Electrical computers and digital processing systems: multicomput – Computer-to-computer data routing – Least weight routing
Reexamination Certificate
1999-01-15
2001-11-27
Butler, Dennis M. (Department: 2182)
Electrical computers and digital processing systems: multicomput
Computer-to-computer data routing
Least weight routing
C713S400000, C713S600000
Reexamination Certificate
active
06324591
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
Generally, the invention relates to the monitoring of the time synchronization between a processing unit, such as a processor core, with exterior processing means, for example other microprocessors or other systems, during an exchange of data between this processing unit and these exterior processing means, by way in particular of memories of the “first-in first-out” type (FIFO memory).
The invention relates in particular to the exchanging of data at high bit rate between processing units.
2. Description of the Related Art
At present, the synchronizing of the exchanging of data between two hardware systems can be carried out with the aid of a master/slave procedure in which one of the two hardware systems, termed the “master”, drives the second, termed the “slave”. However, such a mechanism is slow since it requires interchanges of synchronization cues (“handshake”). Furthermore, when such a procedure uses a DMA-based technique, well known to the person skilled in the art (“direct memory access”), the DMA is in charge of the exchange. It is therefore the guarantor that the correct data will arrive at the correct moment. It relies on an addressed organization of the data to be accessed (each datum being associated with an address). However, such a solution only operates well in simple situations (transfer between a memory and a processing core for example) and an asymmetry is created since it is necessary for the exchange to take place between a master and a slave, thus rendering communication impossible without the intermediary of two DMAs.
Furthermore, the DMA requires time to possibly take into account the alterations in its environment (data unavailable, data originating from several sources, etc.) which crops the passband.
It is also possible to use supervised links in which there is still symmetry between the two systems exchanging the data, but the synchronization between these two systems is taken on board by a supervision unit. Such systems are also slower, adapt less easily to the local context of the two systems exchanging the data. Moreover, the supervisor may become complex in cases where several communication paths exist between several systems. Finally, the exchanges are less flexible.
It is further desired that the process and device which will allow massive exchanges of data (wide passband) whose significance does not depend on the precise instant at which they take place and on the time intervals between the data exchanged, but only on the order of the data in the sequence. It is further desired that the process and device allow an exchange of data at high speed between a processing unit and exterior processing means, this having a considerable advantage for real-time systems.
SUMMARY OF THE INVENTION
In an embodiment, the data exchanges may take place through several data streams, means being provided to check the consistency between these streams, signal the anomalies and correct them in most cases or even in all cases.
The self-synchronization preferably allows the processing unit talking to the exterior processing means to work in tempo with the data to be processed and to adapt itself to this tempo. This significant advantage is obtained in particular by the fact that separators are exchanged which travel in the same direction as the data (and make it possible to “tag” the data flow) and are processed “on the fly”, as opposed to conventional systems for which there is always two-stage working.
More precisely, in these conventional systems, there is either an interchange of cues in the form of a data send followed by a return of an acknowledgement of receipt, this inevitably taking more time, or, prior to the exchange proper, initialization of a particular mode indicating the nature of the exchange which is going to take place. This can consume an ever larger share of the time the smaller the size of the data or should they arrive irregularly.
Contrary to these conventional methods, an embodiment of the process makes it possible to exchange the data “on the fly” and it moreover makes it possible to correct anomalies in the sequence of data and of separators without slowing the whole system down.
In an embodiment, each data stream possesses its own clock outside or may be asynchronous. In all cases, external access to a data stream is completely uncorrelated temporally from the clock of the processing unit. Thus, an “outgoing” data stream may feed the input of the processing unit directly or constitute an incoming stream for another circuit.
Firstly a process for monitoring the time synchronization between a processing unit and an exterior processing means, the said processing unit being clocked by an internal clock signal and the exterior means delivering, by way of at least one input communication means, at least one clocked incoming stream of incoming cues, to at least one input port connected to the said unit, is described in which
incoming cues of data temporally separated by incoming cues for so-called separation (separators), according to a predetermined scheme for incoming separation, are incorporated into the incoming stream,
the processing unit is controlled by a first control mechanism which formulates for the said unit, at each cycle of the internal clock, at least one read request cue representative on the one hand of the absence or of the presence of an actual request for reading an incoming cue and on the other hand of the nature of the incoming cue possibly requested by the said unit having regard to the predetermined scheme for incoming separation. Stated otherwise, the read request cue specifies whether the processing unit is actually requesting the reading of an incoming cue or else is not requesting anything at all. The read request cue also specifies the nature of the cue possibly requested by the unit, that is to say a datum or a separator.
The read-available contents of the input port are analyzed, in the presence of an actual request for reading an incoming cue. For example in the case where the input port (for example a FIFO) is empty, and in the event of an absence of a read-available incoming cue, or in the presence of a read-available incoming cue of a different nature from that defined by the said actual read request (the FIFO delivers a datum while the processing unit is requesting a separator and vice versa), a synchronization anomaly cue is formulated.
This synchronization anomaly cue can be a specific cue or a logic signal prompting disabling of the internal operation of the processing unit. In most cases, the synchronization anomaly will be able to be self-corrected. This being so in certain very particular cases, for example when several types of separators exist and no hierarchization has been provided for between these various types of separator, time self-synchronization is not possible and it is then performed under the control of an exterior supervisor.
When the external means deliver respectively to a plurality of input ports by way of a plurality of digital buses, a plurality of clocked incoming streams, incoming cues of data temporally separated by incoming cues for separation, according to respective predetermined schemes for incoming separation, are incorporated into each incoming stream. The first control mechanism then formulates a plurality of read request cues, each being representative for a corresponding incoming stream, on the one hand of the absence or of the presence of an actual request for reading an incoming cue of the said corresponding incoming stream, and, on the other hand of the nature of the incoming cue possibly requested by the said unit having regard to the corresponding predetermined scheme for incoming separation. The available contents of each relevant input port are analyzed, in the presence of at least one actual request for reading an incoming cue of one at least of the incoming streams. In the event of an absence of available incoming cue in one at least of the said relevant input ports, or in the presence of an available incoming
Dufal Frederic
Remy Michel
Rogel Pierre-René
Butler Dennis M.
Conley & Rose & Tayon P.C.
France Telecom
Meyertons Eric B.
LandOfFree
Method and device providing time synchronization between a... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and device providing time synchronization between a..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and device providing time synchronization between a... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2607326