Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F12/00—Accessing, addressing or allocating within memory systems or architectures
  • G06F12/02—Addressing or allocation; Relocation
  • G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
  • G06F12/12—Replacement control
  • G06F12/121—Replacement control using replacement algorithms
  • G06F12/128—Replacement control using replacement algorithms adapted to multidimensional cache systems, e.g. set-associative, multicache, multiset or multilevel
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F12/00—Accessing, addressing or allocating within memory systems or architectures
  • G06F12/02—Addressing or allocation; Relocation
  • G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
  • G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
  • G06F12/0844—Multiple simultaneous or quasi-simultaneous cache accessing
  • G06F12/0855—Overlapped cache accessing, e.g. pipeline
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F12/00—Accessing, addressing or allocating within memory systems or architectures
  • G06F12/02—Addressing or allocation; Relocation
  • G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
  • G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
  • G06F12/0864—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches using pseudo-associative means, e.g. set-associative or hashing
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F12/00—Accessing, addressing or allocating within memory systems or architectures
  • G06F12/02—Addressing or allocation; Relocation
  • G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
  • G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
  • G06F12/0875—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches with dedicated cache, e.g. instruction or stack
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F12/00—Accessing, addressing or allocating within memory systems or architectures
  • G06F12/02—Addressing or allocation; Relocation
  • G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
  • G06F12/0802—Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
  • G06F12/0893—Caches characterised by their organisation or structure
  • G06F12/0895—Caches characterised by their organisation or structure of parts of caches, e.g. directory or tag array
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
  • G06F2212/10—Providing a specific technical effect
  • G06F2212/1016—Performance improvement
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
  • G06F2212/40—Specific encoding of data in memory or cache
  • G06F2212/401—Compressed data
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
  • G06F2212/60—Details of cache memory
  • G06F2212/6032—Way prediction in set-associative cache
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06F—ELECTRIC DIGITAL DATA PROCESSING
  • G06F2212/00—Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
  • G06F2212/62—Details of cache specific to multiprocessor cache arrangements
  • G06F2212/621—Coherency control relating to peripheral accessing, e.g. from DMA or I/O device

Definitions

• aspects disclosed herein relate to the field of computer microprocessors (also referred to herein as processors). More specifically, aspects disclosed herein relate to cache tag compression.
• Modem processors conventionally rely on caches to improve processing performance.
• Caches work by exploiting temporal and spatial locality in the instruction streams and data streams of the workload.
• a portion of the cache is dedicated to storing cache tag arrays.
• Cache tags store the address of the actual data fetched from the main memory.
• the speed of determining a hit or a miss in the cache is constrained by the width of the tag array, where larger tag arrays conventionally create more timing pressure.
• a wider (or larger) tag has more bits. To determine if there is a hit in the cache, all bits of the tag must be compared against the probe address. More bits in the comparison may therefore require more delay to return a result of the comparison.
• a cache memory comprises a tag compression structure and a tag array.
• the tag compression structure stores a plurality of tag compression entries. Each of the plurality of tag compression entries contains a respective set of higher order bits of a memory tag and are identified by a respective index value.
• the tag array stores a plurality of tag array entries. Each of the plurality of tag array entries contains a respective set of lower order bits of the memory tag and an index value corresponding to one of the index values of the tag compression structure.
• the first memory tag may comprise at least a portion of a memory address of a data element stored in a data array.
• a method comprises storing, in a first tag compression entry of a plurality of tag compression entries in a tag compression structure, a set of higher order bits of a first memory tag. Each of the plurality of tag compression entries are identified by a respective index value.
• the method further comprises storing, in a first tag array entry of a plurality of tag array entries in a tag array, a set of lower order bits of the memory tag and an index value for the first tag array entry.
• the index value of the first tag array entry corresponds to the first tag compression entry in the tag compression structure.
• the first memory tag may comprise at least a portion of a memory address of a data element stored in a data array.
• a non-transitory computer-readable medium stores instructions that, when executed by a processor, cause the processor to perform an operation comprising storing, in a first tag compression entry of a plurality of tag compression entries in a tag compression structure, a set of higher order bits of a first memory tag. Each of the plurality of tag compression entries are identified by a respective index value.
• the operation further comprises storing, in a first tag array entry of a plurality of tag array entries in a tag array, a set of lower order bits of the memory tag and an index value for the first tag array entry.
• the index value of the first tag array entry corresponds to the first tag compression entry in the tag compression structure.
• the first memory tag may comprise at least a portion of a memory address of a data element stored in a data array.
• an apparatus comprises means for storing, in a first tag compression entry of a plurality of tag compression entries, a set of higher order bits of a first memory tag. Each of the plurality of tag compression entries are identified by a respective index value.
• the apparatus further comprises means for storing, in a first tag array entry of a plurality of tag array entries, a set of lower order bits of the memory tag and an index value for the first tag array entry.
• the index value of the first tag array entry corresponds to the first tag compression entry.
• the first memory tag may comprise at least a portion of a memory address of a data element.
• Figure 1 illustrates a processor which provides cache tag compression, according to one aspect.
• Figure 2 is a logical view of components of a processor configured to provide cache tag compression, according to one aspect.
• Figure 3 is a schematic illustrating example values stored in a tag compression cache, tag array, and data array, according to one aspect.
• Figure 4 is a flow chart illustrating a method to provide cache tag compression, according to one aspect.
• Figure 5 is a flow chart illustrating a method to establish a new tag compression cache entry, according to one aspect
• Figure 6 is a block diagram illustrating a computing device integrating a processor configured to provide cache tag compression, according to one aspect.
• caches may be tagged with some set of upper address bits.
• cache tags require storage for a large number of bits, and modern cache architectures conventionally require substantially larger tags than previous generations.
• KB 4-way 32 kilobyte
• the number of lines of such a cache is 2 15 /2 9 , or 512 lines. Each line is associated with a tag.
• the number of sets of the cache is equal to 2 15 /2 6 /2 2 , or 128 sets (with a 7-bit set decode).
• the tag bits do not include the 7-bit set decode or 6-bit byte-offset, meaning the tag bits start at A[7+6], or A[13], where "A" indicates address input to the tag search. Therefore, for an example address of 64 bits, the tag is compared against A[63: 13], and the tag storage is 51 bits per line. The total tag storage for the cache is therefore 51 bits/line * 512 lines, or 26,112 bits, resulting in a storage overhead of 10% for storing the cache tags.
• tags are arbitrary and code-dependent, a subset of the bits of each of the tag values may have little variability, such that the subset of bits has the same values across multiple cache tags. Aspects disclosed herein exploit this tendency by eliminating redundant storage of multiple copies of the same value across multiple cache tags. Specifically, aspects disclosed herein introduce a tag compression cache that is configured to store unique tag values. Doing so removes a number of bits from the cache tag. The bits that are removed from the cache tag may be replaced with a smaller number of bits that point to a tag compression cache entry.
• the tag compression cache may hold any set of bits from the cache tag, which include, without limitation, upper (or higher) address bits, address space identifiers (ASID), security bits, exception level bits, and the like.
• aspects disclosed herein access the tag compression cache and the cache in parallel to determine if an input memory address results in a cache "hit” or "miss.” Specifically, aspects disclosed herein access the tag compression cache using a set of higher order bits of the input memory address. If there is a hit in the tag compression cache (i.e., an entry in the tag compression cache matches the higher order bits of the input memory address), the tag compression cache is configured to retum the index value of matching entry. In parallel, aspects disclosed herein search the cache using a set of lower order bits of the input memory address.
• the cache is configured to return the lower order bits and the index value in the matching entry. Aspects disclosed herein may then compare the index values returned from the cache and the tag compression cache. If the index values do not match, aspects disclosed herein retum an indication of a cache miss. Aspects disclosed herein may further compare the lower bits returned from the cache tag entry to the lower bits of the input memory address. If the lower bits do not match, aspects disclosed herein return an indication of a cache miss. However, if the index values match and the lower bits match, aspects disclosed herein retum an indication of a cache hit (i.e., the requested data is present in the cache).
• a cache is used herein as a reference example to facilitate description of aspects of the disclosure.
• the use of a cache should not be considered limiting of the disclosure, as the techniques described herein apply to other hardware structures that have tag fields, such as processor caches, data caches, instruction caches, line fill buffers, translation lookaside buffers (TLB), and the like. Any reference to a particular hardware structure should not be considered limiting of the disclosure.
• FIG. 1 illustrates a processor 101 that provides cache tag compression, according to one aspect.
• the processor 101 includes an instruction execution pipeline 112 which executes instructions.
• the pipeline 112 may be a superscalar design, with multiple parallel pipelines, each of which includes various non-architected registers (not pictured), and one or more arithmetic logic units (also not pictured).
• the processor 101 also includes a cache 102 (also referred to as a cache memory 102), which stores lines of data from one or more higher levels of memory 108.
• the higher levels of memory 108 may include, without limitation, higher level caches and/or main (system) memory.
• the cache 102 is a Level 1 (LI) data cache.
• the CPU 101 may include numerous variations, and the CPU 101 shown in Figure 2 is for illustrative purposes and should not be considered limiting of the disclosure.
• the CPU 101 may be a graphics processing unit (GPU).
• GPU graphics processing unit
• the cache 102 includes cache logic 103, a tag compression cache 104, a tag array 105, and a data array 106.
• the cache logic 103 generally controls operation of the cache 102, such as determining whether cache hits or misses occur in a particular operation.
• the tag compression cache (TCC) 104 is a hardware structure configured to hold a set of upper (or higher) order bits of memory tags.
• the TCC 104 may hold one or more of the upper address bits of the memory tag, ASID bits, security bits, exception level bits, and the like.
• the TCC 104 is generally configured to store redundant values that would otherwise be redundantly stored in the tag array 105.
• the tag array 105 is generally configured to store the addresses of data stored in the cache 102.
• each entry of the tag array 105 is configured to store a set of lower order bits of memory tags and an index value which corresponds to the index value of an entry in the TCC 104.
• the data aray 106 stores the data of the cache lines.
• the TCC 104 is implemented as a content addressable memory (CAM) structure.
• the tag array 105 and data array 106 may be implemented as a CAM.
• the TCC 104 is a fully- associative, least recently used (LRU) replacement structure.
• the TCC 104 may have any number of entries, such as two, four, or eight.
• Means for storing data in the TCC 104, tag array 105, and data array 106 may include one or more memory cells.
• the processor 101 may seek to determine whether data located in one of the higher levels of memory 108 is present within the cache 102.
• the cache logic 103 may determine whether the data at the input memory address is in the cache 102. To do so, the cache logic 103 may initiate a CAM search of the TCC 104 using the higher order bits of the input memory address. If an entry in the TCC 104 matches higher order bits of the input memory address, the TCC 104 is configured to return the index value of the matching entry in the TCC 104.
• the cache logic 103 may initiate a CAM search of the tag array 105 in parallel with the CAM search of the TCC 104.
• the cache logic 103 may search the tag array 105 using the lower order bits of the input memory address. If the lower order bits of an entry in the tag array 105 match the lower order bits of the input memory address, the tag array 105 is configured to return the lower order bits and the index value of the matching entry. The cache logic 103 may then compare the index value returned by the TCC 104 and the tag array 105. If the index values do not match, the cache logic 103 may return an indication of a cache miss. The cache logic 103 may also compare the lower order bits returned by the tag array 105 to the lower order bits of the input memory address. If the lower order bits match, and the index values match, the cache logic 103 is configured to return an indication of a cache hit.
• the cache logic 103 may be configured to create an entry for the input memory address in the TCC 104. If a free (or otherwise available) entry exists in the TCC 104, the cache logic 103 may store a set of higher order bits of the input memory address to the free entry. If no entries are available in the TCC 104, the cache logic 103 may evict an existing entry in the TCC 104, and invalidate any corresponding entries in the tag array 105 (and data array 106) that specify the index value of the evicted entry.
• the cache logic 103 may apply any suitable replacement policy for the TCC 104, such as LRU, or replacing the TCC 104 entry that is associated with the fewest cache lines (which may be determined by a counter for each entry in the TCC 104).
• the cache logic 103 may invalidate the entries in any feasible manner, such as by a flash invalidation or a line by line invalidation.
• the cache logic 103 may also create an entry in the tag array 105 that specifies the lower order bits of the input memory address and an index value of the TCC 104 entry storing the higher order bits of the input memory address.
• the cache logic 103 may probe the TCC 104 every clock cycle. However, this may draw large amounts of power, and may be in the critical timing path. The more entries that are built into the TCC 104, the more power and time may be required to probe the TCC 104. Therefore, in some aspects, the cache logic 103 may include logic that avoids the need to search the TCC 104 in a majority of cycles. For example, a set of registers may be configured to store the bits returned by a previous access to the TCC 104 (and whether the corresponding TCC 104 entry remains valid).
• the cache logic 103 may determine whether the input memory address bits corresponding to those that are held in the TCC 104 are the same as in the previous access to the TCC 104. The cache logic 103 may then determine whether the associated TCC 104 entry has not been invalidated since the last access. If both determinations are true, the cache logic 103 may determine that the compression cache tag must be in the TCC 104, and that it is safe to reuse the previous cycle's index, rendering it unnecessary to search the TCC 104 in the current cycle. If either determination is not true, the cache logic 103 cannot reuse the TCC 104 index of the previous cycle, the control logic 103 must search the TCC 104.
• Means for searching the TCC 104, the tag array 105, and the data array 106 include logic implemented as hardware and/or software. Similarly, the logic implemented as hardware and/or software may serve as means for reading and/or writing values, returning indications of hits and/or misses, and returning values from the TCC 104, the tag array 105, and the data array 106.
• Example of such means logic includes memory controllers, cache controllers, and data controllers.
• the compression scheme depicted in Figure 1 may be applied to other PC-tagged and/or ASID-tagged microarchitectural structures.
• a virtual address may be transformed into the form ⁇ tag compression index value, partial virtual address ⁇ , which may be used throughout the block. Doing so would allow all data-path elements to operate on this shortened form, including a translation lookaside buffer (TLB), branch predictors, breakpoints, caches, and the like.
• TLB translation lookaside buffer
• branch predictors branch predictors
• breakpoints caches
• the transformed address could be converted into the original address whenever necessary, such as at block boundaries, for cache-content debug facilities, and the like.
• the tag compression technique may be applied independently to instruction addresses, load/store addresses, and physically tagged structures.
• the processor 101 is disposed on an integrated circuit including the instruction execution pipeline 112 and the cache 102.
• the cache 102 may be located on a separate integrated circuit from an integrated circuit including the processor 101.
• FIG. 2 is a logical view of a portion of a cache logic 103 of a processor 101 configured to provide cache tag compression, according to one aspect.
• the processor 101 includes the tag compression cache (TCC) 104 and the cache tag array 105.
• TCC tag compression cache
• the TCC 104 is an indexed hardware structure storing a plurality of TCC entries 202.
• the TCC entries 202 may include a set of higher order address bits of a memory tag.
• the TCC entries 202 may further include bits related to ASIDs, security levels, exception levels, and the like.
• the cache tag array 105 includes a plurality of entries, each of the entries having a tag compression cache (TCC) index value 204 and a cache tag 205.
• TCC tag compression cache
• the TCC index value 204 may correspond to the index value of an entry 202 in the TCC 104.
• the cache tags 205 may include the lower order bits of a memory tag. As shown, the cache tags 205 store the lower order bits of the memory tag, namely bits 0 through M of an N- bit memory tag. Similarly, the TCC 104 stores the higher order bits of the N-bit memory tag, namely bits M+l through N of the memory tag. Therefore, a complete memory tag (bits 0 through N) may be produced by combining the higher order bits stored in the TCC 104 (bits M+l through N) with the lower order bits of the cache tag 205 (bits 0 through M).
• the processor 101 may request data stored at a memory address 201.
• the cache logic 103 may be configured to determine whether the data stored at the memory address 201 is located in the cache 102.
• the memory address 201 may be a virtual memory address having a length of N bits.
• the cache logic 103 may probe the TCC 104 and the cache tag array 105 in parallel. More specifically, as shown, the cache logic 103 may perform a CAM search of the cache tag array 105. If one of the cache tags 205 includes an entry which matches bits 0 through M of the memory address 201, the cache tag array 105 may return a TCC index value 208 and a cache tag 209.
• a splitter 213 may split the memory address 201, which provides bits M+l through N of the memory address 201 to the TCC 104.
• a CAM search on the TCC 104 may be performed using the bits M+l through N of the memory address 201. If an entry 202 in the TCC 104 includes a value that matches the value of the bits M+l through N of the memory address 201, the TCC 104 is configured to return a TCC entry 206 and a TCC index value 207, where the index value 207 identifies the matching TCC entry 206.
• the TCC index value 207 may be any number of bits in length suitable to allow each entry 202 in the TCC 104 to be uniquely identified by a respective TCC index value 207.
• the TCC index value 207 corresponding to each entry may be two bits in length.
• each TCC index 204 stored in the cache tag array 105 would also be two bits in length.
• the TCC index values 207 are implicitly identified based on the location (or address) of a given TCC entry 202.
• a comparator 210 compares the TCC index value 207 returned from the TCC 104 to the TCC index value 208 returned by the cache tag array 105.
• a comparator 211 compares the lower order bits 0:M of the memory address 201 to the lower order bits 0:M stored in a cache tag entry 209 returned from the cache tag array 105.
• the AND logic 212 is configured to perform a logical AND operation on the outputs of the comparators 210, 211. If both comparators 210, 211 indicate a match, then the output of the AND logic 212 is configured to return an indication of a cache hit. Stated differently, if the index values compared by the comparator 210 match, and the lower order bits 0:M compared by the comparator 211 match, then the contents of the requested memory address 201 are stored in the cache 102, and the cache logic 103 may return an indication of a cache hit.
• the cache logic 103 causes a comparison of bits 0:M and index values. Doing so allows the cache logic 103 to compare fewer bits than a traditional tag compare (of bits 0:N), resulting in a faster comparison than the traditional compare.
• the cache logic 103 may be configured to fetch the requested data from a higher order memory, and populate the TCC 104 and the cache tag array 105 as described above. Similarly, if either of the initial CAM searches of the TCC 104 and the cache tag array 105 miss, the cache logic 103 may be configured to fetch the requested data from a higher order memory, and populate the TCC 104 and the cache tag array 105 as described above.
• FIG. 3 is a schematic 300 illustrating example values stored in the tag compression cache (TCC) 104, the cache tag array 105, and the data array 106, according to one aspect.
• each TCC entry 202 in the TCC 104 specifies a portion of a memory tag and is identified by a respective TCC index value 207.
• the respective index values 207 uniquely identify each TCC entry 202.
• the index value 207 is implicitly provided by the location (or address) of each entry in the TCC 104. In such aspects, therefore, the index value 207 is not expressly defined in the TCC 104, and the index values 207 are depicted in Figure 3 to facilitate explanation of such aspects.
• the tag portion of each TCC entry 202 may be the higher order address bits of a memory tag, and may include additional information, such as ASID bits, security bits, exception level bits, and the like.
• the table 301 includes example values of the cache tag array 105 and the data array 106.
• the data array 106 stores a data line 302 for each entry in the cache 102.
• the cache tag array 105 includes index values 204 and cache tags 205.
• the index values 204 in the cache tag array 105 correspond to the index value 207 corresponding to a given TCC entry 202 in the TCC 104, while each corresponding cache tag 205 stores the lower order bits of a memory tag. Therefore, as shown, multiple entries in the table 301 may store the same index values 204.
• the TCC 104 stores a single copy of the full bit pattern of higher order bits of the memory tag, reducing the amount of redundant information stored in the cache tag array 105 (namely the different index values 204, and not the lengthier higher order bits of the memory tag). Therefore, a memory tag is stored across the TCC 104 and the cache tag array 105, with the higher order bits stored in the TCC 104, and the lower order bits stored in the cache tag array 105.
• the tag portion of the TCC entry 202 in the TCC 104 may replace the index value 204 of the cache tag array 105 to produce a complete memory tag. For example, in the first row of the table 301, the bits "00" of the index 204 would be replaced by the tag bits "01... 0100" of the TCC entry 202 corresponding to the index value 207 of "00" in the TCC 104.
• the configuration depicted in Figure 3 may replace a hypothetical instruction cache with an example organization depicted in Table I:
• the hypothetical instruction cache requires 26,624 bits of storage overhead for tag storage, of the total cache storage of 32KB (or 32,768 Bytes, or 262,144 bits).
• the instruction cache tag required in this example configuration would be ⁇ PC[48: 13], ASID ⁇ .
• each TCC entry 202 would store the 21 PC[48:28] bits and the 16 bits of the ASID, for a total of 37 bits.
• the total amount of storage required for the TCC 104 would therefore be 4 entries times 37 bits, or 148 bits.
• the TCC entries 202 may be implicitly indexed by the location (or address) of each respective TCC entry 202 in the TCC 104 (for example, in the four-entry TCC 104, locations 0, 1, 2, and 3 correspond to binary index values of 00, 01, 10, and 11, respectively).
• the indexes 204 of the cache tag array 105 would store the two-bit index values, while the cache tags 205 would store the tag bits PC[27: 13], or 15 bits, for a total of 17 bits.
• the total storage in the cache tag array 105 would be 512 entries, times 17 bits, for a total of 8,704 bits.
• the total storage required across the TCC 104 and the cache tag array 105 would be 8,852 bits (148 + 8,704), which is significantly less than the 26,624 bits required by the instruction cache described in Table I.
• FIG. 4 is a flow chart illustrating a method 400 to provide cache tag compression, according to one aspect.
• the method 400 begins at step 410, where the cache logic 103 may receive a memory address of N bits in length. The memory address may be received as part of a request to determine whether the data stored at the memory address is located in the cache 102. The cache logic 103 may then probe the TCC 104 and the cache tag array 105 in parallel. More specifically, at step 415, the cache logic 103 may perform a CAM search on the cache tag array 105 using lower M bits of the memory address received at step 410.
• step 460 the cache tag array 105 may return the lower M bits of the memory address and the index value stored in the matching entry. If there is no entry in the cache tag array 105 storing a value matching the lower M bits of the memory address, there is a miss in the cache tag array 105, and the method proceeds to step 480.
• the cache logic 103 may probe the TCC 104 using the higher order bits (M+l through N) of the requested memory address. If an entry in the TCC 104 stores a value matching the higher order bits of the requested memory address, the method proceeds to step 460. In the event of a hit in the TCC 104, the TCC 104 may return the index value of the entry in the TCC 104 that matches the higher order bits of the memory address. If there is a miss in the TCC 104, the method proceeds to step 430, where the cache logic 103 may determine whether available entries exist in the TCC 104.
• the method proceeds to step 440, where the cache logic 103 may evict an entry in the TCC 104 and invalidate any corresponding entries in the cache tag array 105.
• the cache logic 103 may use any suitable eviction policy, such as evicting the least recently used entry, or evicting the entry in the TCC 104 that is associated with the fewest entries in the cache tag array 105.
• the cache logic 103 may invalidate the entries in the cache tag array 105 in any feasible manner, such as by a flash invalidation or a line by line invalidation.
• the cache logic 103 may establish a new entry in the TCC 104.
• the cache logic 103 may split the memory address into a lower set of bits 0:M and a higher set of bits M+1 :N. The cache logic 103 may then store the lower set of bits in the cache tag array 105, and store the higher set of bits in the TCC 104.
• the cache logic 103 may compare the index values returned by the TCC 104 and the cache tag array 105. The cache logic 103 may further compare the lower order bits returned by the cache tag array 105 to the lower order bits of the requested memory address. If both comparisons result in a match, the method proceeds to step 470, where the cache logic may return an indication of a cache hit. If one (or both) of the comparisons performed at step 460 do not result in a match, the method proceeds to step 480, where the cache logic 103 returns an indication of a cache miss.
• Figure 5 is a flow chart illustrating a method 500 corresponding to step 450 to establish a new entry in the tag compression cache 104, according to one aspect.
• the method 500 begins at step 510, where the processor 101 may retrieve the data associated with the memory address provided at step 410 from a higher level memory (such as a higher level cache or main memory).
• the cache logic 103 may store the higher order bits (such as bits M+1 :N) of the memory address in a first entry of the TCC 104.
• the cache logic 103 may store the index value of the first entry of the TCC 104 and the lower order bits (such as bits 0:M) of the memory address into a first entry of the tag array 105.
• the cache logic 103 may store the data retrieved at step 510 to an entry in the data array 106 corresponding to the first entry of the tag array 105.
• Figure 6 is a block diagram illustrating a computing device 601 integrating the processor 101 configured to provide cache tag compression, according to one aspect. All of the apparatuses and methods depicted in Figures 1-5 may be included in or performed by the computing device 601.
• the computing device 601 may also be connected to other computing devices via a network 630.
• the network 630 may be a telecommunications network and/or a wide area network (WAN).
• the network 630 is the Internet.
• the computing device 601 may be any device which includes a processor configured to implement cache tag compression, including, without limitation, a desktop computer, a laptop computer, a tablet computer, and a smart phone.
• the computing device 601 generally includes the processor 101 connected via a bus 620 to a memory 608, a network interface device 618, a storage 609, an input device 622, and an output device 624.
• the computing device 601 is generally under the control of an operating system (not shown). Any operating system supporting the functions disclosed herein may be used.
• the processor 101 is included to be representative of a single CPU, multiple CPUs, a single CPU having multiple processing cores, and the like.
• the network interface device 618 may be any type of network communications device allowing the computing device 601 to communicate with other computing devices via the network 630.
• the storage 609 may be a persistent storage device. Although the storage 609 is shown as a single unit, the storage 609 may be a combination of fixed and/or removable storage devices, such as fixed disc drives, solid state drives, SAN storage, NAS storage, removable memory cards or optical storage.
• the memory 608 and the storage 609 may be part of one virtual address space spanning multiple primary and secondary storage devices.
• the input device 622 may be any device for providing input to the computing device 601.
• a keyboard and/or a mouse may be used.
• the output device 624 may be any device for providing output to a user of the computing device 601.
• the output device 624 may be any conventional display screen or set of speakers.
• the output device 624 and input device 622 may be combined.
• a display screen with an integrated touch-screen may be used.
• aspects disclosed herein reduce the amount of space required to store memory tag values.
• a smaller tag array structure may be used to store the lower order bits of the memory tag. Doing so may allow such structures to be created using a smaller overall silicon die, which lowers manufacturing costs, reduces leakage current, and improves circuit timing by allowing other hardware structures to be placed closer together.
• the saved space may be spent on other microarchitectural structures.
• any suitable means capable of performing the operations such as a processor, firmware, application specific integrated circuit (ASIC), gate logic/registers, memory controller, or a cache controller.
• ASIC application specific integrated circuit
• any operations illustrated in the Figures may be performed by corresponding functional means capable of performing the operations.
• the foregoing disclosed devices and functionalities may be designed and configured into computer files (e.g. RTL, GDSII, GERBER, etc.) stored on computer readable media. Some or all such files may be provided to fabrication handlers who fabricate devices based on such files. Resulting products include semiconductor wafers that are then cut into semiconductor die and packaged into a semiconductor chip. Some or all such files may be provided to fabrication handlers who configure fabrication equipment using the design data to fabricate the devices described herein.
• computer files e.g. RTL, GDSII, GERBER, etc.
• Resulting products formed from the computer files include semiconductor wafers that are then cut into semiconductor die (e.g., the processor 101) and packaged, and may be further integrated into products including, but not limited to, mobile phones, smart phones, laptops, netbooks, tablets, ultrabooks, desktop computers, digital video recorders, set- top boxes and any other devices where integrated circuits are used.
• semiconductor die e.g., the processor 101
• Resulting products formed from the computer files include semiconductor wafers that are then cut into semiconductor die (e.g., the processor 101) and packaged, and may be further integrated into products including, but not limited to, mobile phones, smart phones, laptops, netbooks, tablets, ultrabooks, desktop computers, digital video recorders, set- top boxes and any other devices where integrated circuits are used.
• the computer files form a design structure including the circuits described above and shown in the Figures in the form of physical design layouts, schematics, a hardware-description language (e.g., Verilog, VHDL, etc.).
• design structure may be a text file or a graphical representation of a circuit as described above and shown in the Figures.
• Design process preferably synthesizes (or translates) the circuits described below into a netlist, where the netlist is, for example, a list of wires, transistors, logic gates, control circuits, I/O, models, etc. that describes the connections to other elements and circuits in an integrated circuit design and recorded on at least one of machine readable medium.
• the medium may be a storage medium such as a CD, a compact flash, other flash memory, or a hard-disk drive.
• the hardware, circuitry, and method described herein may be configured into computer files that simulate the function of the circuits described above and shown in the Figures when executed by a processor. These computer files may be used in circuitry simulation tools, schematic editors, or other software applications.
• implementations of aspects disclosed herein may also be tangibly embodied (for example, in tangible, computer-readable features of one or more computer-readable storage media as listed herein) as one or more sets of instructions executable by a machine including an array of logic elements (e.g. , a processor, microprocessor, microcontroller, or other finite state machine).
• a machine including an array of logic elements (e.g. , a processor, microprocessor, microcontroller, or other finite state machine).
• the term "computer- readable medium” may include any medium that can store or transfer information, including volatile, nonvolatile, removable, and non-removable storage media.
• Examples of a computer-readable medium include an electronic circuit, a semiconductor memory device, a ROM, a flash memory, an erasable ROM (EROM), a floppy diskette or other magnetic storage, a CD-ROM/DVD or other optical storage, a hard disk or any other medium which can be used to store the desired information, a fiber optic medium, a radio frequency (RF) link, or any other medium which can be used to carry the desired information and can be accessed.
• the computer data signal may include any signal that can propagate over a transmission medium such as electronic network channels, optical fibers, air, electromagnetic, RF links, etc.
• the code segments may be downloaded via computer networks such as the Internet or an intranet. In any case, the scope of the present disclosure should not be construed as limited by such aspects.

Landscapes

• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Engineering & Computer Science (AREA)
• General Physics & Mathematics (AREA)
• Memory System Of A Hierarchy Structure (AREA)

Priority Applications (7)

Applications Claiming Priority (2)

Publications (1)

Family

ID=55755783

Family Applications (1)

Country Status (10)

Families Citing this family (13)

Citations (1)

Family Cites Families (16)

• 2015
  • 2015-05-20 US US14/716,947 patent/US9514061B1/en active Active
• 2016
  • 2016-04-08 JP JP2017559370A patent/JP6724043B2/ja active Active
  • 2016-04-08 AU AU2016265131A patent/AU2016265131B2/en active Active
  • 2016-04-08 CN CN201680026963.6A patent/CN107735773B/zh active Active
  • 2016-04-08 EP EP16717073.7A patent/EP3298493B1/en active Active
  • 2016-04-08 KR KR1020177033433A patent/KR102138697B1/ko active Active
  • 2016-04-08 HU HUE16717073A patent/HUE049200T2/hu unknown
  • 2016-04-08 ES ES16717073T patent/ES2787002T3/es active Active
  • 2016-04-08 WO PCT/US2016/026664 patent/WO2016186747A1/en not_active Ceased
  • 2016-04-19 TW TW105112179A patent/TWI698745B/zh active

Patent Citations (1)

Also Published As

Similar Documents

Legal Events