Sun Microsystems, Inc.

Oracle International Corporation

Oracle America, Inc.

Bruce E Petrick

Dale Robert Greenley

Marc R Tremblay

Kevin B Normoyle

Gary S Goldman

Mukesh K Patel

Bruce J Chang

Damien Walker

Embodiments include systems and methods for providing reliable and precise sample alignment across different clock domains. Some embodiments operate in context of microprocessor power management circuits seeking correlated measurements of voltage droop (VD) and phase delay (PD). For example, a rolling code is generated for each of multiple second clock domain sample times (CDSTs). VD and the rolling code are both sampled according to a first clock domain to generate VD samples and corresponding VCode samples for each of multiple first CDSTs. PD can be sampled according to the second clock domain to generate PD samples for each of the second CDSTs, each associated with the rolling code for its second CDST. For any first CDST, the VD sample for the first CDST can be aligned with a PD sample for a coinciding second CDST by identifying matching associated rolling codes.

Alignment of samples across different clock domains

Systems and methods that allow for dynamically deconfiguring, reconfiguring and/or otherwise configuring nodes (e.g., processors) in a symmetric multiprocessing system (e.g., a symmetric multiprocessor) in a manner that avoids, or at least limits, inefficiencies such as renumbering of node IDs, system reboots, SW configuration handle changes, and the like. In one arrangement, a number of modules, tables and/or the like that are configured to generate node IDs and/or convert node IDs from one form to another form can be intelligently implemented within an SMP to allow the various processes and/or components of an SMP to utilize the node IDs in a more efficient manner. For instance, as SDs in an SMP are often configured to work with CNIDs (e.g., for use in determining at which node a particular requested cache line resides), any node GNIDs that are sent to the SD for processing can first be converted into corresponding CNIDs.

Dynamic node configuration in directory-based symmetric multiprocessing systems

A system may be employed for allowing the synchronous operation of an asynchronous system. The system may be a system that may include multiple clusters. The clusters may include asynchronous clock domains and may also receive a global clock signal through a global clock grid that may overlay the system. Furthermore, a method may be employed for synchronizing asynchronous clock domains within a cluster. The method of synchronizing may include providing a global clock that corresponds to a global clock grid to each cluster. Additionally, the method of synchronizing may include accounting for the mismatch between the asynchronous clock domains by employing logic in a block.

Synchronous operation of a system with asynchronous clock domains

Methods and apparatuses are presented that allow power savings on a processor executing a plurality of threads on a plurality of cores. The method may include providing a first timing signal to a processor, determining the power requirements of the processor, loading a symbol into a shift register, where the symbol may be associated with the power requirements of the processor, providing a second timing signal to the processor, where the second timing signal may include a gated representation of the first timing signal and the symbol.

Method and apparatus for power savings in a multi-threaded processor using a symbol gated with a clock signal

A method of dynamically switching mapping schemes for cache includes a microprocessor, a first mapping scheme, a second mapping scheme and switching circuitry for switching between the first mapping scheme and the second mapping scheme. The microprocessor is in communication with the cache through the switching circuitry and stores information within the cache using one of the first mapping scheme and second mapping scheme. Also, monitoring circuitry for determining whether one of instructions and load/store operations is using the cache is included. Further, the switching circuitry switches between the first mapping scheme and the second mapping scheme based on which one of instructions and load/store operations is using the cache.

Method and apparatus for dynamically switching a cache between direct-mapped and 4-way set associativity

A coprocessor coupled to a hardware processor and capable of performing multimedia operations is provided. The coprocessor includes an instruction fetch and decode unit which is coupled to a plurality of execution units including an integer execution unit and a multimedia execution unit. The coprocessor includes a superscalar architecture and each of the execution units includes a plurality of pipelined stages. Accordingly, the multimedia execution unit has several integer execution units which can be executed in parallel for improved multimedia performance. A visible register set is coupled to the integer execution unit for receiving operands to initialize operation of the coprocessor. Further, a first register file is coupled to the multimedia execution unit and a second register file is coupled to the integer execution unit. A memory bus coupled to memory and the integer execution unit is used for accessing data and multimedia applications in memory as indicated by values in the visible register set.

Processor coupled by visible register set to modular coprocessor including integrated multimedia unit

Embodiments of the present invention provide a stack renaming method and apparatus for stack based processors. Using principles of the present invention, a stack can be accessed simultaneously by one or more functional units in a stack processor. The stack apparatus includes a stack renaming unit capable of renaming a logical stack address to a real stack address. Each logical stack address corresponds to a storage element in the stack renaming unit which stores a real stack address. A circular counter is used in the stack renaming unit to sequentially cycle through each of the logical stack addresses. The real stack addresses corresponding to each of the logical stack addresses can be stored out of order in the stack renaming unit. A stack control unit is coupled to the stack renaming unit and provides one or more control signals to the stack renaming unit and coordinates the operation of the stack renaming unit within the stack apparatus. A stack memory circuit coupled to the stack renaming unit receives the real stack address and generates a data value to be used in a functional unit. The real address is provided to a functional unit by a real address output port coupled to the stack renaming unit. As long as there are no data dependencies or structural dependences, a functional unit is able to operate on the data value associated with the real address while other functional units may be accessing and using other stack values.

System for simultaneously accessing one or more stack elements by

An apparatus for processing a write miss signal from a copy-back data cache includes a load-store unit with an allocating load buffer, a non-allocating store buffer, and a priority control circuit to generate write-after-read hazards and read-after-write hazards to preserve the processing priority of entries within the allocating load buffer and the non-allocating store buffer. A prefetch circuit enqueues a prefetch command in the allocating load buffer and a store command in the non-allocating store buffer upon a write miss to the copy-back data cache. Thus, the priority control circuit forces a write-after-read hazard on the store command in the non-allocating store buffer. As a result, the prefetch command in the allocating load buffer secures an allocated line in the copy-back data cache, allowing the store command of the non-allocating store buffer to write data to the allocated line.

Apparatus and method for write miss processing in a copy-back data cache

A computer processor complex including a hardware processor coupled to a multimedia coprocessor is provided. This computer processor complex is capable of separately processing a stream of non-multimedia instructions in addition to a stream of multimedia instructions such as are used in MPEG audio and video. The computer processor complex includes a visible register set including registers for a program counter and a data pointer. The program counter is used to hold the address in memory where the multimedia instructions are located and the data pointer indicates where the data, corresponding to these multimedia instructions, is located in memory. A hardware processor is coupled to a first bidirectional port on the visible register set and a multimedia coprocessor is coupled to a second bidirectional port on the visible register set. The bidirectional ports allow the hardware processor and the coprocessor to exchange data and status information typically using an interrupt based communication mechanism. A main memory device is also coupled to the hardware processor over bidirectional port and coupled to the multimedia processor over a second bidirectional port. This arrangement allows the hardware processor and the coprocessor to share main memory and load separate instruction streams from main memory.

Processor complex for executing multimedia functions

An innovative method and system of performing multiway branch operations on a microprocessor architecture which supports single instruction multiple data (SIMD) operations is provided. A computer processor includes a branch condition register, a graphic status register, a displacement register, a branch offset register, a program counter register and circuit logic responsive to a multiway branch opcode. Bitwise AND logic coupled to the branch condition register and the graphic status register performs a bitwise logical AND between a mask contained in the branch condition register and multiple comparison results contained in the graphic status register. An output port from bitwise logical AND is coupled to a constant array and selects a set of constant values based on the bitwise logical AND result value. A shifter logic coupled to the branch offset register and the displacement register bitwise left shifts the displacement value stored in the displacement register a predetermined amount based upon the value in the branch offset. The shifter logic is also coupled to receive a constant value from the constant array which is added to the shifted result by concantenating the constant value to the lower order bits shifted into the shifter logic. An adder circuit coupled to the shifter logic and a program counter register adds the results to the program counter value to generate a relative branch address. Finally, a branch logic coupled to the adder circuit responds to the multiway branch opcode and sets the program counter to the relative branch address provided by the adder circuit.

System and method for performing multiway branches using a visual

A 'slow' signal is not sent across chip to be combined with combinatorial logic, but rather, the logic with which it would be combined is partitioned such that there are two outputs, one if the 'slow' signal would be true and a second if the 'slow' signal would be false. Both of these outputs are then provided to a multiplexer. The original 'slow' signal selects the correct signal, thus saving the interconnect time delay. The concepts also apply to combinations of multiple 'slow' signals.

Logic speed-up by selecting true/false combinations with the slowest

A central processing unit is connected to an external memory including system memory and an external cache. The central processing unit includes a First-In-First-Out (FIFO) load buffer configured to generate an access to the external memory in response to a data prefetch command. The access to external memory has an associated data load latency period as data is moved from the system memory into the external cache. Instead of requiring the access to external memory to be completed before another FIFO load buffer address is processed, as is typically required in a FIFO load buffer configuration, the FIFO load buffer responds to the data prefetch command by processing additional stored addresses during the data load latency period.

Growing community of inventors

Sunnyvale, CA, United States of America

Bruce E Petrick