Abstract: According to one embodiment, a method is disclosed. The method includes detecting a load miss at a central processing unit (CPU), stalling a read only buffer (ROB), speculatively retiring an instruction causing the ROB stall and subsequent instructions, keeping registers that have not been renamed in the ROB upon retirement, and flushing the CPU pipeline upon receiving data from the load miss.

Abstract: A clock frequency control unit for an integrated circuit (IC) includes a clock generator, a finite state machine (FSM), and a gating circuit (GC). The FSM has at least first and second states corresponding to non-low workload low workload states, respectively. In the first state, the GC provides a clock signal to functional units of the IC with the same frequency as the clock generator output. In the second state, the GC reduces the frequency of the clock signal. In one embodiment, the GC masks out selected cycles of the clock generator output to reduce the clock signal frequency. The FSM monitors the operation of the IC to transition from the first state to the second state when selected “low workload” conditions are detected (e.g., long latency cache miss). Similarly, the FSM transitions from the second state to the first state when selected “non-low workload” conditions are detected.

Abstract: A clock frequency control unit for an integrated circuit (IC) includes a clock generator, a finite state machine (FSM), and a gating circuit (GC). The FSM has at least first and second states corresponding to non-low workload low workload states, respectively. In the first state, the GC provides a clock signal to functional units of the IC with the same frequency as the clock generator output. In the second state, the GC reduces the frequency of the clock signal. In one embodiment, the GC masks out selected cycles of the clock generator output to reduce the clock signal frequency. The FSM monitors the operation of the IC to transition from the first state to the second state when selected “low workload” conditions are detected (e.g., long latency cache miss). Similarly, the FSM transitions from the second state to the first state when selected “non-low workload” conditions are detected.

Abstract: A method and apparatus for changing the configuration of a multi-core processor is disclosed. In one embodiment, a throttle module (or throttle logic) may determine the amount of parallelism present in the currently-executing program, and change the execution of the threads of that program on the various cores. If the amount of parallelism is high, then the processor may be configured to run a larger amount of threads on cores configured to consume less power. If the amount of parallelism is low, then the processor may be configured to run a smaller amount of threads on cores configured for greater scalar performance.

Abstract: A memory system is provided for storing multiple data types. The memory system includes a main memory, a local cache, and a translation unit. The local cache has multiple entries, each of which includes a data field to store data and a status field to indicate a storage state for the stored data. The translation unit includes a translation lookaside buffer (TLB) and a status-cache (STC). The TLB stores address translations for data in the main memory, and the STC stores storage states for data indicated by the address translations.

Abstract: A method and apparatus for resuming operations from a low latency wake-up low power state. One embodiment provides a system including a processor, an operating system, and a memory subsystem that requires initialization commands to exit a memory low power state. Control logic detects exit from an operating system low latency low power state and responsively generates a plurality of initialization commands to remove the memory subsystem from the memory low power state prior to deasserting a stop clock signal and allowing execution to resume.

Abstract: Apparatus, system and methods are provided for performing speculative data prefetching in a chip multiprocessor (CMP). Data is prefetched by a helper thread that runs on one core of the CMP while a main program runs concurrently on another core of the CMP. Data prefetched by the helper thread is provided to the helper core. For one embodiment, the data prefetched by the helper thread is pushed to the main core. It may or may not be provided to the helper core as well. A push of prefetched data to the main core may occur during a broadcast of the data to all cores of an affinity group. For at least one other embodiment, the data prefetched by a helper thread is provided, upon request from the main core, to the main core from the helper core's local cache.

Abstract: The present invention provides a mechanism for implementing z-compression in a manner that is transparent to the user. Blocks of z-data are associated with storage locations in a z-data buffer in cleared, compressed, or uncompressed data states. Operations to the z-data buffer are monitored for selected operations. These operations may include clear or lock operations. If a selected operation is detected, a modified version of the selected operation is implemented to mask differences between the storage states of the data blocks.

Abstract: A clock frequency control unit for an integrated circuit (IC) includes a clock generator, a finite state machine (FSM), and a gating circuit (GC). The FSM has at least first and second states corresponding to non-low workload low workload states, respectively. In the first state, the GC provides a clock signal to functional units of the IC with the same frequency as the clock generator output. In the second state, the GC reduces the frequency of the clock signal. In one embodiment, the GC masks out selected cycles of the clock generator output to reduce the clock signal frequency. The FSM monitors the operation of the IC to transition from the first state to the second state when selected “low workload” conditions are detected (e.g., long latency cache miss). Similarly, the FSM transitions from the second state to the first state when selected “non-low workload” conditions are detected.

Abstract: A method and apparatus for executing low power validations for high confidence predictions. More particularly, the present invention pertains to using confidence levels of speculative executions to decrease power consumption of a processor without affecting its performance. Non-critical instructions, or those instructions whose prediction, rather than verification, lie on the critical path, can thus be optimized to consume less power.

Abstract: A memory system is provided for storing multiple data types. The memory system includes a main memory, a local cache, and a translation unit. The local cache has multiple entries, each of which includes a data field to store data and a status field to indicate a storage state for the stored data. The translation unit includes a translation lookaside buffer (TLB) and a status-cache (STC). The TLB stores address translations for data in the main memory, and the STC stores storage states for data indicated by the address translations.

Abstract: A method and apparatus for resuming operations from a low latency wake-up low power state. One embodiment provides a system including a processor, an operating system, and a memory subsystem that requires initialization commands to exit a memory low power state. Control logic detects exit from an operating system low latency low power state and responsively generates a plurality of initialization commands to remove the memory subsystem from the memory low power state prior to deasserting a stop clock signal and allowing execution to resume.

Abstract: A graphics system is provided to implement compression of depth or z-data. The graphic system includes a buffer, a status table, and a read/write unit. The buffer stores depth data for multiple blocks of pixels in associated buffer entries. The status table stores status values for the entries of the buffer. The status value for a given entry indicates an access mode for the corresponding depth data according to whether the data is compressed, uncompressed or in a reference state. The read/write unit implements data accesses for a given entry responsive to the status value associated with the entry.

Abstract: A memory system is provided for storing multiple data types. The memory system includes a main memory, a local cache, and a translation unit. The local cache has multiple entries, each of which includes a data field to store data and a status field to indicate a storage state for the stored data. The translation unit includes a translation lookaside buffer (TLB) and a status-cache (STC). The TLB stores address translations for data in the main memory, and the STC stores storage states for data indicated by the address translations.

Abstract: A system and corresponding method use a PAUSE instruction as a low power hint in a single threaded or multithreaded environment using “processor slow mode.” One embodiment actually lowers the frequency of the processor clock. Another embodiment virtually lowers the frequency of the processor clock by gating M clock cycles out of every N clock cycles. When all threads have issued a PAUSE instruction, the processor enters slow mode and remains there for a while. After this while, the processor returns to normal mode. Alternatively, an event, such as an interrupt or an exception, can cause the processor to return to normal mode from slow mode.

Abstract: A method and apparatus for executing different sets of instructions that cause a processor to perform different data type operations on different physical registers files that logically appear to software as a single aliased register file. According to one aspect of the invention, a processor is provided that includes at least two physical register files--one for executing scalar data type operations and the other for executing packed data type operations. In addition, the processor includes a transition unit that is configured to cause the two physical register files to logically appear to software executing on the processor as a single logical register file.

Abstract: The invention provides a method and apparatus for providing operand reads in a processor pipeline. According to one aspect of the invention, a method is described for executing an instruction in a computer pipeline that requires different operands be read from the same register file in different stages of the computer pipeline. According to another aspect of the invention, a method is described for executing an instruction in a processor pipeline. According to this method, at least a first operand is read from a register file in a first stage of the processor pipeline. If execution of the instruction causes the processor to place the first operand in a storage area other than the register file, then the first operand in written to that storage area in a subsequent stage of the processor pipeline. Otherwise, one or more ALU operations are performed on the first operand and at least a second operand in a different subsequent stage of the processor pipeline.

Abstract: The specification discloses a method and apparatus for determining the length of variable-length instructions that appear sequentially in an instruction stream without differentiation. The apparatus may be used to facilitate parallel processing of such variable-length instructions by a computer system.