Abstract: An instruction cache includes instruction storage, a plurality of lock indicators, and control logic. The instruction storage includes a plurality of cache blocks to store instructions. Each of the lock indicators is associated with one of the cache blocks so as to control access to the associated cache block. The control logic is configured to: set to a write disable state, on access of a given one of the cache blocks, a given one of the lock indictors associated with the given one of the cache blocks; to determine whether a given instruction is stored in the instruction storage; and to deny write access to the given one of the cache blocks that is assigned to store the given instruction based on the given one of the block indicators being set to the write disable state.

Abstract: A method and apparatus for executing program loops. A processor, includes an execution unit and an instruction fetch buffer. The execution unit is configured to execute instructions. The instruction fetch buffer is configured to store instructions for execution by the execution unit. The instruction fetch buffer includes a loop buffer configured to store instructions of an instruction loop for repeated execution by the execution unit. The loop buffer includes buffer control logic. The buffer control logic includes pointers, and is configured to predecode a loop jump instruction, identify loop start and loop end instructions using the predecoded loop jump instruction and pointers; and to control non-sequential instruction execution of the instruction loop. The width of the pointers is determined by loop buffer length and is less than a width of an address bus for fetching the instructions stored in the loop buffer from an instruction memory.

Abstract: An instruction cache includes instruction storage, a plurality of lock indicators, and control logic. The instruction storage includes a plurality of cache blocks to store instructions. Each of the lock indicators is associated with one of the cache blocks so as to control access to the associated cache block. The control logic is configured to: set to a write disable state, on access of a given one of the cache blocks, a given one of the lock indictors associated with the given one of the cache blocks; to determine whether a given instruction is stored in the instruction storage; and to deny write access to the given one of the cache blocks that is assigned to store the given instruction based on the given one of the block indicators being set to the write disable state.

Abstract: A method and apparatus for executing program loops. A processor, includes an execution unit and an instruction fetch buffer. The execution unit is configured to execute instructions. The instruction fetch buffer is configured to store instructions for execution by the execution unit. The instruction fetch buffer includes a loop buffer configured to store instructions of an instruction loop for repeated execution by the execution unit. The loop buffer includes buffer control logic. The buffer control logic includes pointers, and is configured to predecode a loop jump instruction, identify loop start and loop end instructions using the predecoded loop jump instruction and pointers; and to control non-sequential instruction execution of the instruction loop. The width of the pointers is determined by loop buffer length and is less than a width of an address bus for fetching the instructions stored in the loop buffer from an instruction memory.

Abstract: In an embodiment of the invention, a method is provided for isolating a ferroelectric memory from a power supply during a write-back cycle or a write cycle of the ferroelectric memory. After it is determined that a write-back cycle or a write cycle will occur in the ferroelectric memory, the power supply is electrically disconnected from the ferroelectric memory before a write-back cycle or a write cycle occurs. Energy during the write-back cycle or the write cycle is provided to the ferroelectric memory by one or more capacitors in this embodiment. After the write-back cycle or the write cycle has ended, the power supply is electrically connected to the ferroelectric memory and the capacitors.

Abstract: A data synchronization arrangement is provided for exchanging clocked data between different clock domains running at the same clock frequency but at an arbitrary relative phase shift. An input data stream synchronized in the first clock domain is written into respective locations of a buffer memory through a write select multiplexer under control of a write select shift register clocked by the first domain clock. An output data stream synchronized in the second clock domain is read from the respective locations of the buffer memory through a real select multiplexer under control of a read select shift register clocked by the second domain clock. A bit synchronization circuit is provided for loading the read select shift register with a bit pattern that has a relative offset relative to the bit pattern of the write select shift register, to correlate for the difference in clock phases.

Abstract: A data synchronization arrangement is provided that is fail-safe at high speed and low power consumption, for exchanging clocked data between different clock domains running in a digital processing equipment at the same clock frequency but at an arbitrary relative phase shift. A register arrangement has a predetermined number of parallel registers, each register having a data input, a write clock input, a read clock input and a data output. A write select multiplexer has an input receiving a write clock signal from a first clock domain, one clock output for each of the parallel registers and connected to a write clock input of a respective register, and one write select input for each clock output. A read select multiplexer has an input receiving a read clock signal from a second clock domain, one clock output for each of the parallel registers and connected to a read clock input of a respective register, and one read select input for each clock output.

Abstract: A data synchronization arrangement is provided that is fail-safe and allows high-speed operation. Clocked data are exchanged between different clock domains running in a digital processing equipment at the same clock frequency but at an arbitrary relative phase shift. The data synchronization arrangement comprises a buffer memory with a predetermined limited number of memory locations each of which has a data write port and a data read port. A write select multiplexer has a data input receiving an input data stream synchronized with the clock from a first clock domain, one data output for each of said memory locations and connected to a respective data write port, and one write select input for each data output. A read select multiplexer has one data input for each of the memory locations and connected to a respective data read port, one read select input for each data input, and a data output supplying an output data stream synchronized with the clock from a second clock domain.

Abstract: A data synchronization arrangement is provided that is fail-safe at high speed and low power consumption, for exchanging clocked data between different clock domains running in a digital processing equipment at the same clock frequency but at an arbitrary relative phase shift. A register arrangement has a predetermined number of parallel registers, each register having a data input, a write clock input, a read clock input and a data output. A write select multiplexer has an input receiving a write clock signal from a first clock domain, one clock output for each of the parallel registers and connected to a write clock input of a respective register, and one write select input for each clock output. A read select multiplexer has an input receiving a read clock signal from a second clock domain, one clock output for each of the parallel registers and connected to a read clock input of a respective register, and one read select input for each clock output.