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

Yannik Stradmann authored 4 years ago and Yannik Stradmann committed 4 years ago

Change-Id: I424e77ae6e6e9071e911debe99c63ec78768393e

Depends-On: 10416,10485,10678,11485
Change-Id: If9a4d0f49cd6d7e837a80e4c52f89e1e45b45bfa

* size(Loopback) == 1
* Loopback encodes halt (0) and tick (1)

Depends-On: 6909
Change-Id: I7b8cdee316a3a54f35873f3cc220387fa97a4f47

Eric Müller authored 4 years ago and Philipp Spilger committed 4 years ago

Change-Id: I025fdc00e2dc1433fd26edc66618e6171100284b

* A connection logs time of encode, decode, commit and
  execute operations
* Add logging to execute_messages, which logs in/out message count
  and time expenditure as difference of the time information before
  and after execution
* change default loglevel from info to warning in order to be able to have
  the time info print in execute_messages on info loglevel without spamming
  the normal user

Depends-On: 11111
Change-Id: I17835a9ae7f089da72a9004a18b967c117b6132d

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

In preparation of the upcoming integration of quiggeldy, some more
test-helper will be added that are also to be used in fisch tests. Hence
properly seperating helpers into different namespaces will prevent
possible clashes in the future.

Change-Id: I5786fde928674435f19323a27259faa09b936e50

* allows mutual exclusion in all higher layers
* mutual exclusion mechanism bound to connection instance instead of placed aside of connection

Change-Id: I2cb74e775aa532ecccf08b3418e16cbd1dd0096b

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

Change-Id: Iaff5843bc979a3672d7ed76c66cfad8ab6ce62eb

Change-Id: I1d4634254d3164499ddddd1a080598c14f2282d5

Depends-On: 10730

Change-Id: Ifcba7e1d4f27d9d7372e0bb6b6ae9d6d358daecc

In order to emulate unique_ptr interface that moves the connection
object from the connection handle. Needed to move the ownership of
connections to hxtorch.

Change-Id: Ib999fdf5a4ca6aa9ecd1085515797f6802fde73e

* As discussed on 2020-04-07 after the Softie's meeting (ECM, YS, PSP,
  CM, OJB), we want to have a streamlined connection interfaces for all
  the possible ways we turn FPGA words into other FPGA words in
  hardware.

* C++-API: Split old Connection objects into two parts:
  * The connection object that the user can only create and destroy but
    which holds state necessary for communication with the backend (hw,
    CoSim, QuiggelyClient, etc) and allows some state to be set.
  * Introduce hxcomm/common/stream.h:
    * The Stream-object featuring a uniform interface handling how to
      feed and read words to/from the backend.
    * The interface is defined via templated `Stream`-class that has to
      be befriended in the corresponding connection implementations
    * The Stream-interface is verified at compile time via the
      Streamable helper-struct.

* Move `execute_words` from fisch as more generic
  `execute_messages`-function that is the entry point for all upper
  layers when communcating with backends.
  * Overloading is possible by fully or partially-specializing the
    detail::ExecutorMessages-helper struct's operator() method (not
    needed once partial function specialization arrives, so never)
  * default implementation in common namespace by adding explicit
    `send_halt_message_type` to ConnectionParameter template argument

* Rework hxcomm/common/{arq,sim}connection.{h,tcc} to use the new Stream
  interface.

* Update all tests to use streams for interacting with the hardware.

* Rename init_parameters_t -> init_parameters_type to be consistent
  (they will be used by quiggeldy but that change is not merged -> no
  API break)

* Provide get_connection_from_env-function that returns a connection
  variant over all possible connections (and is used in pyhxcomm, see
  below).

* Introduce seperate `pyhxcomm`-namespace that is private to hxcomm and
  only to be accessed from Python. It adds convenience context-wrappers
  for connections ensuring connections get closed upon exiting contexts
  in Python
  * pyhxcomm::Managed<Connection> provides enter/exit-methods
  * enter() returns pyhxcom::Handle<Connection> holding the connection
  * if Handle<Connection> gets destroyed, connection gets freed (i.e.
    disconnected)
  * Managed<Connection> also holds a weak-ptr to the returned
    ConnectionHandle object and disconnects the Handle upon exit()-ing
  * Expose union over all Connections as `pyhxcomm_vx.ConnectionHandle`
    that can be used in type annotations in Python.
  * Expose special context manager `pyhxcomm_vx.ManagedConnection` that
    automatically creates a new connection from the environment and
    returns `ConnectionHandle` (which is a concrete `<Connection>Handle`
    in python rather than the cumbersome variant type).

* Introduce hxcomm::visit_connection that allows for applying operations
  on connections wrapped in Handle-objects such as variants for hxcomm
  and pyhxcomm::ConnectionHandles. It is used in
  hxcomm::execute_messages as well as stadls::run-pybind11 bindings.
  * Functions templated over connections can be easily specialized for
    all wrapped connections via:
    * `ConnectionIsPlainGuard<Connection> = 0`
    * `ConnectionIsWrappedGuard<Connection> = 0`
  * Another possibility is to perform compile time checks using
    `ConnectionIsPlain<Connection>::value` (see `run`-implementation in
    haldls for example).

* Add `static char const name[]` attributes for all connection for more
  informative logging.

Depends-On: 10346,10335,10703,10978,11027

Change-Id: I3ff901db53a5cb94a1faa50b18dc36786246f389

Change-Id: I3a003e27bddb3633b7c79ce71db23b2041c3dc22

Change-Id: I193c5b92bb79151bf75a16984c7ee8d1f1461ecb

* needed for run to terminate within gdb

Change-Id: Iacef0cf1cbcafe83b11068885bc1ac520a5a83da

This reverts commit 149edd5b.

Reason for revert: Connection to chip hangs frequently, a script running
for 10 minutes is hard to complete. The problem was introduced in this
change, but there may be better solutions than only reverting.

Change-Id: I463bc0692a265a69b49d71f64714cc52980d233e

Change-Id: I7aca0378ef8b793c1289836f5a0a73bae170fbda

* saves >200MB space on install

Change-Id: Ieca3ddf5ea245bddf6b8577ee3e67a145b97881e

* exponential backoff of waiting until word-data is ready to be
  received
* because of event-based inter-thread signalling leads to blocking
  in both fetching and decoding thread
* 1ms max latency
* solves the problem that two threads are busy waiting

Change-Id: I6419fec729d69cf067d8c614eb8c58109cac1664

Change-Id: I5b025d3a4b74dc9adf6e31d64bf3ac9ed87429d0

Change-Id: I3974261a0c3960115f12a48c5f1a6a348f9a4b54

Change-Id: I75f5fcc76870406696f6b8da1e7041ed041105ff

Change-Id: Ieb8fb41f8a2c174a81000fd4af9446485d0b6292

* Use dls-core for core-software stack
* Drop 'visionary' prefix

Change-Id: I8aacbd7276d87269fd9f21082321ca42eb5303a9

Change-Id: Ieb32f507bc3a58a260f6648ca37224d30cc22c34

Change-Id: I9cc2524a50772b31160c76130bb65ef070e9887d

Change-Id: Ib57bf96b4b2b1991eca81d446d66dd522904b64d

* use rant directly for only case (num-bits of JTAG data)
* use bitset for bit-aligned data in spike/MADC sample from chip

Depends-On: 10242
Change-Id: Ica7ea95caa84d228545ace9ff671f2a50940b39e