during squash

During squash of branch predictor history, RAS recovery mess up the
stack because of function "restore" in RAS (src/cpu/pred/ras.cc). In
restore function, it does not update "usedEntries" variable resulting in
restore failure.

To be specific, in order to remove mispredicted call, it uses pop() and
it updates tos. However in order to restore mispredicted ret
instruction, it uses restore() but it does not update tos. This pair of
function call mess up the RAS resulting in many misspeculation.

The solution is to update usedEntries variable as “push” function does.
This is possible because restoration is done with reverse order of push
and pop.

Jira Issue: https://gem5.atlassian.net/browse/GEM5-732

Change-Id: Ia14e71c26d20b2795fd55a6a0dd3284c03570614
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33214
Reviewed-by: Trivikram Reddy <tvreddy@ucdavis.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>

tests: Removed author info from insttest test.py

We no longer include author information directly in gem5 source.

Change-Id: I460d399f25ea955e7cf3bf5ed002246199ef4436
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33976
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>

tests,arch-arm: Pass gem5_root as an arg in run.py

Previously `tests/gem5/fs/linux/arm/run.py` contained an ugly,
hard-coded `gem5_root` variable. In this commit we pass `gem5_root`
as an argument from `tests/gem5/fs/linux/arm/test.py`, utilizing
`config.base_dir`.

Change-Id: I2b1e3369b1078cce9375fadb7c39fa4292648658
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33955
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

tests: Standardized test resources download dir

We were downloading resources to various different locations, for no
real reason. This standardizes the process. From this commit onwards,
all testing resources are downloaded to `tests/gem5/resources` by
default. This may be overriden via the `--bin-path` TestLib argument.

Note: In order to do this I have changed the meaning of the `bin-path`
TestLib argument slightly. Previously the `bin-path` assumed a flat
(non-existant) hierarchy. A simple directory of local resources. This
new bin-path functionality maintains logical sub-directories. This is
technically an API change and will be noted in the release notes.

Change-Id: I4df85c121fa65f787fd71f03d74361afea121380
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33145
Reviewed-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

tests: Removed learning-gem5 tests/configs

These configs are currently unused, and only served as redirects. They
are therefore being removed.

Change-Id: I12e5c2adb3a9c63ffc3177472eaa698b1cf80e41
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33144
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

tests: Moved realview config files

This is part of a process of getting rid of the `tests/config`
directory, and placing these configs either where they are used,
removing them if unneeded, or moving them to `configs/example`.

These config files, in this patchset, are part of the realview tests
found in `tests/gem5/fs/linux/arm/`. They have been moved to
`tests/gem5/configs`.

Change-Id: I7706b59c58da6413f5f3dd816a1e5cd54a834a58
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33143
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>

tests: Removed test-progs/asmtest

The insttests source is now found in gem5-resources:
https://gem5.googlesource.com/public/gem5-resources/+/refs/heads/master/src/asmtest

The pre-compiled binaries are pulled from dist.gem5.org.

There is no reason to keep these here. They are therefore being
removed.

Change-Id: Ic7869677278248f77b2703497ae7fc808d8f767a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33142
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

tests: Removed test-progs/pthread

The Pthread source is now found in gem5-resources:
https://gem5.googlesource.com/public/gem5-resources/+/refs/heads/master/src/pthreads

The pre-compiled binaries are pulled from dist.gem5.org.

There is no reason to keep these here. They are therefore being
removed.

Change-Id: Id07dc72c62d19920fb304eea505deaa9ebc5599f
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33141
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

tests: Removed the ignoring of tests

This commit reverts
https://gem5-review.googlesource.com/c/public/gem5/+/23023.

It has proven to be an unpopular piece of functionality which makes it
too easy to silently ignore failing tests. The new policy will be to
remove/comment-out failing tests in the testing source and tag with Jira
issues as to why.

Change-Id: I17d69fc57a9171ce3702e019615390a9aa3da250
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33139
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

tests: Removed SPARC Insttests from .testignore

Due to the fixing of the SPARC insttest binary, recorded here:
https://gem5-review.googlesource.com/c/public/gem5-resources/+/33396,
these tests now pass.

Change-Id: I4dca4504476f6d388e607a1075d44e9be69b5259
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33138
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

tests: Removed tests/configs/tgen files

These files were needed by the memory tests found in `tests/memory`, but
this directory already has local copies.

Change-Id: I04628b151d2ab60b1127990dbe8baaab77e768b1
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33136
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

tests: Removed `tests/halt.sh`

`tests/halt.sh` is a duplicate of `configs/tests/halt.sh`. The one
instance of using `halt.sh`, `tests/gem5/fs/linux/arm/run.py`, has been
updated to use `configs/tests/halt.sh` and `tests/halt.sh` has been
removed.

Change-Id: I42f77a22b5985959a579c75541e4a0fdc135aa0c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33135
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

tests: Removed realview tests from .testignore

The tests still fail, as recorded in
https://gem5.atlassian.net/browse/GEM5-364, though they have been
removed from the .testignore file as part of our goal of removing the
.testignore directory: https://gem5.atlassian.net/browse/GEM5-361.

Change-Id: I74f8a6c86e24835acbb4891ab4c88320baf12346
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33134
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

util: add dev-hsa commit message tag

The dev-hsa commit message tag was originally an option, but
appears to have been removed during the merge of the AMD GCN3
staging branch.  This commit adds it back.

Change-Id: Ie755b5ebe6ca1e5e92583b1588fd7aaeddcb5b00
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34095
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Set ContextId on request from trace CPU

Adds a contextId to the trace CPU in one more case that was missing.
Without this a panic is triggered in the cache.

Change-Id: I78bd70ad1e3657c9a6a1d56c234c007c2e2b586c
Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34035
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>

util: Install scons 3.1 from pip in gcn-gpu dockerfile

A previous commit updated the minimum required version of scons to 3.0

The gcn Dockerfile previously installed scons from apt, which installed
scons 2.4, as the Dockerfile is based on Ubuntu 16

This patch installs scons through pip, which installs scons 3.1

Change-Id: I4f731b301f97e25c730df26afde20ae1cdfaa1b3
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/34075
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Daniel Gerzhoy <daniel.gerzhoy@gmail.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Matthew Poremba <matthew.poremba@amd.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: convert thread_state to new style stats

Change-Id: Ib8cc8633ca5fced63918a7a6d10e15126f7c7459
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33400
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: convert tage_base to new style stats

Change-Id: If03102af545855125e87782c77ff5b43da8ac73b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33937
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: convert statistical_corrector to new style stats

Change-Id: Id9e075fb45babeeafe65105679c8bf2135823d41
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33936
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: convert loop_predictor to new style stats

Change-Id: Ib0383fc6d5f884fd6c020bcd938eee2f802ad412
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33935
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: convert bpred_unit to new style stats

Change-Id: Ife80b2df3cb900a73a4f0c1d6925d9ed2d625dd0
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33797
Reviewed-by: Trivikram Reddy <tvreddy@ucdavis.edu>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>

arch: Use a fault to trigger system calls in SE mode.

When the system call happens during the execution of the system call
instruction, it can be ambiguous what state takes precedence, the state
update from the instruction or the system call. These may be tracked
differently and found in an unpredictable order in, for example, the O3
CPU. An instruction can avoid updating any state explicitly, but
implicitly updated state (specifically the PC) will always update,
whether the instruction wants it to or not.

If the system call can be deferred by using a Fault object, then it's no
longer ambiguous. The PC update will be discarded, and the system call
can set the PC however it likes. Because there is no implicit PC update,
the PC needs to be walked forward, either to what it would have been
anyway, or to what the system call set in NPC.

In addition, because of the existing semantics around handling Faults,
the instruction no longer needs to be marked as serializing,
non-speculative, etc.

The "normal", aka architectural, aka FS version of the system call
instructions don't return a Fault artificially.

Change-Id: I72011a16a89332b1dcfb01c79f2f0d75c55ab773
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33281
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>

cpu-kvm: convert kvm base to new style stats

Change-Id: Iab2e99720cf9ac58edfcbdcedc944264eb12b7e1
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33796
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu-o3: convert fetch to new style stats

Change-Id: Ib50a303570ac1dd45ff11a32a823f47a6c4c02cd
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33815
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

ext: Link gem5 libelf to ext/libelf/libelf.a

Currently, gem5 might use system's libelf library instead of
the one compiled from ext/libelf.
This commit tells scons to use ext/libelf version.

JIRA: https://gem5.atlassian.net/browse/GEM5-756

Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Change-Id: I8dc4555c32a956e9f5249288c71982fa6a3678f7
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33941
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

ext: Revert "base: Use system libelf instead of ext"

This reverts commit bbb32ca1ef8eb8249cdca72be6171b94f61bd62e,
which tells scons to use the system's libelf instead of
ext/libelf.

Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Change-Id: I3bb3e62f2ef0fbc72983c221d5570edb4b35d157
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33940
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

ext: Revert "ext: remove libelf"

This reverts commit fa13042b5a1b4120bb7d96d15170bbd5d5068fad,
which removes ext/libelf.

Signed-off-by: Hoa Nguyen <hoanguyen@ucdavis.edu>
Change-Id: Id0ffb480fa5f5fe8faa4816d367b580ebe4c38d7
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33939
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem,ext: Integrating DRAMSim3 with gem5

Adding DRAMSim3 source code to the gem5 source code, the original
code was taken from umd-memsys github at https://github.com/umd-memsys/

Change-Id: I32c982206f33b0acf2121f322d15baa064c412c4
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/31757
Reviewed-by: Ayaz Akram <yazakram@ucdavis.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>

mem,ext: Fixed DRAMSim2 Integration

Fixed the way callbacks were used due to changes in
src/sim/callback.hh. Removed author line in SConsript.

Change-Id: I2c2b8dbe13e4f58680806126cd9cf209748e788a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33938
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem: Relax packet limit in packet queue

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: I4ac24bf18a0aff08a5b33c48179b882b27ef910c
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30317
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch: Add uReset helper to UPCState

This allows to reset without advancing the pc

https://gem5.atlassian.net/browse/GEM5-587

Change-Id: Ied566f4cd5efed5eb500447d3f14388482435475
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30315
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Alexandru Duțu <alexandru.dutu@amd.com>

arch, mem: Initial Hardware Transactional Memory implementation

Gem5 Hardware Transactional Memory (HTM)

Here we provide a brief note describing HTM support in Gem5 at
a high level.

HTM is an architectural feature that enables speculative concurrency in
a shared-memory system; groups of instructions known as transactions are
executed as an atomic unit. The system allows that transactions be
executed concurrently but intervenes if a transaction's
atomicity/isolation is jeapordised and takes corrective action. In this
implementation, corrective active explicitely means rolling back a
thread's architectural state and reverting any memory updates to a point
just before the transaction began.

This HTM implementation relies on--
(1) A checkpointing mechanism for architectural register state.
(2) Buffering speculative memory updates.

This patch is focusing on the definition of the HTM checkpoint (1)

The checkpointing mechanism is architecture dependent. Each ISA
leveraging HTM support can define a class HTMCheckpoint inhereting from
the generic one (GenericISA::HTMCheckpoint).

Those will need to save/restore the architectural state by overriding
the virtual HTMCheckpoint::save (when starting a transaction) and
HTMCheckpoint::restore (when aborting a transaction).

Instances of this class live in O3's ThreadState and Atomic's
SimpleThread.  It is up to the ISA to populate this instance when
executing an instruction that begins a new transaction.

JIRA: https://gem5.atlassian.net/browse/GEM5-587

Change-Id: Icd8d1913d23652d78fe89e930ab1e302eb52363d
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30314
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

misc: Remove the "fault" parameter from syscall functions.

This parameter was never set or used, just plumbed everywhere,
occasionally with a dummy value. This change removes all of that
plumbing.

Change-Id: I9bc31ffd1fbc4952c5d3096f7f21eab30102300b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33277
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>

arch-arm,arch-x86: Added missing overrides

These overrides were missing, causing compilation errors when
compiling with clang 9. Noted here:
https://www.mail-archive.com/gem5-dev@gem5.org/msg35657.html

Change-Id: I6d09a0e57af3131b9172d01468d2cdcf4b444c5d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33817
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Bobby R. Bruce <bbruce@ucdavis.edu>
Tested-by: kokoro <noreply+kokoro@google.com>

sim: Add missing overrides in the *Fault classes

Change-Id: I7a74df78f0f85ccf7fd896f98b301c1f998c1497
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33777
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

base: Make the Value stat's functor method accept lambdas.

This class can already accept a proxy variable and a "functor" which is
a pointer to either a function or an instance of a class with the ()
operator overloaded.

This change adds a FunctorProxy partial specialization which accepts
anything that can be used to construct a std::function<Result()>. The
constructor argument is copied and stored in the proxy which makes it
possible to define a lambda inline without having to keep a copy of it
around for the proxy to point to.

Also, the ValueBase stat's functor method now has a second version which
accepts a const reference rather than just a reference to its argument.
We need both because when accepting a reference to a lambda it needs to
be a const reference, but when accepting a pointer to a functor object,
we don't want it to be const because that would force the () operator to
also be const.

Change-Id: Icb1b3682d51b721f6e16614490ed0fe289cee094
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32901
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>

scons: Make scons aware of changes to cxx_type in param types.

If the cxx_type value changes, then the way the parameter is declared
in the param struct will also change, and the header needs to be
updated. scons would miss this sort of change before because it was only
checking the module the SimObject's source came from. The python names
and types of the parameters could stay the same, but the C++
representation might have changed because of edits somewhere else.

This CL assumes that cxx_type is the only thing that will change and
transparently affect the params struct. I tried making scons sensitive
to the entire ptype which would capture other parameters, but that
didn't work for some reason. This should be pretty safe in general, but
not 100% safe

Issue-on: https://gem5.atlassian.net/browse/GEM5-753
Change-Id: I06774889e60b987f727799f55d7ea2a775b6a319
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33695
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

test: Fix unittest cprintftime's build.

This test, which measures the performance of cprintf vs. sprintf, was
missing a couple of includes which were needed for the alarm() and
signal() functions, as well as the SIGALRM constant.

Also, it was using %#x to print the value of a pointer which gcc
complained about when compiling sprintf. This is fixed by changing that
format specifier to %p, the specifier to use when printing pointers.
Apparently either the implicit conversion to an integer value (which %#x
expects) or the size of the type it was converted to weren't good enough
for gcc any more.

Change-Id: I8eca3479bef2c2fa79f8ef4881bb3ff35d7c54ca
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33897
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

test: Remove refcnttest from the unittest SConscript.

The test itself was removed, but it was left in the SConscript. If you
tell scons to build everything in that directory, it will try to build
that test and fail.

Change-Id: I1e3923b0de12e891f53dab6f4e6e3e2b6975dc45
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33896
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-gcn3: Added missing header to hsa_driver.cc

`TypedBufferArg`, used in `src/dev/hsa/hsa_driver.cc` is defined in
`src/sim/syscall_emul_buf.hh` yet was not included. This commit adds
this missing header.

Change-Id: I3239a097eb71b6ebdad045eab6525a888a970f08
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33816
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Fix copy ellision on base compressor

Newer compiler versions have a problem with this move as
it prevents copy elision.

Change-Id: I802703df12e171d6a377b673d0ad7e202456b516
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33835
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch: Remove unused variable pcbb from ThreadInfo

Change-Id: Ib9e46934f1613c98758662cba26a46fcc2a76146
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33776
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

scons: Avoid the unsupported option -Wno-c99-designator in MacOS

Change-Id: I4d95c75915b17531bdd6d9161eb266bb91cd7bef
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33775
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

python: Import reduce function in FileSystemConfig

Not sure if this is required due to python3 or something else, but I got
the error "NameError: name 'reduce' is not defined". This fixes that
error.

Change-Id: I2dd71674306abcad1a90311664b18b9eee29b9ac
Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32374
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby: Change request to response in MOESI_AMD_Base-dir.sm

The responseToDMA MessageBuffer in MOESI_AMD_Base-dir.sm
transmits both data and acks, but it's vnet_type is currently
set as request. This should be changed to response.

Signed-off-by: Sampad Mohapatra <sampad.mohapatra@gmail.com>
Change-Id: I0eb9e8fc8e25111849605a710a5150ce5fc3b83b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33755
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Srikant Bharadwaj <srikant.bharadwaj@amd.com>
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

scons: Remove the AddLocalOption workaround.

The "append" option of the Help() scons method can be used to avoid
clobbering the built in and local option help.

This has the nice side effect of making it easier to add options in
other files since you now only need the built in AddOption provided by
scons itself, not the custom AddLocalOption version.

Change-Id: Ifa566087797d578df0c90f8f4fca70c8152fbf63
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32115
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Reviewed-by: Bobby R. Bruce <bbruce@ucdavis.edu>
Maintainer: Gabe Black <gabeblack@google.com>

mem-cache: Use cache's max CR on perfect compressor

Use cache's max_compression_ratio to setup the max_compression_ratio
of the PerfectCompressor.

Change-Id: Ib44aa61975fb2cc52f27f64a86c9df9c5531aa1a
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33387
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Explicitly define threshold of BDI's sub-compressors

Allow all sub-compressors of BDI to be successful as long as
they are able to compress. Then, BDI's actual size threshold
acts as the cutting point.

This situation arises on any multi compressor; yet, generalizing
this assumption might be too bold.

Change-Id: Iec5057d16d4a7ba5fb573133a30ea10869bd67e0
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33386
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Make compression size threshold a percentage

By changing the parameter into a percentage, changing the block
size will automatically reconfigure the size threshold. Also,
change the default percentage to 50% to avoid storing blocks
unlikely to co-allocate in compressed format.

Change-Id: I1458f19db39becc2d40c00269132fea01770016f
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33385
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Add stats for failed compressions

Add statistics to keep track of the number of times compression
has failed to provide blocks whose compressed size passes the
size threshold.

Also, update the compressed data's size if compression fails.

Change-Id: If3479572bf114f07911238c602ffef3a90b6a931
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33384
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Handle zero sizes on compression

The size can be zero in special occasions, which would
generate divisions by zero. This patch expands the
stats to support them. It also fixes the compression
factor calculation in the Multi compressor.

As a side effect, now that zero sizes are handled, allow
the Zero compressor to generate it.

Change-Id: I9f7dee76576b09fdc9bef3e1f3f89be3726dcbd9
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33383
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Add an extra decomp lat to multi compressor

There is extra hardware required when dealing with multi
compressors. As such, add a parameter to allowing increasing
their decompression latency to account for any extra delay.

Change-Id: I153e4c5ab6927ac092e2ebd767fe88974597bb20
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33382
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Store BDI's encoding in tags

According to the original paper the compressors' encodings are
stored in the tag-store (Storage cost analysis section).

Change-Id: I4c34f86022eea6d1ba0ae29dd74d5714bbad367a
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33381
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Add encoding bits to the data of multi compressors

When compressing using a multi-compressor, one must be able to
identify which sub-compressor should be used to decompress data.
This can be achieved by either adding encoding bits to block's
tag or data entry.

It was previously assumed that these encoding bits would be added
to the tag, but now make it a parameter that defaults to the data
entry.

Change-Id: Id322425e7a6ad59cb2ec7a4167a43de4c55c482c
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33380
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Standardize data parsing in compressors

The compressors are not able to process a whole line at once,
so they must divide it into multiple same-sized chunks. This
patch makes the base compressor responsible for this division,
so that the derived classes are mostly agnostic to this
translation.

This change has been coupled with a change of the signature
of the public compress() to avoid introducing a temporary
function rename. Previously, this function did not return
the compressed data, under the assumption that everything
related to the compressed data would be handled by the
compressor. However, sometimes the units using the compressor
could need to know or store the compressed data.

For example, when sharing dictionaries the compressed data
must be checked to determine if two blocks can co-allocate
(DISH, Panda et al. 2016).

Change-Id: Id8dbf68936b1457ca8292cc0a852b0f0a2eeeb51
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33379
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Allow inheriting from DitionaryCompressor's comp data

Previously either the compression data was the one declared within
DictionaryCompressor, or the derived class would have to override
the compress() to use a derived compression data.

With this change, the instantiation can be overridden, and thus
any derived class can choose the compression data pointer type
they need to use.

Change-Id: I387936265a3de6785a6096c7a6bd21774202b1c7
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33378
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Upgrade Compressor::Multi's stats

Use new style stats API for Compressor::Multi's stats.

Change-Id: Ia0313704cae4e7bd6bc675c71ea75b42a8e542f2
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33377
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Upgrade BaseDictionaryCompressor's stats

Upgrade this compressor's stats to match current stats API.

Change-Id: I1cb69230f8deca053bc860cedafc9e6e78446df7
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33376
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Fix RepeatedQwords compressor

This compressor does not allocate dictionary entries when there
is a match. This was causing the compressor to always fail.

Change-Id: I50eb56fa284854f3ee87f33af2c6e0a5c5248d7c
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33375
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Fix integer promotion of mask

When applying the bitwise not to a short integer the compiler
automatically promotes it to an integer. For example, if a 8-bit
mask=0xFF, and the compiler decides to promote the mask to 32-bit
to apply the bitwise not, ~mask=0xFFFFFF00, which will yield wrong
results for popcount(): expected=0, got=24.

Change-Id: I95efba5532c27ca004ff6947d4b51a8a14f09741
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33374
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

misc: Use VPtr in hsa_driver.cc

This change updates HSADriver::allocateQueue to take in a ThreadContext
pointer as opposed to a PortProxy ref. This allows the TypedBufferArg
to be replaced with VPtr.

This also fixes building GCN3_X86

Change-Id: I1fea26b10c7344daf54a0cb05337e961f834a5fd
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33655
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Matthew Poremba <matthew.poremba@amd.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem: convert base prefetcher and queued to new style stats

Base and Queued inside src/mem/cache/prefetch converted

Change-Id: I3d5907b58efefc4d8522b89f073507f2548bff2f
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33475
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

base: avoid recreating socket at each call to listen()

A new socket was created each time listen() is called,
which is problematic when the bind or listen operation
on it are not successful (mostly because the associated port is
already in use). It can lead gem5 to open too many files and crash
for multicores configurations, a socket being created
for remote GDB for each core. The other way to deal with
this problem would be to close the socket in the case the
function return false. But I find the proposed solution
simpler.

Change-Id: I848955a10c89e1da033bf773c83556a5dc5ef9a2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32994
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

scons: Set the minimum scons version to 3.0.

Change-Id: Id57a93e819588d2231d2d2d8b28cd62b05fbbe9b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33675
Reviewed-by: Steve Reinhardt <stever@gmail.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

configs: Add parameter for GPU scalar cache mandatory queue size

There was a missing option (--buffers-size) used to set the mandatory
queue size for the scalar controllers. This patch renames the option to
be more clear, and adds it to the argument parser.

Default of 128 taken from the implementation on the GCN staging branch

Change-Id: I58b6b57be07498cdf6e39c0bb85982674ec4caa6
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32676
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

misc: Fix db_offset calculation

db_offset used to be calculated through pointer arithmetic. Pointer
arithmetic increments the address by the size of the data type the
pointer is pointing at. In the previous db_offset calculation, that
was a uint32_t, which means the input was multiplied by 4, which is
sizeof(uint32_t)

This patch multiplies the input value by sizeof(uint32_t) before
assigning it to db_offset.

Change-Id: I9042560303ae6b8b1054b98e9a16a9da27843bb2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/32678
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Alexandru Duțu <alexandru.dutu@amd.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

configs: set hsaTopology properties from options

This change sets the properties in hsaTopology to the proper values
specified by the user through command-line arguments. This ensures
that if the properties file is read by a program, it will return
the correct values for the simulated hardware.

This change also adds in a command-line argument for the lds size, as
it was the only other property used in hsaTopology that didn't have
a command-line argument. The default value (65536) is taken from
src/gpu-compute/LdsState.py

Change-Id: I17bb812491708f4221c39b738c906f1ad944614d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/31995
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Alexandru Duțu <alexandru.dutu@amd.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Maintainer: Matt Sinclair <mattdsinclair@gmail.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

ext: remove libelf

Change-Id: I52eb9a7bb7d9f3522ff565498dc2821a0d1ed0d6
Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33318
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>

base: Use system libelf instead of ext

The only change needed is to remove EM_SPARC64, which from what I can
tell was removed from elf.h in 1998.

https://sources.debian.org/src/glibc/2.24-11+deb9u1/ChangeLog.8/#L6134

Change-Id: I0dd7e23ea44b19c2ebd9c6eff7cbaedfe69d821b
Signed-off-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33317
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Gabe Black <gabeblack@google.com>

util: Updated Dockfiles with the libelf-dev dep

This is required if we eventually remove `ext/libelf` (
https://gem5.atlassian.net/browse/GEM5-752), otherwise our tests will
fail.

The corresponding Docker images have been built and uploaded to:
https://gcr.io/gem5-test/

Change-Id: I1bd069dfb968b56eac4c4da33929b5ff895eaa6f
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33596
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

util: Added GCN-GPU to cloudbuild_create_images

This adds the GCN-GPU Docker image to this cloudbuild yaml file. We
use this file to create our Docker images:

```
gcloud builds submit --config \
util/cloudbuild/cloudbuild_create_images.yaml
```

The GCN-GPU docker image can be obtained at:
https://gcr.io/gem5-test/gcn-gpu

Change-Id: I7564db78a9f00507f5acc6cf6098de35f98b6fb1
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33595
Reviewed-by: Matt Sinclair <mattdsinclair@gmail.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

configs: Update starter_fs.py for latest Arm FS binaries.

Updated the default kernel and root device names to match the latest
Arm full-system binaries available for download on the gem5 website.

Also added a command line option to allow the root device to be
specified as an optional command line argument.

Change-Id: I27f90ffaf0f4b35c5dcc4c22ac2fbd34f8a040a4
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/30814
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Fix coding style in addressTranslation methods

armFault -> arm_fault

Change-Id: I6263b105f8757b34dd15a06b16abe7289073614d
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33434
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Check if PAC is implemented before executing insts

If Armv8.3-PAuth (PAC) extension is not supported, most instrucions
will trigger an Undefined Instruction fault; except for a group of
them living in the HINT space; those should be treated as NOP.

Change-Id: Idec920ed15e0310ec9132a3cb3701cdb7e7cf9d1
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33455
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Introduce HavePACExt helper

This will check for presence of pointer authentication extension.
According to the reference manual, Pointer authentication is
implemented if the value of at least one of

ID_AA64ISAR1_EL1.{APA, API, GPA, GPI}

is not 0b0000.

Change-Id: I4e98e65758e8edc953794e5b618d2c6c3f6000ae
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33454
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

python: Add support for introspecting scalar stats

This change adds a wrapper for the ScalarInfo stat type to enable
introspection of scalar stats from Python. Due to the slightly
confusing use of proxy objects in the stat system, PyBind11 fails to
automatically cast to the right wrapper type. This is worked around in
the by explicitly casting to the relevant type's Python wrapper.

To make the interface more Python-friendly, this change also changes
the semantics of resolveStat to raise an exception if the stat can't
be found.

Change-Id: If1fc6fe238fc9d69d4e22369a4988a06407d2f7c
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/33176
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
Maintainer: Jason Lowe-Power <power.jg@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>