Data processing in which concurrently executed processes communicate via a FIFO buffer
View Patent ↗A processing circuit executes a data producing process and a data consuming process. The data producing process produces a stream of data, the data consuming process consumes the stream of data concurrently with production of the stream. A first-in first-out buffer passes data from the stream between the data producing process and the data consuming process. The buffer comprises buffer memory, the buffer writes data-items from the stream in circular fashion into the buffer memory. A consuming process interface is arranged to process a command for making a data grain from the stream available to the data consuming process. The interface responds to the command by testing whether addresses of data within the grain to which access has to be gained wrap around in the circular FIFO buffer. The interface copies the grain from the FIFO buffer to the auxiliary memory region, on condition that the addresses wrap around, so that the wrap around is eliminated in the copied grain, The interface returns an indication to the consuming process to read the grain from the FIFO buffer when the addresses do not wrap around inside the grain, or an indication to read from the auxiliary memory region, when the addresses wrap around.
1. A data processing apparatus, comprising:
a processing circuit arranged to execute a data producing process and a data consuming process, the data producing process producing a stream of data and the data consuming process consuming the stream of data concurrently with production of the stream;
processing memory accessible to the data consuming process;
a first-in first-out circular buffer unit for passing data from the stream between the data producing process and the data consuming process, the circular buffer unit comprising buffer memory and writing data-item from the stream in a circular fashion into the buffer memory; and
a consuming process application program interface (API) of the circular buffer unit, the consuming process API being arranged to:
access an auxiliary memory region accessible to the data consuming process,
process a command for making a data grain from the stream available to the data consuming process,
respond to the command by testing whether addresses of data within the grain to which access has to be gained wrap around in the buffer memory,
copy, in response to detection that the addresses wrap around, the entire grain from the buffer memory to the auxiliary memory region, so that the wrap around is eliminated in the copied grain, and
return a pointer to the consuming process indicating a location in the buffer memory from which to read the entire grain when the addresses do not wrap around inside the grain, or a pointer indicating a location of the auxiliary memory region from which to read the entire grain when the addresses wrap around.
2. A data processing apparatus according to claim 1 , wherein the consuming process is arranged to select an address of the auxiliary memory and a grain size as part of the command.
3. A data processing apparatus according to claim 1 , wherein the data producing process and the data consuming process are arranged to use a first and second grain size for sending data to and receiving data from circular buffer unit respectively, the first and second grain size differing from one another.
4. A data processing apparatus according to claim 1 , wherein the data producing process s arranged to use variable grain sizes for sending data.
5. A data processing apparatus according to claim 4 , wherein a first grain size and a size of the circular buffer unit are selected so that addresses of the data in the circular buffer unit always wrap around between successive grains of the first gain size.
6. a data processing apparatus according to claim 1 , further comprising:
a further processing memory accessible to the data producing process for producing the data stream; and
a producing process application program interface (API) of the circular buffer unit, arranged to:
receive a further auxiliary memory region selection for the data producing process,
process a further command for outputting an output data grain from the stream,
respond to the further command by testing whether addresses of data within the output grain will wrap around in the buffer memory,
return a pointer indicating a location in the buffer memory to which the producing process writes the entire grain when the addresses do not wrap around inside the grain, or a pointer indicating a location of the further auxiliary memory region to which the producing process writes the entire grain when the addresses wrap around, and
copy, in response to detection that the addresses wrap around, the entire grain from the further auxiliary memory region to the buffer memory so that the wrap around is created in the copied grain.
7. A data processing apparatus, comprising:
a processing circuit arranged to execute a data producing process and a data consuming process, the data producing process producing a stream of data and the data consuming process consuming the stream of data concurrently with production of the steam;
processing memory accessible to the data producing process;
a first-in first-out circular buffer unit for passing data the stream between the data producing process and the data consuming process, the circular buffer unit comprising buffer memory and writing data-items from the stream in a circular fashion into the buffer memory; and
a producing process application program interface (API) of the circular buffer unit, the producing process API being arranged to:
access an auxiliary memory region accessible to the data producing process,
process a command for making memory available to the data producing process for writing a produced grain, respond to the command by testing whether addresses of data within the grain for which memory has to be made available wrap around in the circular buffer memory, and
return a pointer to the producing process indicating a location in the buffer memory to which to write the entire grain when the addresses do not wrap around inside the grain, or a pointer indicating a location of auxiliary memory region to which to write the entire grain when the addresses wrap around.
8. A machine implemented method for implementation of a signal processing task comprising concurrently executing processes between which a data stream is communicated via a circular buffer memory, the method comprising:
providing an application program interface (API) that provides for selectable definition of a grain size and an auxiliary memory region, the API including a function to be called by a data consuming one of the processing to gain access to a grain of data stored in the buffer memory, wherein the function is arranged to:
test whether addresses of the grain to which access has to be gained wrap around in the buffer memory,
copy the entire grain from the buffer memory to the auxiliary memory region when the addresses wrap around in the grain, so that the wrap around is eliminated in the copied grain, and
return, as a result of the call, a pointer to the consuming one of the processing indicating a location in the buffer memory from which to read the entire grain when the addresses do no wrap around inside the grain, or a pointer indicating a location of the auxiliary memory region from which to read the entire grain when the addresses wrap around in the grain;
receiving a specification of the signal processing task;
identifying a call to said function in the specification of the signal processing task; and
implementing the call using the function from the API.
9. A machine implemented method according to claim 8 wherein the API hides a distribution of processes over processing elements from the specification of the processing task, the implementation of the function being selected according to the distribution.
10. A machine implemented method according to claim 9 the method further comprising:
generating integrated circuit manufacturing control information for implementing the machine implementation; and
manufacturing an integrated circuit under control of the integrated circuit manufacturing control information.
11. A computer readable medium comprising an application program interface (API) for accessing a circular FIFO buffer, the API providing for selection of a grain size for an application program that accesses data from the circular FIFO buffer and for definition of an auxiliary memory region for the application program, the API comprising a function to be called from the application program for gaining access to a region from the circular FIFO buffer, the function being arranged to:
test whether addresses of data within the grain to which access has to be gained wrap around in the circular FIFO buffer;
copy the entire grain from the FIFO buffer to an auxiliary memory region when the addresses wrap around inside the grain, so that the wrap around is eliminated in the copied grain; and
return, as a result of the call, a pointer indicating a location in the FIFO buffer from which to read the entire grain when the addresses do not wrap around inside the grain, or a pointer indicating a location of the auxiliary memory region from which to read the entire grain when the addresses wrap around inside the grain.
12. a computer readable medium containing a program of instructions for executing the method of claim 7 .