Cache memory hierarchy having a large write through first level that allocates for CPU read misses only and a small write back second level that allocates for CPU write misses only

Claim 2

In a data processing system, first and second levels of a cache memory hierarchy comprising; (a) a write through (WT) cache memory, at the first level of the cache memory hierarchy for (i) caching a first subset of data stored in a main memory: (ii) allowing read accesses by a processor and write accesses by said processor, as well as snoop accesses by direct memory access (DMA) devices; and (iii) modifying the contents of said WT cache memory only in response to indications of read misses of said read accesses, by implementing a first allocation policy wherein for each of said read miss indications, said first subset of data stored in the WT cache memory is modified by replacing a line of data in the first subsist with a new line of data from the main memory, said new line of data from the main memory having been the subject of the missed read access; and (b) a write back (WB) cache memory, at the second level of the cache memory hierarchy, for (i) caching a second subset of data stored in said main memory; (ii) allowing read accesses by said processor and write accesses by said processor, as well as read accesses by said DMA devices and write accesses by said DMA devices; and (iii) modifying the contents of said WB cache memory only in response to indications of write misses of said write accesses by said processor, by implementing a second allocation policy wherein for each of said write miss indications, said second subset of data stored in the WB cache memory is modified by replacing a line of data in the second subset with a new line data from the processor, said new line of data from the processor having been the subject of the missed write access.

Claim 3

The cache memory hierarchy as set forth in claim 1, wherein said WT cache memory at the first level of the cache memory hierarchy having a first cache size designed to provide a high cache hit rate, whereas said WB cache memory at the second level of the cache memory hierarchy having a second cache size, complementary to said first cache size, designed merely to reduce write accesses to said main memory.

Claim 4

The cache memory hierarchy as set forth in claim 2, wherein said first cache size is substantially larger than said second cache size.

Claim 5

The cache memory hierarchy as set forth in claim 3, wherein said first cache size is about eight times larger than said second cache size.

Claim 6

The cache memory hierarchy as set forth in claim 1, wherein said WT cache memory at the first level of the cache memory hierarchy and said WB cache memory at the second level of cache memory hierarchy are n-way set associative cache memory.

Claim 7

A computer system comprising: a) a central processing unit (CPU) for executing instructions; b) a main memory for storing instructions and data; c) a peripheral device for providing periphery functions that makes direct memory accesses; d) first and second levels of a cache memory hierarchy coupled to said CPU, said main memory and said peripheral device having 1) a write through (WT) cache memory at the first level of the cache memory hierarchy for (i) caching a first subset of instructions and data stored in said main memory; (ii) allowing read accesses by said CPU and write accesses by said CPU as well as snoop accesses by said peripheral device; and (iii) modifying the contents of said WT cache memory only in response to indications of read misses of said read accesses by said CPU, by implementing a first allocation policy wherein for each of said read miss indications, said first subset of data stored in the WT cache memory is modified by replacing a line of data in the first subset with a new line of data from the main memory, said new line of data from the main memory having been the subject of the missed read access; and 2) a write back (WB) cache memory at the second level of the cache memory hierarchy for i) caching a second subset of instructions and data stored in said main memory; (ii) allowing read accesses by said CPU and write accesses by said CPU as well as read accesses by said peripheral device and write accesses by said peripheral device; and (iii) modifying the contents of said WB cache memory in response to indications of write misses of said write accesses by said processor, by implementing a second allocation policy wherein for each of said write miss indications, said second subset of data stored in the WB cache memory is modified by replacing a line of data in the second subset with a new line data from the processor, said new line of data from the processor having been the subject of the missed write access.

Claim 8

A computer system comprising: a) a central processing unit (CPU) for executing instructions; b) a main memory for storing instructions and data; c) a peripheral device for providing periphery functions that makes direct memory accesses; d) first and second levels of a cache memory hierarchy coupled to said CPU, said main memory and said peripheral device having 1) a write through (WT) cache memory at the first level of the cache memory hierarchy for (i) caching a first subset of instructions and data stored in said main memory; (ii) allowing read accesses by said CPU and write accesses by said CPU as well as snoop accesses by said peripheral device; and (iii) modifying the contents of said WT cache memory only in response to indications of read misses of said read accesses by said CPU, by implementing a first allocation policy wherein for each of said read miss indications, said first subset of data stored in the WT cache memory is modified by replacing a line of data in the first subset with a new line of data from the main memory, said new line of data from the main memory having been the subject of the missed read access; and 2) a write back (WB) cache memory at the second level of the cache memory hierarchy for i) caching a second subset of instructions and data stored in said main memory; (ii) allowing read accesses by said CPU and write accesses by said CPU as well as read accesses by said peripheral device and write accesses by said peripheral device; and (iii) modifying the contents of said WB cache memory in response to indications of write misses of said write accesses by said processor, by implementing a second allocation policy wherein for each of said write miss indications, said second subset of data stored in the WB cache memory is modified by replacing a line of data in the second subset with a new line data from the processor, said new line of data from the processor having been the subject of the missed write access.

Claim 9

The computer system as set forth in claim 6, wherein said WT cache memory at the first level of the cache memory hierarchy having a first cache size designed to provide a high cache hit rate, whereas said WB cache memory at the second level of the cache memory hierarchy having a second cache size, complementary to said first cache size, designed merely to reduce write accesses to said main memory.

Claim 10

The computer system as set forth in claim 7, wherein said first cache size is substantially larger than said second cache size.

Claim 11

The computer system as set forth in claim 8, wherein said first cache size is about eight times larger than said second cache size.

Claim 12

The computer as set forth in claim 6, wherein said WT cache memory at the first level of the cache memory hierarchy and said WB cache memory at the second level of the cache memory hierarchy are n-way set associative cache memory.