Sun Microsystems, Inc.

Advanced Micro Devices Corporation

Nexgen Microsystems

Nexgen, Inc.

Dale Robert Greenley

John Gregory Favor

Korbin S Van Dyke

Robert Alan Cargnoni

David R Stiles

Harold L McFarland

Shrenik M Mehta

Leslie Dean Kohn

Bruce E Petrick

Greg Williams

Ming Yeh

Marc R Tremblay

Gary S Goldman

Arthur T Leung

Chitresh Chandra Narasimhaiah

Lakshminarasimhan Varadadesikan

Le Tu Quach

Ken Okin

Senthilkumar Diraviam

A method and apparatus for stopping the internal clock of a microprocessor synchronously with the execution of an instruction is provided. A stop instruction is placed in a sequence of instructions to be executed by the microprocessor. The execution of the stop execution may store a stop value into a stop register of the microprocessor. Clock stop logic detects when the stop value has been stored into the stop register. The clock stop logic instructs a clock generation component, of the microprocessor, to cease generation of an internal clock signal, thereby preventing the microprocessor from changing state. As further instructions are not executed by the microprocessor, the state of the microprocessor reflects the execution of the instruction immediately prior to the stop instruction. The processing state of the microprocessor may be obtained for use in debugging the design of the microprocessor or the instructions executed thereby.

Mechanism to stop instruction execution at a microprocessor

Hold time methods, systems, and computer program products are implemented to ensure that storage elements in a circuit have sufficient hold times without detrimentally affecting cycle times for other paths in the circuit. Electric circuits are designed by determining which storage elements have hold-time deficiencies, and by inserting an appropriate time delay element in a selected path preceding the storage element, or at a source or destination storage element, without exceeding a predetermined cycle time in a second path that overlaps the first path, for at least one storage element in an electrical circuit. The invention simulates insertion of the time delay element before or after a logic element that precedes a storage element that has a hold-time deficiency. The invention determines which storage elements and which circuit paths between adjacent storage elements are subject to hold time deficiencies, assembles a hold time deficiency list, and determines whether a selected hold-time-deficient storage element can be preceded with a delay element without violating a cycle time constraint in another cycle path. The invention further includes determining the placement of a delay element in a storage system having a network of storage elements interconnected with logic elements by using a criterion that protects cycle times in other storage element paths with slower signal propagation. The method further includes modifying a net list of storage and logic element interconnections to include insertion of one or more time delay elements.

Automatic delay element insertion system for addressing holdtime problems

An integrated circuit having a normal mode for operating under normal operating conditions and a debug mode for operating to test and debug the integrated circuit. The integrated circuit includes a plurality of output pins that carry a first plurality of signals in the normal mode and carry a second plurality of signals in the debug mode. In one embodiment, the integrated circuit embodies a microprocessor. The microprocessor may include logic circuitry for enabling the second plurality of signals to be output from a multiplexer to the output pins in response to a predetermined event, such as a hit in an associated memory unit.

Method and apparatus for debugging an integrated circuit

A pipeline control system for implementing a virtual architecture having a complex instruction set is distributed over RISC-like semi-autonomous functional units in a processor. Decoder logic fetches instructions of the target architecture and translates them into simpler RISC-like operations. These operations, each having an associated tag, are issued to the functional units. Address processing unit computes addresses of the instructions and operands, performs segment relocation, and manages the processor's memory. Operations are executed by the units in a manner that is generally independent of operation processing by the other units. The units report termination information back to the decoder logic, but do not irrevocably change the state of the machine. Based on the termination information, the decoder logic retires normally terminated operations in order. Thus, the functional units enable multiple operations to be executed in a speculative and out-of-order manner to fully utilize the resources of the processor.

Method and apparatus for executing string instructions

A synchronizing circuit receives an external signal and yields an output that is synchronized with the system clock and operates at the frequency of the external signal. The signal output from the synchronizing circuit is fed into the clock-enable input of the storage element, and the system clock signal is fed into the clock input of the storage element. Because the clock-enable signal triggers the storage element, the storage element is driven at the external signal frequency. Clock skew is eliminated because the system clock used for the clock input to the storage element is skew-controlled.

Methods and apparatus for synchronizing asynchronous test structures and eliminating clock skew considerations

A separate cacheable-in-virtual-cache attribute bit (CV) is maintained for each page of memory in the translation table maintained by the operating system. The CV bit indicates whether the memory addresses on the page to which the translation table entry refers are cacheable in virtually indexed caches. According to a first embodiment, when there are two or more aliases which are not offset by multiples of the virtual cache size, all of the aliases are made non-cacheable in virtually indexed caches by deasserting the CV bits for all aliases. With regards to the contents of the translation lookaside buffer (TLB), the translations for all aliases may simultaneously coexist in the TLB because no software intervention is required to insure data coherency between the aliases. According to second and third embodiments of the present invention, when there are two or more aliases which are not offset by multiples of the virtual cache size, only one of those aliases may remain cacheable in virtual caches. For the other aliases, the CV bits for the translation pages containing those aliases are deasserted. The operating system has the responsibility of flushing data from the virtually indexed internal cache before deasserting the CV attribute for a page. The second embodiment allows the newer mapping to a physical address to remain in the first-level cache, while the third embodiment allows the older alias to remain in the first-level cache when a newer alias is mapped.

Cachability attributes of virtual addresses for optimizing performance

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 method and apparatus for dynamically switching the relative priorities of the load buffer and store buffer with respect to external memory resources in a superscalar processor. According to a first embodiment, a protocol dictates that the load buffer always prevails until the store buffer reaches a certain 'high water mark,' (an upper threshold) at which time the store buffer gains priority. After the store buffer has gained priority, it continues to access the memory until it is depleted to a 'low water mark,' (a lower threshold) at which time the load buffer regains priority. Whenever the store buffer reaches the high water mark, it gains priority until it drains down to the low water mark. This reduces the tendency for the store buffer to become full and block the processor. According to a second embodiment, the load buffer prevails if it is above its high water mark. If the load buffer is below its high water mark, the load buffer prevails until the store buffer reaches its high water mark, at which time the store buffer gains priority. After the store buffer has gained priority, it maintains priority until the load buffer reaches its high water mark. This reduces the tendency for either buffer to become full and block the processor.

Dynamic priority switching of load and store buffers in superscalar

A pipeline control system is distributed over the functional units (15, 17, 20, 25) in a processor (10). Decoder logic (12) issues operations, each with an associated tag, to the functional units, with up to n operations allowed to be outstanding. The units execute the operations and report termination information back to the decoder logic, but do not irrevocably change the state of the machine. Based on the termination information, the decoder logic retires normally terminated operations in order. If an operation terminates abnormally, the decoder logic instructs the units to back out of those operations that include and are later than the operation that terminated abnormally.

Computer processor with distributed pipeline control that allows

A dual-ported tag array of a cache allows simultaneous access of the tag array by miss data of older LOAD instructions being returned during the same cycle that a new LOAD instruction is accessing the tag array to check for a cache hit. Because a load buffer queues LOAD instructions, the cache tags for older LOAD instructions which missed the cache return later when new LOAD instructions are accessing a tag array to check for cache hits. A method and apparatus for calculating and maintaining a hit bit in a load buffer perform the determination of whether or not a newly dispatched LOAD will hit the cache after it has been queued into the load buffer and waited for all older LOADs to be processed. A load buffer data entry includes the hit bit and all information necessary to process the LOAD instruction and calculate the hit bits for future LOAD instructions which must be buffered. A method and apparatus for servicing LOAD instructions, in which the access of the data array portion of a cache and the tag array portion are decoupled, allows the delayed access of the data array after a LOAD has been delayed in the load buffer without reaccessing the tag array.

Hit bit for indicating whether load buffer entries will hit a cache when

Semi-autonomous RISC pipelines for overlapped execution of RISC-like

A pipeline control system for implementing a virtual architecture having complex instruction set is distributed over RISC-like semi-autonomous functional units in a processor. Decoder logic fetches instructions of the target architecture and translates them into simpler RISC-like operations. These operations, each having an associated tag, are issued to the functional units. Address processing unit computes addresses of the instructions and operands, performs segment relocation, and manages the processor's memory. Operations are executed by the units in a manner that is generally independent of operation processing by the other units. The units report termination information back to the decoder logic, but do not irrevocably change the state of the machine. Based on the termination information, the decoder logic retires normally terminated operations in order. Thus, the functional units enable multiple operations to be executed in a speculative and out-of-order manner to fully utilize the resources of the processor.

Semi-Autonomous RISC pipelines for overlapped execution of RISC-like

A method and apparatus for improving the performance of pipelined computer processors having fast clock speeds and processing cycles for performing signed and unsigned load operations to memory. A load instruction includes signed and unsigned variants. Signed load operations include accessing signed load data from memory, aligning and sign extending the value of the accessed data. The present invention provides two different schemes for scheduling the processor's pipelines to handle accessing signed and unsigned data from memory. The two schemes include a data dependent scheme that accesses data from memory where the sign of a load data is not known until the data has been accessed from memory, and an opcode dependent scheme when the sign value of data being accessed is known prior to accessing memory. In the data dependent scheme, the processor is scheduled to process a signed LOAD in two cycles if the load data is negative, and one cycle if the data is positive. In the opcode dependent scheme, signed load instructions are scheduled to execute in the processor's pipeline�s! in two cycles regardless of whether the data fetched from memory is negative or positive. The present invention further includes a delay return scheme in which the processor is able to process multiple consecutive load instructions by transitioning into a scheduling mode of handling all loads as signed.

Method and apparatus for optimizing signed and unsigned load processing

A dual-ported tag array of a cache allows simultaneous access of the tag array by miss data of older LOAD instructions being returned during the same cycle that a new LOAD instruction is accessing the tag array to check for a cache hit. Because a load buffer queues LOAD instructions, the cache tags for older LOAD instructions which missed the cache return later when new LOAD instructions are accessing a tag array to check for cache hits. A method and apparatus for calculating and maintaining a hit bit in a load buffer perform the determination of whether or not a newly dispatched LOAD will hit the cache after it has been queued into the load buffer and waited for all older LOADs to be processed. A load buffer data entry includes the hit bit and all information necessary to process the LOAD instruction and calculate the hit bits for future LOAD instructions which must be buffered. A method and apparatus for servicing LOAD instructions, in which the access of the data array portion of a cache and the tag array portion are decoupled, allows the delayed access of the data array after a LOAD has been delayed in the load buffer without reaccessing the tag array. A method and apparatus allow access to the first level cache and second level cache to occur simultaneously for two separate LOADs in the load buffer.

Methods and apparatuses for servicing load instructions

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

Los Gatos, CA, United States of America

Dale Robert Greenley