International Business Machines Corporation

James H Pomerene

Rudolph N Rechtschaffen

Frank J Sparacio

Philip George Emma

Joshua W Knight

Thomas R Puzak

Charles Franklin Webb

David Meltzer

Kevin Patrick McAuliffe

Frederick J Aichelmann, Jr

Kimming K So

Keith N Langston

Howard E Sachar

Philip L Rosenfeld

Gururaj S Rao

Rex H Blumberg

A method for addressing data in a cache unit which has a plurality of congruence classes, following a failure which disables one or more of the congruence classes in the cache unit. A plurality of synonym classes are established. A subset of the congruence classes is assigned to each of the synonym classes. Any disabled congruence classes are identified. The synonym class to which the disabled congruence class belongs is identified. An alternate congruence class is selected which belongs to the same synonym class as the disabled congruence class. When a request is received by the cache to store a line of data into the disabled congruence class, the line is stored into the alternate congruence class in response to the request.

Cache remapping using synonym classes

System and method for predicting a multiplicity of future branches simultaneously (parallel) from an executing program, to enable the simultaneous fetching of multiple disjoint program segments. Additionally, the present invention detects divergence of incorrect branch predictions and provides correction for such divergence without penalty. By predicting an entire sequence of branches in parallel, the present invention removes restrictions that decoding of multiple instructions in a superscalar environment must be limited to a single branch group. As a result, the speed of today's superscalar processors can be significantly increased. The present invention includes three main embodiments: (1) the first embodiment is directed to a simplex multibranch prediction device, that can predict a plurality of branch groups in one cycle and provide early detection of wrong predictions; (2) the second embodiments is directed to a duplex multibranch prediction device that can detect divergence in a predicted stream, and provide redirection (correction) within the stream; and (3) the third embodiment is directed to an n-plex multibranch prediction device, that can predict n multiplicity of branch predictions simultaneously and provide an early detection of wrong predictions as well as correction of wrong predictions.

Simultaneous prediction of multiple branches for superscalar processing

A digital computer includes a main and an auxiliary pipeline processor which are configured to concurrently execute contiguous groups of instructions taken from a single instruction sequence. The instructions in a sequence may be divided into groups by using either taken-branch instructions or certain instructions which may change the contents of the general purpose registers as group delimiters. Both methods of grouping the instructions use a branch history table to predict the sequence in which the instructions will be executed.

Multiple sequence processor system

A system is provided for management of data in cache memories in a multiprocessor environment which allows portions of lines to be valid and exclusive, while other portions are valid, but not exclusive, or invalid. A processor may store into portions of a line under its exclusive control without invalidating copies of the line held in the cache memories of the other processors. The system includes at least two processors, a shared main memory and a system control element, and each processor has a corresponding cache memory, a modified line stack and a sectored line directory. The modified line stack identifies data lines which have been changed since being made resident in cache memory. It also identifies the status of change of each word within those lines. A 'shared exclusive' flag in the system control element identifies each line for which portions of the line are under exclusive control of more than one processor. The sectored line directory identifies the control status and change status of individual words within a line flagged as 'shared exclusive.' If a line is shared exclusive, an entry for that line is recorded in the sectored line directory. For those lines with entries in the sectored line directory, a processor may store into words within its exclusive control, and fetch words within its exclusive or read-only control. Remote processors may fetch words which are held read-only by the local processor, and store into words which are marked invalid in the cache memory of the local processor.

Method and apparatus for dynamic cache line sectoring in multiprocessor

Method and apparatus for correctly predicting an outcome of a branch instruction in a system of the type that includes a Branch History Table (BHT) and branch instructions that implement non-explicit subroutine calls and returns. Entries in the BHT have two additional stage fields including a CALL field to indicate that the branch entry corresponds to a branch that may implement a subroutine call and a PSEUDO field. The PSEUDO field represents linkage information and creates a link between a subroutine entry and a subroutine return. A target address of a successful branch instruction is used to search the BHT. The branch is known to be a subroutine return if a target quadword contains an entry prior to a target halfword that has the CALL field set. The entry with the CALL bit set is thus known to be the corresponding subroutine call, and the entry point to the subroutine is given by the target address stored within the entry. A PSEUDO entry is inserted into the BHT at the location corresponding to the entry point of the subroutine, the PSEUDO entry being designated as such by having the PSEUDO field asserted. The PSEUDO entry contains the address of the returning branch instruction in place of the target address field.

Subroutine return through branch history table

A cache memory system develops an optimum sequence for transferring data values between a main memory and a line buffer internal to the cache. At the end of a line transfer, the data in the line buffer is written into the cache memory as a block. Following an initial cache miss, the cache memory system monitors the sequence of data requests received for data in the line that is being read in from main memory. If the sequence being used to read in the data causes the processor to wait for a specific data value in the line, a new sequence is generated in which the specific data value is read at an earlier time in the transfer cycle. This sequence is associated with the instruction that caused the first miss and is used for subsequent misses caused by the instruction. If, in the process of handling a first miss related to a specific instruction, a second miss occurs which is caused by the same instruction but which is for data in a different line of memory, the sequence associated with the instruction is marked as an ephemeral miss. Data transferred to the line buffer in response to an ephemeral miss is not stored in the cache memory and limited to that portion of the line accessed within the line buffer.

Cache miss facility with stored sequences for data fetching

Methods and apparatus are described for processing branch instructions using a history based branch prediction mechanism (such as a branch history table) in combination with a data dependent branch table (DDBT), where the branch instructions can vary in both outcome and test operand location. The novel methods and apparatus are sensitive to branch mispredictions and to operand addresses used by the DDBT, to identify irrelevant DDBT entries. Irrelevant DDBT entries are identified within the prediction mechanism using state bits which, when set, indicate that: (1) a given entry in the prediction mechanism was updated by the DDBT and (2) subsequent to such update a misprediction occurred making further DDBT updates irrelevant. Once a DDBT entry is determined to be irrelevant, it is prevented from updating the prediction mechanism. The invention also provides methods and apparatus for locating and removing irrelevant entries from the DDBT. The update packet, sent by the DDBT to the history based prediction mechanism, is expanded to include the test operand address actually used by the DDBT. If the state bits indicate the update is irrelevant, then the operand address can be used to locate and delete the offending DDBT entry since the DDBT is organized based on operand addresses. Additionally, the invention provides for inhibiting creation of further DDBT entries when a Branch Wrong Guess event occurs subsequent to a DDBT update to a given prediction mechanism entry.

Methods and apparatus for insulating a branch prediction mechanism from

A store through cache environment managed exclusively grants exclusivity on a large granularity basis. A cross-invalidate is realized for all changed lines via a single transmission when exclusivity is released. A dynamic table that operates in conjunction with a directory look-aside table (DLAT) determines a number of pages that can be held exclusive simultaneously. For adequate operating speed, the special table must be either fully associative or at least set associative. Alternatively, the table can be incorporated into the DLAT. Each DLAT entry is also extended to include a set of 'resident' bits and a 'valid nonresident' bit. When exclusively is released, the set of local change bits is broadcast to all processors. Upon receipt of such broadcast, the appropriate action is to change the 'valid nonresident' indication to read-only and to clear residence bits whose corresponding local change bit is set.

Cache management for multi-processor systems utilizing bulk

A fast queue mechanism is provided which keeps a queue of changes (i.e. store actions) issued by each processor, which queue is accessible by all processors. When any processor issues a store action to a line of memory in the queue, the old data is overwritten with the new data. If the queue does not currently have a corresponding entry, a new entry is activated. Room for the new entry is made by selecting some existing entry, either the oldest or the least recently used, to be removed. An entry that is to be removed is first used to update the line corresponding to it in main memory. After the changes held in the entry to be removed are applied to the old value of the line (from main memory) and the updated value is put back into main memory, the entry in the queue is removed by marking it 'empty'. When a processor accesses a line of data not in its cache, a cache miss occurs and it is necessary to fetch the line from main memory. Such fetches are monitored by the queue mechanism to see if it is holding changes to the line being fetched. If so, the changes are applied to the line coming from main memory before the line is sent to the requesting processor. After a new entry is made in the queue mechanism, other store actions to the same entry by any processor may occur and usually a number of store actions will occur to the entry before it is removed to make room for another.

Data processing system with fast queue store interposed between

In an approach to reducing delays resulting from resolution of conditional branch instructions, such instructions are pre-executed in a coprocessor which precedes a pipeline processor and prepared an instruction stream for input to the pipeline processor. Because of this pre-execution, the input instruction stream has fewer conditional branches for the pipeline processor to resolve. Also, the coprocessor may handle address generation interlock situations which also cause execution delays in the pipeline processor.

Pipeline processing apparatus for executing instructions in three

Monitoring apparatus is provided to allow out-of-sequence fetching of operands while preserving the appearance of in-sequence fetching to the processor of a computer. The key elements include a stack (119) of N entries holding the addresses of the last M, where M is less than or equal to N, out-of-sequence fetches. A comparator (103) is provided for comparing addresses in the stack with a test address. This test address is supplied via an OR gate (107) as either store addresses or cross-invalidate addresses, the latter being for a multiprocessor system. The addresses in the stack that compare with the test address are set as invalid. In addition, all addresses in the stack are set as invalid on the occurrence of a cache miss or serializing instruction. Finally, a select and check entry function (113) associates an address in the stack with the instruction it represents and deletes the address from the stack when the instruction is handled in its proper sequence.

Posting out-of-sequence fetches

Apparatus for fetching instructions in a computing system. A broadband branch history table is organized by cache line. The broadband branch history table determines from the history of branches the next cache line to be referenced and uses that information for prefetching lines into the cache.

Multiple branch analyzer for prefetching cache lines

A method and apparatus for controlling access to its general purpose registers (GPRs) by a high end machine configuration including a plurality of execution units within a single CPU. The invention allows up to 'N' execution units to be concurrently executing up to 'N' instructions using the GPR sequentially or different GPR's concurrently as either SINK or SOURCE while at the same time preserving the logical integrity of the data supplied to the execution units. The use of the invention allows a higher degree of parallelism in the execution of the instructions than would otherwise be possible if only sequential operations were performed. A series of special purpose tags are associated with each GPR and execution unit. These tags are used together with control circuitry both within the GPR's, within the individual execution units and within the instruction decode unit, which permit the multiple use of the registers to be accomplished while maintaining the requisite logical integrity.

Method and apparatus for guaranteeing the logical integrity of data in

A high speed buffer store arrangement for use in a data processing system having multiple cache buffer storage units in a hierarchial arrangement permits fast transfer of wide data blocks. On each cache chip, input and output latches are integrated thus avoiding separate intermediate buffering. Input and output latches are interconnected by 64-byte wide data buses so that data blocks can be shifted rapidly from one cache hierarchy level to another and back. Chip-internal feedback connections from output to input latches allow data blocks to be selectively reentered into a cache after reading. An additional register array is provided so that data blocks can be furnished again after transfer from cache to main memory or CPU without accessing the respective cache. Wide data blocks can be transferred within one cycle, thus tying up caches much less in transfer operations, so that they have increased availability.

High speed buffer store arrangement for quick wide transfer of data

A prefetching mechanism for a system having a cache has, in addition to the normal cache directory, a two-level shadow directory. When an information block is accessed, a parent identifier derived from the block address is stored in a first level of the shadow directory. The address of a subsequently accessed block is stored in the second level of the shadow directory, in a position associated with the first-level position of the respective parent identifier. With each access to an information block, a check is made whether the respective parent identifier is already stored in the first level of the shadow directory. If it is found, then a descendant address from the associated second-level position is used to prefetch an information block to the cache if it is not already resident therein. This mechanism avoids, with a high probability, the occurrence of cache misses.

Growing community of inventors

Chappaqua, NY, United States of America

James H Pomerene