US 4,755,930 AGranted

Hierarchical cache memory system and method

AssigneeEncore Computer Corporation

InventorsAndrew W. Wilson, Jr., Steven J. Frank

Life of this patent

Estimating the term…

Loading patent…

Abstract

A caching system for a shared bus multiprocessor which includes several processors each having its own private cache memory. Each private cache is connected to a first bus to which a second, higher level cache memory is also connected. The second, higher level cache in turn is connected either to another bus and higher level cache memory or to main system memory through a global bus. Each higher level cache includes enough memory space so as to enable the higher level cache to have a copy of every memory location in the caches on the level immediately below it. In turn, main memory includes enough space for a copy of each memory location of the highest level of cache memories. The caching can be used with either write-through or write-deferred cache coherency management schemes.

Claims14

Dependent claims

A method of maintaining cache coherency in a hierarchical cache memory system designed for use with a multiprocessor computer system including a plurality of processors, each having an associated private cache memory, and a central memory, said method comprising the steps of:

providing a shared cache memory connected to said private cache memories through a first bus shared by at least two of said processors and to said central memory through a second global bus;
when a processor performs a write operation to a memory location of its associated private cache memory thereby changing the contents of that memory location, copying the contents of said memory location into said shared cache memory connected to said first bus and into said central memory connected to said second global bus;
invalidating cache memory locations of cache memories, other than the private cache memory to which the write operation is performed and the cache memory into which the location was copied, which contain a copy of the written-to memory location;
placing said private cache memory location to which the write operation is performed in a reserved state indicating that the private cache containing said reserved memory location is the only private cache with a valid copy of said memory location;
enabling said processor, having an associated private cache memory means with a memory location in a reserved state, to perform additional write operations only to said associated private cache memory means, said shared cache memory means and said central memory retaining the contents of said reserved memory location as it existed prior to said additional write operations.

2 more elements

An hierarchical cache memory system for use with a multiprocessor computer system including a plurality of processors and a central memory, said cache memory system comprising:

at least two groups of first level cache memory means, each of said first level cache memory means being associated with only one of the plurality of processors, all memory locations of said first level cache memory being writable;
first bus means associated with each of said at least two groups of first level cache memory means for connecting only said first level cache memory means in a group with each other;
one unique second level cache memory means connected to each of said first bus means, each of said second level cache memory means containing a copy of the contents of each memory location of all of the first level cache memory means connected to its first bus means;
second bus means connected to all of said second level cache memory means and to said central memory, said central memory containing a copy of the contents of each memory locations of all of said second level cache memory means;
means for writing the contents of a memory location of a first level cache memory means into a corresponding memory location of the second level cache memory means connected to its first bus means and into a corresponding memory location of said central memory each time a write operation changing the contents of a memory location is performed to a memory location of a first level cache memory means;
and
means for invalidating the contents of a memory location of a first level cache memory means or of a second level cache memory means corresponding to a location of another first level cache memory means to which a write operation changing the contents of the memory location has been performed.

4 more elements

A multiprocessor system comprising:

first multiprocessor system including at least one processor means, each of said at least one processor means being connected to an associated private cache memory means for storing frequently accessed memory locations, each of said at least one private cache memory means being connected to the others of said at least one private cache memory means through a first common bus means;
first shared cache memory means connected to said first common bus means, said first shared cache memory means including a copy of the contents of each memory location of each of said private cache memory means connected to said first common bus means;
second multiprocessor system including at least one processor means, each of said at least one processor means being connected to an associated private cache memory means for storing frequently accessed memory locations, each of said at least one private cache memory means being connected to the others of said at least one private cache memory means of said second multiprocessor system through a second common bus means;
second shared cache memory means connected to said second common bus means, said second cache memory means including a copy of the contents of each memory location of each of said private cache memory means connected to said second common bus means;
central memory means connected to a global bus means to which said first and second cache memory means are also connected, said central memory means including a copy of the contents of each memory location of said first and second shared cache memory means;
means for writing the contents of a memory location of a private cache memory means into a corresponding memory location of said first or second shared cache memory means and into a corresponding memory location of said central memory means each time a write operation changing the contents of a memory location is performed to a memory location of a cache memory means;
and
means for invalidating the contents of a memory location of a private cache memory means or of said first or second shared cache memory means which include a memory location corresponding to a location of another private cache means to which a write operation has been performed.

5 more elements

A multiprocessor system comprising:

first multiprocessor system including at least one processor means, each of said at least one processor means being connected to an associated private cache memory means for storing frequently accessed memory locations, each of said at least one private cache memory means being connected to the others of said at least one private cache memory means through a first common bus means;
first shared cache memory means connected to said first common bus means, said first shared cache memory means including a copy of the contents of each memory location of each of said private cache memory means connected to said first common bus means;
second multiprocessor system including at least one processor means, each of said at least one processor means being connected to an associated private cache memory means for storing frequently accessed memory locations, each of said at least one private cache memory means being connected to the others of said at least one private cache memory means of said second multiprocessor system through a second common bus means;
second shared cache memory means connected to said second common bus means, said second cache memory means including a copy of the contents of each memory location of each of said private cache memory means connected to said second common bus means;
central memory means connected to a global bus means to which said first and second cache memory means are also connected, said central memory means including a copy of the contents of each memory location of said first and second shared cache memory means;
and
means for placing a private cache memory location in a reserved state when the processor associated with the private cache performs a write operation to said memory location thereby changing the contents of the memory locations, said reserved state indicating that the private cache containing said reserved memory location is the only private cache with a valid copy of said memory location.

4 more elements

Dependencies are inferred from the claim text.