摘要:
A method and apparatus is provided for testing the logic functionality and electrical continuity of a ring oscillator comprising an odd number of inverters connected to form a closed loop. In the method and apparatus, a known value is forced through the ring oscillator, to test the complete circuit path thereof. Thus, a low overhead deterministic test of the functionality of the ring oscillator is provided. In a useful embodiment of the invention, a method is provided for testing functionality and electrical continuity in a ring oscillator, wherein a first test device is inserted between the input of a first inverter and the output of an adjacent second inverter. The first test device is then operated to apply first and second test bits as input test signals to the first inverter input. The embodiment further comprises detecting the response to the applied first and second test bit signals at the output of the second inverter, and using the detected responses in providing an evaluation of functionality of the ring oscillator.
摘要:
An apparatus for controlling rounding modes in a single instruction multiple data (SIMD) floating-point unit is disclosed. The SIMD floating-point unit includes a floating-point status-and-control register (FPSCR) having a first rounding mode bit field and a second rounding mode bit field. The SIMD floating-point unit also includes means for generating a first slice and a second slice. During a floating-point operation, the SIMD floating-point unit concurrently performs a first rounding operation on the first slice and a second rounding operation on the second slice according to a bit in the first rounding mode bit field and a bit in the second rounding mode bit field within the FPSCR, respectively.
摘要:
The present invention provides for controlling the power consumption of an element. A first power control command is issued by software for the element. It is determined if the power control command corresponds to an allowable power control state for that element as defined by the hardware. If the power control command is not an allowable power control state for that element, the hardware sets the power control at a higher level than the power control state issued by the software. The software is real time software, and the software also sets minimally acceptable activity control states. A hierarchy of power consumption is defined for different elements of a chip by software, which provides the minimum level of power consumption by any element or sub-element on a chip.
摘要:
A system for using a plurality of heterogeneous processors in a common computer system is presented. Each processor type in the heterogeneous group handles a particular instruction set. The processors share a common memory using a common bus. In one embodiment, one of the processor types accesses the memory using DMA instructions. In another embodiment, a cache for each type of processor is stored in the common memory pool. In one embodiment, one or more PowerPC processors shares a memory with one or more Synergistic Processing Complex (SPC). A common table is used to track and maintain memory for the various processors.
摘要:
Disclosed is a coherent cache system that operates in conjunction with non-homogeneous processing units. A set of processing units of a first configuration has conventional cache and directly accesses common or shared system physical and virtual address memory through the use of a conventional MMU (Memory Management Unit). Additional processors of a different configuration and/or other devices that need to access system memory are configured to store accessed data in compatible caches. Each of the caches is compatible with a given protocol coherent memory management bus interspersed between the caches and the system memory.
摘要:
The present invention provides for authenticating code and/or data and providing a protected environment for execution. The present invention provides for dynamically partitioning and un-partitioning a local store for the authentication of code or data. The local store is partitioned into an isolated and non-isolated section. Code or data is loaded into the isolated section. The code or data is authenticated in the isolated section of the local store. After authentication, the code is executed. After execution, the memory within the isolated region of the attached processor unit is erased, and the attached processor unit de-partitions the isolated section within the local store.
摘要:
A system and method for maintaining circuit delay characteristics during power management mode. The method for maintaining circuit delay characteristics during power management mode continually toggles the clock distribution circuits at a frequency sufficiently low that it does not significantly impact chip power dissipation. The clock frequency used to toggle the clock distribution circuits is high enough to minimize the asymmetrical stress on the clock buffer transistors so that both P and N device characteristics equally change over time.
摘要:
A circuit topology which can be used to create an array of individually tuned oscillators operating at different frequencies determined by common control inputs and an easily managed variation in design dimensions of several components is provided. An array of oscillators are provided arranged in columns and rows. Each oscillator in a column is unique from the other oscillators in the column based on number of stages in the oscillator and fanout so that each oscillator will operate at a unique frequency. Oscillators of different columns within the array may differ by a common setting of the selects to these oscillators and the physical ordering of the oscillators in the column to further reduce the possibility of injection locking. A base delay cell provides selects to each column of oscillators such that each column may be programmed to operate at a different frequency from its neighbors.
摘要:
Systems and methods for distributing thread instructions in the pipeline of a multi-threading digital processor are disclosed. More particularly, hardware and software are disclosed for successively selecting threads in an ordered sequence for execution in the processor pipeline. If a thread to be selected cannot execute, then a complementary thread is selected for execution.
摘要:
A System and method for hiding memory latency in a multi-thread environment is presented. Branch Indirect and Set Link (BISL) and/or Branch Indirect and Set Link if External Data (BISLED) instructions are placed in thread code during compilation at instances that correspond to a prolonged instruction. A prolonged instruction is an instruction that instigates latency in a computer system, such as a DMA instruction. When a first thread encounters a BISL or a BISLED instruction, the first thread passes control to a second thread while the first thread's prolonged instruction executes. In turn, the computer system masks the latency of the first thread's prolonged instruction. The system can be optimized based on the memory latency by creating more threads and further dividing a register pool amongst the threads to further hide memory latency in operations that are highly memory bound.