Change-Id: If5c2561b2d2587adc63333d57131b2065c4c174d

Based on a discussion between PSP and OJB:

We only want one `friend` declaration in each connection; and that is
for `Stream`.

Stream then has a nested Streamable struct, that performs that checks
because it has a right to access private/protected members. It is
implemented as a series of static_asserts that give a helpful error
message indicating what failed exactly. But, that means that when we
instantiate the default-constructed Stream it is effectively a
declaration of intend (i.e., "we intend this connection instance to
support the full Stream interface and hard-fail if it doesn't"). We
cannot perform a "check" if the Connection is even supposed to fulfill
the full Stream interface beforehand. A separate checker would require
to also be friends with all connections..

Hence, up until now we provide a manual hint
("has_full_stream_interface") that can be used to separate the
connections into those that are supposed to fulfill the full stream
interface (and are then used for in-depth hxcomm-tests) and those that
aren't.

Better:
We should have the indication if a Connection is supposed to be
Streamable in a separate type so that Stream only ever gets instantiated
via its default implementation where a failure to fulfil Streamable
should result in a compile error with helpful error message.

This struct is supports_full_stream_interface.

Change-Id: I402df2115373d7becfd3d5a3c76c60b3f1e94c66

Change-Id: I716c34a9f48755c6b1b35fa1cddb9e2a097c6e48

Change-Id: I4a09a8111a6ad57bf2a94c4be8b6dff91c1544bf

Depends-On: 10801
Change-Id: I527b9abff8913165c8370ab548604a377b883464

Depends-On: 12737

Change-Id: I7375eea63e4272b06282835f133e8478ac2afd57

* replace tbb::concurrent_queue by std::vector and mutex
  because the concurrent queue does not support thread-safe swap
* allows fast acquisition of _all_ messages from the connection
* in the typical use case of sending, waiting until halt and receiving
  this leads to less time spend with exclusive access to the connection
  and therefore more possible parallelism
* remove single-message receive() and try_receive() interface

Depends-On: 11113
Change-Id: Ifd6cebfcea3b9bc56699aaf9db741579ca074a81

* Stream::add(), Encoder::operator(), Decoder::operator()

Change-Id: Icda158123eed0f9c7af0a7aa2ef06ace4856cb43

* same performance
* newer implementation, full C++17 support, constexpr variant & visit
* less non-std dependencies

Depends-On: 11711
Change-Id: I721a383221492ef1370ddc1fa06a1097f1eb1657

* defaulted operator was implicitely deleted

Change-Id: I554c9d709d6b1d6b79d945606135a21180608286

* use std::showbase, remove logging in Connection::add(message)
* change PHY word logging loglevel to trace

Change-Id: Ic485583ddf4eceddb760315d572126df21358d04

Depends-On: 11747
Change-Id: I74653af54fb322644156a9d985316a4c8cb87a93

* this was very useful during debugging to verify the client not choking
  on received packets

* logger needs to be initialized prior to working thread starting ->
  reorder members.

Change-Id: I6779ccc98a017efd4af6cfc179c9e4dbaa1847da

* A send queue consisting of a single packet of HostARQ is used
* full packets are sent automatically
* this involves the implementation change, that add() is allowed
  to transfer to the hardware and commit only guarantees all
  messages added until now are sent
* beforehand during we did never communicate to the hardware
* no heap allocation used as part of encoding and sending in
  the connection
* maximal encoding latency is the time to encode a single packet
* since encoding speed at the moment is around 800MB/s, maximal
  throughput without this change is 1/(1/800 + 1/125) = 108 MB/s,
  which is only 86% of full gigabit
* with this change if encoding is faster than the transfer,
  the full transfer bandwidth can be used

Change-Id: Ic1f7d3f663197bb06a98df1025123d002f21d4b1

* fix reference type of argument to visitor to be constant reference,
  which is guaranteed by the iterable being constant reference

Change-Id: I0df700bd1059ee522dc589e8457588cb2aae7b03

Change-Id: I0936486e41d158d8d015652ae673d7fc487281b9

* remove unecessary assertion (condition is guaranteed by operation)
* place update to m_buffer_filling_level above block, where value
  is not needed anymore
* speed improvement of 20% from 666MB/s to 800MB/s on RyzenHost

Change-Id: Ibe4eac7ae3ef763ff2821f10f9781dee905bca57

Change-Id: I062b8e2bb639120044e80749d88e6619b6dabe70

Depends-On: 12215
Change-Id: I131ffb0a189e157029b3631f235ffe66f0939039

Depends-On: 10581
Change-Id: Ied502779cb2981be1696cea02d453f5a8c7485df

Depends-On: 12005
Change-Id: I7a4a0df165988e2356c35b2cd1773f672a03df76

Depends-On: 11170,11696
Change-Id: If6bef2b51e7696c9368a35216d215f206886ca1b

Change-Id: Id6b6587af1f784303fe90e17a87b182b3f9a1f70

Not all connections will support the full direct-access interface that
the default implementation of Stream provides but rather only the subset
needed for execute_messages()-invocations.

* Introduce `ConnectionFullStreamInterfaceVariant` that is merely a
  filtered version of the full `ConnectionVariant` with those types
  removed that do not support the full(= default) Stream interface.

* Introduce `get_connection_full_stream_interface_from_env()` that gets
  the connection defined by the environment but checks if it can support
  the full_stream_interface. If so the connection is wrapped in an
  `optional`, otherwise nullopt is returned.

* Tests are then skipped at runtime via `GTEST_SKIP` if the optional is
  empty.

Change-Id: Ib3e196723e9c2beb8a0b00f57372f5b7cb359369

Change-Id: Ifce435b71d581094bbb42b3313ac70ddc0601938

Since we cannot use aliases for unspecified templates (see point 2.2
here:
http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2002/n1406.pdf),
having templated template parameters is of less use than anticipated
because we cannot have a single source of truth in regards to what
sequence-like type is used throughout the stack.

Solution:
* Replace all `template <typename> class Sequence` parameters with
  helper structs that map the connection-type to return-/arugment-types
  for execute_messages.
  If we ever decide to switch from vector to some other sequence, only
  this mapping will have to be updated.

* Extend `GetMessageTypes`-helper struct so that it maps
  `ConnectionParameters` to `MessageTypes` directly besides extracting
  `message_types`-argument from connection-types.
  This is important so that the RCF-interface in xiggeldy can derive its
  interface from the execute_messages-signature.

Change-Id: I65045d361ad10a97e06fce3178d081c621fc02c7

* For the ARQConnection the hwdb is queried for unique identifier bits

Depends-On: 11708,11930,11946
Change-Id: I39ea8ce989534bb410094a5eaf37d68e3eb438e6

Change-Id: I5765e4204b458a1dce017564e461f67600b05a85

Change-Id: I8b1271440ffd76af981b472b8caf4544e300f04b

Philipp Spilger authored 4 years ago and Oliver Breitwieser committed 4 years ago

* autogeneration is not possible if there exist connections which
  can't generically be defined as a templated class which only takes
  the specific ConnectionParameter, which is the case e.g. if
  specializations are to be made depending on the actual usage

Depends-On: 11765

Change-Id: I1ff982ae4f03e1f99e926fe9a91a2235f3cdcab3

Depends-On: 11361
Change-Id: I5ed7189f793495f99e0bc1e9e7d9373f11a126f8

Change-Id: Iacac09a15f5a7debac5116de999dd21e7dfb0b61

* API-wise almost the same, but easier to understand, use, provide

Change-Id: Icf448f1e5902191c4cdf2d1f19f8a9ce41e02bec

Depends-On: 11627
Change-Id: If99e9d3fcdce0cc21889e7d2054c741b667500c2

* has_leading_comma(word) is now static
* message-type specific decode_message lambda is separate templated member function
* code common to all message-types split out of specific function -> less code size

Change-Id: I8b41e4af9aa6efab88963f2fa519f36c3e3f9ed0

* use logger instead of direct cout in throughputtests

Change-Id: I343e55338499c827e8afab79c7ffd687413e8bdb

* use get_connection_from_env followed by a visit
* remove unused dispatch headers

Change-Id: I4b69d959443e4d1a6a57284ae8e96ffad69a6519

* no performance drop
* less code complexity

Change-Id: I8c7513ca5cd883d5f4420a7812983356a4fdbe98

Change-Id: I8c934d68f775412d681681a07d24f2542bc271ec

Oliver Breitwieser authored 4 years ago and Philipp Spilger committed 4 years ago

* Add support for filling standard-layouted data types with random data
  prior to serialization so that the test fails if not all members are
  serialized.

* Performed manual verification that test breaks if not all members of
  ConnectionTimeInfo are serialized.

Change-Id: I0b7e257e776ec71e7b5e7131375fb069c6388dcf