CA1236588A - Dynamically allocated local/global storage system - Google Patents
Dynamically allocated local/global storage systemInfo
- Publication number
- CA1236588A CA1236588A CA000491267A CA491267A CA1236588A CA 1236588 A CA1236588 A CA 1236588A CA 000491267 A CA000491267 A CA 000491267A CA 491267 A CA491267 A CA 491267A CA 1236588 A CA1236588 A CA 1236588A
- Authority
- CA
- Canada
- Prior art keywords
- storage
- interleaving
- processor
- dynamically
- processors
- 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.)
- Expired
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
- G06F12/0223—User address space allocation, e.g. contiguous or non contiguous base addressing
- G06F12/0292—User address space allocation, e.g. contiguous or non contiguous base addressing using tables or multilevel address translation means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
- G06F12/06—Addressing a physical block of locations, e.g. base addressing, module addressing, memory dedication
- G06F12/0646—Configuration or reconfiguration
- G06F12/0692—Multiconfiguration, e.g. local and global addressing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
- G06F12/08—Addressing or allocation; Relocation in hierarchically structured memory systems, e.g. virtual memory systems
Abstract
ABSTRACT OF THE DISCLOSURE
Method and Apparatus for dynamically parti-tioning a storage system into a global storage ef-ficiently accessible by a number of processors connected to a network, and local storage effi-ciently accessible by individual processors, in-cluding means for interleaving storage references by a processor; means under the control of each processor for controlling the means for interleav-ing storage references and means for dynamically directing storage references to first or second portions of storage.
Method and Apparatus for dynamically parti-tioning a storage system into a global storage ef-ficiently accessible by a number of processors connected to a network, and local storage effi-ciently accessible by individual processors, in-cluding means for interleaving storage references by a processor; means under the control of each processor for controlling the means for interleav-ing storage references and means for dynamically directing storage references to first or second portions of storage.
Description
DYNAMICALLY ALLOCAT~D I,OCAL/GLOBAL STORAGE SYSTEM
BACKGROUND OF THE INVE~TION
1. Field of the Invention The present invention relates to data processor storage systems and more particularly to dynamic storage systems for multiprocessor systems.
BACKGROUND OF THE INVE~TION
1. Field of the Invention The present invention relates to data processor storage systems and more particularly to dynamic storage systems for multiprocessor systems.
2. Description or the Prior Art The following are systems representative of the prior art.
U.S. Patent 4,365,295 shows a multiprocessor system including a memory system in which the memory of each processor module is divided into four logical address areas. The memory system includes a map which trans-lates logical addresses to physical addresses and which coacts with the multiprocessor system to bring pages from secondary memory into primary main memory as re-quired to implement a virtual memory system.
l~is patent which describes a conventional memory mapping system, does not address the efficient access of memory by single or multiple processors including interleaving storage references by a processor and dy-namically directing storage rererences to global or lo-cal portions of each storage module.
, ~
r ~L~3~S8~3 U.S. Patent 4,228,496 shows a multiprocessor system including a memory system as above to implement a vir-tual memory system.
However, this patent which describes a conventional memory mapping system, does not address thé efficient access of memory by single or multiple processors in-cluding interleaving storage references by a processor and dynamically directing storage references to global or local portions of each storage module.
U.S. Patent 4,174,514 shows apparatus for perform-ing neighborhood transformations on data matrices for image processing and the like achieving processing speeds greater than serial processors within a economy of memory through use of a plurality of serial neigh-borhood processors that simultaneously operate upon ad-joining partitioned segments of a single data matrix.
This patent shows a multiprocessor system without any provision for access by all processors to a common ~lobal storage.
U.S. Patent 4,121,286 shows apparatus for allocat-ing and deallocating memory space in a multiprocessor environment.
This patent which describes a conventional memory mapping system, does not address the efficient access of memory by single or multiple processors including 8~3 interleaving storage references by a processor and dy-namically directing storage references to global or lo-cal portions of each storage module.
U.S. Patent 3,916,383 shows a resource allocation circuit selectively activating individual processors by time slice basis where a time slice has approximately the same time duration as the system storage time. The resource allocation circuit includes a priorty network which receives real time common resource utilizacion requests from the processors according to the individual processor needs, assigns a priorty rating to the re-ceived request and alters in response thereto the oth-erwise sequential activation of the processors. The patent shows a system with several indspendent data processors within a single central processor which is not a true multiprocessor system in the usual sense.
The present invention relates to a system having one or more independent processors forming a multi-processor in which a storage system is dynamically partitioned into global storage and local storage.
V.S. Patent 3,820,079 shows a multiprocessing com-puter structured in modular form around a common control and data bus. Control functions for the various modules are distributed among the modules to facilitate system "
., f ~2365l3~
flexibility. The patent shows a system including con-ventional memory mapping and interleaving.
Unlike the present invention, the memory mapping does not controi the interleaving and the interleaving S is the same over all modules for all data.
U.S. Patent 3,641,505 shows a multiprocessor com-puting system in which a number of processing units, program storage units, variable storage units and input/output units may be selectively combined to form one or more independent data processing systems. System partitioning into more than one independent system is controlled alternatively by manual switching or program directed partitioning signals.
This patent which describes a conventional memory mapping system, does not address the efficient access of memory by single or multiple processors including interleaving storage references by a processor and dy-namically directing storage reforences to global or lo-cal portions of each storage module.
U.S. Patent 3,601,812 shows a memory sys~em for buffering several computers to a central storage unit or a compu~er to several small memory units and a par-titioned address scheme for the efficient use thereof.
The digits of the address are decomposed in~o two dis-joint subsets one of which is used as a buffer memory address and the other of which is stored with data word to effect identification thereof.
The patent deals with buffering memory data in a multiprocessor and does not show a dynamically parti- -tioned storage system including interleaving storage references by a processor and directing dynamically storage references to global or local portions of stor-age.
The prior art discussed above does not teach nor suggest the present invention as disclosed and claimed herein.
SUMMARY OF THE INVENTION
It is an object of the present invention to dynam-ically parti-tion a storage system into a global storage efficiently accessible by a plurality of processorsl and local storage efficiently accessible by individual processors, by method and apparatus comprising: means for interleaving storage references by a processor;
means under the control of each processor for control-ling the means for interleaving storage references;
means for dynamically directing storage references to first or second portions of storage.
It is another object of the present invention to dynamically partition a storage system as above by ~ .
6~
method and apparatus further including, assigning a first portion of storage to a referencing processor and a second portion of storage is assigned to another of the processors.
; It is another object of the present invention to dynamically partition a storage system as above by method and apparatus further including, a first means for allocating storage on page boundaries.
It is another object of the present invention to dynamically partition a storage system as above by method and apparatus further including, a second means for dynamically allocating storage on variable segment boundaries.
It is another object of the present invention to dynamically parti-tion a storage system as above by method and apparatus further including, means for con-trolling storage intsrleaving by said first and second means for allocating storage.
It is another object of the present invention to dynamically parlition a storage system as above by method and apparatus further including, means for interleaving storage by a factor equal to any power of 2 between 0 and the number of processing nodes of the System.
'~ YO984-026 - 6 -6~
It is another object of the present invention to dynamically partition a storage system as above by method and apparatus further including, a variable amount right rotate of a variable-width bit-fieId means S for limiting a number of storage modules over hhich interleaving is performed to a number less than a pre-determined maximum.
It is another object of the present in~-ention to dynamically partition a storage system as above by method and apparatus further including, means to re-map an interleaving sweep across memories to provide dif-ferent sequences of memory module access for different successive interleaving sweeps.
Accordingly, the present invention includes method and apparatus for dynamically partitioning a storage system into a global`storage efficiently accessible by a number of processors connected to a network, and local storage efficiently accessible by individual processors, including 0eans for interleaving storage references by a processor; means under the control of each processor for controlling the means for interleaving storage ref-erences and means for dynamically directing storage references to first or second portions of storage.
The foregoing and other objects, features and ad-vantages of the invention will be apparent from the more particular description of the preferred embodiments of the invention. as illustrated in the accompanving draw-ing.
BRIEF DESCRIPTION OF THE D~'ING
FIG. l is a bloc~ diagram of a multiprocessor sys-tem according to the present invention.
FIG. 2 is a block diagram of a processing node ac-cording to the present invention.
FIG. 3 is a network address chart showing the ad-dress organization according to the present invention.
FIG. 4 is a chart of a page of s~quentially mapped addresses in accordance with the present invention.
FIG. 5 is a chart of a page of interleaved mapped addresses in accordance with the present invention.
FIG. 6 is a chart showing interleaved pages of global and local storage.
FIG. 7 is a block diagram of a Map/Interleave block shown in FIG. 2 according to the presen~ invention.
FIG. 8 is a block diagram of a Network/Storage Interface block shown in FIG. 2 according to the present invention.
In the drawing, like elements are designated with similar reference numbers, and identical elements in Y0~84-026 - 8 ~.~3~
different specific embodiments are designated by iden-tical reference nnmbers.
~3~
DESCRIPTION OF PREFERRED E~IBODI~IENTS OF T~IE INVENTION
I ntroduction The present invention allows the main store of a multiple processor computer to be dynamically parti-tioned, at run time, between storage iocal to each processor and storage globally accessible by all processors.
Prior art multiprocessor systems provide either l.only local, and no global storage 2.only global, and no local storage
U.S. Patent 4,365,295 shows a multiprocessor system including a memory system in which the memory of each processor module is divided into four logical address areas. The memory system includes a map which trans-lates logical addresses to physical addresses and which coacts with the multiprocessor system to bring pages from secondary memory into primary main memory as re-quired to implement a virtual memory system.
l~is patent which describes a conventional memory mapping system, does not address the efficient access of memory by single or multiple processors including interleaving storage references by a processor and dy-namically directing storage rererences to global or lo-cal portions of each storage module.
, ~
r ~L~3~S8~3 U.S. Patent 4,228,496 shows a multiprocessor system including a memory system as above to implement a vir-tual memory system.
However, this patent which describes a conventional memory mapping system, does not address thé efficient access of memory by single or multiple processors in-cluding interleaving storage references by a processor and dynamically directing storage references to global or local portions of each storage module.
U.S. Patent 4,174,514 shows apparatus for perform-ing neighborhood transformations on data matrices for image processing and the like achieving processing speeds greater than serial processors within a economy of memory through use of a plurality of serial neigh-borhood processors that simultaneously operate upon ad-joining partitioned segments of a single data matrix.
This patent shows a multiprocessor system without any provision for access by all processors to a common ~lobal storage.
U.S. Patent 4,121,286 shows apparatus for allocat-ing and deallocating memory space in a multiprocessor environment.
This patent which describes a conventional memory mapping system, does not address the efficient access of memory by single or multiple processors including 8~3 interleaving storage references by a processor and dy-namically directing storage references to global or lo-cal portions of each storage module.
U.S. Patent 3,916,383 shows a resource allocation circuit selectively activating individual processors by time slice basis where a time slice has approximately the same time duration as the system storage time. The resource allocation circuit includes a priorty network which receives real time common resource utilizacion requests from the processors according to the individual processor needs, assigns a priorty rating to the re-ceived request and alters in response thereto the oth-erwise sequential activation of the processors. The patent shows a system with several indspendent data processors within a single central processor which is not a true multiprocessor system in the usual sense.
The present invention relates to a system having one or more independent processors forming a multi-processor in which a storage system is dynamically partitioned into global storage and local storage.
V.S. Patent 3,820,079 shows a multiprocessing com-puter structured in modular form around a common control and data bus. Control functions for the various modules are distributed among the modules to facilitate system "
., f ~2365l3~
flexibility. The patent shows a system including con-ventional memory mapping and interleaving.
Unlike the present invention, the memory mapping does not controi the interleaving and the interleaving S is the same over all modules for all data.
U.S. Patent 3,641,505 shows a multiprocessor com-puting system in which a number of processing units, program storage units, variable storage units and input/output units may be selectively combined to form one or more independent data processing systems. System partitioning into more than one independent system is controlled alternatively by manual switching or program directed partitioning signals.
This patent which describes a conventional memory mapping system, does not address the efficient access of memory by single or multiple processors including interleaving storage references by a processor and dy-namically directing storage reforences to global or lo-cal portions of each storage module.
U.S. Patent 3,601,812 shows a memory sys~em for buffering several computers to a central storage unit or a compu~er to several small memory units and a par-titioned address scheme for the efficient use thereof.
The digits of the address are decomposed in~o two dis-joint subsets one of which is used as a buffer memory address and the other of which is stored with data word to effect identification thereof.
The patent deals with buffering memory data in a multiprocessor and does not show a dynamically parti- -tioned storage system including interleaving storage references by a processor and directing dynamically storage references to global or local portions of stor-age.
The prior art discussed above does not teach nor suggest the present invention as disclosed and claimed herein.
SUMMARY OF THE INVENTION
It is an object of the present invention to dynam-ically parti-tion a storage system into a global storage efficiently accessible by a plurality of processorsl and local storage efficiently accessible by individual processors, by method and apparatus comprising: means for interleaving storage references by a processor;
means under the control of each processor for control-ling the means for interleaving storage references;
means for dynamically directing storage references to first or second portions of storage.
It is another object of the present invention to dynamically partition a storage system as above by ~ .
6~
method and apparatus further including, assigning a first portion of storage to a referencing processor and a second portion of storage is assigned to another of the processors.
; It is another object of the present invention to dynamically partition a storage system as above by method and apparatus further including, a first means for allocating storage on page boundaries.
It is another object of the present invention to dynamically partition a storage system as above by method and apparatus further including, a second means for dynamically allocating storage on variable segment boundaries.
It is another object of the present invention to dynamically parti-tion a storage system as above by method and apparatus further including, means for con-trolling storage intsrleaving by said first and second means for allocating storage.
It is another object of the present invention to dynamically parlition a storage system as above by method and apparatus further including, means for interleaving storage by a factor equal to any power of 2 between 0 and the number of processing nodes of the System.
'~ YO984-026 - 6 -6~
It is another object of the present invention to dynamically partition a storage system as above by method and apparatus further including, a variable amount right rotate of a variable-width bit-fieId means S for limiting a number of storage modules over hhich interleaving is performed to a number less than a pre-determined maximum.
It is another object of the present in~-ention to dynamically partition a storage system as above by method and apparatus further including, means to re-map an interleaving sweep across memories to provide dif-ferent sequences of memory module access for different successive interleaving sweeps.
Accordingly, the present invention includes method and apparatus for dynamically partitioning a storage system into a global`storage efficiently accessible by a number of processors connected to a network, and local storage efficiently accessible by individual processors, including 0eans for interleaving storage references by a processor; means under the control of each processor for controlling the means for interleaving storage ref-erences and means for dynamically directing storage references to first or second portions of storage.
The foregoing and other objects, features and ad-vantages of the invention will be apparent from the more particular description of the preferred embodiments of the invention. as illustrated in the accompanving draw-ing.
BRIEF DESCRIPTION OF THE D~'ING
FIG. l is a bloc~ diagram of a multiprocessor sys-tem according to the present invention.
FIG. 2 is a block diagram of a processing node ac-cording to the present invention.
FIG. 3 is a network address chart showing the ad-dress organization according to the present invention.
FIG. 4 is a chart of a page of s~quentially mapped addresses in accordance with the present invention.
FIG. 5 is a chart of a page of interleaved mapped addresses in accordance with the present invention.
FIG. 6 is a chart showing interleaved pages of global and local storage.
FIG. 7 is a block diagram of a Map/Interleave block shown in FIG. 2 according to the presen~ invention.
FIG. 8 is a block diagram of a Network/Storage Interface block shown in FIG. 2 according to the present invention.
In the drawing, like elements are designated with similar reference numbers, and identical elements in Y0~84-026 - 8 ~.~3~
different specific embodiments are designated by iden-tical reference nnmbers.
~3~
DESCRIPTION OF PREFERRED E~IBODI~IENTS OF T~IE INVENTION
I ntroduction The present invention allows the main store of a multiple processor computer to be dynamically parti-tioned, at run time, between storage iocal to each processor and storage globally accessible by all processors.
Prior art multiprocessor systems provide either l.only local, and no global storage 2.only global, and no local storage
3.global storage and a fixed amount of local storage Some of the systems of type 2 have a fixed amount of local storage in the form of a cache to effectively reduce global storage latency; as will be noted, the present invention does not preclude the use of a cache or, in general, the use of a storage hierarchy.
Unlike the above systems, the invention described here allows the storage configuratlon to be dynamically altered to fit the needs of the user resulting in sub-stantially improved performance over a wider range of applications. Efficient passing of messages batween processors, achieved in systems of type 1 above by special hardware, is also supported by this invention.
Y~984-026 - 10 -,, ~3 f 12365B~3 Configuration ~ s shown in Fig. 1, the ~achine organization needed consists of ~ processing nodes 20 connected bv some communications network 10. The processors and main s~orage of the system are contained in the nodcs. (see Fig. 2) ~ny ne~work providing communication among all the nodes may be used.
~'etwork Description Fig. l shows an interconnection network (ICN) 10 which connects the ~arious nodes 20 together. This invention does not require any specific interconnection network design, but such network must necessarily have as a minimum the following capabilities:
Messages which originate at any ons node 20 can be reliably routed through network 10 to any other node 20.
The routing of a message is based upon addressing information con~ained within a "~ode 1,''' field of the message.
The message-routing functions of the IC~ 10, when coupled with those of the various nodes ~0, must enable any processor to access any memory location at an~- node Yo984-026 - 11 -20 merely by specifying the correct absolute address.
The memory-mapping mechanisms of this invention provide each processor with the capability of ~ene~ating such absolute addresses.
Fig. ~ shows the contents of a node. ~ddresses ~or storage references issued by the processor (PROC) 2Z are mapped by the MAP/INTERLEAVE (M/I) 24 as described be-low.
A cache 26 is used to satisfy some storage refer-ences after mapping. The invention described here does not require the use of a cache nor does it restrict the placement of the cache. For example the cache 26 could reside between'the processor 22 and M/I block 24.
References not satisfied by the cache 26 (or all references, if there is no cache) are directed by the network/storage interface (NET/STORE INTF. (NSI)) 28 to either the portion of main store 30 at that node or through the network 10 to store 30 of another node.
The NSI 28 also receives reference requests from other nodes and directs them to the storage of a node to be satisfied. This effectively makes the node's storage 30 dual-ported. Close to the same increase in efficiency, at lower cost, can be obtained by locally interleaving a node's storage 30 and overlapping the processing of interleaved requests.
.
~3~
Local/Global Mapping ~ 4 performs the usual two-lcvel segment/page mapping of virtùal addresses produced bv processor ~2 to real addresses, under the direction of some form of segment/page tables held in the ~ain s~ore 30. ~le real addresses produced uniquely iden~ify every word or byte in all the nodes' stores: the high-order bits specify the node number, and the lo~--order bits specify the word or byte withi.n a node's store. This is illustra~ed in Fig. 3.
In this invention, ~/I 24 may also perform an interleaving transformation on the address. Whether it does so or not is specified by an additional field, unique to this invention, that is added to entries in the segment and/or page tables. The effect of this transformation is to make a page of real storage a se-quential block of addresses completely contained within a node (see Fig. 4); or a block of addresses that is scattered across several nodes' stores (see Fig. 5).
A sequential page can thus be guaran~eed ~o be in a node's own store 30, local to that processor 2~ and ..
quickly accessible, providing the function of a local storage. Since an interleaved page is spread across many storage blocks, the probability of storage con-~5 flicts ~hen multiple processors reference it i5 greatly Y~98~-026 - 13 -.
reduced; this provides efficient globally-accessible storage.
To further reduce the probability of conflicts, the interlea~ing transformation may also "hash" the node number por~ion of the address, for example, by ,~OR-ing (e~clusive-OR-ing) the node number portion of the ad-dress with other address bits. This would reduce the probability of conflict when regular patterns OL access occur.
The degree of interleaving used -- the number of nodes across which an interleaved page is spread -- may be specified by the additional field added to the seg-ment and/or page tables. This field may also specify characteristics of the 'Ihashing'' used.
lS By having some pages mapped sequentially, and some interleaved, part of main store 30 may be "local" and part "global." The amount that is local vs. global is under control of the storage mapping tables, and thus may be changed at run time to match the requirements of applications.
An example of the kind of main store use that this invention makes possible is illustrated in Fig. 6. This shows global storage allocated from one end of all nodes' storage 30, local storage from the other. While this is not the only way of using the invention described ~;~36~
here, it illustraces how ~he invention allows the pro-portions Oc stor-~ge used for ~lobal and local s~orage to chan~e in the collrse of running 3DDlications.
Message Passing In aiaition to ~e commllnic3tion af~ordea b- global storage, direct inter-processor message passing is sup-ported by this invention: Direct main storage data movement ir.stru_;ions (e.g., ".~'CL" IB'I*System 3iO
Principles of ~per~_ion~ can be used to move data from a sçquentiai page in one processor to a sequenlial page in another processor, without disturbing or requiring use of any other node's storage.
Description of Storage Mapping Tables The storage mapping tables are used by the M/I.
They define the mapping performed by the ~I/I between the address issued by the processor and the address accessed in memory. Specifically, and unique to this invention, they define whether an interleaving transformation is to be applied to an address or not, and may specify ~hat interleaving trans~ormation if any is to be applied.
The tables themselves may reside in the M/I itself; or in the main memorv of the system (either global or local storage), referenced by the M/I; or,in both. Wherever they reside, they are modifiable by software running on the system's processors. It will often be convenient -* Trade Mark f ~3~
to combine the definition of interleaving in these ta-bles with a virtual memorv mapping of some fo.m, e.~.~
page mappingS segment mapping, or two-level segment and page mappi~g ((reference: Baer, J., "Computer Systems Architecture", ~omputer Science Pre~ss, Rockville. MD, 1980)) by extending the usual contenls of such tables to include a field of at least one bit containing in-formation determining the interleaving and/or remapping.
This h~s been done in the preferred embodiment described here, but is not required by this invention, which.only requires that the existence and/or amount of the inter-leave be controlled by each processor. Other mechanisms for doing this include: extending the processors' in~
struction set to have interleaved and non-interleaved lS data access instructions; by instruction set extension or I/O instruction control, have instructions that turn interleaving on or off for data and/or instruction fetch.
Description of the Operation of the M/I 24 Fi~. 7 illustrates the operation of the Map/Interleave (M/I) for the case where memory mapping and low-order remapping are both incorporated. The top of the figure shows a virtual address as received from the processor and stored in VAR 242. This i5 subdivided, as shown, into a segment and~or page index (S/P I) 244, YO98/~-026 - 16 -~:3~S8~
a page offset (P0) 246, and a word offset (~iiO) 248.
These fields have the conventional meanings in memory mapping systems. The ~0, which specifies which byte in an addressed word (or word in a larger minimal unit of addressing) is to be accessed is passed through the en-tire mapping process unchanged (as shown~, and will not be mentioned further.
The S/P I is used in a conventional way as an index into the storage mapping tables, as shown. From the storage mapping tables, the real Segment/Page offset (S/P 0) 250 is derived in a conventional way by Table Lookup to form a Real Address as shown. Unique to this invention, the Table Lookup also produces an interleave amoun~ ~as shown) associated with each segment and/or page specified in the storage mapping tables.
After the Real Address is derived, the low-order Remap 252 may be applied to produce a Remapped Address in RAR 254. This may also be applied as par~t of the variable amount variable-width right rotate described below, or may be omitted, in which case the Real Address is passed through unchanged to the next stage. The low-order Remap operates on a field LR to produce a new address field LR' of the same width, using the rest or the Real Address (field labelled HR) as shown. The wid-th of LR (and LR') may be any value between two extremes:
~;~3~
at largest, it is equal in width to the page offset (PO); at smallest, it is the maximum allowed interleave amount, i.e., if the width is N, the maximum number of memory modules i5 2**N. Fig. 7 shows it at an intermediate point between these two extremes. The purpose of the low-order Remap is to randomize successive addresses that are to be interleaved across a subset of memory modules so that they are accessed in different sequences. This lowers the probability of many processors accessing the same memory module simultaneously when the data structures being accessed have a size that is an integer multiple of the amount of storage in one inter-leaved sweep across all the memories. The maximum size of LR arises from the need to keep pages addressed in contigu-ously-addresqed blocks; the minimum size is the minimum needed to effectively perform the function described above.
The low-order Remap is one-to-one, i.e., every possible value of LR must be mapped into a different value of LR'.
One possible low-order Remap is the following: Let the bits of LR be named LRO, LRl, ... LRn from right to left; and the bits of HR and LR' be named similarly. Then, using "xor" to represent the conventional exclusive-or logic function, a suitable low-order remap is: LR'O = LRO xor HRO; LR'l = LRl xor HRl; ... LR'n = LRn xor HRn.
~2~658~3 The actual interleaving transformation is then performed by a variable amount ri~ht rotnte on a Variable-l~idth bit-field device ~56, producing the ac-tual Absolute Address used to access ~he system's stor-S age modules. This uses the Interlea~e Amollnt derived earlier, and operates on the real address after r~map-ping (if remapping is done) excluding the word offset (~0). The width of the field to be rotated and the amount tne field is to be rotated are specified by the interleave amount. The operation of the right rotate is as follows: Let HS be numbered similarly as LR above.
Given an interleave amount of Q, the width of the field to be rotated is HSq-l through HSO. The number of bit positions the field is rotated is Q. Instead of a var-iable amount Variable-Width right rotate, a conventional bitwise rotation of the combined HS, CS, and LS fields by Q could be used. However, the scheme presented allows systems to be constructed with fewer than the ma~imum number of processing nodes because it retains, in the Absolute Address Reg 258, high-order (leftmost) Os that appeared in the Remapped Address in RAR 254. Conventional rotation would not do this, and therefore the fact that all possible values of LS must be allowed forces ad-dressing of all possible nodes 2; Y098'1-026 - l9 -..
3~
In the absolute address, the final HS' field des-ignates the processing node whose storage module con-tains the data to be accessed (~ode #); the combined CS
and LS' fields indicate the offset in that storage mod-ule where the data word is to be found (Storage Ofrset);
and the W0 field indicates which byte or sub-word is desired.
Note that when the interleave amount is 0, the variable amount Variable~ idth righ~ rotate leaves HS' equal to HS, and ~S' equal to LS. This leaves the Ab-solute Address the same as the Remapped Address, thus providing direct sequential addressing. This provides the sequential addressing described above. Appropriate values in the Storage Mapping Tables allow this to be l; storage local to the node generating the addresses, or storage entirely contained in other nodes (the latter useful for message passing and other operations).
Note also that the use of less than the maximum possible interleaving effectively restricts the processors across which global memory is allocated.
This can be used in several ways, e.g.: (a) -to allow the system to continue to operate, although in a de-graded mode, if some of the storage modules are inoper-ative due to their failure, the failure of the network, etc.; (b) to effectively partition the system, allowing ~:3~
parts to have their own global and local memory allo-cation independent of other parts, thus educing inter-ference between those parts -- either to run several independent problems, or a well-partitioned single problem.
Operation of the Cache ~6 The invention as described above can function with or without a private cache memory 26. The c3che can be positioned as indicated in Figure 2 or between the processor and NSI. The function of cache memory is ~o reduce memory access time for those memory accesses which occur repeatedly in time or at contiguous memory addresses. For cache coherence to be maintained in a multiprocessor configuration, it is necessary for such a cache to have an additional capability which would not ordinarily be implemented on a un.iprocessor cache. If for example one processor can read one memory location at approximately the same time that another processar is writing in the same location, it is required ~hat neither processor satisfy such memory references in its own cache. This additional capability can be provided by a variety of different means, such as cross-interrogation between different cache memories, or by specifying Cerlain memory locations to be non-cacheable.
2~ YO984-0~6 - 21 -~L~3~
Any such cacheing scheme (or none at all) can be applied in conjunction ~ith this invcntion.
Network/Storage interface 28 e invention includes a l~'etwor~-Storage interface ~'SI) 2S whose operation is illustrated in Figure ~. ThA
routing functions of this unit (as described belowj are necessary for the proper functioning of this invention.
Any hardware configuration which provides these same message-routinO functions can be employed in this in-vention, and its implementation should be straightfor-ward for anyone skilled in the art. Such a unit is associated with each processor node, as illustrated in Figure 2. The function of this unit is to route messages between the associated processor, the associated memory-controller, and other processor-nodes on the network. The types of messages sent include, but are not limited to:
o ~oad requests issued by the local processor.
Store requests issued by the local processor.
Cache-line load raquests issued by the local cacbe, resulting from cache-misses on storage raquests by the local processor:
`6Si~
Cache-line store requests issued by the local cache, resulting from cache-misses on storage requests by the local processor.
Responses ~o storage load or store requests b~ ~he local processor and/or cache.
Load or store requests issued bv other processors or caches, referencing memory locations contained in the memory of the local processor node.
Responses to storage requests issued by other processors or caches, being returned from the memory of the local processor node.
r Messages from the local processor to remote processors, or from remote processor nodes to the local processor.
Synchronization requests (such as test-and-set~
etc.) issued by the local processor, to be per-formed at the local memory or at remote memory lo-cations.
Responses to synchronization requests.
~0 Y0904-026 - 23 -.
5B~;3 All such messages must contain information suffi-cient to identi~y the type of the message.
In addition, all such messages arri~ing at the .~SI
28 must contain information sufficient to determine w11ether the message is to be routed to the local processor/cache 26, the local store 30, or to the interconnection network lO. In the case of storage re-quests, by a processor or cache, such information is contained in the "Node "" field of the memory address.
If the value of the "Node #" field coincides with ~he number of the local node, such requests are routed to the local memory 30; otherwise they are routed to the interconnection network lO. The memory-mapping scheme described above ensures that the required interleaving is thereby performed. Similarly, responses to storage requests are routed either to the local processor 22 or to the interconnection network lO, so as to return to the processor ~ode which originated the message. Other messages also must con-tain "Node #" fields and message type identifying codes, which uniquely identify such messages in order to be properly routed by NSI 28. The NSI is capable of routing messages from any of the three sources to any of the other two outputs, based on in-formation contained in fields within the messages. In ~5 Y0984-026 - 24 -a6~88 particular, the devices shown in the figure can operate to perform such routing as follows:
The PE router (PE RTE) 282 receives messages from the PE ~ If the "Node ,'" indicates the curren~
node, the PE RTE 282 sends the message to the local store 30 via the local memory concentrator (LM CON) 284; other~ise, it sends it to the network via the networ~ concentrator (NET CON) 286.
The local memory router (LM RTE) 288 receives re-sponse messages from the local store 30. If the "Node #" indicates the current node, the LM RTE 288 sends the message to the local PE 22 via the PE
concentrator (PE CON) 290; otherwise, it sends it to the network via the network concentrator (NET
lS CON) 286.
The network router (NET RTE) 92 receives messages from the network, and on the basis of the type of each message determines whether it is (a) a reauest from another processor for access to the local mem-ory module; or (b) a reply from another node con-taining information requested by the current node from another node's local memory. In case (a), the Y0984-026 - 2~ -~3~S~3~
message is sent to the local memory via the L~ CON
~84; otherwise, it is sent to the locnl PE 22 via the PE CON 290.
The networ~ concentrator 286 receives messages (ei-S ther requests or replies) from either the PE 22, via tha PE RTE 282; or the LM 30, via the ~I RTE 288.
It passes both to the network 10 for routing to an-other node based on the message's "~ode i,".
The PE concentrator 290 receives reply messages from either the local store 30, via the ~M RTE 288; or the network 10, via NET RTE 292. It passes them to the PE 22 (and/or cache 26).
The local memory concentrator 284 receives request rnessages from either the local PE 22, via the PE RTE
282; or network 10, via NET RTE 292. It passes them to local store 30.
In addition to paths for data communication, the routers and concentrators indicated above must communi-cale control information indicating when data is valid ~0 (from the router to the concentrator and when it can be accepted from the concentrator to the router).
A two-ported memory could be used instead of the L~ RTE ~88 and L~l CON ~84 devices.
Thus, while the invention has been descri~ed ~ith reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without depart-ing from the scope of the invention.
Unlike the above systems, the invention described here allows the storage configuratlon to be dynamically altered to fit the needs of the user resulting in sub-stantially improved performance over a wider range of applications. Efficient passing of messages batween processors, achieved in systems of type 1 above by special hardware, is also supported by this invention.
Y~984-026 - 10 -,, ~3 f 12365B~3 Configuration ~ s shown in Fig. 1, the ~achine organization needed consists of ~ processing nodes 20 connected bv some communications network 10. The processors and main s~orage of the system are contained in the nodcs. (see Fig. 2) ~ny ne~work providing communication among all the nodes may be used.
~'etwork Description Fig. l shows an interconnection network (ICN) 10 which connects the ~arious nodes 20 together. This invention does not require any specific interconnection network design, but such network must necessarily have as a minimum the following capabilities:
Messages which originate at any ons node 20 can be reliably routed through network 10 to any other node 20.
The routing of a message is based upon addressing information con~ained within a "~ode 1,''' field of the message.
The message-routing functions of the IC~ 10, when coupled with those of the various nodes ~0, must enable any processor to access any memory location at an~- node Yo984-026 - 11 -20 merely by specifying the correct absolute address.
The memory-mapping mechanisms of this invention provide each processor with the capability of ~ene~ating such absolute addresses.
Fig. ~ shows the contents of a node. ~ddresses ~or storage references issued by the processor (PROC) 2Z are mapped by the MAP/INTERLEAVE (M/I) 24 as described be-low.
A cache 26 is used to satisfy some storage refer-ences after mapping. The invention described here does not require the use of a cache nor does it restrict the placement of the cache. For example the cache 26 could reside between'the processor 22 and M/I block 24.
References not satisfied by the cache 26 (or all references, if there is no cache) are directed by the network/storage interface (NET/STORE INTF. (NSI)) 28 to either the portion of main store 30 at that node or through the network 10 to store 30 of another node.
The NSI 28 also receives reference requests from other nodes and directs them to the storage of a node to be satisfied. This effectively makes the node's storage 30 dual-ported. Close to the same increase in efficiency, at lower cost, can be obtained by locally interleaving a node's storage 30 and overlapping the processing of interleaved requests.
.
~3~
Local/Global Mapping ~ 4 performs the usual two-lcvel segment/page mapping of virtùal addresses produced bv processor ~2 to real addresses, under the direction of some form of segment/page tables held in the ~ain s~ore 30. ~le real addresses produced uniquely iden~ify every word or byte in all the nodes' stores: the high-order bits specify the node number, and the lo~--order bits specify the word or byte withi.n a node's store. This is illustra~ed in Fig. 3.
In this invention, ~/I 24 may also perform an interleaving transformation on the address. Whether it does so or not is specified by an additional field, unique to this invention, that is added to entries in the segment and/or page tables. The effect of this transformation is to make a page of real storage a se-quential block of addresses completely contained within a node (see Fig. 4); or a block of addresses that is scattered across several nodes' stores (see Fig. 5).
A sequential page can thus be guaran~eed ~o be in a node's own store 30, local to that processor 2~ and ..
quickly accessible, providing the function of a local storage. Since an interleaved page is spread across many storage blocks, the probability of storage con-~5 flicts ~hen multiple processors reference it i5 greatly Y~98~-026 - 13 -.
reduced; this provides efficient globally-accessible storage.
To further reduce the probability of conflicts, the interlea~ing transformation may also "hash" the node number por~ion of the address, for example, by ,~OR-ing (e~clusive-OR-ing) the node number portion of the ad-dress with other address bits. This would reduce the probability of conflict when regular patterns OL access occur.
The degree of interleaving used -- the number of nodes across which an interleaved page is spread -- may be specified by the additional field added to the seg-ment and/or page tables. This field may also specify characteristics of the 'Ihashing'' used.
lS By having some pages mapped sequentially, and some interleaved, part of main store 30 may be "local" and part "global." The amount that is local vs. global is under control of the storage mapping tables, and thus may be changed at run time to match the requirements of applications.
An example of the kind of main store use that this invention makes possible is illustrated in Fig. 6. This shows global storage allocated from one end of all nodes' storage 30, local storage from the other. While this is not the only way of using the invention described ~;~36~
here, it illustraces how ~he invention allows the pro-portions Oc stor-~ge used for ~lobal and local s~orage to chan~e in the collrse of running 3DDlications.
Message Passing In aiaition to ~e commllnic3tion af~ordea b- global storage, direct inter-processor message passing is sup-ported by this invention: Direct main storage data movement ir.stru_;ions (e.g., ".~'CL" IB'I*System 3iO
Principles of ~per~_ion~ can be used to move data from a sçquentiai page in one processor to a sequenlial page in another processor, without disturbing or requiring use of any other node's storage.
Description of Storage Mapping Tables The storage mapping tables are used by the M/I.
They define the mapping performed by the ~I/I between the address issued by the processor and the address accessed in memory. Specifically, and unique to this invention, they define whether an interleaving transformation is to be applied to an address or not, and may specify ~hat interleaving trans~ormation if any is to be applied.
The tables themselves may reside in the M/I itself; or in the main memorv of the system (either global or local storage), referenced by the M/I; or,in both. Wherever they reside, they are modifiable by software running on the system's processors. It will often be convenient -* Trade Mark f ~3~
to combine the definition of interleaving in these ta-bles with a virtual memorv mapping of some fo.m, e.~.~
page mappingS segment mapping, or two-level segment and page mappi~g ((reference: Baer, J., "Computer Systems Architecture", ~omputer Science Pre~ss, Rockville. MD, 1980)) by extending the usual contenls of such tables to include a field of at least one bit containing in-formation determining the interleaving and/or remapping.
This h~s been done in the preferred embodiment described here, but is not required by this invention, which.only requires that the existence and/or amount of the inter-leave be controlled by each processor. Other mechanisms for doing this include: extending the processors' in~
struction set to have interleaved and non-interleaved lS data access instructions; by instruction set extension or I/O instruction control, have instructions that turn interleaving on or off for data and/or instruction fetch.
Description of the Operation of the M/I 24 Fi~. 7 illustrates the operation of the Map/Interleave (M/I) for the case where memory mapping and low-order remapping are both incorporated. The top of the figure shows a virtual address as received from the processor and stored in VAR 242. This i5 subdivided, as shown, into a segment and~or page index (S/P I) 244, YO98/~-026 - 16 -~:3~S8~
a page offset (P0) 246, and a word offset (~iiO) 248.
These fields have the conventional meanings in memory mapping systems. The ~0, which specifies which byte in an addressed word (or word in a larger minimal unit of addressing) is to be accessed is passed through the en-tire mapping process unchanged (as shown~, and will not be mentioned further.
The S/P I is used in a conventional way as an index into the storage mapping tables, as shown. From the storage mapping tables, the real Segment/Page offset (S/P 0) 250 is derived in a conventional way by Table Lookup to form a Real Address as shown. Unique to this invention, the Table Lookup also produces an interleave amoun~ ~as shown) associated with each segment and/or page specified in the storage mapping tables.
After the Real Address is derived, the low-order Remap 252 may be applied to produce a Remapped Address in RAR 254. This may also be applied as par~t of the variable amount variable-width right rotate described below, or may be omitted, in which case the Real Address is passed through unchanged to the next stage. The low-order Remap operates on a field LR to produce a new address field LR' of the same width, using the rest or the Real Address (field labelled HR) as shown. The wid-th of LR (and LR') may be any value between two extremes:
~;~3~
at largest, it is equal in width to the page offset (PO); at smallest, it is the maximum allowed interleave amount, i.e., if the width is N, the maximum number of memory modules i5 2**N. Fig. 7 shows it at an intermediate point between these two extremes. The purpose of the low-order Remap is to randomize successive addresses that are to be interleaved across a subset of memory modules so that they are accessed in different sequences. This lowers the probability of many processors accessing the same memory module simultaneously when the data structures being accessed have a size that is an integer multiple of the amount of storage in one inter-leaved sweep across all the memories. The maximum size of LR arises from the need to keep pages addressed in contigu-ously-addresqed blocks; the minimum size is the minimum needed to effectively perform the function described above.
The low-order Remap is one-to-one, i.e., every possible value of LR must be mapped into a different value of LR'.
One possible low-order Remap is the following: Let the bits of LR be named LRO, LRl, ... LRn from right to left; and the bits of HR and LR' be named similarly. Then, using "xor" to represent the conventional exclusive-or logic function, a suitable low-order remap is: LR'O = LRO xor HRO; LR'l = LRl xor HRl; ... LR'n = LRn xor HRn.
~2~658~3 The actual interleaving transformation is then performed by a variable amount ri~ht rotnte on a Variable-l~idth bit-field device ~56, producing the ac-tual Absolute Address used to access ~he system's stor-S age modules. This uses the Interlea~e Amollnt derived earlier, and operates on the real address after r~map-ping (if remapping is done) excluding the word offset (~0). The width of the field to be rotated and the amount tne field is to be rotated are specified by the interleave amount. The operation of the right rotate is as follows: Let HS be numbered similarly as LR above.
Given an interleave amount of Q, the width of the field to be rotated is HSq-l through HSO. The number of bit positions the field is rotated is Q. Instead of a var-iable amount Variable-Width right rotate, a conventional bitwise rotation of the combined HS, CS, and LS fields by Q could be used. However, the scheme presented allows systems to be constructed with fewer than the ma~imum number of processing nodes because it retains, in the Absolute Address Reg 258, high-order (leftmost) Os that appeared in the Remapped Address in RAR 254. Conventional rotation would not do this, and therefore the fact that all possible values of LS must be allowed forces ad-dressing of all possible nodes 2; Y098'1-026 - l9 -..
3~
In the absolute address, the final HS' field des-ignates the processing node whose storage module con-tains the data to be accessed (~ode #); the combined CS
and LS' fields indicate the offset in that storage mod-ule where the data word is to be found (Storage Ofrset);
and the W0 field indicates which byte or sub-word is desired.
Note that when the interleave amount is 0, the variable amount Variable~ idth righ~ rotate leaves HS' equal to HS, and ~S' equal to LS. This leaves the Ab-solute Address the same as the Remapped Address, thus providing direct sequential addressing. This provides the sequential addressing described above. Appropriate values in the Storage Mapping Tables allow this to be l; storage local to the node generating the addresses, or storage entirely contained in other nodes (the latter useful for message passing and other operations).
Note also that the use of less than the maximum possible interleaving effectively restricts the processors across which global memory is allocated.
This can be used in several ways, e.g.: (a) -to allow the system to continue to operate, although in a de-graded mode, if some of the storage modules are inoper-ative due to their failure, the failure of the network, etc.; (b) to effectively partition the system, allowing ~:3~
parts to have their own global and local memory allo-cation independent of other parts, thus educing inter-ference between those parts -- either to run several independent problems, or a well-partitioned single problem.
Operation of the Cache ~6 The invention as described above can function with or without a private cache memory 26. The c3che can be positioned as indicated in Figure 2 or between the processor and NSI. The function of cache memory is ~o reduce memory access time for those memory accesses which occur repeatedly in time or at contiguous memory addresses. For cache coherence to be maintained in a multiprocessor configuration, it is necessary for such a cache to have an additional capability which would not ordinarily be implemented on a un.iprocessor cache. If for example one processor can read one memory location at approximately the same time that another processar is writing in the same location, it is required ~hat neither processor satisfy such memory references in its own cache. This additional capability can be provided by a variety of different means, such as cross-interrogation between different cache memories, or by specifying Cerlain memory locations to be non-cacheable.
2~ YO984-0~6 - 21 -~L~3~
Any such cacheing scheme (or none at all) can be applied in conjunction ~ith this invcntion.
Network/Storage interface 28 e invention includes a l~'etwor~-Storage interface ~'SI) 2S whose operation is illustrated in Figure ~. ThA
routing functions of this unit (as described belowj are necessary for the proper functioning of this invention.
Any hardware configuration which provides these same message-routinO functions can be employed in this in-vention, and its implementation should be straightfor-ward for anyone skilled in the art. Such a unit is associated with each processor node, as illustrated in Figure 2. The function of this unit is to route messages between the associated processor, the associated memory-controller, and other processor-nodes on the network. The types of messages sent include, but are not limited to:
o ~oad requests issued by the local processor.
Store requests issued by the local processor.
Cache-line load raquests issued by the local cacbe, resulting from cache-misses on storage raquests by the local processor:
`6Si~
Cache-line store requests issued by the local cache, resulting from cache-misses on storage requests by the local processor.
Responses ~o storage load or store requests b~ ~he local processor and/or cache.
Load or store requests issued bv other processors or caches, referencing memory locations contained in the memory of the local processor node.
Responses to storage requests issued by other processors or caches, being returned from the memory of the local processor node.
r Messages from the local processor to remote processors, or from remote processor nodes to the local processor.
Synchronization requests (such as test-and-set~
etc.) issued by the local processor, to be per-formed at the local memory or at remote memory lo-cations.
Responses to synchronization requests.
~0 Y0904-026 - 23 -.
5B~;3 All such messages must contain information suffi-cient to identi~y the type of the message.
In addition, all such messages arri~ing at the .~SI
28 must contain information sufficient to determine w11ether the message is to be routed to the local processor/cache 26, the local store 30, or to the interconnection network lO. In the case of storage re-quests, by a processor or cache, such information is contained in the "Node "" field of the memory address.
If the value of the "Node #" field coincides with ~he number of the local node, such requests are routed to the local memory 30; otherwise they are routed to the interconnection network lO. The memory-mapping scheme described above ensures that the required interleaving is thereby performed. Similarly, responses to storage requests are routed either to the local processor 22 or to the interconnection network lO, so as to return to the processor ~ode which originated the message. Other messages also must con-tain "Node #" fields and message type identifying codes, which uniquely identify such messages in order to be properly routed by NSI 28. The NSI is capable of routing messages from any of the three sources to any of the other two outputs, based on in-formation contained in fields within the messages. In ~5 Y0984-026 - 24 -a6~88 particular, the devices shown in the figure can operate to perform such routing as follows:
The PE router (PE RTE) 282 receives messages from the PE ~ If the "Node ,'" indicates the curren~
node, the PE RTE 282 sends the message to the local store 30 via the local memory concentrator (LM CON) 284; other~ise, it sends it to the network via the networ~ concentrator (NET CON) 286.
The local memory router (LM RTE) 288 receives re-sponse messages from the local store 30. If the "Node #" indicates the current node, the LM RTE 288 sends the message to the local PE 22 via the PE
concentrator (PE CON) 290; otherwise, it sends it to the network via the network concentrator (NET
lS CON) 286.
The network router (NET RTE) 92 receives messages from the network, and on the basis of the type of each message determines whether it is (a) a reauest from another processor for access to the local mem-ory module; or (b) a reply from another node con-taining information requested by the current node from another node's local memory. In case (a), the Y0984-026 - 2~ -~3~S~3~
message is sent to the local memory via the L~ CON
~84; otherwise, it is sent to the locnl PE 22 via the PE CON 290.
The networ~ concentrator 286 receives messages (ei-S ther requests or replies) from either the PE 22, via tha PE RTE 282; or the LM 30, via the ~I RTE 288.
It passes both to the network 10 for routing to an-other node based on the message's "~ode i,".
The PE concentrator 290 receives reply messages from either the local store 30, via the ~M RTE 288; or the network 10, via NET RTE 292. It passes them to the PE 22 (and/or cache 26).
The local memory concentrator 284 receives request rnessages from either the local PE 22, via the PE RTE
282; or network 10, via NET RTE 292. It passes them to local store 30.
In addition to paths for data communication, the routers and concentrators indicated above must communi-cale control information indicating when data is valid ~0 (from the router to the concentrator and when it can be accepted from the concentrator to the router).
A two-ported memory could be used instead of the L~ RTE ~88 and L~l CON ~84 devices.
Thus, while the invention has been descri~ed ~ith reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without depart-ing from the scope of the invention.
Claims (16)
1. Apparatus for dynamically- partitioning a storage system into a global storage, efficiently accessi-ble by a plurality of processors, and local stor-age, efficiently accessible by individual processors, comprising: means for interleaving storage references by a processor; means under the control of each processor for controlling said means for interleaving storage references; means for dynamically directing storage references to first or second portions of storage.
2. Apparatus according to claim 1, wherein said first portion of storage is assigned to a referencing processor and said second portion of storage is assigned to another of said plurality of processors.
3. Apparatus according to claim 1, further comprising: first means for allocating storage on page boundaries.
4. Apparatus according to claim 1, further comprising: second means for dynamically allocat-ing storage on variable segment boundaries.
5. Apparatus according to claim 1, further comprising: means for controlling storage inter-leaving by said first and second means for allo-cating storage.
6. Apparatus according to claim 1, further comprising: means for interleaving storage by a factor equal to any power of 2 between 0 and a number of processing nodes of the system.
7. Apparatus according to claim 6, further comprising: means for limiting a number of storage modules over which interleaving is performed to a number less than a predetermined maximum by a var-iable amount right rotate of a variable-width bit-field means.
8. Apparatus according to claim 1, further comprising: means to re-map an interleaving sweep across storage modules to provide different se-quences of storage module access for different successive interleaving sweeps.
9. Method for dynamically partitioning a storage system into a global storage, efficiently accessible by a plurality of processors, and local storage, efficiently accessible by individual processors, comprising the steps of: interleaving storage references by a processor; controlling a means for interleaving storage references under the control of each processor; and dynamically direct-ing storage references to first or second portions of storage.
10. A method according to claim 9, further comprising the steps of: assigning said first portion of storage to a referencing processor; and assigning said second portion of storage to another of said plurality of processors.
11. A method according to claim 9, further comprising the step of: allocating storage on page boundaries.
12. A method according to claim 9, further comprising the step of: dynamically allocating storage on variable segment boundaries.
13. A method according to claim 9, further comprising the step of: controlling storage interleaving by first and second means for allo-cating storage.
14. A method according to claim 9, further comprising the step of: interleaving storage by a factor equal to any power of 2 between 0 and the number of processing nodes of the system.
15. A method according to claim 14, further comprising the step of: limiting a number of storage modules over which interleaving is per-formed to a number less than a predetermined maxi-mum by a variable amount right rotate of a variable-width bit-field means.
16. A method according to claim 9, further comprising the step of: remapping an interleaving sweep across storage modules to provide different sequences of storage module access for different successive interleaving sweeps.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US06/664,131 US4754394A (en) | 1984-10-24 | 1984-10-24 | Multiprocessing system having dynamically allocated local/global storage and including interleaving transformation circuit for transforming real addresses to corresponding absolute address of the storage |
| US664,131 | 1984-10-24 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA1236588A true CA1236588A (en) | 1988-05-10 |
Family
ID=24664672
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA000491267A Expired CA1236588A (en) | 1984-10-24 | 1985-09-20 | Dynamically allocated local/global storage system |
Country Status (11)
| Country | Link |
|---|---|
| US (2) | US4754394A (en) |
| EP (1) | EP0179401B1 (en) |
| JP (1) | JPS61103258A (en) |
| KR (1) | KR910001736B1 (en) |
| CN (1) | CN1004307B (en) |
| CA (1) | CA1236588A (en) |
| DE (1) | DE3586389T2 (en) |
| GB (1) | GB2165975B (en) |
| HK (2) | HK23690A (en) |
| IN (1) | IN166397B (en) |
| PH (1) | PH25478A (en) |
Families Citing this family (136)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5093913A (en) * | 1986-12-22 | 1992-03-03 | At&T Laboratories | Multiprocessor memory management system with the flexible features of a tightly-coupled system in a non-shared memory system |
| US4811216A (en) * | 1986-12-22 | 1989-03-07 | American Telephone And Telegraph Company | Multiprocessor memory management method |
| US5136706A (en) * | 1987-04-30 | 1992-08-04 | Texas Instruments Incorporated | Adaptive memory management system for collection of garbage in a digital computer |
| US5111389A (en) * | 1987-10-29 | 1992-05-05 | International Business Machines Corporation | Aperiodic mapping system using power-of-two stride access to interleaved devices |
| JPH063589B2 (en) * | 1987-10-29 | 1994-01-12 | インターナシヨナル・ビジネス・マシーンズ・コーポレーシヨン | Address replacement device |
| US5226039A (en) * | 1987-12-22 | 1993-07-06 | Kendall Square Research Corporation | Packet routing switch |
| US5761413A (en) * | 1987-12-22 | 1998-06-02 | Sun Microsystems, Inc. | Fault containment system for multiprocessor with shared memory |
| US5055999A (en) * | 1987-12-22 | 1991-10-08 | Kendall Square Research Corporation | Multiprocessor digital data processing system |
| US5341483A (en) * | 1987-12-22 | 1994-08-23 | Kendall Square Research Corporation | Dynamic hierarchial associative memory |
| US5251308A (en) * | 1987-12-22 | 1993-10-05 | Kendall Square Research Corporation | Shared memory multiprocessor with data hiding and post-store |
| US5276826A (en) * | 1988-01-04 | 1994-01-04 | Hewlett-Packard Company | Apparatus for transforming addresses to provide pseudo-random access to memory modules |
| JPH0291747A (en) * | 1988-09-29 | 1990-03-30 | Hitachi Ltd | Information processor |
| JPH0833799B2 (en) * | 1988-10-31 | 1996-03-29 | 富士通株式会社 | Data input / output control method |
| US5117350A (en) * | 1988-12-15 | 1992-05-26 | Flashpoint Computer Corporation | Memory address mechanism in a distributed memory architecture |
| AU615084B2 (en) * | 1988-12-15 | 1991-09-19 | Pixar | Method and apparatus for memory routing scheme |
| FR2642252A1 (en) * | 1989-01-26 | 1990-07-27 | Centre Nat Rech Scient | Circuit interconnection unit, especially of the crossbar type, process for employing a circuit interconnection unit and uses of a circuit interconnection unit |
| IT1228728B (en) * | 1989-03-15 | 1991-07-03 | Bull Hn Information Syst | MULTIPROCESSOR SYSTEM WITH GLOBAL DATA REPLICATION AND TWO LEVELS OF ADDRESS TRANSLATION UNIT. |
| EP0389151A3 (en) * | 1989-03-22 | 1992-06-03 | International Business Machines Corporation | System and method for partitioned cache memory management |
| US5072369A (en) * | 1989-04-07 | 1991-12-10 | Tektronix, Inc. | Interface between buses attached with cached modules providing address space mapped cache coherent memory access with SNOOP hit memory updates |
| US5144692A (en) * | 1989-05-17 | 1992-09-01 | International Business Machines Corporation | System for controlling access by first system to portion of main memory dedicated exclusively to second system to facilitate input/output processing via first system |
| US5301327A (en) * | 1989-06-30 | 1994-04-05 | Digital Equipment Corporation | Virtual memory management for source-code development system |
| JPH03150637A (en) * | 1989-11-08 | 1991-06-27 | Oki Electric Ind Co Ltd | System for assigning register correspondingly to pipeline |
| US5715419A (en) * | 1989-12-05 | 1998-02-03 | Texas Instruments Incorporated | Data communications system with address remapping for expanded external memory access |
| US5682202A (en) * | 1989-12-08 | 1997-10-28 | Fuji Photo Film Co., Ltd. | Apparatus for recording/reproducing video data in a memory card on a cluster basis |
| WO1991010204A1 (en) * | 1989-12-26 | 1991-07-11 | Eastman Kodak Company | Image processing apparatus having disk storage resembling ram memory |
| AU645785B2 (en) * | 1990-01-05 | 1994-01-27 | Maspar Computer Corporation | Parallel processor memory system |
| US5161156A (en) * | 1990-02-02 | 1992-11-03 | International Business Machines Corporation | Multiprocessing packet switching connection system having provision for error correction and recovery |
| US5153595A (en) * | 1990-03-26 | 1992-10-06 | Geophysical Survey Systems, Inc. | Range information from signal distortions |
| JP2653709B2 (en) * | 1990-04-20 | 1997-09-17 | 富士写真フイルム株式会社 | Image / audio data playback device |
| JPH0418638A (en) * | 1990-05-11 | 1992-01-22 | Fujitsu Ltd | Static memory allocation processing method |
| JPH0430231A (en) * | 1990-05-25 | 1992-02-03 | Hitachi Ltd | Main storage addressing system |
| US5230051A (en) * | 1990-09-04 | 1993-07-20 | Hewlett-Packard Company | Distributed messaging system and method |
| US5265207A (en) * | 1990-10-03 | 1993-11-23 | Thinking Machines Corporation | Parallel computer system including arrangement for transferring messages from a source processor to selected ones of a plurality of destination processors and combining responses |
| JPH04246745A (en) * | 1991-02-01 | 1992-09-02 | Canon Inc | Memory access system |
| AU1587592A (en) | 1991-03-18 | 1992-10-21 | Echelon Corporation | Networked variables |
| DE69232169T2 (en) | 1991-03-18 | 2002-07-18 | Echelon Corp | PROGRAMMING LANGUAGE STRUCTURES FOR A NETWORK FOR TRANSMITTING, SCANING AND CONTROLLING INFORMATION |
| US6493739B1 (en) | 1993-08-24 | 2002-12-10 | Echelon Corporation | Task scheduling in an event driven environment |
| US5269013A (en) * | 1991-03-20 | 1993-12-07 | Digital Equipment Corporation | Adaptive memory management method for coupled memory multiprocessor systems |
| US5341485A (en) * | 1991-05-07 | 1994-08-23 | International Business Machines Corporation | Multiple virtual address translation per computer cycle |
| US5630098A (en) * | 1991-08-30 | 1997-05-13 | Ncr Corporation | System and method for interleaving memory addresses between memory banks based on the capacity of the memory banks |
| CA2078312A1 (en) | 1991-09-20 | 1993-03-21 | Mark A. Kaufman | Digital data processor with improved paging |
| CA2078310A1 (en) * | 1991-09-20 | 1993-03-21 | Mark A. Kaufman | Digital processor with distributed memory system |
| US5263003A (en) * | 1991-11-12 | 1993-11-16 | Allen-Bradley Company, Inc. | Flash memory circuit and method of operation |
| CA2073516A1 (en) * | 1991-11-27 | 1993-05-28 | Peter Michael Kogge | Dynamic multi-mode parallel processor array architecture computer system |
| US5434992A (en) * | 1992-09-04 | 1995-07-18 | International Business Machines Corporation | Method and means for dynamically partitioning cache into a global and data type subcache hierarchy from a real time reference trace |
| US5359730A (en) * | 1992-12-04 | 1994-10-25 | International Business Machines Corporation | Method of operating a data processing system having a dynamic software update facility |
| US5845329A (en) * | 1993-01-29 | 1998-12-01 | Sanyo Electric Co., Ltd. | Parallel computer |
| US5598568A (en) * | 1993-05-06 | 1997-01-28 | Mercury Computer Systems, Inc. | Multicomputer memory access architecture |
| US5584042A (en) * | 1993-06-01 | 1996-12-10 | International Business Machines Corporation | Dynamic I/O data address relocation facility |
| JPH06348584A (en) * | 1993-06-01 | 1994-12-22 | Internatl Business Mach Corp <Ibm> | Data processing system |
| US5638527A (en) * | 1993-07-19 | 1997-06-10 | Dell Usa, L.P. | System and method for memory mapping |
| GB9320982D0 (en) * | 1993-10-12 | 1993-12-01 | Ibm | A data processing system |
| US5583990A (en) * | 1993-12-10 | 1996-12-10 | Cray Research, Inc. | System for allocating messages between virtual channels to avoid deadlock and to optimize the amount of message traffic on each type of virtual channel |
| US5613067A (en) * | 1993-12-30 | 1997-03-18 | International Business Machines Corporation | Method and apparatus for assuring that multiple messages in a multi-node network are assured fair access to an outgoing data stream |
| JP3687990B2 (en) * | 1994-01-25 | 2005-08-24 | 株式会社日立製作所 | Memory access mechanism |
| SE515344C2 (en) * | 1994-02-08 | 2001-07-16 | Ericsson Telefon Ab L M | Distributed database system |
| US5530837A (en) * | 1994-03-28 | 1996-06-25 | Hewlett-Packard Co. | Methods and apparatus for interleaving memory transactions into an arbitrary number of banks |
| US5537635A (en) * | 1994-04-04 | 1996-07-16 | International Business Machines Corporation | Method and system for assignment of reclaim vectors in a partitioned cache with a virtual minimum partition size |
| US5907684A (en) * | 1994-06-17 | 1999-05-25 | International Business Machines Corporation | Independent channel coupled to be shared by multiple physical processing nodes with each node characterized as having its own memory, CPU and operating system image |
| US5727184A (en) * | 1994-06-27 | 1998-03-10 | Cirrus Logic, Inc. | Method and apparatus for interfacing between peripherals of multiple formats and a single system bus |
| JP2625385B2 (en) * | 1994-06-30 | 1997-07-02 | 日本電気株式会社 | Multiprocessor system |
| US5500852A (en) * | 1994-08-31 | 1996-03-19 | Echelon Corporation | Method and apparatus for network variable aliasing |
| US5812858A (en) * | 1994-09-16 | 1998-09-22 | Cirrus Logic, Inc. | Method and apparatus for providing register and interrupt compatibility between non-identical integrated circuits |
| US5685005A (en) * | 1994-10-04 | 1997-11-04 | Analog Devices, Inc. | Digital signal processor configured for multiprocessing |
| US6182121B1 (en) * | 1995-02-03 | 2001-01-30 | Enfish, Inc. | Method and apparatus for a physical storage architecture having an improved information storage and retrieval system for a shared file environment |
| US5850522A (en) * | 1995-02-03 | 1998-12-15 | Dex Information Systems, Inc. | System for physical storage architecture providing simultaneous access to common file by storing update data in update partitions and merging desired updates into common partition |
| US5860133A (en) * | 1995-12-01 | 1999-01-12 | Digital Equipment Corporation | Method for altering memory configuration and sizing memory modules while maintaining software code stream coherence |
| US6745292B1 (en) | 1995-12-08 | 2004-06-01 | Ncr Corporation | Apparatus and method for selectively allocating cache lines in a partitioned cache shared by multiprocessors |
| US5708790A (en) * | 1995-12-12 | 1998-01-13 | International Business Machines Corporation | Virtual memory mapping method and system for address translation mapping of logical memory partitions for BAT and TLB entries in a data processing system |
| US5896543A (en) * | 1996-01-25 | 1999-04-20 | Analog Devices, Inc. | Digital signal processor architecture |
| US5954811A (en) * | 1996-01-25 | 1999-09-21 | Analog Devices, Inc. | Digital signal processor architecture |
| US5892945A (en) | 1996-03-21 | 1999-04-06 | Oracle Corporation | Method and apparatus for distributing work granules among processes based on the location of data accessed in the work granules |
| US5784697A (en) * | 1996-03-27 | 1998-07-21 | International Business Machines Corporation | Process assignment by nodal affinity in a myultiprocessor system having non-uniform memory access storage architecture |
| US5940870A (en) * | 1996-05-21 | 1999-08-17 | Industrial Technology Research Institute | Address translation for shared-memory multiprocessor clustering |
| US6134601A (en) * | 1996-06-17 | 2000-10-17 | Networks Associates, Inc. | Computer resource management system |
| US5933852A (en) * | 1996-11-07 | 1999-08-03 | Micron Electronics, Inc. | System and method for accelerated remapping of defective memory locations |
| EP0931290A1 (en) * | 1997-03-21 | 1999-07-28 | International Business Machines Corporation | Address mapping for system memory |
| US6065045A (en) * | 1997-07-03 | 2000-05-16 | Tandem Computers Incorporated | Method and apparatus for object reference processing |
| US6002882A (en) * | 1997-11-03 | 1999-12-14 | Analog Devices, Inc. | Bidirectional communication port for digital signal processor |
| US6101181A (en) * | 1997-11-17 | 2000-08-08 | Cray Research Inc. | Virtual channel assignment in large torus systems |
| US5970232A (en) * | 1997-11-17 | 1999-10-19 | Cray Research, Inc. | Router table lookup mechanism |
| US6230252B1 (en) | 1997-11-17 | 2001-05-08 | Silicon Graphics, Inc. | Hybrid hypercube/torus architecture |
| US6085303A (en) * | 1997-11-17 | 2000-07-04 | Cray Research, Inc. | Seralized race-free virtual barrier network |
| US6061779A (en) * | 1998-01-16 | 2000-05-09 | Analog Devices, Inc. | Digital signal processor having data alignment buffer for performing unaligned data accesses |
| US6085254A (en) * | 1998-04-10 | 2000-07-04 | Telefonaktiebolaget Lm Ericsson (Publ) | Dynamic size alteration of memory files |
| US6401189B1 (en) | 1998-08-05 | 2002-06-04 | Michael J. Corinthios | General base state assignment for optimal massive parallelism |
| US6216174B1 (en) | 1998-09-29 | 2001-04-10 | Silicon Graphics, Inc. | System and method for fast barrier synchronization |
| US6275900B1 (en) | 1999-01-27 | 2001-08-14 | International Business Machines Company | Hybrid NUMA/S-COMA system and method |
| US7472215B1 (en) | 1999-03-31 | 2008-12-30 | International Business Machines Corporation | Portable computer system with thermal enhancements and multiple power modes of operation |
| US6345306B1 (en) * | 1999-05-05 | 2002-02-05 | International Business Machines Corporation | Packager apparatus and method for physically and logically packaging and distributing items in a distributed environment |
| US6609131B1 (en) | 1999-09-27 | 2003-08-19 | Oracle International Corporation | Parallel partition-wise joins |
| US6549931B1 (en) | 1999-09-27 | 2003-04-15 | Oracle Corporation | Distributing workload between resources used to access data |
| US6751698B1 (en) * | 1999-09-29 | 2004-06-15 | Silicon Graphics, Inc. | Multiprocessor node controller circuit and method |
| US6674720B1 (en) | 1999-09-29 | 2004-01-06 | Silicon Graphics, Inc. | Age-based network arbitration system and method |
| US6643754B1 (en) * | 2000-02-15 | 2003-11-04 | International Business Machines Corporation | System and method for dynamically allocating computer memory |
| US7260543B1 (en) | 2000-05-09 | 2007-08-21 | Sun Microsystems, Inc. | Automatic lease renewal with message gates in a distributed computing environment |
| US7111163B1 (en) | 2000-07-10 | 2006-09-19 | Alterwan, Inc. | Wide area network using internet with quality of service |
| DE10049498A1 (en) * | 2000-10-06 | 2002-04-11 | Philips Corp Intellectual Pty | Digital home network with distributed software system having virtual memory device for management of all storage devices within network |
| US7401161B2 (en) * | 2000-12-18 | 2008-07-15 | Sun Microsystems, Inc. | High performance storage array interconnection fabric using multiple independent paths |
| US7072976B2 (en) * | 2001-01-04 | 2006-07-04 | Sun Microsystems, Inc. | Scalable routing scheme for a multi-path interconnection fabric |
| US7007189B2 (en) * | 2001-05-07 | 2006-02-28 | Sun Microsystems, Inc. | Routing scheme using preferred paths in a multi-path interconnection fabric in a storage network |
| US20030037061A1 (en) * | 2001-05-08 | 2003-02-20 | Gautham Sastri | Data storage system for a multi-client network and method of managing such system |
| US6832301B2 (en) * | 2001-09-11 | 2004-12-14 | International Business Machines Corporation | Method for recovering memory |
| US6901491B2 (en) * | 2001-10-22 | 2005-05-31 | Sun Microsystems, Inc. | Method and apparatus for integration of communication links with a remote direct memory access protocol |
| US7124410B2 (en) * | 2002-01-09 | 2006-10-17 | International Business Machines Corporation | Distributed allocation of system hardware resources for multiprocessor systems |
| GB2419006B (en) * | 2002-04-22 | 2006-06-07 | Micron Technology Inc | Providing a register file memory with local addressing in a SIMD parallel processor |
| US7346690B1 (en) | 2002-05-07 | 2008-03-18 | Oracle International Corporation | Deferred piggybacked messaging mechanism for session reuse |
| US7797450B2 (en) * | 2002-10-04 | 2010-09-14 | Oracle International Corporation | Techniques for managing interaction of web services and applications |
| US7085897B2 (en) * | 2003-05-12 | 2006-08-01 | International Business Machines Corporation | Memory management for a symmetric multiprocessor computer system |
| US7379424B1 (en) | 2003-08-18 | 2008-05-27 | Cray Inc. | Systems and methods for routing packets in multiprocessor computer systems |
| US7921262B1 (en) | 2003-12-18 | 2011-04-05 | Symantec Operating Corporation | System and method for dynamic storage device expansion support in a storage virtualization environment |
| US7685354B1 (en) * | 2004-06-30 | 2010-03-23 | Sun Microsystems, Inc. | Multiple-core processor with flexible mapping of processor cores to cache banks |
| US7873776B2 (en) * | 2004-06-30 | 2011-01-18 | Oracle America, Inc. | Multiple-core processor with support for multiple virtual processors |
| KR100591371B1 (en) * | 2005-03-23 | 2006-06-20 | 엠텍비젼 주식회사 | Method for varying size of partitioned blocks of shared memory and portable terminal having shared memory |
| US7493400B2 (en) | 2005-05-18 | 2009-02-17 | Oracle International Corporation | Creating and dissolving affinity relationships in a cluster |
| US8037169B2 (en) * | 2005-05-18 | 2011-10-11 | Oracle International Corporation | Determining affinity in a cluster |
| US20070011557A1 (en) | 2005-07-07 | 2007-01-11 | Highdimension Ltd. | Inter-sequence permutation turbo code system and operation methods thereof |
| US7856579B2 (en) | 2006-04-28 | 2010-12-21 | Industrial Technology Research Institute | Network for permutation or de-permutation utilized by channel coding algorithm |
| US7797615B2 (en) | 2005-07-07 | 2010-09-14 | Acer Incorporated | Utilizing variable-length inputs in an inter-sequence permutation turbo code system |
| US7814065B2 (en) | 2005-08-16 | 2010-10-12 | Oracle International Corporation | Affinity-based recovery/failover in a cluster environment |
| US9176741B2 (en) * | 2005-08-29 | 2015-11-03 | Invention Science Fund I, Llc | Method and apparatus for segmented sequential storage |
| US20160098279A1 (en) * | 2005-08-29 | 2016-04-07 | Searete Llc | Method and apparatus for segmented sequential storage |
| US8468356B2 (en) | 2008-06-30 | 2013-06-18 | Intel Corporation | Software copy protection via protected execution of applications |
| US9086913B2 (en) | 2008-12-31 | 2015-07-21 | Intel Corporation | Processor extensions for execution of secure embedded containers |
| US9171044B2 (en) * | 2010-02-16 | 2015-10-27 | Oracle International Corporation | Method and system for parallelizing database requests |
| US8793429B1 (en) * | 2011-06-03 | 2014-07-29 | Western Digital Technologies, Inc. | Solid-state drive with reduced power up time |
| US9747287B1 (en) * | 2011-08-10 | 2017-08-29 | Nutanix, Inc. | Method and system for managing metadata for a virtualization environment |
| US9009106B1 (en) | 2011-08-10 | 2015-04-14 | Nutanix, Inc. | Method and system for implementing writable snapshots in a virtualized storage environment |
| US8601473B1 (en) | 2011-08-10 | 2013-12-03 | Nutanix, Inc. | Architecture for managing I/O and storage for a virtualization environment |
| US8549518B1 (en) | 2011-08-10 | 2013-10-01 | Nutanix, Inc. | Method and system for implementing a maintenanece service for managing I/O and storage for virtualization environment |
| US9652265B1 (en) | 2011-08-10 | 2017-05-16 | Nutanix, Inc. | Architecture for managing I/O and storage for a virtualization environment with multiple hypervisor types |
| US9772866B1 (en) | 2012-07-17 | 2017-09-26 | Nutanix, Inc. | Architecture for implementing a virtualization environment and appliance |
| US9268707B2 (en) | 2012-12-29 | 2016-02-23 | Intel Corporation | Low overhead paged memory runtime protection |
| US9417903B2 (en) | 2013-06-21 | 2016-08-16 | International Business Machines Corporation | Storage management for a cluster of integrated computing systems comprising integrated resource infrastructure using storage resource agents and synchronized inter-system storage priority map |
| US20170078367A1 (en) * | 2015-09-10 | 2017-03-16 | Lightfleet Corporation | Packet-flow message-distribution system |
| US20170108911A1 (en) * | 2015-10-16 | 2017-04-20 | Qualcomm Incorporated | System and method for page-by-page memory channel interleaving |
Family Cites Families (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3820079A (en) * | 1971-11-01 | 1974-06-25 | Hewlett Packard Co | Bus oriented,modular,multiprocessing computer |
| US3796996A (en) * | 1972-10-05 | 1974-03-12 | Honeywell Inf Systems | Main memory reconfiguration |
| GB1411182A (en) * | 1973-01-04 | 1975-10-22 | Standard Telephones Cables Ltd | Data processing |
| US4228496A (en) * | 1976-09-07 | 1980-10-14 | Tandem Computers Incorporated | Multiprocessor system |
| US4174514A (en) * | 1976-11-15 | 1979-11-13 | Environmental Research Institute Of Michigan | Parallel partitioned serial neighborhood processors |
| US4149242A (en) * | 1977-05-06 | 1979-04-10 | Bell Telephone Laboratories, Incorporated | Data interface apparatus for multiple sequential processors |
| US4285040A (en) * | 1977-11-04 | 1981-08-18 | Sperry Corporation | Dual mode virtual-to-real address translation mechanism |
| US4254463A (en) * | 1978-12-14 | 1981-03-03 | Rockwell International Corporation | Data processing system with address translation |
| US4280176A (en) * | 1978-12-26 | 1981-07-21 | International Business Machines Corporation | Memory configuration, address interleaving, relocation and access control system |
| US4371929A (en) * | 1980-05-05 | 1983-02-01 | Ibm Corporation | Multiprocessor system with high density memory set architecture including partitionable cache store interface to shared disk drive memory |
| US4442484A (en) * | 1980-10-14 | 1984-04-10 | Intel Corporation | Microprocessor memory management and protection mechanism |
| US4509140A (en) * | 1980-11-10 | 1985-04-02 | Wang Laboratories, Inc. | Data transmitting link |
| US4414624A (en) * | 1980-11-19 | 1983-11-08 | The United States Of America As Represented By The Secretary Of The Navy | Multiple-microcomputer processing |
| JPS57162056A (en) * | 1981-03-31 | 1982-10-05 | Toshiba Corp | Composite computer system |
| JPS58149551A (en) * | 1982-02-27 | 1983-09-05 | Fujitsu Ltd | Storage controlling system |
| JPS58154059A (en) * | 1982-03-08 | 1983-09-13 | Omron Tateisi Electronics Co | Memory access system of parallel processing system |
| US4648035A (en) * | 1982-12-06 | 1987-03-03 | Digital Equipment Corporation | Address conversion unit for multiprocessor system |
| US4577274A (en) * | 1983-07-11 | 1986-03-18 | At&T Bell Laboratories | Demand paging scheme for a multi-ATB shared memory processing system |
| US4591975A (en) * | 1983-07-18 | 1986-05-27 | Data General Corporation | Data processing system having dual processors |
-
1984
- 1984-10-24 US US06/664,131 patent/US4754394A/en not_active Expired - Lifetime
-
1985
- 1985-08-20 PH PH32816A patent/PH25478A/en unknown
- 1985-09-20 CA CA000491267A patent/CA1236588A/en not_active Expired
- 1985-09-24 JP JP60208985A patent/JPS61103258A/en active Granted
- 1985-10-14 CN CN85107534.7A patent/CN1004307B/en not_active Expired
- 1985-10-15 KR KR1019850007590A patent/KR910001736B1/en not_active IP Right Cessation
- 1985-10-17 DE DE8585113174T patent/DE3586389T2/en not_active Expired - Fee Related
- 1985-10-17 EP EP85113174A patent/EP0179401B1/en not_active Expired - Lifetime
- 1985-10-21 GB GB08525903A patent/GB2165975B/en not_active Expired
- 1985-10-24 IN IN838/MAS/85A patent/IN166397B/en unknown
-
1988
- 1988-03-16 US US07/168,721 patent/US4980822A/en not_active Expired - Lifetime
-
1990
- 1990-03-29 HK HK236/90A patent/HK23690A/en unknown
-
1995
- 1995-06-08 HK HK89995A patent/HK89995A/en not_active IP Right Cessation
Also Published As
| Publication number | Publication date |
|---|---|
| GB2165975B (en) | 1988-07-20 |
| US4754394A (en) | 1988-06-28 |
| KR860003553A (en) | 1986-05-26 |
| IN166397B (en) | 1990-04-28 |
| HK23690A (en) | 1990-04-06 |
| DE3586389T2 (en) | 1993-03-04 |
| GB8525903D0 (en) | 1985-11-27 |
| KR910001736B1 (en) | 1991-03-22 |
| GB2165975A (en) | 1986-04-23 |
| EP0179401A2 (en) | 1986-04-30 |
| HK89995A (en) | 1995-06-16 |
| DE3586389D1 (en) | 1992-08-27 |
| EP0179401B1 (en) | 1992-07-22 |
| JPS61103258A (en) | 1986-05-21 |
| US4980822A (en) | 1990-12-25 |
| EP0179401A3 (en) | 1989-09-13 |
| CN1004307B (en) | 1989-05-24 |
| PH25478A (en) | 1991-07-01 |
| CN85107534A (en) | 1987-04-15 |
| JPH0520776B2 (en) | 1993-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA1236588A (en) | Dynamically allocated local/global storage system | |
| US5581765A (en) | System for combining a global object identifier with a local object address in a single object pointer | |
| EP0817059B1 (en) | Auxiliary translation lookaside buffer for assisting in accessing data in remote address spaces | |
| US5684993A (en) | Segregation of thread-specific information from shared task information | |
| US5897664A (en) | Multiprocessor system having mapping table in each node to map global physical addresses to local physical addresses of page copies | |
| EP2510444B1 (en) | Hierarchical translation tables control | |
| EP0737338B1 (en) | Address translation for massively parallel processing systems | |
| US7577816B2 (en) | Remote translation mechanism for a multinode system | |
| US5694567A (en) | Direct-mapped cache with cache locking allowing expanded contiguous memory storage by swapping one or more tag bits with one or more index bits | |
| US5940870A (en) | Address translation for shared-memory multiprocessor clustering | |
| EP1096385B1 (en) | A method and apparatus for forming an entry address | |
| US4991088A (en) | Method for optimizing utilization of a cache memory | |
| US6055617A (en) | Virtual address window for accessing physical memory in a computer system | |
| US4758946A (en) | Page mapping system | |
| CA2083634C (en) | Method and apparatus for mapping page table trees into virtual address space for address translation | |
| US5897660A (en) | Method for managing free physical pages that reduces trashing to improve system performance | |
| JPH03220644A (en) | Computer apparatus | |
| US20190018777A1 (en) | Address translation cache partitioning | |
| EP0745940B1 (en) | An apparatus and method for providing a cache indexing scheme less susceptible to cache collisions | |
| JPH0812636B2 (en) | Virtual memory control type computer system | |
| RU2396592C2 (en) | Method for arrangement of globally adressed common memory in multiprocessor computer | |
| Henskens et al. | A capability-based distributed shared memory | |
| RU2487398C1 (en) | Method of creating virtual memory and device for realising said method | |
| EP1396790A2 (en) | Remote translation mechanism of a virtual address from a source a node in a multi-node system | |
| US20130262790A1 (en) | Method, computer program and device for managing memory access in a multiprocessor architecture of numa type |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| MKEX | Expiry |