Data streaming between peer subsystems of a computer system

• US 5,568,614 A
• Filed: 07/29/1994
• Issued: 10/22/1996
• Est. Priority Date: 07/29/1994
• Status: Expired due to Fees

• Current Assignee: International Business Machines SA (International Business Machines Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Pipitone, Ralph M., Mendelson, Richard N.
• Primary Examiner(s): Geckil, Mehmet B.
• Application Number:
  US08/282,993
• Time in Patent Office: 816 Days
• Field of Search: 395/800, 395/890, 395/200.05, 395/842, 395/290, 395/474, 395/200.8, 395/479
• US Class Current: 709/231
• CPC Class Codes: G06F 13/28 using burst mode transfer, e.g. direct memory access DMA, cycle steal G06F13/32 takes precedence
• CPC Class Codes: G06F 9/542 Event management; Broadcasting; Multicasting; Notifications
• CPC Class Codes: G06F 9/544 Buffers; Shared memory; Pipes
• CPC Class Codes: H04N 19/42 characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation H04N19/635 takes precedence
• CPC Class Codes: H04N 19/423 characterised by memory arrangements H04N19/433 takes precedence
• CPC Class Codes: H04N 19/61 in combination with predictive coding
• CPC Class Codes: H04N 21/4143 embedded in a Personal Computer [PC]
• CPC Class Codes: H04N 21/4302 Content synchronisation processes, e.g. decoder synchronisation
• CPC Class Codes: H04N 21/443 OS processes, e.g. booting an STB, implementing a Java virtual machine in an STB or power management in an STB
• CPC Class Codes: H04N 21/4435 Memory management allocation of memory to service a request G06F9/5016; addressing or allocating within memory systems or architectures G06F12/02

Abstract

A data handling arrangement for a computer system, with particular application to multimedia systems, allows device adapters (control units) attached to the system to autonomously (without interrupting the system processor) control processing of a data stream of arbitrary length through memory buffers which are smaller than the stream. In this (stream processing) operation, data constituting the data stream flows through devices controlled by the adapter in succession, and is held in shared memory buffers as it passes between devices. The adapters are prepared for the stream processing operation by a system processor, indicate their preparation status to the processor, receive an initiating signal from the processor, and then proceed to direct the operation to completion without further assistance from the processor. In the operation, the adapters coordinate usage of the memory buffers by respective devices. Coordination is accomplished using a notification signalling protocol defined during preparation of the adapters.

Claims

1. For a computer system--that contains a processor subsystem, a memory subsystem, peripheral devices, device control units controlling said peripheral devices in response to commands received from said processor subsystem, and a channel linking said device control units to said processor and memory subsystems, said channel having a peer communication feature enabling said control units to communicate directly with each other without intervention of said processor subsystem and without intermediate buffer storage in said memory subsystem of information being communicated between said control units--a mechanism for enabling one of said control units to direct processing of a data stream of arbitrary length, by a device controlled by said one control unit, in cooperation with first and second other said control units respectively controlling first and second other devices, said mechanism comprising:

means in said one control unit for receiving setup command information defining;

(1) input and output buffer storage spaces in said memory subsystem that are to be managed by said one control unit in cooperation respectively with said first and second other control units;

(2) a notification protocol, defining signals to be directly transferred between said one control unit and said first and second other control units via said peer communication feature of said channel, said signals to be used by said control units for managing use of said input and output buffer storage spaces; and

(3) a process to be conducted, on a said data stream of arbitrary length in said input buffer storage space, by said device controlled by said one control unit, said data to be written to said input buffer storage space by said first other device, said process to produce result data to be written to said output buffer storage space for further handling by said second other device, and said process to be conducted to completion without interruption of said processor subsystem;

said input and output buffer storage spaces being characterized in that each has insufficient storage capacity to simultaneously store all of the data potentially contained in said data stream;

means in said one control unit responsive to said setup command information for preparing said one control unit and said device controlled by said one control unit to;

(1) perform said process defined by said command information on a said data stream to be conveyed through said input buffer storage space;

(2) forward data resulting from performance of said process to said second other device via said output buffer storage space; and

(3) manage usage of said input and output buffer storage space, in cooperation respectively with said first and second other control units, via said notification protocol and said peer communication feature of said channel; and

means in said one control unit, responsive to an initiating signal from said processor subsystem and to conditions established by said preparing means, for directing a streaming operation in which;

(1) said process is performed on data of said data stream forwarded through said input buffer storage space;

(2) data resulting from performance of said process is forwarded to said second other device via said output buffer storage space;

(3) transport of said data stream through said input buffer storage space, and transport of said resulting data through said output buffer storage space, are jointly managed by said one control unit and said first and second other control units using said notification protocol and said peer communication feature of said channel; and

(4) all of the foregoing functions are carried out to completion, i.e., until all of the data in the data stream has been processed and forwarded, without interruption of said processor subsystem.

2. The mechanism defined in claim 1 wherein said setup command information received by said one control unit is in the form of a plurality of variously sized requests, each request containing a parameters field including one or more descriptors, each descriptor defining a function essential to said streaming operation.

3. The mechanism in accordance with claim 2 wherein said descriptors include information defining dimensions of said input and output buffer storage spaces in said memory subsystem.

4. The mechanism in accordance with claim 3 wherein said descriptors include information defining an input buffer group constituting said input buffer storage space, said input buffer group containing at least two discretely separate variably sized storage subspaces, information defining dimensions of each of said subspaces, and information allowing for simultaneous access to said subspaces of said input group to enable said first other device to be writing data to one of said subspaces while said device controlled by said one control unit is reading data from another one of said subspaces.

5. The mechanism of claim 4 wherein said descriptors include information defining a communication protocol between said one control unit and said first other control unit, to be conducted via said peer communication feature of said channel;

said protocol enabling said one control unit and said first other control unit to control time interleaved access to said subspaces in said input buffer group, by said device controlled by said one and said first other device, in a manner ensuring that data in said data stream is moved efficiently through said input buffer group.

6. The mechanism of claim 3 wherein said descriptors include information defining an output buffer group constituting said output buffer storage space, said output buffer group consisting of at least two discretely separate storage subspaces, information defining dimensions of individual storage subspaces in said output buffer group, and information configuring said one control unit to manage access to said output buffer group in coordination with said second other control unit so as to allow said second other device to read and process data occupying one of said subspaces while data produced by said device controlled by said one control unit is being written to another one of said subspaces.

7. The mechanism of claim 6 wherein said notification protocol is of a form enabling said one control unit and said second other control unit to jointly manage access to said subspaces in said output buffer group, by said device controlled by said one control unit and said device controlled by said second other control unit, in a manner ensuring that said result data is moved efficiently through said output buffer group.

8. A mechanism in accordance with claim 3 wherein:

said input buffer storage space is configured for use as a FIFO (first in first out) buffer device, receiving blocks of data sequentially from said first other device and providing access to said blocks of data to said device controlled by said one control unit in the sequence in which the respective blocks were received.

9. The mechanism in accordance with claim 1 including:

means in said one control unit for providing a signal to said processor subsystem when said means for preparing said one control unit to conduct said streaming operation concludes its actions;

said signal to said processor subsystem serving to indicate either that said one control unit is prepared for said streaming operation or that said one control unit can not be prepared for said streaming operation due to error;

said signal to said processor subsystem preceding issuance of said initiating signal by said processor subsystem, and being used by said processor subsystem as a condition precedent to issuance of said initiating signal.

10. A mechanism in accordance with claim 1 wherein:

said setup command information includes synchronization information enabling said one control unit to synchronize said streaming operation to another streaming operation occurring external to said one control unit.

11. For use in a computer system, a streaming device adapter controlling a device performing operations requiring processing of arbitrarily long data streams, said adapter comprising:

means enabling said adapter to communicate directly with other device adapters through a channel of said computer system having a peer communication feature permitting such direct communication;

means for receiving setup command information from said computer system defining;

(1) an output buffer store having dimensions chosen by said computer system;

(2) a source of data to be supplied sequentially to said device controlled by said adapter, said data constituting a data stream of arbitrary length;

(3) a process to be conducted by said device controlled by said adapter, on all of the data in said data stream, with data resulting from said process to be written into said output buffer for further process handling by another device controlled by another adapter; and

(4) a protocol for communications to be conducted between said adapter and said another adapter, during the performance of said process on said data stream, to coordinate the use of said output buffer by said device and said another device;

wherein the length of said data stream exceeds the capacity of said output buffer, requiring repeated use of said output buffer to convey portions of said resulting data to said another device through said output buffer, and wherein said protocol requires use of said direct communication feature of said channel to indicate when data in said output buffer is available for further process handling by said another device; and

means responsive to an initiating signal issued by said computer system after said setup command information has been received by said receiving means, and to said setup command information, for directing said device controlled by said adapter to perform said process, and for using said protocol to notify said another adapter when data representing discrete portions of said data stream is available in said output buffer store;

said means responsive to an initiating signal causing said process to be performed continuously, until all of the data in said data stream has been processed, without intermediate interruption of or further instructions from said computer system.

12. An adapter according to claim 11 contained in a unit that is separate from said computer system and peripherally attached to said computer system through said channel.

13. A device adapter according to claim 11 implemented in a surrogate unit contained in a program executed by said computer system.

14. An adapter according to claim 11 wherein said output buffer store contains at least two sections, each section capable of storing a data block of predetermined length, and said communication protocol allows for said device controlled by said adapter and said another device to have concurrent access to different said sections;

whereby said another device can be processing a block of said resulting data in one of said sections while another block of resulting data is being entered into another one of said sections.

15. An adapter according to claim 11 wherein:

said output buffer store is contained in a memory subsystem of said computer system which is accessible to said device controlled by said adapter through said channel;

said information in said setup commands includes information defining the number of said sections in said output buffer store and dimensions of each section; and

said setup command information defining said communications protocol includes information defining events occurring in said sections as conditions precedent to initiating communications from said adapter to said another adapter, and the form of signals to be contained in said communications.

16. An adapter according to claim 11 wherein said computer system is capable of initiating concurrent processing of multiple different data streams, by devices controlled by said adapter and other adapters, and the setup command information furnished to said adapter includes information for synchronizing the processing of one data stream, in said device controlled by said adapter, with the processing of a different data stream in a different device.

17. An adapter according to claim 16 wherein said synchronizing information includes information defining the location of a timing function in the memory of said computer system which provides a time reference to the processing of said different data stream.

18. An adapter according to claim 11 which controls at least one device capable of performing a plurality of different transformational processes on data, and said setup command information includes information defining a selected one of said plurality of transformational processes as said process to be performed on said data stream.

19. A computer system comprising:

a processor subsystem;

a memory subsystem;

a plurality of device adapters;

a plurality of devices each controlled by one of said adapters; and

an I/O channel linking said subsystems, said adapters, and said devices;

wherein;

said channel includes a peer communication feature enabling said adapters to directly exchange control information without involvement of any of said subsystems;

said processor subsystem includes means for concurrently preparing at least two of said adapters, and respective devices controlled by said at least two adapters, to autonomously process data in a data stream of arbitrary length without interrupting said processor subsystem until all of the data has been processed;

said preparing means includes;

means for furnishing said at least two adapters with information defining;

buffer storage spaces in said memory subsystem to be cooperatively managed by said adapters for forwarding the data in said data stream between devices controlled by said at least two adapters;

a process to be conducted by devices controlled by said at least two adapters on the data in said data stream; and

a communication protocol to be observed by said at least two adapters, via said peer communication feature of said channel, for indicating when said buffer storage spaces contain data that is available to be read and processed by said devices controlled by said at least two adapters;

said buffer storage spaces having insufficient storage capacity to simultaneously store all of the data in said data stream;

means for receiving indications from said at least two adapters when said at least two adapters are respectively prepared for controlling processing of data in said data stream; and

means for signaling said at least two adapters to begin controlling processing of said data stream;

and said at least two adapters include means responsive to said signal to begin to control respective devices to process data in said data stream;

said adapters cooperating via said communication protocol to manage access to said buffer storage spaces and control processing of data in said data stream autonomously, without interrupting said processor subsystem, until all of the data in said data stream has been processed.

20. A computer system containing processor and memory subsystems, a channel linked to said subsystems, said channel having a peer communication feature, and plural device control units and data handling devices linked to said subsystems via said channel, said control units being able to intercommunicate directly with each other via said peer communication feature, wherein said system comprises:

a source of data representing a data stream of arbitrary length to be processed sequentially by a plurality of said devices;

means in said processor subsystem for preparing at least two of said control units to control processing of said data stream by devices respectively controlled by said at least two control units;

said preparation including furnishing said control units with information defining variably sized and variably located data buffers in said memory subsystem that are to be shared by pairs of devices controlled by said at least two control units, said buffers being filled with data by one device of each pair and emptied by the other device of the respective pair, said buffers shared by any said pair of devices having insufficient capacity to store the entire data stream at one time;

said preparation further including indications by said at least two control units to said processor subsystem when respective control units are prepared to control processing of said data stream by respective devices, and issuance of an initiating signal from said processor subsystem to at least one of said at least two control units when all of said at least two control units have indicated their preparation; and

means in said at least two control units, responsive to said initiating signal and information received during said preparation, for controlling respective devices to sequentially process all of the data in said data stream without interruption of said processor subsystem;

said at least two control units cooperating in pairs to control forwarding of data between devices controlled by each pair of units;

said means responsive to said initiating signal in each said pair of control units using said notification protocol to indicate when a buffer shared by a pair of devices controlled by the respective pair of units, contains data written by one device of the pair that is available for processing by the other device of the same pair.

21. A data handling system comprising:

processor and memory subsystems;

at least two control units;

at least two data handling devices respectively controlled by said at least two control units; and

a channel linking said control units and devices to said processor and memory subsystems;

said channel having a peer communication feature allowing for said at least two control units to intercommunicate directly with each other without involvement of said processor subsystem; and

wherein;

said processor subsystem includes means for setting up and initiating a data streaming operation, to be autonomously controlled by said at least two control units, in which;

a data stream of arbitrary length is to be handled in succession by said at least two devices, said data is to be passed to successive ones of said at least two devices through a shared memory buffer allocated by said processor subsystem, and said devices are each to perform a data transformation function selectively specified by said processor subsystem; and

each of said at least two control units includes means conditioned by said setting up and initiating means for directing the performance of said data streaming operation by a respective one of said at least two devices in an autonomous manner, requiring no interruption of said processor subsystem prior to completion of the operation, and in a manner which effectively isolates the respective device from conditions and events occurring external to that device;

said directing means including means for managing repeated use of said shared memory buffers allocated by said processor subsystem to supply data in said data stream to the respective device and carry transformed data in said data stream away from the respective device;

said repeated use being managed in coordination with other said at least two control units, via notification communications conveyed between said control units through said channel, and in a manner which isolates the respective device from buffer reuse events occurring as a result of said notification communications.