International Business Machines Corporation

Hong Lam Hua

Robert H Bell, Jr

Mysore Sathyanaranyana Srinivas

Binh K Hua

Men-Chow Chiang

William A Maron

Wen Xiong

Ram Raghavan

Mysore Sathyanarayana Srinivas

Sivarama K Kodukula

James R Gallagher

Wen-Tzer Thomas Chen, Jr

Diane Garza Flemming

Matthew Accapadi

Bret Ronald Olszewski

Dirk Michel

Mathew Accapadi

Herman Dietrich Dierks

Elizabeth Silvia

Anh Tuan Dang

Nasr-Eddine Walehiane

Techniques for preserving memory affinity in a computer system is disclosed. In response to a request for memory access to a page within a memory affinity domain, a determination is made if the request is initiated by a processor associated with the memory affinity domain. If the request is not initiated by a processor associated with the memory affinity domain, a determination is made if there is a page ID match with an entry within a page migration tracking module associated with the memory affinity domain. If there is no page ID match, an entry is selected within the page migration tracking module to be updated with a new page ID and a new memory affinity ID. If there is a page ID match, then another determination is made whether or not there is a memory affinity ID match with the entry with the page ID field match. If there is no memory affinity ID match, the entry is updated with a new memory affinity ID; and if there is a memory affinity ID match, an access counter of the entry is incremented.

Technique for preserving memory affinity in a non-uniform memory access data processing system

According to one aspect of the present disclosure, a method and technique for performance-driven cache line memory access is disclosed. The method includes: receiving, by a memory controller of a data processing system, a request for a cache line; dividing the request into a plurality of cache subline requests, wherein at least one of the cache subline requests comprises a high priority data request and at least one of the cache subline requests comprises a low priority data request; servicing the high priority data request; and delaying servicing of the low priority data request until a low priority condition has been satisfied.

Performance-driven cache line memory access

According to one aspect of the present disclosure a system and technique for performance-driven cache line memory access is disclosed. The system includes: a processor, a cache hierarchy coupled to the processor, and a memory coupled to the cache hierarchy. The system also includes logic executable to, responsive to receiving, a request for a cache line: divide the request into a plurality of cache subline requests, wherein at least one of the cache subline requests comprises a high priority data request and at least one of the cache subline requests comprises a low priority data request; service the high priority data request; and delay servicing of the low priority data request until a low priority condition has been satisfied.

Some embodiments of the inventive subject matter are directed to operations that include determining that an access request to a computer memory results in a cache miss. In some examples, the operations further include determining an amount of cache resources used to service additional cache misses that occurred within a period prior to the cache miss. Furthermore, in some examples, the operations further include servicing the access request to the computer memory based, at least in part, on the amount of the cache resources used to service the additional cache misses within the period prior to the cache miss.

Dynamic prioritization of cache access

A method, system and computer-usable medium are disclosed for managing transient instruction streams. Transient flags are defined in Branch-and-Link (BRL) instructions that are known to be infrequently executed. A bit is likewise set in a Special Purpose Register (SPR) of the hardware (e.g., a core) that is executing an instruction request thread. Subsequent fetches or prefetches in the request thread are treated as transient and are not written to lower-level caches. If an instruction is non-transient, and if a lower-level cache is non-inclusive of the L1 instruction cache, a fetch or prefetch miss that is obtained from memory may be written in both the L1 and the lower-level cache. If it is not inclusive, a cast-out from the L1 instruction cache may be written in the lower-level cache.

Performance of emerging applications in a virtualized environment using transient instruction streams

A storage subsystem combining solid state drive (SSD) and hard disk drive (HDD) technologies provides low access latency and low complexity. Separate free lists are maintained for the SSD and the HDD and blocks of file system data are stored uniquely on either the SSD or the HDD. When a read access is made to the subsystem, if the data is present on the SSD, the data is returned, but if the block is present on the HDD, it is migrated to the SSD and the block on the HDD is returned to the HDD free list. On a write access, if the block is present in the either the SSD or HDD, the block is overwritten, but if the block is not present in the subsystem, the block is written to the HDD.

Hybrid storage subsystem with mixed placement of file contents

According to one aspect of the present disclosure, a method and technique for variable cache line size management is disclosed. The method includes: determining whether an eviction of a cache line from an upper level sectored cache to an unsectored lower level cache is to be performed, wherein the upper level cache includes a plurality of sub-sectors, each sub-sector having a cache line size corresponding to a cache line size of the lower level cache; responsive to determining that an eviction is to be performed, identifying referenced sub-sectors of the cache line to be evicted; invalidating unreferenced sub-sectors of the cache line to be evicted; and storing the referenced sub-sectors in the lower level cache.

Variable cache line size management

According to one aspect of the present disclosure, a system and technique for variable cache line size management is disclosed. The system includes a processor and a cache hierarchy, where the cache hierarchy includes a sectored upper level cache and an unsectored lower level cache, and wherein the upper level cache includes a plurality of sub-sectors, each sub-sector having a cache line size corresponding to a cache line size of the lower level cache. The system also includes logic executable to, responsive to determining that a cache line from the upper level cache is to be evicted to the lower level cache: identify referenced sub-sectors of the cache line to be evicted; invalidate unreferenced sub-sectors of the cache line to be evicted; and store the referenced sub-sectors in the lower level cache.

A cache within a computer system receives a partial write request and identifies a cache hit of a cache line. The cache line corresponds to the partial write request and includes existing data. In turn, the cache receives partial write data and merges the partial write data with the existing data into the cache line. In one embodiment, the existing data is 'modified' or 'dirty.' In another embodiment, the existing data is &#x201c;shared.&#x201d; In this embodiment, the cache changes the state of the cache line to indicate the storing of the partial write data into the cache line.

Partial line cache write injector for direct memory access write

Some embodiments of the inventive subject matter are directed to determining that a memory access request results in a cache miss and determining an amount of cache resources used to service cache misses within a past period in response to determining that the memory access request results in the cache miss. Some embodiments are further directed to determining that servicing the memory access request would increase the amount of cache resources used to service cache misses within the past period to exceed a threshold. In some embodiments, the threshold corresponds to reservation of a given amount of cache resources for potential cache hits. Some embodiments are further directed to rejecting the memory access request in response to the determining that servicing the memory access request would increase the amount of cache resources used to service cache misses within the past period to exceed the threshold.

Some embodiments of the inventive subject matter are directed to a cache comprising a tracking unit and cache state machines. In some embodiments, the tracking unit is configured to track an amount of cache resources used to service cache misses within a past period. In some embodiments, each of the cache state machines is configured to, determine whether a memory access request results in a cache miss or cache hit, and in response to a cache miss for a memory access request, query the tracking unit for the amount of cache resources used to service cache misses within the past period. In some embodiments, the each of the cache state machines is configured to service the memory access request based, at least in part, on the amount of cache resources used to service the cache misses within the past period according to the tracking unit.

A method for minimizing cache conflict misses is disclosed. A translation table capable of facilitating the translation of a virtual address to a real address during a cache access is provided. The translation table includes multiple entries, and each entry of the translation table includes a page number field and a hash value field. A hash value is generated from a first group of bits within a virtual address, and the hash value is stored in the hash value field of an entry within the translation table. In response to a match on the entry within the translation table during a cache access, the hash value of the matched entry is retrieved from the translation table, and the hash value is concatenated with a second group of bits within the virtual address to form a set of indexing bits to index into a cache set.

Method and apparatus for minimizing cache conflict misses

A method, system and computer-usable medium are disclosed for managing prefetch streams in a virtual machine environment. Compiled application code in a first core, which comprises a Special Purpose Register (SPR) and a plurality of first prefetch engines, initiates a prefetch stream request. If the prefetch stream request cannot be initiated due to unavailability of a first prefetch engine, then an indicator bit indicating a Prefetch Stream Dispatch Fault is set in the SPR, causing a Hypervisor to interrupt the execution of the prefetch stream request. The Hypervisor then calls its associated operating system (OS), which determines prefetch engine availability for a second core comprising a plurality of second prefetch engines. If a second prefetch engine is available, then the OS migrates the prefetch stream request from the first core to the second core, where it is initiated on an available second prefetch engine.

Growing community of inventors

Austin, TX, United States of America

Hong Lam Hua