Abstract:
Techniques are disclosed relating to configuring an interlock memory system. In one embodiment, a method includes determining a sequence of memory access requests for a program and generating information specifying memory access constraints based on the sequence of memory accesses, where the information is usable to avoid memory access hazards for the sequence of memory accesses. In this embodiment, the method further includes configuring first circuitry using the information, where the first circuitry is included in or coupled to a memory. In this embodiment, after the configuring, the first circuitry is operable to perform memory access requests to the memory corresponding to the sequence of memory accesses while avoiding the memory access hazards, without receiving other information indicating the memory access hazards.
Abstract:
Techniques are disclosed relating to self-addressing memory. In one embodiment, an apparatus includes a memory and addressing circuitry coupled to or comprised in the memory. In this embodiment, the addressing circuitry is configured to receive memory access requests corresponding to a specified sequence of memory accesses. In this embodiment, the memory access requests do not include address information. In this embodiment, the addressing circuitry is further configured to assign addresses to the memory access requests for the specified sequence of memory accesses. In some embodiments, the apparatus is configured to perform the memory access requests using the assigned addresses.
Abstract:
Techniques are disclosed relating to resolving memory access hazards. In one embodiment, an apparatus includes a memory and circuitry coupled to or comprised in the memory. In this embodiment, the circuitry is configured to receive a sequence of memory access requests for the memory, where the sequence of memory access requests is configured to access locations associated with entries in a matrix. In this embodiment, the circuitry is configured with memory access constraints for the sequence of memory access requests. In this embodiment, the circuitry is configured to grant the sequence of memory access requests subject to the memory access constraints, thereby avoiding memory access hazards for a sequence of memory accesses corresponding to the sequence of memory access requests.
Abstract:
Techniques relating to LDPC encoding. A set of operations is produced that is usable to generate an encoded message based on an input message. The set of operations corresponds to operations for entries in a smaller matrix representation that specifies locations of non-zero entries in an LDPC encoding matrix. A mobile device is configured with the set of operations to perform LDPC encoding. Circuitry configured with the set of operations performs LDPC encoding with high performance, relatively small area and/or low power consumption.
Abstract:
Techniques are disclosed relating to implementation of LDPC encoding circuitry on a single integrated circuit (IC). In some embodiments, circuitry on a single IC includes message circuitry configured to receive or generate a message to be encoded, encode circuitry configured to perform low density parity check (LDPC) encoding on the message, noise circuitry configured to apply noise to the encoded message, and decode circuitry configured to perform LDPC decoding of the message. In some embodiments, the disclosed techniques may reduce production costs (e.g., by reducing overall chip area), facilitate LDPC testing, and/or provide multiple different functions relating to message transmission on a single chip.
Abstract:
System and method for performing correlation analysis. A program that includes multiple program structures and one or more data objects is stored. Each data object is shared by at least two of the program structures. For each program structure, decomposition effects on each of the data objects shared by the program structure resulting from each of a respective one or more optimizing transforms applied to the program structure are analyzed. One or more groups of correlated structures are determined based on the analyzing. Each group includes two or more program structures that share at least one data object, and at least one optimizing transform that is compatible with respect to the two or more program structures and the shared data object. For at least one group, the at least one optimizing transform is usable to transform the two or more program structures to meet a specified optimization objective.
Abstract:
Techniques are disclosed relating to reordering sequences of memory accesses. In one embodiment, a method includes storing a specified sequence of memory accesses that corresponds to a function to be performed. In this embodiment, the specified sequence of memory accesses has first memory access constraints. In this embodiment, the method further includes reordering the specified sequence of memory accesses to create a reordered sequence of memory accesses that has second, different memory access constraints. In this embodiment, the reordered sequence of memory accesses is usable to access a memory to perform the function. In some embodiments, performance estimates are determined for a plurality of reordered sequences of memory accesses, and one of the reordered sequences is selected based on the performance estimates. In some embodiments, the reordered sequence is used to compile a program usable to perform the function.
Abstract:
Techniques are disclosed relating to self-addressing memory. In one embodiment, an apparatus includes a memory and addressing circuitry coupled to or comprised in the memory. In this embodiment, the addressing circuitry is configured to receive memory access requests corresponding to a specified sequence of memory accesses. In this embodiment, the memory access requests do not include address information. In this embodiment, the addressing circuitry is further configured to assign addresses to the memory access requests for the specified sequence of memory accesses. In some embodiments, the apparatus is configured to perform the memory access requests using the assigned addresses.
Abstract:
Techniques are disclosed relating to encoding communications using low-density parity check codes, which may be based on an LDPC encoding matrix.
Abstract:
System and method for compiling a program, including determining one or more program structures containing one or more variables at the entry and exit of each program structure, wherein each variable specifies a value transfer operation of one or more source variables to a destination variable between outside the program structure and inside the program structure. A subset of the destination variables may be determined for which assigning the destination variable to a memory resource of a corresponding source variable does not disrupt the functionality of the program. Implementation of the value transfer operations may be executable to map each of the determined subset of destination variables to a respective memory resource. The mapping may be dynamically changed, thereby transferring the value from the first source variable to the destination variable without copying the value between the memory resources.