Single-thread cache miss rate estimation
US-2016246722-A1 · Aug 25, 2016 · US
US9645927B2 · US · B2
| Field | Value |
|---|---|
| Publication number | US-9645927-B2 |
| Application number | US-201514744139-A |
| Country | US |
| Kind code | B2 |
| Filing date | Jun 19, 2015 |
| Priority date | Jun 20, 2014 |
| Publication date | May 9, 2017 |
| Grant date | May 9, 2017 |
A practical reading order for non-experts. Skip the full description unless you need deep technical detail.
What the patent document calls the invention.
A short plain-language summary of the technical disclosure.
Who owns or filed the patent and who is credited as inventor.
Filing, priority, publication, and grant dates set the timeline.
The legal scope of protection — read this for what is actually claimed.
Technology tags used to group this patent with similar filings.
Prior art links and similar publications in this corpus.
Official abstract text for this publication.
A module of cache coherence management by directory, in which each datum stored in cache memory is associated with a state, at least one of which indicates data sharing among a plurality of processors, the module including a storage unit to store a directory containing a list of cache memory addresses, each address possibly associated with a state corresponding to the state of the datum available at this address, and a processing unit configured to update said list, said processing unit being configured so as not to list the address lines related to data associated with the first state.
Opening claim text (preview).
The invention claimed is: 1. A module of cache coherence management by directory in a multiprocessor system, in which each datum stored in cache memory is associated with one state among a plurality of states, at least a first state of which indicates data sharing among a plurality of processors of the multiprocessor system, the module comprising: a storage unit configured to store a directory containing a list of cache memory addresses, each address associated with one state among a plurality of states, said one state corresponding to the state of the datum available at the address, and a processing unit configured to update said list, said processing unit being configured so as not to list in said list address lines related to a first datum associated with the first state indicating data sharing among the plurality of processors notwithstanding the first datum being currently found in the first state and so as to list in said list address lines related to second datum associated with at least a second state different than the first state. 2. The Module according to claim 1 , wherein the processing unit is configured to list only the addresses related to data associated with a state indicating an exclusivity of said data to the respective processors. 3. The module according to claim 1 , wherein the processing unit is configured to operate selectively according to at least two modes: a first mode of operation in which the address lines associated with said first state are not listed in said list, and a second mode of operation in which the address lines associated with said first state are listed in said list. 4. The module according to claim 3 , wherein in the first mode of operation, the only addresses stored are those related to data associated with a state indicating an exclusivity of said data to the respective processors. 5. A multiprocessor device comprising: a plurality of processors; a plurality of modules of cache memory respectively associated with said processors, and at least one proxy module according to claim 1 , configured to manage cache coherence among said modules of cache memory. 6. A process of cache coherence management by directory in a multiprocessor system, in which each datum stored in cache memory is associated with one state among a plurality of states, at least a first state of which indicates data sharing among a plurality of processors, the process comprising: receiving a request relative to a datum in cache memory; determining a state of the datum; launching an update of a directory containing a list of cache memory addresses, each address associated with one state among a plurality of states corresponding to the state of the datum available at the address, if the determined state is at least a second state different from the first state, wherein the update comprises listing in said list an address line related to the datum; and launching a query of cache memory to obtain the datum, if said determined state is at least the first state indicating data sharing among the plurality of processors, without update of said directory that would otherwise list in said list the address line related to the datum. 7. The process according to claim 6 , wherein the second state is a state indicating the exclusivity of the datum to a processor. 8. The process according to claim 7 , wherein the update is launched only if the second state is a state indicating the exclusivity of the datum to a processor. 9. The process according to claim 6 , wherein the launching of the update includes a step of eviction of an address line of the directory. 10. The process according to claim 6 , further comprising: receiving a request relative to a datum in cache memory; and prior to the determination of the state of the datum, the launching of the update, and the launching of the query, determining a mode of operation to determine whether it is necessary to list in the list the address lines related to data associated with the first state indicating data sharing among a plurality of processors.
for multiprocessing or multitasking · CPC title
State-only directory, i.e. not recording identity of sharing or owning nodes · CPC title
Associative directories (G06F12/0822 takes precedence) · CPC title
of operating mode, e.g. cache mode or local memory mode · CPC title
Distributed directories, e.g. linked lists of caches · CPC title
Related publications grouped by family.
Answers are generated from the same data shown on this page.