Electrical computers and digital processing systems: processing – Processing architecture – Data driven or demand driven processor
Reexamination Certificate
2001-03-14
2004-09-14
Treat, William M. (Department: 2183)
Electrical computers and digital processing systems: processing
Processing architecture
Data driven or demand driven processor
C712S026000
Reexamination Certificate
active
06792522
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to data driven information processors, and particularly to a data driven information processor that carries out processing using a data packet in which a plurality of operand data are stored.
2. Description of the Background Art
FIGS. 18A-18D
show the structure of a conventional data driven information processor and a data packet disclosed in Japanese Patent Laying-Open No. 9-114664.
Referring to
FIG. 18A
, a conventional data driven information processor includes an input unit
100
to which a port Pi (i=1, 2, 3, . . . n) is connected to input external data, a junction unit
101
, a paired data detection unit
102
, an operation apparatus
103
, a program storage unit
104
and a branch unit
105
.
FIG. 18B
is a block diagram of a portion of operation apparatus
103
.
FIGS. 18C and 18D
show respective examples of the field structure of a conventional data packet.
The data packet of
FIG. 18C
includes a field F
1
storing instruction information OP, a field F
2
storing destination information DN, a field F
3
storing a generation number GN, and a field F
4
storing at least one data Di (i=1, 2, 3, . . . n).
In program storage unit
104
of
FIG. 18A
, a data flow program is prestored. The data flow program includes a plurality of program data corresponding to a set of instruction information OP and destination information DN.
Instruction information OP serves to identify the operation executed on the contents of corresponding data field F
4
. Destination information DN is used to read out the subsequent program data from program storage unit
104
. Reading out program data from program storage unit
104
is referred to as “program fetch” hereinafter. Generation number GN serves to identify a plurality of data packets having the same destination information DN. Data Di implies a plurality of data applied to input unit
100
via port Pi, corresponding in number to the number of ports Pi. In this example, the plurality of data Di in field F
4
share the same instruction information OP, destination information DN and generation number GN.
When data Di which is a portion of data in field F
4
of the data packet is to be used in the operation in the data driven information processor of
FIG. 18A
, the field structure of the data packet shown in
FIG. 18D
is employed. In the data packet of
FIG. 18D
, a data field F
50
is provided instead of data field F
4
of FIG.
18
C. The remaining field structure of
FIG. 18D
is similar to that of
FIG. 18C
, and description thereof will not be repeated. Field F
50
includes a flag V corresponding to each data Di. Flag V indicates whether the corresponding data Di is to be used in the operation, i.e., valid data, or not used in the operation, i.e. invalid data.
Input unit
100
of
FIG. 18A
receives plurality of data individually through port Pi serving to input externally applied data to the data driven information processor. The data is stored in field F
4
in one data packet of FIG.
18
C. Then, that data packet is output. Junction unit
101
arbitrates and inputs the data packets output from input unit
100
and branch unit
105
, which is sequentially output to paired data detection unit
102
. Paired data detection unit
102
effects the matching operation of the data packets applied from junction unit
100
. As to the data packets forming a pair, i.e. two different data packets with matching destination information DN, the contents in data field F
4
of one data packet is stored in data field F
4
of the other data packet, and that other data packet is output. Accordingly, data Di that is to become the two operand data for a dyadic operation, for example, is detected as the paired data. Here, one of the paired data is referred to as LDi (i=1, 2, 3, . . . n) and the other of the paired data is referred to as RDi (i=1, 2, 3, . . . n).
Operation apparatus
103
receives the data packet output from paired data detection unit
102
to decode instruction information OP in the received data packet and applies a predetermined processing on data Di stored in data field F
4
in the received data packet according to the decoded result. The result is stored in data field F
4
of the received data packet to be output.
Program storage unit
104
receives the data packet output from operation apparatus
103
to read out the subsequent destination information DN and the subsequent instruction information OP from the data flow program according to addressing based on destination information DN of the received data packet. The readout information are stored in fields F
2
and F
1
, respectively, of the received data packet. Then, that data packet is output. Branch unit
105
receives the data packet output from program storage unit
104
to provide that data packet outside of the data driven information processor or to junction unit
101
according to destination information DN in the received data packet.
Operation apparatus
103
includes a plurality of operation units OPUi (i=1, 2, 3, . . . n) to which the paired data (LDi and RDi) and instruction information OP of the data packet input to operation apparatus
103
are applied. Each operation unit (OPUi) simultaneously processes each corresponding paired data according to instruction information OP. The processed result is output as data Di. Output data Di is each stored in field F
4
of the input data packet. Then, that data packet is output.
In the above-described conventional data driven information processor, a plurality of operation units (OPUi) that carry out the same operation on each data Di in the data packet are prepared. These operation units (OPUi) carry out processing based on common instruction information OP. This means that all data Di in the data packet are unconditionally subjected to the same operation. In order to apply an operation only on desired data Di in data field F
4
, the data packet must be divided prior to the operation to carry out a predetermined operation on the desired data Di by a corresponding operation unit (OPUi), followed by a combining process for the divided data packets. Thus, the process becomes redundant in the conventional data driven information processor for an operation that is to be applied only on desired data Di in data field F
4
.
To apply an operation only on desired data Di in the conventional data driven information processor, the data packet of
FIG. 18D
is used to effect the operation selectively on data Di having a flag V that indicates valid data, i.e. on only the desired data Di. By employing this method, redundancy of the process is eliminated since the dividing and combining processes of the data packets are not required. However, the size of the data packet becomes larger since the region to store flag V is additionally required for each data Di in the data packet. The width of the data transmission path of the data driven information processor is increased to obviate reduction in the size of the processor.
SUMMARY OF THE INVENTION
An object of the present invention is to provide a data driven information processor that can improve the efficiency of computing operation using a data packet in which a plurality of operand data are stored.
To achieve the above object, the data driven information processor of the present invention includes a paired data detection unit, a processing unit, and a control unit. The paired data detection unit sequentially receives a data packet including at least a destination field to store destination information, an instruction field to store instruction information, and a data field to store a plurality of data independent of each other to detect two different received data packets forming a pair. The contents of the data field of one received data packet detected as a pair are stored into the data field of the other input packet detected as the pair. Then, that other data packet is output.
The processing unit receives the data packet output from the paired data detection unit to app
Hatakeyama Kouichi
Takamatsu Kisho
Birch & Stewart Kolasch & Birch, LLP
Sharp Kabushiki Kaisha
Treat William M.
LandOfFree
Data driven information processor carrying out processing... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Data driven information processor carrying out processing..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Data driven information processor carrying out processing... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3218149