US20050210201A1 - System and method for controlling prefetching - Google Patents

System and method for controlling prefetching Download PDF

Info

Publication number
US20050210201A1
US20050210201A1 US11/134,094 US13409405A US2005210201A1 US 20050210201 A1 US20050210201 A1 US 20050210201A1 US 13409405 A US13409405 A US 13409405A US 2005210201 A1 US2005210201 A1 US 2005210201A1
Authority
US
United States
Prior art keywords
prefetch
prefetch buffer
buffer system
programmed
pcrs
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Abandoned
Application number
US11/134,094
Inventor
Kimming So
Chengfuh Tang
Eric Tsang
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Avago Technologies International Sales Pte Ltd
Original Assignee
Kimming So
Tang Chengfuh J
Eric Tsang
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Kimming So, Tang Chengfuh J, Eric Tsang filed Critical Kimming So
Priority to US11/134,094 priority Critical patent/US20050210201A1/en
Publication of US20050210201A1 publication Critical patent/US20050210201A1/en
Assigned to BANK OF AMERICA, N.A., AS COLLATERAL AGENT reassignment BANK OF AMERICA, N.A., AS COLLATERAL AGENT PATENT SECURITY AGREEMENT Assignors: BROADCOM CORPORATION
Assigned to AVAGO TECHNOLOGIES GENERAL IP (SINGAPORE) PTE. LTD. reassignment AVAGO TECHNOLOGIES GENERAL IP (SINGAPORE) PTE. LTD. ASSIGNMENT OF ASSIGNORS INTEREST (SEE DOCUMENT FOR DETAILS). Assignors: BROADCOM CORPORATION
Assigned to BROADCOM CORPORATION reassignment BROADCOM CORPORATION TERMINATION AND RELEASE OF SECURITY INTEREST IN PATENTS Assignors: BANK OF AMERICA, N.A., AS COLLATERAL AGENT
Abandoned legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F12/00Accessing, addressing or allocating within memory systems or architectures
    • G06F12/02Addressing or allocation; Relocation
    • G06F12/08Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
    • G06F12/0802Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches
    • G06F12/0862Addressing of a memory level in which the access to the desired data or data block requires associative addressing means, e.g. caches with prefetch
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F2212/00Indexing scheme relating to accessing, addressing or allocation within memory systems or architectures
    • G06F2212/60Details of cache memory
    • G06F2212/6022Using a prefetch buffer or dedicated prefetch cache

Definitions

  • a control processor e.g., central processing unit (CPU)
  • CPU central processing unit
  • the CPU may, for example, handle interrupts, manage other functional resources and interact with users. To perform these tasks in a timely manner, the execution speed of the CPU is a substantial factor with respect to the overall system performance. Memory latency, in turn, is a substantial factor with respect to the execution speed.
  • the CPU may tend to access short streams of sequencing addresses. It is difficult to build a shared memory system that satisfies these different types of requests. Thus, the memory latency of the CPU may be long (e.g., tens of cycles) even if the memory bandwidth is high.
  • Prefetching may include, for example, loading particular data to storage close to the CPU in anticipation that the CPU may use the data in the near future.
  • coverage and accuracy of a particular prefetching scheme can vary with different programs and applications.
  • effectiveness of a particular prefetching scheme can even vary with respect to the memory region being accessed by the CPU. In fact, there are some circumstances in which a particular prefetching scheme would be more effective if it were turned off.
  • conventional prefetching schemes and controls may not be changed in real time (i.e., on the fly) to accommodate dynamic environments.
  • the present invention may provide a system that controls prefetching and may include, for example, a prefetch buffer system coupled to a processing unit and to a memory.
  • the prefetch buffer system may include, for example, a prefetch controller that is adapted to be programmable such that prefetch control features can be selected.
  • the present invention may provide a method that controls a prefetch buffer system.
  • the method may include one or more of the following: programming values into particular registers of the prefetch buffer system; selecting particular prefetch scheme features based upon the programmed values in the particular registers; and performing a particular prefetch scheme according to the selected particular prefetch scheme features.
  • the present invention may provide a method that switches between prefetching schemes.
  • the method may include one or more of the following: performing a first prefetching scheme; writing values into particular registers of a prefetch buffer system; and switching to a second prefetching scheme based on the values written into the particular registers of the prefetch buffer system.
  • the present invention may provide a system that controls a prefetch buffer system.
  • the system may include, for example, a processor coupled to one or more control registers of the prefetch buffer system.
  • software being executed by the processor may change prefetching features of the prefetch buffer system.
  • FIG. 1 shows a block diagram illustrating an embodiment of a system that caches data streams according to the present invention.
  • FIGS. 2 A-B show a flowchart illustrating an embodiment of a process that controls prefetching when a central processing unit (CPU) attempts to access a particular cache line according to the present invention.
  • CPU central processing unit
  • FIG. 3 shows an embodiment of a prefetch buffer system according to the present invention.
  • FIG. 1 shows a block diagram illustrating an embodiment of a system that controls prefetching according to the present invention.
  • the system 10 may include, for example, a central processing unit (CPU) 20 , a prefetch buffer system 70 , a system interconnect and memory controller (SIMC) 80 and a memory 40 .
  • the CPU 20 may also include, for example, a cache 60 .
  • the cache 60 may include one or more cache, for example, a level one (L1) cache and a level two (L2) cache which may store cache lines.
  • the SIMC 80 may include, for example, a memory controller 30 and other system functional units 50 .
  • the memory controller 30 may include, for example, a memory scheduler (not shown).
  • the CPU 20 may be coupled to the SIMC 80 which, in turn, may be coupled to the memory 40 via, for example, the memory controller 30 .
  • the prefetch buffer system 70 may be coupled both to the CPU 20 and to the SIMC 80 .
  • the prefetch buffer system 70 may be closer to the CPU 20 than the memory 40 and may be coupled, for example, to a bus, which may couple the CPU 20 to the SIMC 80 .
  • the prefetch buffer system 70 may be in communications with the CPU 20 and the SIMC 80 using other communication means.
  • the prefetch buffer system 70 may include, for example, a cache or a buffer.
  • the prefetch buffer system 70 may include a prefetch buffer (e.g., a spatial cache, a fully functional cache, a buffer, etc.)
  • the prefetch buffer system 70 may also include, for example, a set of controls that may be adapted to prefetch streams of data into the prefetch buffer system 70 to reduce the memory latency of the CPU 20 .
  • the set of controls may be adapted to be programmable such that particular prefetch features or schemes may be, for example, enabled, disabled or selected. In one example, when a particular prefetch scheme becomes ineffective or less effective for a particular environment, then the set of controls of the prefetch buffer system 70 may be programmed and implemented in real time (e.g., on the fly).
  • software running, for example, at least in part, in the CPU 20 may access the prefetch buffer system 70 to enable, disable or select particular prefetch features or schemes by changing values stored in the prefetch buffer system 70 .
  • the enabling, disabling or selecting may occur during, for example, the normal operation of the CPU 20 .
  • the CPU 20 may monitor the effectiveness of a particular prefetch scheme and, to improve the effectiveness of the prefetch scheme, may select a new prefetch scheme that is supported by the prefetch buffer system 70 .
  • the set of controls may also be adapted to be programmable such that the contents stored in the prefetch buffer system 70 (e.g., in a prefetch buffer of the prefetch buffer system 70 ) may be manipulated. For example, values may be written into the set of controls which invalidate one or more blocks stored in the prefetch buffer system 70 .
  • the prefetch buffer system 70 may be adapted to send one or more kinds of block requests to the memory 40 .
  • a block may include a single cache line or a plurality of cache lines.
  • the prefetch buffer system 70 may be adapted to request a missing block from the memory 40 when, for example, a CPU cache line miss (i.e., the CPU 20 was unable to find a particular cache line in its cache 60 ) also misses the prefetch buffer system 70 .
  • the prefetch buffer system 70 may also request a prefetch block from the memory 40 when, for example, a particular block is determined by the prefetch buffer system 70 to be needed by the CPU 20 ahead of the actual need of the CPU 20 .
  • a request for a prefetch block may be generated by the prefetch buffer system 70 regardless of whether or not the CPU 20 may find the missing cache line in the prefetch buffer system 70 .
  • a request for a particular prefetch block might not be generated if the particular prefetch block already resides in the prefetch buffer system 70 .
  • FIGS. 2 A-B show a flowchart illustrating an embodiment of a process that controls prefetching when the CPU 20 attempts to access a particular cache line according to the present invention.
  • the CPU 20 may be looking for a particular cache line (e.g., data in a particular cache line) of the CPU cache 60 .
  • query 100 it may be determined whether the particular cache line is presently in the CPU cache 60 . If the particular cache line is presently stored in the CPU cache 60 (i.e., if it is not a CPU cache miss), then the CPU may access (e.g., read) the particular cache line in step 110 .
  • the process may then, for example, loop back to step 90 in which the CPU 20 may look for a CPU cache line (e.g., another CPU cache line).
  • a CPU cache line e.g., another CPU cache line.
  • query 100 if the particular cache line is not presently stored in the CPU cache 60 (i.e., if it is a CPU cache miss), then, in query 105 , it may be determined whether the prefetch buffer system 70 is enabled.
  • the prefetch buffer system 70 may be programmed by the software to be in a disabled mode. If the prefetch buffer system 70 is in a disabled mode (i.e., it is not enabled), then, in step 115 , the CPU cache miss may be passed on to the memory 40 via the memory controller 30 . The memory 40 may then directly provide the missing cache line or missing block to the CPU 20 . In step 110 , the CPU may access the cache line and, in step 90 , the CPU 20 may look for another cache line as described above. If the prefetch buffer system 70 is enabled, then, in step 120 , the CPU 20 may check the prefetch buffer system 70 .
  • the prefetch buffer system 70 may send the particular cache line to the CPU 20 and possibly perform a programmed prefetch based on, for example, information relating to the cache line requested by the CPU 20 .
  • the prefetch buffer system 70 may request a prefetch block or a prefetch cache line from the memory 40 and may store the prefetch block or the prefetch cache line in, for example, a prefetch buffer of the prefetch buffer system 70 .
  • the type of prefetch scheme or the parameters related to a particular prefetch scheme may be programmed and stored in the set of controls of the prefetch buffer system 70 .
  • Software may program the prefetch buffer system 70 to perform different prefetch schemes or variations of some prefetch schemes.
  • the process may, for example, loop back to step 110 in which the CPU 20 may access the cache line received from the memory 40 or the prefetch buffer system 70 and, in step 90 , may look for a cache line (e.g., another cache line) in the CPU cache 60 .
  • the prefetch request in step 140 need not be received by the memory 40 or be completed before the start of step 110 in the loop back to step 90 .
  • the prefetch buffer system 70 may request a missing block from the memory 40 .
  • the requested missing block may include a single cache line (e.g., the particular cache line that was missed) or a plurality of cache lines including the particular cache line that was missed.
  • the prefetch buffer system 70 may receive the requested missing block and may forward the particular cache line of the requested missing block to the CPU 20 .
  • the particular cache line of the requested missing block may be directly forwarded to the CPU 20 .
  • the prefetch buffer system 70 may store the entire missing block or just the remaining data (e.g., the cache lines other than the particular cache line) of the requested missing block in the prefetch buffer system 70 (e.g., in a prefetch buffer).
  • the prefetch buffer system 70 may perform a programmed prefetch based on, for example, information relating to the cache line requested by the CPU 20 or information relating to the requested missing block.
  • the type of prefetch scheme or the parameters related to a particular prefetch scheme may be programmed and stored in the set of controls of the prefetch buffer system 70 .
  • Software may program the prefetch buffer system 70 to perform different prefetch schemes or variations of some prefetch schemes.
  • the software may program differently in different prefetching environments.
  • the prefetch request may include requesting a prefetch block which may include a single cache line or a block request including a plurality of cache lines.
  • the prefetch buffer system 70 may give priority to the missing block request and send the missing block request before the prefetch block request to the memory controller 30 .
  • the prefetch buffer system 70 may not generate a prefetch block request if the requested prefetch block already resides in the prefetch buffer system 70 .
  • FIG. 3 shows an embodiment of a prefetch buffer system 70 according to the present invention.
  • the prefetch buffer system 70 may be coupled, for example, to the memory controller 30 of the SIMC 80 .
  • the prefetch buffer system 70 may include, for example, a CPU interface 230 , a selector 240 , a read buffer 250 , a prefetch controller 260 , a selector 270 , a prefetch buffer 280 , a decision block 310 , a buffer 320 , a prefetch request buffer (PRB) 330 and a selector 340 .
  • the prefetch controller 260 may include, for example, prefetch control registers (PCRs) 350 .
  • the components may be coupled as illustrated in FIG. 3 . However, some couplings have not been illustrated.
  • the CPU interface 230 may be coupled to the CPU 20 or to a bus that is coupled to the CPU 20 .
  • the memory controller 30 may be coupled, for example, to the memory 40 and to other
  • the prefetch buffer 280 may include, for example, a data array 290 and a directory 300 (e.g., a tag array).
  • the data array 290 may be adapted to store data prefetched from the memory 40 .
  • the data array 290 may be arranged into blocks in which each block may store a single cache line or multiple cache lines.
  • the data array 290 and the directory 300 may be configured as a fully-functional cache (e.g., an L2 cache).
  • the line size of an L1 cache in the CPU 20 may be 32 bytes and the block size of the prefetch buffer 280 may be 128 bytes.
  • each block of the prefetch buffer 280 may store four cache lines.
  • the directory 300 may be adapted to store the addresses of the data.
  • Each entry in the directory 300 may include, for example, a block address and several valid bits.
  • each valid bit may correspond to each unit of cache line belonging to the same block.
  • the block may be invalidated if, for example, all of its corresponding valid bits are in an off state.
  • An invalidated block may become a candidate location for refilling prefetched data.
  • an invalidated block may be an indication of a preferred location for refilling prefetched data.
  • the prefetch controller 260 may include, for example, one or more PCRs 350 (e.g., memory-mapped registers).
  • a PCR 350 may be accessed by the CPU 20 through, for example, load and store instructions.
  • a special physical address may be allocated for each PCR 350 and the location may not fall in any cacheable memory segment of the CPU 20 .
  • the control of the operations of the prefetch buffer system 70 including, for example, the prefetch buffer 280 may be set or reset at run time (e.g., during system initialization or other run times), for example, by software through the use of a programmable prefetch controller 260 .
  • each PCR 350 will be referenced by a respective address: PCR- 0 , PCR- 1 , PCR- 2 , PCR- 3 and PCR- 4 .
  • the present invention also using a different number of PCRs. For example, there may be more or less than five PCRs with more or less than five PCR addresses.
  • the present invention may include one or more of the programmable prefetch features discussed below. The present invention also may contemplate controlling other aspects or features of the prefetch buffer system 70 that may be known to one of ordinary skill in the art, although not explicitly described herein.
  • the prefetch buffer system 70 may be disabled or enabled according to a value programmed into, for example, the first register PCR- 0 of the prefetch controller 260 . For example, if the value 0 is stored in the register PCR- 0 , then the prefetch buffer system 70 may be in a disabled mode. A CPU read may then pass through the prefetch buffer system 70 to the memory controller 30 . The requested data may be provided directly by the memory 40 to the CPU 20 via the memory controller 30 . If the value 1 is stored in the register PCR- 0 , then the prefetch buffer system 70 may be in a no-prefetch mode.
  • the prefetch buffer system 70 may only refill a block when it is missing, but may not prefetch additional blocks. In one example, if the block may include a plurality of cache lines including the missing cache line, then the prefetch buffer system 70 may only refill the block into the prefetch buffer 280 , but may not prefetch additional blocks. If the value 2 is stored in the register PCR- 0 , then the prefetch buffer system 70 may be in an active prefetch mode.
  • the prefetch buffer system 70 may employ, for example, a default prefetch scheme or other prefetch scheme.
  • the prefetch controller 260 may cause the prefetch buffer system 70 to prefetch an additional block according to a particular prefetch scheme.
  • the prefetch controller 260 may provide a selection of different prefetching schemes as well as variations of particular prefetching schemes from which the software may choose.
  • Writing a value to register PCR- 1 may indicate to the prefetch controller 260 to switch to a particular predefined prefetch scheme or to switch to a variation of a particular prefetch scheme. For example, if the value 0 is stored in the register PCR- 1 , then the prefetch scheme may be a sequential prefetch.
  • the prefetch controller may control the prefetch buffer system 70 to prefetch the next sequential block (e.g., block b+1) into the prefetch buffer 280 .
  • the prefetch scheme may perform strides or jumps. For example, if the CPU 20 accesses block b, then the prefetch controller 260 may cause the prefetch buffer system 70 to prefetch block b+4.
  • the stride number may be changed and stored in another PCR so that the prefetch scheme may easily be adapted (e.g., at run time) according to the circumstance.
  • the prefetch scheme may be a backwards sequential prefetch scheme. For example, if the CPU 20 accesses block b, then the prefetch controller 260 may cause the prefetch buffer system 70 to prefetch block b ⁇ 1. If the value 3 is stored in the register PCR- 1 , then the prefetch scheme may be a backwards prefetch scheme with a particular backwards stride or jump. The prefetch controller 260 may be made to perform other types of prefetch schemes or variations of those schemes by storing other values in the register PCR- 1 .
  • the software may manipulate the contents of the prefetch buffer 280 by manipulating the contents of the registers of the prefetch controller 260 . For example, if the value 1 is stored in the register PCR- 2 of the prefetch controller 260 (e.g., the software writes the value 1 into register PCR- 2 ), then the prefetch controller 260 may invalidate all of the blocks in the prefetch buffer 260 . If the prefetch buffer contains no modified data, then the invalidation may be accomplished in one cycle or, at the most, a very few cycles. In another example, if a particular address is stored in the register PCR- 3 , then the prefetch controller 260 may invalidate the block in the prefetch buffer 280 with a tag ID that is the same as the block tag associated with the address.
  • the software may also control the scope of a particular prefetch scheme by writing a particular value in the register PCR- 4 of the prefetch controller 260 . For example, if the value 0 is stored in the register PCR- 4 , then every block under the prefetch scheme selected according to register PCR- 1 may be prefetched. If the value 1 is stored in the register PCR- 4 , then only CPU instruction misses may be prefetched. An additional CPU signal may be employed on the CPU bus to indicate if a CPU request is for instructions or is for data. If the value 2 is stored in the register PCR- 4 , then the prefetch scheme may only be active if a CPU 20 access falls within a particular range of addresses. The range may be defined by parameters stored in other PCRs.
  • the prefetch scheme may also prefetch blocks in the non-cacheable memory segments of the CPU 20 to the prefetch buffer 280 . If the blocks of the prefetch buffer 280 are shared with other devices in the system, then the blocks may be easily invalidated with programmable control as described above.
  • the PRB 330 may be adapted to store memory requests that are waiting to be sent or that have been sent but not yet completed.
  • the memory requests may include, for example, block address information.
  • Status bits may be used to indicate, for example, a valid request, a prefetch block request, a missing block request, a request already sent to memory, etc.
  • the PRB 330 may include, for example, one or more data structures.
  • the PRB 330 may include a first data structure (e.g., a queue) that may store missing block requests and a second data structure (e.g., a stack) that may store prefetch block requests.
  • the capacity of the PRB 330 may be determined, for example, by the maximum number of CPU misses that the prefetch controller may support.
  • a data structure may include one or more last-in-first-out (LIFO) buffers, first-in-first-out (FIFO) buffers or any other type of buffers.
  • the PRB 330 may also employ a single data structure that may store both missing block requests and prefetch block requests.
  • a CPU access may cause a look up in the prefetch buffer 280 .
  • a CPU read request may be generated in response, for example, to an L1 cache miss or a non-cacheable read.
  • the CPU cache miss may result in a look up in the prefetch buffer 280 (assuming that the prefetch buffer system 70 has not been programmed into a disabled mode).
  • the prefetch buffer system 70 may be programmed such that non-cacheable read requests may be directly provided by the memory 40 .
  • the prefetch buffer 280 may stage non-cacheable data as long as the software knows how to maintain the data coherency among the devices accessing the memory 40 .
  • the look up in the prefetch buffer 280 may be passed from the CPU buffer 230 to the prefetch controller 260 .
  • the prefetch controller 260 may determine whether or not the prefetch buffer 280 contains the missing CPU cache line.
  • the prefetch controller 260 may, via the selector 270 , use address information from the directory 300 in determining whether or not the missing cache line is present in the data array 290 . If the missing cache line is present in the data array 290 , then the decision block 310 may determine that it is a prefetch buffer hit and may send the missing cache line to the CPU 20 via the buffer 320 , the selector 240 and the CPU interface 230 . In one embodiment, once the particular cache line in the data array 290 has been sent to the CPU 20 , then the particular cache line in the data array 290 may be invalidated.
  • the prefetch buffer system 70 may also generate a prefetch request in light of information concerning the missing cache line. For example, the prefetch buffer system 70 may check whether or not the next sequential block before or after the block containing the missing cache line resides in the prefetch buffer 280 . If it does not, then the prefetch buffer system 70 may generate a prefetch request (e.g., in the prefetch controller 260 or in the PRB 330 ) and store it in PRB 330 . The prefetch request may be transmitted to the memory controller 30 via the selector 340 . Priority may be given to missing requests over prefetch requests stored in the PRB 330 .
  • a prefetch request e.g., in the prefetch controller 260 or in the PRB 330
  • the prefetch buffer system 70 may then generate a missing block request in which the requested missing block may contain the missing CPU cache line.
  • the missing block request may be stored in the PRB 330 (e.g., in a queue). Since the prefetch buffer system 70 may be able to handle multiple miss requests, a particular missing block request may have to wait for its opportunity in the queue to be transmitted to the memory controller 30 via the selector 340 . In addition to the missing block request, depending upon its programming, the prefetch buffer system 70 may also generate a prefetch block request.
  • the prefetch block request may not be generated if the block requested is already residing in the prefetch buffer 280 .
  • the prefetch block request may be modified if the block requested is already residing in the prefetch buffer 280 .
  • the requested prefetch block may be the next sequential block before or after the block in the missing block request depending upon the programming of the PCRs 350 of the prefetch controller 260 .
  • the prefetch block may be part of a set of blocks (e.g., aligned blocks or other sets of blocks).
  • the memory 40 may be partitioned into sets of blocks.
  • a set of blocks may include block b and block b+1 or block b and block b ⁇ 1, in which b ⁇ 1, b and b+1 are block addresses in the memory 40 .
  • the prefetch block request may depend upon which of the blocks in a set of blocks is not the missing block.
  • the prefetch block request may be stored in the PRB 330 (e.g., in a stack). In one example, the prefetch block request may not be sent to the memory controller 30 until the respective memory block request or any memory block request has been completed.
  • the missing cache line may be sent to the CPU 20 via the read buffer 250 , the selector 240 and the CPU interface 230 .
  • the rest of the data in the requested missing block (e.g., the other cache lines in the missing block) may be refilled into the prefetch buffer 280 .
  • the entire missing block is stored in the prefetch buffer 280 , however, the cache line used by the CPU 20 (which was also sent directly to the CPU 20 ) may be invalidated.
  • the entire missing block may be received by the prefetch buffer system 70 before the missing cache line in the missing block is forwarded to the CPU 20 . Then, the prefetch buffer system 70 may remove the missing block request from the PRB 330 .
  • the entire prefetched block may be stored in the prefetch buffer 280 .
  • Information carried by the prefetch block (or a requested missing block) may be referenced with respect to information stored in the PRB 330 .
  • the respective memory request or an entry associated with the respective memory request may be used to determine where in the prefetch buffer 280 to refill the block of data.
  • the PRB 330 may include information relating to the least recently used block that may used to refill the block of data. After storing the data, the PRB entry may be removed.
  • a CPU write request may also result in a look up request in the prefetch buffer 280 .
  • a write request and write data may be sent directly to the memory controller 30 .
  • information about the write request may also be sent to prefetch buffer system 70 which may determine whether or not there is a write hit in the prefetch buffer 280 (i.e., whether the cache line being written to the memory 40 already resides in a block of the prefetch buffer 280 ). If the cache line resides in the prefetch buffer 280 then, since it may be an older version of the cache line being written to the memory 40 , it may be invalidated.
  • the prefetch buffer system 70 may invalidate the cache line or the entire block containing the cache line.
  • the PCRs 350 (e.g., PCR- 2 and PCR- 3 ) of the prefetch controller 260 may be programmed, for example, to invalidate all the blocks or particular blocks (e.g., a particular cache line or particular cache lines) in the prefetch buffer 280 . If a write miss occurs (i.e., the cache line being written does not reside in the prefetch buffer 280 ), then the request may be completed. In one embodiment, invalidation is but one programmable option. Other options may include, for example, updating the contents of the line and/or the block stored in the prefetch buffer 280 .

Abstract

Systems and methods that control prefetching are provided. In one embodiment, a system may include, for example, a prefetch buffer system coupled to a processing unit and to a memory. The prefetch buffer system may include, for example, a prefetch controller that is adapted to be programmable such that prefetch control features can be selected.

Description

    RELATED APPLICATIONS
  • This application makes reference to, claims priority to and claims benefit from U.S. Provisional Patent Application Ser. No. 60/409,256, entitled “System and Method for Controlling Prefetching,” filed on Sep. 9, 2002; U.S. Provisional Patent Application Ser. No. 60/409,240, entitled “System and Method for Caching,” filed on Sep. 9, 2002; U.S. Provisional Patent Application Ser. No. 60/409,361, entitled “System and Method for Directional Prefetching,” filed on Sep. 9, 2002.
  • INCORPORATION BY REFERENCE
  • The above-referenced United States patent applications are hereby incorporated herein by reference in their entirety.
  • BACKGROUND OF THE INVENTION
  • In many fields and applications, a control processor (e.g., central processing unit (CPU)) shares a memory with multiple devices via a memory controller. The CPU may, for example, handle interrupts, manage other functional resources and interact with users. To perform these tasks in a timely manner, the execution speed of the CPU is a substantial factor with respect to the overall system performance. Memory latency, in turn, is a substantial factor with respect to the execution speed. Unlike media processors, for example, that access memory in long data streams, the CPU may tend to access short streams of sequencing addresses. It is difficult to build a shared memory system that satisfies these different types of requests. Thus, the memory latency of the CPU may be long (e.g., tens of cycles) even if the memory bandwidth is high.
  • One solution to the memory latency problem employs the technique of prefetching. Prefetching may include, for example, loading particular data to storage close to the CPU in anticipation that the CPU may use the data in the near future. However, the coverage and accuracy of a particular prefetching scheme can vary with different programs and applications. In addition, the effectiveness of a particular prefetching scheme can even vary with respect to the memory region being accessed by the CPU. In fact, there are some circumstances in which a particular prefetching scheme would be more effective if it were turned off. However, conventional prefetching schemes and controls may not be changed in real time (i.e., on the fly) to accommodate dynamic environments.
  • Further limitations and disadvantages of conventional and traditional approaches will become apparent to one of ordinary skill in the art through comparison of such systems with some aspects of the present invention as set forth in the remainder of the present application with reference to the drawings.
  • BRIEF SUMMARY OF THE INVENTION
  • Aspects of the present invention may be found, for example, in systems and methods that control prefetching. In one embodiment, the present invention may provide a system that controls prefetching and may include, for example, a prefetch buffer system coupled to a processing unit and to a memory. The prefetch buffer system may include, for example, a prefetch controller that is adapted to be programmable such that prefetch control features can be selected.
  • In another embodiment, the present invention may provide a method that controls a prefetch buffer system. The method may include one or more of the following: programming values into particular registers of the prefetch buffer system; selecting particular prefetch scheme features based upon the programmed values in the particular registers; and performing a particular prefetch scheme according to the selected particular prefetch scheme features.
  • In yet another embodiment, the present invention may provide a method that switches between prefetching schemes. The method may include one or more of the following: performing a first prefetching scheme; writing values into particular registers of a prefetch buffer system; and switching to a second prefetching scheme based on the values written into the particular registers of the prefetch buffer system.
  • In still yet another embodiment, the present invention may provide a system that controls a prefetch buffer system. The system may include, for example, a processor coupled to one or more control registers of the prefetch buffer system. At run time, software being executed by the processor may change prefetching features of the prefetch buffer system.
  • These and other features and advantages of the present invention may be appreciated from a review of the following detailed description of the present invention, along with the accompanying figures in which like reference numerals refer to like parts throughout.
  • BRIEF DESCRIPTION OF THE DRAWINGS
  • FIG. 1 shows a block diagram illustrating an embodiment of a system that caches data streams according to the present invention.
  • FIGS. 2A-B show a flowchart illustrating an embodiment of a process that controls prefetching when a central processing unit (CPU) attempts to access a particular cache line according to the present invention.
  • FIG. 3 shows an embodiment of a prefetch buffer system according to the present invention.
  • DETAILED DESCRIPTION OF THE INVENTION
  • FIG. 1 shows a block diagram illustrating an embodiment of a system that controls prefetching according to the present invention. The system 10 may include, for example, a central processing unit (CPU) 20, a prefetch buffer system 70, a system interconnect and memory controller (SIMC) 80 and a memory 40. The CPU 20 may also include, for example, a cache 60. The cache 60 may include one or more cache, for example, a level one (L1) cache and a level two (L2) cache which may store cache lines. The SIMC 80 may include, for example, a memory controller 30 and other system functional units 50. The memory controller 30 may include, for example, a memory scheduler (not shown). The CPU 20 may be coupled to the SIMC 80 which, in turn, may be coupled to the memory 40 via, for example, the memory controller 30. The prefetch buffer system 70 may be coupled both to the CPU 20 and to the SIMC 80. In one example, the prefetch buffer system 70 may be closer to the CPU 20 than the memory 40 and may be coupled, for example, to a bus, which may couple the CPU 20 to the SIMC 80. However, the prefetch buffer system 70 may be in communications with the CPU 20 and the SIMC 80 using other communication means.
  • The prefetch buffer system 70 may include, for example, a cache or a buffer. In one embodiment, the prefetch buffer system 70 may include a prefetch buffer (e.g., a spatial cache, a fully functional cache, a buffer, etc.) The prefetch buffer system 70 may also include, for example, a set of controls that may be adapted to prefetch streams of data into the prefetch buffer system 70 to reduce the memory latency of the CPU 20. The set of controls may be adapted to be programmable such that particular prefetch features or schemes may be, for example, enabled, disabled or selected. In one example, when a particular prefetch scheme becomes ineffective or less effective for a particular environment, then the set of controls of the prefetch buffer system 70 may be programmed and implemented in real time (e.g., on the fly).
  • In one embodiment, software running, for example, at least in part, in the CPU 20 may access the prefetch buffer system 70 to enable, disable or select particular prefetch features or schemes by changing values stored in the prefetch buffer system 70. The enabling, disabling or selecting may occur during, for example, the normal operation of the CPU 20. For example, the CPU 20 may monitor the effectiveness of a particular prefetch scheme and, to improve the effectiveness of the prefetch scheme, may select a new prefetch scheme that is supported by the prefetch buffer system 70. The set of controls may also be adapted to be programmable such that the contents stored in the prefetch buffer system 70 (e.g., in a prefetch buffer of the prefetch buffer system 70) may be manipulated. For example, values may be written into the set of controls which invalidate one or more blocks stored in the prefetch buffer system 70.
  • The prefetch buffer system 70 may be adapted to send one or more kinds of block requests to the memory 40. A block may include a single cache line or a plurality of cache lines. The prefetch buffer system 70 may be adapted to request a missing block from the memory 40 when, for example, a CPU cache line miss (i.e., the CPU 20 was unable to find a particular cache line in its cache 60) also misses the prefetch buffer system 70. The prefetch buffer system 70 may also request a prefetch block from the memory 40 when, for example, a particular block is determined by the prefetch buffer system 70 to be needed by the CPU 20 ahead of the actual need of the CPU 20. A request for a prefetch block may be generated by the prefetch buffer system 70 regardless of whether or not the CPU 20 may find the missing cache line in the prefetch buffer system 70. A request for a particular prefetch block might not be generated if the particular prefetch block already resides in the prefetch buffer system 70.
  • FIGS. 2A-B show a flowchart illustrating an embodiment of a process that controls prefetching when the CPU 20 attempts to access a particular cache line according to the present invention. In step 90, the CPU 20 may be looking for a particular cache line (e.g., data in a particular cache line) of the CPU cache 60. In query 100, it may be determined whether the particular cache line is presently in the CPU cache 60. If the particular cache line is presently stored in the CPU cache 60 (i.e., if it is not a CPU cache miss), then the CPU may access (e.g., read) the particular cache line in step 110. The process may then, for example, loop back to step 90 in which the CPU 20 may look for a CPU cache line (e.g., another CPU cache line). In query 100, if the particular cache line is not presently stored in the CPU cache 60 (i.e., if it is a CPU cache miss), then, in query 105, it may be determined whether the prefetch buffer system 70 is enabled.
  • The prefetch buffer system 70 may be programmed by the software to be in a disabled mode. If the prefetch buffer system 70 is in a disabled mode (i.e., it is not enabled), then, in step 115, the CPU cache miss may be passed on to the memory 40 via the memory controller 30. The memory 40 may then directly provide the missing cache line or missing block to the CPU 20. In step 110, the CPU may access the cache line and, in step 90, the CPU 20 may look for another cache line as described above. If the prefetch buffer system 70 is enabled, then, in step 120, the CPU 20 may check the prefetch buffer system 70.
  • In query 130, it may be determined whether the particular cache line is residing in the prefetch buffer system 70. If the particular cache line is presently stored in the prefetch buffer system 70 (i.e., if it is not a prefetch buffer miss), then, in step 140, the prefetch buffer system 70 may send the particular cache line to the CPU 20 and possibly perform a programmed prefetch based on, for example, information relating to the cache line requested by the CPU 20. In a prefetch, the prefetch buffer system 70 may request a prefetch block or a prefetch cache line from the memory 40 and may store the prefetch block or the prefetch cache line in, for example, a prefetch buffer of the prefetch buffer system 70. The type of prefetch scheme or the parameters related to a particular prefetch scheme may be programmed and stored in the set of controls of the prefetch buffer system 70. Software may program the prefetch buffer system 70 to perform different prefetch schemes or variations of some prefetch schemes. The process may, for example, loop back to step 110 in which the CPU 20 may access the cache line received from the memory 40 or the prefetch buffer system 70 and, in step 90, may look for a cache line (e.g., another cache line) in the CPU cache 60. The prefetch request in step 140 need not be received by the memory 40 or be completed before the start of step 110 in the loop back to step 90.
  • In query 130, if the particular cache line is not presently stored in the prefetch buffer system 70 (i.e., if it is a prefetch buffer miss), then, in step 150, the prefetch buffer system 70 may request a missing block from the memory 40. The requested missing block may include a single cache line (e.g., the particular cache line that was missed) or a plurality of cache lines including the particular cache line that was missed. In one example, in step 160, the prefetch buffer system 70 may receive the requested missing block and may forward the particular cache line of the requested missing block to the CPU 20. In another example, the particular cache line of the requested missing block may be directly forwarded to the CPU 20. In step 170, the prefetch buffer system 70 may store the entire missing block or just the remaining data (e.g., the cache lines other than the particular cache line) of the requested missing block in the prefetch buffer system 70 (e.g., in a prefetch buffer). In step 180, the prefetch buffer system 70 may perform a programmed prefetch based on, for example, information relating to the cache line requested by the CPU 20 or information relating to the requested missing block. The type of prefetch scheme or the parameters related to a particular prefetch scheme may be programmed and stored in the set of controls of the prefetch buffer system 70. Software may program the prefetch buffer system 70 to perform different prefetch schemes or variations of some prefetch schemes. The software may program differently in different prefetching environments. The prefetch request may include requesting a prefetch block which may include a single cache line or a block request including a plurality of cache lines. In one example, when the prefetch buffer system 70 has generated a missing block request and a prefetch block request, the prefetch buffer system 70 may give priority to the missing block request and send the missing block request before the prefetch block request to the memory controller 30. In one example, the prefetch buffer system 70 may not generate a prefetch block request if the requested prefetch block already resides in the prefetch buffer system 70.
  • FIG. 3 shows an embodiment of a prefetch buffer system 70 according to the present invention. The prefetch buffer system 70 may be coupled, for example, to the memory controller 30 of the SIMC 80. The prefetch buffer system 70 may include, for example, a CPU interface 230, a selector 240, a read buffer 250, a prefetch controller 260, a selector 270, a prefetch buffer 280, a decision block 310, a buffer 320, a prefetch request buffer (PRB) 330 and a selector 340. The prefetch controller 260 may include, for example, prefetch control registers (PCRs) 350. The components may be coupled as illustrated in FIG. 3. However, some couplings have not been illustrated. For example, the CPU interface 230 may be coupled to the CPU 20 or to a bus that is coupled to the CPU 20. The memory controller 30 may be coupled, for example, to the memory 40 and to other functional devices 50.
  • The prefetch buffer 280 may include, for example, a data array 290 and a directory 300 (e.g., a tag array). The data array 290 may be adapted to store data prefetched from the memory 40. The data array 290 may be arranged into blocks in which each block may store a single cache line or multiple cache lines. In one example, the data array 290 and the directory 300 may be configured as a fully-functional cache (e.g., an L2 cache). In another example, the line size of an L1 cache in the CPU 20 may be 32 bytes and the block size of the prefetch buffer 280 may be 128 bytes. Thus, each block of the prefetch buffer 280 may store four cache lines. The directory 300 may be adapted to store the addresses of the data. Each entry in the directory 300 may include, for example, a block address and several valid bits. In one example, each valid bit may correspond to each unit of cache line belonging to the same block. The block may be invalidated if, for example, all of its corresponding valid bits are in an off state. An invalidated block may become a candidate location for refilling prefetched data. In one example, an invalidated block may be an indication of a preferred location for refilling prefetched data.
  • The prefetch controller 260 may include, for example, one or more PCRs 350 (e.g., memory-mapped registers). A PCR 350 may be accessed by the CPU 20 through, for example, load and store instructions. A special physical address may be allocated for each PCR 350 and the location may not fall in any cacheable memory segment of the CPU 20.
  • The control of the operations of the prefetch buffer system 70 including, for example, the prefetch buffer 280 may be set or reset at run time (e.g., during system initialization or other run times), for example, by software through the use of a programmable prefetch controller 260. For clarity, each PCR 350 will be referenced by a respective address: PCR-0, PCR-1, PCR-2, PCR-3 and PCR-4. Although discussed with respect to the below-listed five PCRs, the present invention also using a different number of PCRs. For example, there may be more or less than five PCRs with more or less than five PCR addresses. In addition, the present invention may include one or more of the programmable prefetch features discussed below. The present invention also may contemplate controlling other aspects or features of the prefetch buffer system 70 that may be known to one of ordinary skill in the art, although not explicitly described herein.
  • The prefetch buffer system 70 may be disabled or enabled according to a value programmed into, for example, the first register PCR-0 of the prefetch controller 260. For example, if the value 0 is stored in the register PCR-0, then the prefetch buffer system 70 may be in a disabled mode. A CPU read may then pass through the prefetch buffer system 70 to the memory controller 30. The requested data may be provided directly by the memory 40 to the CPU 20 via the memory controller 30. If the value 1 is stored in the register PCR-0, then the prefetch buffer system 70 may be in a no-prefetch mode. In the no-prefetch mode, the prefetch buffer system 70 may only refill a block when it is missing, but may not prefetch additional blocks. In one example, if the block may include a plurality of cache lines including the missing cache line, then the prefetch buffer system 70 may only refill the block into the prefetch buffer 280, but may not prefetch additional blocks. If the value 2 is stored in the register PCR-0, then the prefetch buffer system 70 may be in an active prefetch mode. The prefetch buffer system 70 may employ, for example, a default prefetch scheme or other prefetch scheme. For example, if, due to a CPU 20 request to the prefetch buffer system 70, a missing block is refilled into the prefetch buffer 280, then the prefetch controller 260 may cause the prefetch buffer system 70 to prefetch an additional block according to a particular prefetch scheme.
  • A second register PCR-1 of the prefetch controller 260 may be used in selecting a particular type of prefetching scheme if the prefetch buffer system 70 is in active prefetch mode (i.e., PCR-0=2). The prefetch controller 260 may provide a selection of different prefetching schemes as well as variations of particular prefetching schemes from which the software may choose. Writing a value to register PCR-1 may indicate to the prefetch controller 260 to switch to a particular predefined prefetch scheme or to switch to a variation of a particular prefetch scheme. For example, if the value 0 is stored in the register PCR-1, then the prefetch scheme may be a sequential prefetch. If the CPU 20 requests block b such that block b may be brought into the prefetch buffer 280, then the prefetch controller may control the prefetch buffer system 70 to prefetch the next sequential block (e.g., block b+1) into the prefetch buffer 280. If the value 1 is stored in the register PCR-1, then the prefetch scheme may perform strides or jumps. For example, if the CPU 20 accesses block b, then the prefetch controller 260 may cause the prefetch buffer system 70 to prefetch block b+4. The stride number may be changed and stored in another PCR so that the prefetch scheme may easily be adapted (e.g., at run time) according to the circumstance. If the value 2 is stored in the register PCR-1, then the prefetch scheme may be a backwards sequential prefetch scheme. For example, if the CPU 20 accesses block b, then the prefetch controller 260 may cause the prefetch buffer system 70 to prefetch block b−1. If the value 3 is stored in the register PCR-1, then the prefetch scheme may be a backwards prefetch scheme with a particular backwards stride or jump. The prefetch controller 260 may be made to perform other types of prefetch schemes or variations of those schemes by storing other values in the register PCR-1.
  • The software may manipulate the contents of the prefetch buffer 280 by manipulating the contents of the registers of the prefetch controller 260. For example, if the value 1 is stored in the register PCR-2 of the prefetch controller 260 (e.g., the software writes the value 1 into register PCR-2), then the prefetch controller 260 may invalidate all of the blocks in the prefetch buffer 260. If the prefetch buffer contains no modified data, then the invalidation may be accomplished in one cycle or, at the most, a very few cycles. In another example, if a particular address is stored in the register PCR-3, then the prefetch controller 260 may invalidate the block in the prefetch buffer 280 with a tag ID that is the same as the block tag associated with the address.
  • The software may also control the scope of a particular prefetch scheme by writing a particular value in the register PCR-4 of the prefetch controller 260. For example, if the value 0 is stored in the register PCR-4, then every block under the prefetch scheme selected according to register PCR-1 may be prefetched. If the value 1 is stored in the register PCR-4, then only CPU instruction misses may be prefetched. An additional CPU signal may be employed on the CPU bus to indicate if a CPU request is for instructions or is for data. If the value 2 is stored in the register PCR-4, then the prefetch scheme may only be active if a CPU 20 access falls within a particular range of addresses. The range may be defined by parameters stored in other PCRs. If the value 3 is stored in the register PCR-4, then, in addition to the prefetches defined under PCR-4=0, the prefetch scheme may also prefetch blocks in the non-cacheable memory segments of the CPU 20 to the prefetch buffer 280. If the blocks of the prefetch buffer 280 are shared with other devices in the system, then the blocks may be easily invalidated with programmable control as described above.
  • The PRB 330 may be adapted to store memory requests that are waiting to be sent or that have been sent but not yet completed. The memory requests may include, for example, block address information. Status bits may be used to indicate, for example, a valid request, a prefetch block request, a missing block request, a request already sent to memory, etc. The PRB 330 may include, for example, one or more data structures. In one embodiment, the PRB 330 may include a first data structure (e.g., a queue) that may store missing block requests and a second data structure (e.g., a stack) that may store prefetch block requests. The capacity of the PRB 330 may be determined, for example, by the maximum number of CPU misses that the prefetch controller may support. A data structure may include one or more last-in-first-out (LIFO) buffers, first-in-first-out (FIFO) buffers or any other type of buffers. The PRB 330 may also employ a single data structure that may store both missing block requests and prefetch block requests.
  • In operation, a CPU access may cause a look up in the prefetch buffer 280. A CPU read request may be generated in response, for example, to an L1 cache miss or a non-cacheable read. In one example, the CPU cache miss may result in a look up in the prefetch buffer 280 (assuming that the prefetch buffer system 70 has not been programmed into a disabled mode). The prefetch buffer system 70 may be programmed such that non-cacheable read requests may be directly provided by the memory 40. In another example, the prefetch buffer 280 may stage non-cacheable data as long as the software knows how to maintain the data coherency among the devices accessing the memory 40.
  • The look up in the prefetch buffer 280 may be passed from the CPU buffer 230 to the prefetch controller 260. The prefetch controller 260 may determine whether or not the prefetch buffer 280 contains the missing CPU cache line. The prefetch controller 260 may, via the selector 270, use address information from the directory 300 in determining whether or not the missing cache line is present in the data array 290. If the missing cache line is present in the data array 290, then the decision block 310 may determine that it is a prefetch buffer hit and may send the missing cache line to the CPU 20 via the buffer 320, the selector 240 and the CPU interface 230. In one embodiment, once the particular cache line in the data array 290 has been sent to the CPU 20, then the particular cache line in the data array 290 may be invalidated.
  • In addition to sending the missing CPU cache line to the CPU 20, depending upon its programming, the prefetch buffer system 70 may also generate a prefetch request in light of information concerning the missing cache line. For example, the prefetch buffer system 70 may check whether or not the next sequential block before or after the block containing the missing cache line resides in the prefetch buffer 280. If it does not, then the prefetch buffer system 70 may generate a prefetch request (e.g., in the prefetch controller 260 or in the PRB 330) and store it in PRB 330. The prefetch request may be transmitted to the memory controller 30 via the selector 340. Priority may be given to missing requests over prefetch requests stored in the PRB 330.
  • If the look up in the prefetch buffer 280 shows that the missing cache line is not in the prefetch buffer 280, then a prefetch buffer miss may have occurred. The prefetch buffer system 70 may then generate a missing block request in which the requested missing block may contain the missing CPU cache line. The missing block request may be stored in the PRB 330 (e.g., in a queue). Since the prefetch buffer system 70 may be able to handle multiple miss requests, a particular missing block request may have to wait for its opportunity in the queue to be transmitted to the memory controller 30 via the selector 340. In addition to the missing block request, depending upon its programming, the prefetch buffer system 70 may also generate a prefetch block request. The prefetch block request may not be generated if the block requested is already residing in the prefetch buffer 280. Alternatively, the prefetch block request may be modified if the block requested is already residing in the prefetch buffer 280. The requested prefetch block may be the next sequential block before or after the block in the missing block request depending upon the programming of the PCRs 350 of the prefetch controller 260. In one example, the prefetch block may be part of a set of blocks (e.g., aligned blocks or other sets of blocks). The memory 40 may be partitioned into sets of blocks. For example, a set of blocks may include block b and block b+1 or block b and block b−1, in which b−1, b and b+1 are block addresses in the memory 40. Thus, the prefetch block request may depend upon which of the blocks in a set of blocks is not the missing block. The prefetch block request may be stored in the PRB 330 (e.g., in a stack). In one example, the prefetch block request may not be sent to the memory controller 30 until the respective memory block request or any memory block request has been completed.
  • When the memory controller 30 fetches the requested missing block, the missing cache line may be sent to the CPU 20 via the read buffer 250, the selector 240 and the CPU interface 230. The rest of the data in the requested missing block (e.g., the other cache lines in the missing block) may be refilled into the prefetch buffer 280. In one example, the entire missing block is stored in the prefetch buffer 280, however, the cache line used by the CPU 20 (which was also sent directly to the CPU 20) may be invalidated. In another example, the entire missing block may be received by the prefetch buffer system 70 before the missing cache line in the missing block is forwarded to the CPU 20. Then, the prefetch buffer system 70 may remove the missing block request from the PRB 330.
  • When a prefetched block is received by the prefetch buffer system 70 from the memory 40, the entire prefetched block may be stored in the prefetch buffer 280. Information carried by the prefetch block (or a requested missing block) may be referenced with respect to information stored in the PRB 330. The respective memory request or an entry associated with the respective memory request may be used to determine where in the prefetch buffer 280 to refill the block of data. For example, the PRB 330 may include information relating to the least recently used block that may used to refill the block of data. After storing the data, the PRB entry may be removed.
  • A CPU write request may also result in a look up request in the prefetch buffer 280. During a write request, a write request and write data may be sent directly to the memory controller 30. However, information about the write request may also be sent to prefetch buffer system 70 which may determine whether or not there is a write hit in the prefetch buffer 280 (i.e., whether the cache line being written to the memory 40 already resides in a block of the prefetch buffer 280). If the cache line resides in the prefetch buffer 280 then, since it may be an older version of the cache line being written to the memory 40, it may be invalidated. The prefetch buffer system 70 may invalidate the cache line or the entire block containing the cache line. The PCRs 350 (e.g., PCR-2 and PCR-3) of the prefetch controller 260 may be programmed, for example, to invalidate all the blocks or particular blocks (e.g., a particular cache line or particular cache lines) in the prefetch buffer 280. If a write miss occurs (i.e., the cache line being written does not reside in the prefetch buffer 280), then the request may be completed. In one embodiment, invalidation is but one programmable option. Other options may include, for example, updating the contents of the line and/or the block stored in the prefetch buffer 280.
  • While the present invention has been described with reference to certain embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the scope of the present invention. In addition, many modifications may be made to adapt a particular situation or material to the teachings of the present invention without departing from its scope. Therefore, it is intended that the present invention not be limited to the particular embodiment disclosed, but that the present invention will include all embodiments falling within the scope of the appended claims.

Claims (22)

1-29. (canceled)
30. A system for controlling prefetching, comprising:
a prefetch buffer system adapted to be communicatively coupled to a processing unit and to a memory, wherein the prefetch buffer system comprises a prefetch controller that is adapted to be programmable such that prefetch control features may be independently specified without specifying other prefetch control features.
31. The system of claim 30, wherein:
the prefetch controller comprises a plurality of prefetch control registers (PCRs);
the PCRs are adapted to be programmed with values relating to prefetch control features; and
there is a one-to-one correspondence between the PCR, or portion thereof, and a prefetch control feature.
32. The system of claim 31, wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to operate in a disabled state.
33. The system of claim 31, wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to operate in a no-prefetch state.
34. The system of claim 31, wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to operate in an active prefetch state.
35. The system of claim 31, wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to operate in accordance with a sequential prefetch scheme.
36. The system of claim 31, wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to operate in accordance with a striding prefetch scheme.
37. The system of claim 31, wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to operate in accordance with a backward prefetch scheme.
38. The system of claim 31, wherein:
the prefetch buffer system comprises a prefetch buffer, and
a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to manipulate contents stored in the prefetch buffer.
39. The system of claim 31, wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to narrow a scope of a programmed prefetch scheme.
40. The system of claim 31, wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to invalidate information stored in a prefetch buffer in response to a write command from the processing unit.
41. The system of claim 30, wherein the prefetch controller is programmed by the processing unit as a function of prefetching effectiveness monitored during run-time.
42. The system of claim 30, wherein the prefetch controller is adapted to switch between predefined selectable prefetch schemes in accordance with programming of the prefetch controller.
43. A method for controlling a prefetch buffer system, comprising:
programming a particular value into a particular register of the prefetch buffer system, wherein there is a one-to-one mapping between the particular register, or a portion thereof, and a prefetch control feature;
selecting a particular prefetch scheme feature based upon the programmed particular value in the particular register; and
performing a particular prefetch scheme characterized, at least in part, by the selected particular prefetch scheme feature.
44. The method of claim 43, further comprising manipulating contents of a prefetch buffer of the prefetch buffer system in a predetermined manner associated with the programmed particular value and the particular register.
45. The method of claim 44, wherein manipulating contents of a prefetch buffer of the prefetch buffer system in a predetermined manner associated with the programmed particular value and the particular register comprises invalidating one or more blocks stored in the prefetch buffer system based, at least in part, upon the programmed particular value and the particular register.
46. The method of claim 43, wherein programming a particular value into a particular register comprises programming the particular value into the particular register in response to prefetching effectiveness monitored during run-time.
47. A prefetch buffer system comprising:
a prefetch buffer; and
a prefetch controller comprising prefetch control registers (PCRs), wherein a particular one of the PCRs, or a portion thereof, may be programmed with a particular value, thereby causing the prefetch buffer system to manipulate contents stored in the prefetch buffer in a predetermined manner associated with the particular value.
48. The prefetch buffer system of claim 47, wherein the PCRs are adapted to be programmable such that the prefetch control features can be independently specified without specifying other prefetch control features.
49. A method for controlling prefetching, the method comprising:
programming a particular value into a particular register of a prefetch buffer system; and
manipulating contents of a prefetch buffer of the prefetch buffer system in a predetermined manner associated with the programmed particular value and the particular register.
50. The method of claim 49, wherein there is a one-to-one mapping between the particular register, or a portion thereof, and a prefetch control feature, and further comprising selecting a particular prefetch feature based, at least in part, upon the programmed particular value in the particular register.
US11/134,094 2002-09-09 2005-05-20 System and method for controlling prefetching Abandoned US20050210201A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US11/134,094 US20050210201A1 (en) 2002-09-09 2005-05-20 System and method for controlling prefetching

Applications Claiming Priority (5)

Application Number Priority Date Filing Date Title
US40936102P 2002-09-09 2002-09-09
US40925602P 2002-09-09 2002-09-09
US40924002P 2002-09-09 2002-09-09
US10/294,539 US6957306B2 (en) 2002-09-09 2002-11-14 System and method for controlling prefetching
US11/134,094 US20050210201A1 (en) 2002-09-09 2005-05-20 System and method for controlling prefetching

Related Parent Applications (1)

Application Number Title Priority Date Filing Date
US10/294,539 Continuation US6957306B2 (en) 2002-09-09 2002-11-14 System and method for controlling prefetching

Publications (1)

Publication Number Publication Date
US20050210201A1 true US20050210201A1 (en) 2005-09-22

Family

ID=43706196

Family Applications (6)

Application Number Title Priority Date Filing Date
US10/294,091 Expired - Fee Related US7167954B2 (en) 2002-09-09 2002-11-14 System and method for caching
US10/294,415 Expired - Fee Related US6931494B2 (en) 2002-09-09 2002-11-14 System and method for directional prefetching
US10/294,539 Expired - Fee Related US6957306B2 (en) 2002-09-09 2002-11-14 System and method for controlling prefetching
US11/101,768 Expired - Fee Related US7627720B2 (en) 2002-09-09 2005-04-08 System and method for directional prefetching
US11/128,831 Expired - Fee Related US7711906B2 (en) 2002-09-09 2005-05-13 System and method for caching
US11/134,094 Abandoned US20050210201A1 (en) 2002-09-09 2005-05-20 System and method for controlling prefetching

Family Applications Before (5)

Application Number Title Priority Date Filing Date
US10/294,091 Expired - Fee Related US7167954B2 (en) 2002-09-09 2002-11-14 System and method for caching
US10/294,415 Expired - Fee Related US6931494B2 (en) 2002-09-09 2002-11-14 System and method for directional prefetching
US10/294,539 Expired - Fee Related US6957306B2 (en) 2002-09-09 2002-11-14 System and method for controlling prefetching
US11/101,768 Expired - Fee Related US7627720B2 (en) 2002-09-09 2005-04-08 System and method for directional prefetching
US11/128,831 Expired - Fee Related US7711906B2 (en) 2002-09-09 2005-05-13 System and method for caching

Country Status (1)

Country Link
US (6) US7167954B2 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7143251B1 (en) * 2003-06-30 2006-11-28 Data Domain, Inc. Data storage using identifiers
US20070226462A1 (en) * 2006-03-21 2007-09-27 Freescale Semiconductor, Inc. Data processor having dynamic control of instruction prefetch buffer depth and method therefor
US20080229072A1 (en) * 2007-03-14 2008-09-18 Fujitsu Limited Prefetch processing apparatus, prefetch processing method, storage medium storing prefetch processing program
US20090119488A1 (en) * 2006-06-15 2009-05-07 Sudarshan Kadambi Prefetch Unit

Families Citing this family (71)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7266725B2 (en) 2001-09-03 2007-09-04 Pact Xpp Technologies Ag Method for debugging reconfigurable architectures
DE19654595A1 (en) 1996-12-20 1998-07-02 Pact Inf Tech Gmbh I0 and memory bus system for DFPs as well as building blocks with two- or multi-dimensional programmable cell structures
US6542998B1 (en) 1997-02-08 2003-04-01 Pact Gmbh Method of self-synchronization of configurable elements of a programmable module
US8686549B2 (en) 2001-09-03 2014-04-01 Martin Vorbach Reconfigurable elements
DE19861088A1 (en) 1997-12-22 2000-02-10 Pact Inf Tech Gmbh Repairing integrated circuits by replacing subassemblies with substitutes
US8230411B1 (en) 1999-06-10 2012-07-24 Martin Vorbach Method for interleaving a program over a plurality of cells
EP2226732A3 (en) 2000-06-13 2016-04-06 PACT XPP Technologies AG Cache hierarchy for a multicore processor
US8058899B2 (en) 2000-10-06 2011-11-15 Martin Vorbach Logic cell array and bus system
US9037807B2 (en) 2001-03-05 2015-05-19 Pact Xpp Technologies Ag Processor arrangement on a chip including data processing, memory, and interface elements
US7844796B2 (en) 2001-03-05 2010-11-30 Martin Vorbach Data processing device and method
US7444531B2 (en) 2001-03-05 2008-10-28 Pact Xpp Technologies Ag Methods and devices for treating and processing data
US7996827B2 (en) 2001-08-16 2011-08-09 Martin Vorbach Method for the translation of programs for reconfigurable architectures
US7434191B2 (en) 2001-09-03 2008-10-07 Pact Xpp Technologies Ag Router
US8686475B2 (en) 2001-09-19 2014-04-01 Pact Xpp Technologies Ag Reconfigurable elements
EP1304619A1 (en) * 2001-10-22 2003-04-23 STMicroelectronics Limited Cache memory operation
AU2003208266A1 (en) 2002-01-19 2003-07-30 Pact Xpp Technologies Ag Reconfigurable processor
US8914590B2 (en) 2002-08-07 2014-12-16 Pact Xpp Technologies Ag Data processing method and device
US20110161977A1 (en) * 2002-03-21 2011-06-30 Martin Vorbach Method and device for data processing
AU2003286131A1 (en) 2002-08-07 2004-03-19 Pact Xpp Technologies Ag Method and device for processing data
US7394284B2 (en) 2002-09-06 2008-07-01 Pact Xpp Technologies Ag Reconfigurable sequencer structure
US7167954B2 (en) * 2002-09-09 2007-01-23 Broadcom Corporation System and method for caching
US7165146B2 (en) * 2003-01-28 2007-01-16 Sun Microsystems, Inc. Multiprocessing computer system employing capacity prefetching
US7600058B1 (en) * 2003-06-26 2009-10-06 Nvidia Corporation Bypass method for efficient DMA disk I/O
US7051159B2 (en) * 2003-06-30 2006-05-23 International Business Machines Corporation Method and system for cache data fetch operations
US7228387B2 (en) * 2003-06-30 2007-06-05 Intel Corporation Apparatus and method for an adaptive multiple line prefetcher
US20050086435A1 (en) * 2003-09-09 2005-04-21 Seiko Epson Corporation Cache memory controlling apparatus, information processing apparatus and method for control of cache memory
US8683132B1 (en) 2003-09-29 2014-03-25 Nvidia Corporation Memory controller for sequentially prefetching data for a processor of a computer system
US8005794B2 (en) * 2003-10-31 2011-08-23 Oracle America, Inc. Mechanism for data aggregation in a tracing framework
US8356142B1 (en) 2003-11-12 2013-01-15 Nvidia Corporation Memory controller for non-sequentially prefetching data for a processor of a computer system
US8700808B2 (en) * 2003-12-01 2014-04-15 Nvidia Corporation Hardware support system for accelerated disk I/O
US7383418B2 (en) * 2004-09-01 2008-06-03 Intel Corporation Method and apparatus for prefetching data to a lower level cache memory
CN101031970B (en) * 2004-09-28 2013-03-27 皇家飞利浦电子股份有限公司 Method and device for storing data on a record medium and for transferring information
US8356143B1 (en) * 2004-10-22 2013-01-15 NVIDIA Corporatin Prefetch mechanism for bus master memory access
US7395375B2 (en) * 2004-11-08 2008-07-01 International Business Machines Corporation Prefetch miss indicator for cache coherence directory misses on external caches
US8122193B2 (en) * 2004-12-21 2012-02-21 Samsung Electronics Co., Ltd. Storage device and user device including the same
EP1974265A1 (en) 2006-01-18 2008-10-01 PACT XPP Technologies AG Hardware definition method
US7685255B2 (en) * 2006-04-24 2010-03-23 Blue Coat Systems, Inc. System and method for prefetching uncacheable embedded objects
US7934058B2 (en) * 2006-12-14 2011-04-26 Microsoft Corporation Predictive caching of assets to improve level load time on a game console
US9208095B2 (en) * 2006-12-15 2015-12-08 Microchip Technology Incorporated Configurable cache for a microprocessor
US7877537B2 (en) * 2006-12-15 2011-01-25 Microchip Technology Incorporated Configurable cache for a microprocessor
US7966457B2 (en) * 2006-12-15 2011-06-21 Microchip Technology Incorporated Configurable cache for a microprocessor
US7702888B2 (en) * 2007-02-28 2010-04-20 Globalfoundries Inc. Branch predictor directed prefetch
US8051250B2 (en) * 2007-03-14 2011-11-01 Hewlett-Packard Development Company, L.P. Systems and methods for pushing data
US20090112975A1 (en) * 2007-10-31 2009-04-30 Microsoft Corporation Pre-fetching in distributed computing environments
JP2011503733A (en) * 2007-11-17 2011-01-27 トーマス リヒター Reconfigurable floating point level and bit level data processing unit
WO2009068014A2 (en) * 2007-11-28 2009-06-04 Pact Xpp Technologies Ag On data processing
EP2235627A1 (en) * 2007-12-07 2010-10-06 Krass, Maren Using function calls as compiler directives
US9311085B2 (en) * 2007-12-30 2016-04-12 Intel Corporation Compiler assisted low power and high performance load handling based on load types
US8458170B2 (en) * 2008-06-30 2013-06-04 Yahoo! Inc. Prefetching data for document ranking
US8316187B2 (en) * 2008-07-08 2012-11-20 International Business Machines Corporation Cache memory including a predict buffer
US8356128B2 (en) * 2008-09-16 2013-01-15 Nvidia Corporation Method and system of reducing latencies associated with resource allocation by using multiple arbiters
US8370552B2 (en) * 2008-10-14 2013-02-05 Nvidia Corporation Priority based bus arbiters avoiding deadlock and starvation on buses that support retrying of transactions
US8364901B2 (en) * 2009-02-13 2013-01-29 Micron Technology, Inc. Memory prefetch systems and methods
US8918588B2 (en) * 2009-04-07 2014-12-23 International Business Machines Corporation Maintaining a cache of blocks from a plurality of data streams
US8698823B2 (en) 2009-04-08 2014-04-15 Nvidia Corporation System and method for deadlock-free pipelining
US8892708B2 (en) * 2011-07-11 2014-11-18 Cisco Technology, Inc. Placement of service delivery locations of a distributed computing service based on logical topology
US8886752B2 (en) * 2011-11-21 2014-11-11 Sony Computer Entertainment America System and method for optimizing transfers of downloadable content
KR102069273B1 (en) * 2013-03-11 2020-01-22 삼성전자주식회사 System on chip and operating method thereof
US9569385B2 (en) 2013-09-09 2017-02-14 Nvidia Corporation Memory transaction ordering
US9645934B2 (en) * 2013-09-13 2017-05-09 Samsung Electronics Co., Ltd. System-on-chip and address translation method thereof using a translation lookaside buffer and a prefetch buffer
JP2015176245A (en) 2014-03-13 2015-10-05 株式会社東芝 Information processing apparatus and data structure
JP6252348B2 (en) * 2014-05-14 2017-12-27 富士通株式会社 Arithmetic processing device and control method of arithmetic processing device
JP2016057763A (en) * 2014-09-08 2016-04-21 株式会社東芝 Cache device and processor
US9684602B2 (en) 2015-03-11 2017-06-20 Kabushiki Kaisha Toshiba Memory access control device, cache memory and semiconductor device
US9792224B2 (en) * 2015-10-23 2017-10-17 Intel Corporation Reducing latency by persisting data relationships in relation to corresponding data in persistent memory
US10073775B2 (en) * 2016-04-01 2018-09-11 Intel Corporation Apparatus and method for triggered prefetching to improve I/O and producer-consumer workload efficiency
US10621095B2 (en) 2016-07-20 2020-04-14 International Business Machines Corporation Processing data based on cache residency
US10452395B2 (en) 2016-07-20 2019-10-22 International Business Machines Corporation Instruction to query cache residency
US10169239B2 (en) * 2016-07-20 2019-01-01 International Business Machines Corporation Managing a prefetch queue based on priority indications of prefetch requests
US10521350B2 (en) 2016-07-20 2019-12-31 International Business Machines Corporation Determining the effectiveness of prefetch instructions
US11099995B2 (en) 2018-03-28 2021-08-24 Intel Corporation Techniques for prefetching data to a first level of memory of a hierarchical arrangement of memory

Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5721865A (en) * 1995-01-20 1998-02-24 Hitachi, Ltd. Information processing apparatus with prefetch control for prefetching data structure from memory through cache memory
US5784711A (en) * 1990-05-18 1998-07-21 Philips Electronics North America Corporation Data cache prefetching under control of instruction cache
US5802569A (en) * 1996-04-22 1998-09-01 International Business Machines Corp. Computer system having cache prefetching amount based on CPU request types
US5829042A (en) * 1996-02-15 1998-10-27 Hewlett-Packard Company Prefetch operation for network peripheral device having shared memory
US5941981A (en) * 1997-11-03 1999-08-24 Advanced Micro Devices, Inc. System for using a data history table to select among multiple data prefetch algorithms
US6131145A (en) * 1995-10-27 2000-10-10 Hitachi, Ltd. Information processing unit and method for controlling a hierarchical cache utilizing indicator bits to control content of prefetching operations
US6401192B1 (en) * 1998-10-05 2002-06-04 International Business Machines Corporation Apparatus for software initiated prefetch and method therefor
US6460115B1 (en) * 1999-11-08 2002-10-01 International Business Machines Corporation System and method for prefetching data to multiple levels of cache including selectively using a software hint to override a hardware prefetch mechanism
US6532521B1 (en) * 1999-06-30 2003-03-11 International Business Machines Corporation Mechanism for high performance transfer of speculative request data between levels of cache hierarchy
US6574712B1 (en) * 1999-11-08 2003-06-03 International Business Machines Corporation Software prefetch system and method for predetermining amount of streamed data
US6782454B1 (en) * 2000-09-29 2004-08-24 Sun Microsystems, Inc. System and method for pre-fetching for pointer linked data structures
US6848029B2 (en) * 2000-01-03 2005-01-25 Dirk Coldewey Method and apparatus for prefetching recursive data structures
US6957306B2 (en) * 2002-09-09 2005-10-18 Broadcom Corporation System and method for controlling prefetching

Family Cites Families (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4980823A (en) * 1987-06-22 1990-12-25 International Business Machines Corporation Sequential prefetching with deconfirmation
US5261066A (en) * 1990-03-27 1993-11-09 Digital Equipment Corporation Data processing system and method with small fully-associative cache and prefetch buffers
JP2881049B2 (en) * 1991-07-30 1999-04-12 株式会社日立製作所 Prefetch buffer
WO1993018459A1 (en) * 1992-03-06 1993-09-16 Rambus Inc. Prefetching into a cache to minimize main memory access time and cache size in a computer system
US5507028A (en) * 1992-03-30 1996-04-09 International Business Machines Corporation History based branch prediction accessed via a history based earlier instruction address
US5588128A (en) * 1993-04-02 1996-12-24 Vlsi Technology, Inc. Dynamic direction look ahead read buffer
US5689679A (en) * 1993-04-28 1997-11-18 Digital Equipment Corporation Memory system and method for selective multi-level caching using a cache level code
US5790823A (en) * 1995-07-13 1998-08-04 International Business Machines Corporation Operand prefetch table
US5734881A (en) * 1995-12-15 1998-03-31 Cyrix Corporation Detecting short branches in a prefetch buffer using target location information in a branch target cache
KR100221028B1 (en) * 1996-07-23 1999-09-15 윤종용 Graphic accelerator and memory-prefetching method of it
JP3175675B2 (en) * 1997-12-04 2001-06-11 日本電気株式会社 Prefetch control device
US6092149A (en) * 1997-05-28 2000-07-18 Western Digital Corporation Disk drive cache system using a dynamic priority sequential stream of data segments continuously adapted according to prefetched sequential random, and repeating types of accesses
US6317810B1 (en) * 1997-06-25 2001-11-13 Sun Microsystems, Inc. Microprocessor having a prefetch cache
US5951678A (en) * 1997-07-25 1999-09-14 Motorola, Inc. Method and apparatus for controlling conditional branch execution in a data processor
US20020042861A1 (en) * 1997-11-07 2002-04-11 Kavipurapu Gautam Nag Apparatus and method for implementing a variable block size cache
US6134643A (en) * 1997-11-26 2000-10-17 Intel Corporation Method and apparatus for cache line prediction and prefetching using a prefetch controller and buffer and access history
US6484239B1 (en) * 1997-12-29 2002-11-19 Intel Corporation Prefetch queue
US6216208B1 (en) * 1997-12-29 2001-04-10 Intel Corporation Prefetch queue responsive to read request sequences
US6643745B1 (en) * 1998-03-31 2003-11-04 Intel Corporation Method and apparatus for prefetching data into cache
US6212603B1 (en) * 1998-04-09 2001-04-03 Institute For The Development Of Emerging Architectures, L.L.C. Processor with apparatus for tracking prefetch and demand fetch instructions serviced by cache memory
GB2348024B (en) * 1999-03-16 2003-06-25 Ibm Cache memory systems
US6594730B1 (en) * 1999-08-03 2003-07-15 Intel Corporation Prefetch system for memory controller
JP3969009B2 (en) * 2001-03-29 2007-08-29 株式会社日立製作所 Hardware prefetch system
US6848030B2 (en) * 2001-07-20 2005-01-25 Freescale Semiconductor, Inc. Method and apparatus for filling lines in a cache
US20030154349A1 (en) * 2002-01-24 2003-08-14 Berg Stefan G. Program-directed cache prefetching for media processors

Patent Citations (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5784711A (en) * 1990-05-18 1998-07-21 Philips Electronics North America Corporation Data cache prefetching under control of instruction cache
US5721865A (en) * 1995-01-20 1998-02-24 Hitachi, Ltd. Information processing apparatus with prefetch control for prefetching data structure from memory through cache memory
US6131145A (en) * 1995-10-27 2000-10-10 Hitachi, Ltd. Information processing unit and method for controlling a hierarchical cache utilizing indicator bits to control content of prefetching operations
US5829042A (en) * 1996-02-15 1998-10-27 Hewlett-Packard Company Prefetch operation for network peripheral device having shared memory
US5802569A (en) * 1996-04-22 1998-09-01 International Business Machines Corp. Computer system having cache prefetching amount based on CPU request types
US5941981A (en) * 1997-11-03 1999-08-24 Advanced Micro Devices, Inc. System for using a data history table to select among multiple data prefetch algorithms
US6401192B1 (en) * 1998-10-05 2002-06-04 International Business Machines Corporation Apparatus for software initiated prefetch and method therefor
US6532521B1 (en) * 1999-06-30 2003-03-11 International Business Machines Corporation Mechanism for high performance transfer of speculative request data between levels of cache hierarchy
US6460115B1 (en) * 1999-11-08 2002-10-01 International Business Machines Corporation System and method for prefetching data to multiple levels of cache including selectively using a software hint to override a hardware prefetch mechanism
US6574712B1 (en) * 1999-11-08 2003-06-03 International Business Machines Corporation Software prefetch system and method for predetermining amount of streamed data
US6848029B2 (en) * 2000-01-03 2005-01-25 Dirk Coldewey Method and apparatus for prefetching recursive data structures
US6782454B1 (en) * 2000-09-29 2004-08-24 Sun Microsystems, Inc. System and method for pre-fetching for pointer linked data structures
US6957306B2 (en) * 2002-09-09 2005-10-18 Broadcom Corporation System and method for controlling prefetching

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7143251B1 (en) * 2003-06-30 2006-11-28 Data Domain, Inc. Data storage using identifiers
US20070226462A1 (en) * 2006-03-21 2007-09-27 Freescale Semiconductor, Inc. Data processor having dynamic control of instruction prefetch buffer depth and method therefor
US9304773B2 (en) * 2006-03-21 2016-04-05 Freescale Semiconductor, Inc. Data processor having dynamic control of instruction prefetch buffer depth and method therefor
US20090119488A1 (en) * 2006-06-15 2009-05-07 Sudarshan Kadambi Prefetch Unit
US7779208B2 (en) * 2006-06-15 2010-08-17 Apple Inc. Prefetch unit
US7996624B2 (en) 2006-06-15 2011-08-09 Apple Inc. Prefetch unit
US8316188B2 (en) 2006-06-15 2012-11-20 Apple Inc. Data prefetch unit utilizing duplicate cache tags
US20080229072A1 (en) * 2007-03-14 2008-09-18 Fujitsu Limited Prefetch processing apparatus, prefetch processing method, storage medium storing prefetch processing program
US8006041B2 (en) * 2007-03-14 2011-08-23 Fujitsu Limited Prefetch processing apparatus, prefetch processing method, storage medium storing prefetch processing program

Also Published As

Publication number Publication date
US6931494B2 (en) 2005-08-16
US20040049641A1 (en) 2004-03-11
US7167954B2 (en) 2007-01-23
US20050182904A1 (en) 2005-08-18
US7711906B2 (en) 2010-05-04
US20040049639A1 (en) 2004-03-11
US7627720B2 (en) 2009-12-01
US20050210200A1 (en) 2005-09-22
US20040049640A1 (en) 2004-03-11
US6957306B2 (en) 2005-10-18

Similar Documents

Publication Publication Date Title
US6957306B2 (en) System and method for controlling prefetching
US6219760B1 (en) Cache including a prefetch way for storing cache lines and configured to move a prefetched cache line to a non-prefetch way upon access to the prefetched cache line
US5829025A (en) Computer system and method of allocating cache memories in a multilevel cache hierarchy utilizing a locality hint within an instruction
JP3577331B2 (en) Cache memory system and method for manipulating instructions in a microprocessor
EP1710693B1 (en) Apparatus and method for supporting execution of prefetch threads
US6292871B1 (en) Loading accessed data from a prefetch buffer to a least recently used position in a cache
US8595443B2 (en) Varying a data prefetch size based upon data usage
JPH04324546A (en) Computer system
US11740930B2 (en) Global coherence operations
US11392498B2 (en) Aliased mode for cache controller
US8266381B2 (en) Varying an amount of data retrieved from memory based upon an instruction hint
JPH06202951A (en) Cash memory system
US6934810B1 (en) Delayed leaky write system and method for a cache memory
JPH0477344B2 (en)
US6874067B2 (en) Eliminating unnecessary data pre-fetches in a multiprocessor computer system

Legal Events

Date Code Title Description
STCB Information on status: application discontinuation

Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION

AS Assignment

Owner name: BANK OF AMERICA, N.A., AS COLLATERAL AGENT, NORTH CAROLINA

Free format text: PATENT SECURITY AGREEMENT;ASSIGNOR:BROADCOM CORPORATION;REEL/FRAME:037806/0001

Effective date: 20160201

Owner name: BANK OF AMERICA, N.A., AS COLLATERAL AGENT, NORTH

Free format text: PATENT SECURITY AGREEMENT;ASSIGNOR:BROADCOM CORPORATION;REEL/FRAME:037806/0001

Effective date: 20160201

AS Assignment

Owner name: AVAGO TECHNOLOGIES GENERAL IP (SINGAPORE) PTE. LTD., SINGAPORE

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BROADCOM CORPORATION;REEL/FRAME:041706/0001

Effective date: 20170120

Owner name: AVAGO TECHNOLOGIES GENERAL IP (SINGAPORE) PTE. LTD

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BROADCOM CORPORATION;REEL/FRAME:041706/0001

Effective date: 20170120

AS Assignment

Owner name: BROADCOM CORPORATION, CALIFORNIA

Free format text: TERMINATION AND RELEASE OF SECURITY INTEREST IN PATENTS;ASSIGNOR:BANK OF AMERICA, N.A., AS COLLATERAL AGENT;REEL/FRAME:041712/0001

Effective date: 20170119