CN103984599B - Method for improving utilization rate of large pages of operating system - Google Patents
Method for improving utilization rate of large pages of operating system Download PDFInfo
- Publication number
- CN103984599B CN103984599B CN201410146873.3A CN201410146873A CN103984599B CN 103984599 B CN103984599 B CN 103984599B CN 201410146873 A CN201410146873 A CN 201410146873A CN 103984599 B CN103984599 B CN 103984599B
- Authority
- CN
- China
- Prior art keywords
- heap
- heap top
- top position
- variable
- page
- 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.)
- Active
Links
Landscapes
- Memory System Of A Hierarchy Structure (AREA)
Abstract
The invention discloses a method for improving the utilization rate of large pages of an operating system. The method comprises the steps that (1) the system adds a variable a in a virtual address space data structure of each process, and each variable a is recorded as a heap top position of a distributed virtual address of the corresponding process; (2) in the process starting, the system initiates a to 0; when any process calls a heap top setting function for the application of an internal storage, a heap top parameter b is transmitted into the system; (3) the system calculates a value c after the alignment of upward large pages of the heap top positions according to a previously-requested heap top position and a currently-requested heap top position b of the process; (4) the system assigns the current value of a to a highest heap address variable of the recorded and distributed internal storage of the process, and compares the current value with c; if the current value is smaller than c, a stack space of the process is increased according to a difference value between the current value and c; if the current value is smaller than c, the release of the internal storage is implemented; if the current value is equal to c, the calling of the internal storage is not implemented. According to the method, the utilization rate of the large pages is fully improved, so that the performance of a program is improved.
Description
Technical field
The present invention relates to a kind of method for improving operating system big page utilization rate, belongs to operating system memory management technique neck
Domain.
Technical background
Paging system is the common features of modern CPU.It establishes a virtual memory space from program to machine
Physical memory between mapping.Internal memory is managed in units of page.Tie including many CPU systems including Intel x86
Structure realizes a kind of big page mechanism, supports various different size of Physical Page granularities such as 4k, 2M, 4M.
Current operating system (such as Linux) is for simplicity, it is common to use the page of 4K sizes, this way is caused
The scope of the virtual address that each mapping is covered only has 4K.If necessary to access internal memory of the segment length for 2M, operating system
Need to process 512 page faults, each page faults distribute 4K internal memories.
By taking (SuSE) Linux OS as an example, Linux starts to support transparent big page from 2.6.38.Operating system is set up
Size is the virtual address of 2M (using being 4M during 32-bit operating system) to physical address map, referred to as big page mapping (abbreviation
Big page).Calculating and the Cache expenses that TLB failures and CPU occur when addressing can be reduced using big page.The mechanism contains several
Individual function, such as arrangement of Physical Page fragment, to the physical memory of big page alignment that ensures to have enough for using.Occurring to lack
Page is judged when interrupting (page fault), is set up using big page if conditions permit and mapped.It is a set of in big page
Deposit operation instrument, there is provided the base memory such as copy-on-write is operated.
The big page module mainly for program in privately owned anonymous internal memory be optimized that (internal memory that program is used is big absolutely
Part is used in the way of private anonymity.And big page is not appropriate for using in other circumstances.Shared drive uses big
Page can increase the expense of copy-on-write, and being mapped to the internal memory of file can increase the burden of caching using big page).The optimization of big page is not
It is compulsory, but adopts the logic " if meeting condition, then use big page ".Linux is weighed with " big page utilization rate "
The usage degree of big page module.It is equal to：In the privately owned anonymous internal memory of one process, with the interior poke that big page form is distributed, remove
With the total privately owned anonymous internal memory usage amount of program.
Distribution of the operating system to internal memory is with page fault drivings.Each process have oneself address space (
Correspondence mm_struct structures in linux system), the virtual address and virtual address of an address space maintenance process is to thing
The mapping of reason address.When one process has just been set up or applied for internal memory, process can call function application some virtual memorys.
At this time operating system can't at once distribute physical memory, but set up a virtual memory area (hereinafter referred to as
Vma), represent and say, this section of virtual address has been used.Each mm_struct includes several vma, and each vma
Some mm_struct must be belonged to.When process is actual reads and writes this section of virtual address, because without storage allocation,
It is bound to page fault.Operating system navigates to correspondence according to the address that page fault occur by mm_struct
Vma, according to occur page fault the reason for and vma attribute determine how to process page fault.If for example accessed
The virtual address for arriving is not inside any vma, or the region for accessing is not readable, then be exactly to have accessed illegal ground
Location, returns segment fault, and kills process.If write operation have accessed a vma indicate it is writeable, but page table subscript
When remembering not writeable address, then be the page for having write copy-on-write, operating system can be a by page copy and be set to again
The writeable page.If the position for accessing is not set up virtual address to the mapping of physical address in vma, that explanation is also
Without storage allocation, now according to the type of vma, distribute physical memory, set up mapping.
Setting up when internal memory maps can decide whether to use big page according to the situation of vma, in general, if on reference address
The vma that one section of 2M Virtual Space is located that lower 2M alignment is constituted is included, and this section of 2M virtual address spaces the inside is not set up
Cross mapping (two grades of page tables are sky), then linux kernel can distribute big page for this section of virtual address.If upper strata application internal memory
When do not account for the demand of big page alignment, the vma Jing of application often will not 2M alignment, cannot thus use big page.
Due to the inherent limitations of big page Paging system, it is desirable to which all of physical address and virtual address must be alignment.
If 64 bit manipulation systems are mapped using the page of 2M, then the necessary 2M alignment of the head and the tail of virtual address, the head of physical address
Tail is also required to 2M alignment.For 32-bit operating system, then could use in the case where physical address and virtual address 4M align
Big page.The management of Physical Page is that operating system oneself is safeguarded, defragmentation work(is already had in the realization of transparent big page
Can, it is ensured that the Physical Page for having sufficiently large page alignment is used to set up big page mapping.Virtual address is usually user's application, no
Can take into account whether bottom has the support of transparent big page, therefore be frequently not big page alignment during virtual address application.For example it is existing
This problem is there is in the transparent big pages of Linux.The application of application program virtual address is unsatisfactory under many circumstances big page pair
Neat requirement, this causes linux kernel not distribute big page, then using little page.This way is caused in many cases,
Linux kernel will not distribute big page to program, and big page utilization rate is not high.
The content of the invention
For technical problem present in prior art, it is an object of the invention to provide a kind of improve operating system big page
The method of utilization rate, can effectively lift the utilization rate of big page, and then improve the performance of program.
The technical scheme is that：
A kind of method for improving operating system big page utilization rate, its step is：
1) system increases by a variable a in the virtual address space data structure of each process, for being recorded as process
The heap top position of the allocated virtual address；And change heap starting and function is set so as to the big page alignment of heap initial address of return；
2) during process initiation, variable a is initialized as 0 by system；Calling heap top to arrange function when the process carries out internal memory Shen
Please when, to incoming a pile top parameter b of system, b be request arrange new heap top position；
3) system is calculated to meet and is somebody's turn to do according to the heap top position of the last request of the process and the heap top position b of current request
The heap top position of process current memory demand value c upwards after big page alignment；
4) currency of process variable a is assigned to system the record storage allocation highest heap address change of the process
Amount, and it is compared with c：If less than c, then system increases the stack space of the process according to the two difference；If big
In c, then internal memory release is carried out according to the two difference, reduce the stack space of the process, if equal, do not carry out internal memory tune
With.
Further, if the heap top position of the last request of the process more than current request heap top position b, from upper
Choose a smaller value between the heap top position once asked and c, systems inspection b to the virtual address between the smaller value whether
Physical memory is assigned, if be allocated, by this section of physical memory clear 0.
Further, if the value of variable a is 0, i.e. system also to the Heap Allocation of the process does not cross internal memory, then system should
The heap top position value of the last request of process is the initial address of the heap of the process.
Further, the heap original position that the heap starting setting function is returned sets to deduct one again after big page alignment upwards
Definite value.
Further, the variable a is long shaping address variable.
The present invention solves the problems, such as that alignment core concept is：When the system for receiving internal memory application in operating system is called, repair
The logic of changed handling internal memory application, on the premise of user's application is met, a part of internal memory of additional allocation, so as to set up alignment
vma.When distribution extra memory, effectively utilizes these extra memorys are required consideration for how, reduce internal memory and waste；For can not
The waste for avoiding, needs to weigh between systematic function and memory cost, with optimum scheme comparison.Invention needs modification operation system
The function of system kernel so that even if the internal memory application on upper strata is unsatisfactory for alignment requirements, kernel can create the vma of big page alignment,
Meet upper strata application, and realize big page.During process, needs make done modification and existing mechanism compatible, and consider
The overhead being likely to result in.
Fig. 1 is the schematic diagram that process uses internal memory under (SuSE) Linux OS.Arrow is represented and called, when textual representation is called
The parameter for providing is provided.One C programmer, mainly has these places using internal memory：Code segment and data segment, global variable
Section, heap, stack.In process initiation, code segment, data segment (initialized global variable), bss sections (no initializtion it is complete
Office variable storage where) and stack just have been set up.They have the function for being responsible for each setting up, wherein the part for using exists
Indicate in figure.
If necessary to the dynamic application internal memory in program operation, there are two methods to use.
A kind of method is to call sys_brk functions.Each process has a heap space, for quickly applying in new
Deposit.It is a vma to pile up operating system the inside, is responsible for by a set of special function, and the low address of vma is referred to as heap bottom,
High address is referred to as heap top, and the internal memory occupation mode of this section of vma is privately owned anonymous, and user can arbitrarily visit in this section of region
Ask.User only has a kind of method operation heap, and (position at heap bottom initializes when process is set up, and whole exactly to arrange heap top
Can not change in individual process running).Internal memory if necessary to apply for K bytes, it is original heap top+K just to arrange new heap top.
Contrary, if it is desired to releasing memory, then the internal memory at heap top point can only be discharged, must be than original heap by the way that new heap top is arranged
Top is little, and operating system will be automatically releasable the internal memory above new heap top.The function for arranging heap top that operating system is provided is sys_
brk。
The privately owned anonymous internal memory that one program is used all is distributed by the two functions.This patent is changed by brk letters
The internal memory of number application.Purpose is, by modification, to make it produce the vma of big page alignment.During so that there are page faults, operating system
Big page can be distributed in this section of vma, so as to lift big page utilization rate.The difficult point of this problem is the upper procedure such as malloc Shen
Big page alignment please not be accounted for during internal memory, their parameter is not necessarily big page alignment.We need to be modified in bottom,
On the one hand meet the demand of upper strata internal memory application, on the other hand safeguard that vma meets the restriction of big page alignment.
For operating system, the modification of sys_brk functions is a specific vma, as long as we ensure that this vma begins
Eventually big page alignment to be ensured that and can all use big page by the internal memory of sys_brk function applications.Vma is changed by sys_brk
When, the low address of vma will not be changed, only the high address of vma can be moved back and forth.The calculation that we optimize for this characteristic Design
Method.The low address of heap correspondence vma is arranged to into big page alignment when heap is initialized, so as to ensure that side is alignd.Another
Side, the heap top pointer that user is arranged is not necessarily big page alignment.Using method be up to extend to the high address of vma greatly
Page alignment, and record the position on the actually used heap top of user.The big of high address is realized by allocating the method for internal memory in advance
Page alignment.Simultaneously the actually used position of user is have recorded, what user saw, be oneself actually used position, and is not known
More internal memories may be had been allocated for.Later during user's application internal memory, application can be started from the position on last time heap top, if one
Partial memory has been carried out distribution, then can directly use, not result in waste.This mechanism be to upper strata it is transparent, can
To realize in the case where upper layer application need not be changed.
Generally speaking, by controlling the length that heap initial address and each heap extend, it is ensured that the vma corresponding to heap is complete
Big page is used entirely.
Compared with prior art, the positive effect of the present invention is：
The performance of the computer system that the invention is lifted with less cost.It is virtual that the cost of optimization mainly expands big page
During address, extra memory cost.It can effectively lift the performance of program.The restriction of original Linux modules causes big page
Utilization rate is relatively low, in some cases, or even will not use big page completely.By the optimization, the use of big page is fully improved
Rate, and then lift the performance of program.The optimization is the optimization for kernel memory management, therefore the program limit of optimization impact
Very wide, almost all of program can be transferred through the optimization and be benefited.The advantage that will be invented with description of test below.
Using SPECCPU2006 as test program set, Linux big pages module and optimization are tested, as a result such as
Table 1：
Table 1 is test result contrast table
Perlbench to xalancbmk is 12 groups of test programs of spec2006 in table 1, when the right is to run accordingly
Between.We estimate the coverage rate of big page by the processing mode of page fault in detection linux kernel.Native is to open
Ruuning situation during the Transparenet Hugepage Support functions that Linux is carried, optimized is only to change
The effect of linux kernel, as with reference to fortune during also Linux disabling Transparenet Hugepage Support functions
Row data None.
Data display, optimization can further lift on the original basis program feature, for bzip2, gcc the two samples
Original mechanism is improved can hardly program feature, but be with the addition of after optimization, and performance is improved.For
Omnetpp, astar, xalancbmk these three examples, original big page module can have some to be lifted, increased my this portion
After the optimization for dividing, the amplitude of lifting is bigger, has exceeded original amplitude.Other examples, though there is the optimization of big page alignment not
Make performance get a promotion, but also do not allow hydraulic performance decline, it is possible to conclude, big page alignment this optimization will not bring volume
Outer performance cost.
Page faults are an important indicators for affecting program feature.System carries out the process of page faults and will produce volume
Outer expense, using big page page faults can be efficiently reduced.Here each single test program page faults is measured
Generation number of times.
Vertical coordinate is the number of times of page faults in Fig. 2, because difference benchmark memory access characteristic is different, there is larger difference
It is different, therefore vertical coordinate is exponential increasing.Abscissa is the statistics respectively of all benchmark.Three are counted altogether, are not had respectively
Using page faults number during linux kernel Transparent huge page modules；After the module；After optimization
Page faults number.As shown by data, after optimization, the number of page faults is substantially reduced, substantially 10000 times with
It is interior, can effectively reduce the number of times that program produces page faults.
Above analytic explanation, the optimization pin of the invention has good effect, can efficiently reduce the number of times of page faults, carries
High big page utilization rate, and then lift program feature.
Description of the drawings
Fig. 1 is internal memory application function call schematic diagram.
Fig. 2 is that in front and back page faults number compares after big page optimization.
Specific embodiment
The method for realizing that big page optimizes on (SuSE) Linux OS is described below.The method is based on linux kernel version
3.6.3, the modification to glibc is based on glibc-2.17, it is adaptable to the machine of 64 x86 architectures.The method is to realize greatly
One instantiation of page optimization.
Data structure mm_struct of modification record the process virtual address space, increases member variable in structure
Longallocate_brk, is recorded as the heap top position of process the allocated virtual address.It is initial in process virtual address space
During change, 0 is set the values to.
In linux system, the include/linux/mm_types.h document definitions data structure related to internal memory,
Including mm_struct.Kernel/fork.c files realize the correlation function of newly-built process in linux system, wherein
Mm_struct structures can be initialized in mm_init functions, sentence mm- is increased in this structure>allocated_
Brk=0, can be initialized as 0 by newly-increased heap top location variable.
Make the big page alignment of vma initial addresses corresponding to heap：Find the function of initialization brk segment bases.It is general right
In different architectures, there are different initialization functions.For x86 frameworks, architecture spy's functional text is being realized
In part arch/x86/kernel/process.c, the power function arch_randomize_brk of initialization brk sections is realized.
In the function, the return value for calling randomize_range functions to obtain big page alignment upwards can be caused corresponding to heap
The big page alignment of vma initial addresses.
Modification performs the logic of sys_brk functions：The operation of managing internal memory defined in mm/mmap.c files, wherein
Function SYSCALL_DEFINE1 (brk, unsigned long, the brk) system of realizing of definition calls the function of sys_brk.
When the function can calculate current heap top position and call function, user requires the heap top position for reaching.Obtain two
Behind the position of heap top, operating system determines to extend up heap or contraction heap space and releasing memory.
In sentence
Newbrk=PAGE_ALIGN (brk)；
Oldbrk=PAGE_ALIGN (mm->brk)；
In calculate original heap top position and new heap top position.After increasing the restriction of big page alignment, need to count again
Calculate the position on heap top.Add following sentence after above-mentioned sentence：
Newbrk=(brk+SUPERPAGE_MASK) ＆PMD_MASK；
if(likely(mm->Allocated_brk)) oldbrk=mm->allocated_brk；
New heap top position is the desired heap top position of user position upwards after big page alignment, original heap top position
The heap top position of sys_brk functions storage allocation was called for last time.
If calling sys_brk functions to change heap success, function can enter the sentence that label set_brk are administered, and use
In the associated member's variable for arranging mm_struct, sentence mm- is increased here>Allocated_brk=newbrk updates heap top
Pointer.
Also need to increase logic in label set_brk：Judge that current sys_brk calls whether user will reduce heap
Top, if it is, check that user reduces heap top and needs whether the virtual memory of release is assigned with physical memory, if be assigned with, will
This partial memory clear 0.Increase this step and be because that upper level applications are given tacit consent to：Call the internal memory meeting of sys_brk newly applications all
It is initialized as 0.In unmodified version, because all of operation is in units of page, during each do_munmap, can be this
All virtual addresses are all discharged to the mapping of physical address in individual region, and if the newly assigned regions of do_brk access,
Be bound to occur pagefault, into linux kernel Memory Allocation program distribute physical memory when, linux kernel will can be treated
The physical memory clear 0 of distribution.In version after the modification, because big page alignment, has partial memory not to be released, inspection is needed
Look into whether this part is allocated physical memory, if be allocated, need clear 0.
There is provided an implementation for reference：Code below is added in label set_brk
This section of code first checks for the whether situation that heap top is shunk.If meeting condition, into if sentences, in if sentences
Find_vma find by primitive justice should releasing memory office of vma areas, in units of page, using follow_page functions check
Whether virtual address is assigned with physical memory, if being assigned with internal memory, clear 0.
After increasing big page optimization, operational process description：
When process goes out beginningization, additionally to mm->Allocated_brk variables clear 0.Then arch_randomize_ is called
Brk functions initialize the initial address of heap.Modification heap starting arranges function arch_randomize_brk, when function is returned,
After by original return address upwards big page alignment, then 16K is deducted, using calculated new address as function return value.With
When afterwards routine call heap top arranges function sys_brk, incoming heap top parameter brk represents the heap top position that schemes call is arranged.It is logical
Such adjustment is crossed, if the heap memory that function is used is little, then this partial memory is less than big page aligned position, can use
Little page.Can be prevented effectively from and cause to waste using little excessive internal memory of program distribution to internal memory.It is larger for internal memory usage amount
Program, remainder can distribute big page, generally ensure preferable big page utilization rate.
Make mm->Brk represents the heap top position that a front schemes call is arranged.
Make heap top position that newbrk represents that schemes call arranges value upwards after big page alignment.
Variable oldbrk records the highest heap address of storage allocation.If mm->Allocated_brk variate-values are
0, represent that heap did not also distribute internal memory, make the initial address that oldbrk is heap；Otherwise make oldbrk=mm->allocated_
brk。
Compare the value of oldbrk and newbrk：
If oldbrk<Newbrk explanations need to expand vma, call do_brk functions to increase vma.
If oldbrk>Newbrk explanations need releasing memory, call do_munmap functions to discharge newbrk to oldbrk
Between internal memory.
Relatively brk and mm->The value of brk, if brk<mm->Brk, checks brk to min (mm->Brk, newbrk) between
Virtual address, if physical memory is assigned, then by this section of internal memory clear 0.
System is called successfully, updates the data structure, makes mm->Brk=brk, mm->Allocated_brk=newbrk, letter
Number terminates.
The big page optimization of heap can be realized by above step.
For other situations, as long as architecture supports that different size of page maps, and system realizes Dram
Distribution, it is possible to be optimized using method proposed by the present invention.Every storage allocation by way of the pointer of setting heap top
System, it is possible to utilize the method for the present invention, arranges an auxiliary variable, records the position of storage allocation；Arrange application heap to rise
The big page alignment in beginning position；And the address after alignment is recalculated when increasing and reclaiming internal memory, and the optimization is can be achieved with, reach
Lift the purpose of big page utilization rate.
It is proposed that optimization operation big page utilization rate scheme, it is mainly technically characterized by by giving birth in an operating system
Into the mode of the vma of big page alignment, it is ensured that the utilization rate of big page.Its technical scheme can be used in all support Paging systems
In operating system.Its main method is to limit heap growth and the granularity for discharging in heap part, reaches and makes operating system generate big page
Alignment vma, so as to the purpose of lift system performance.The method selectively Extended RAM so that in lift system performance
Meanwhile, effectively prevent the waste that Extended RAM may be brought.It is every by reasonable Extended RAM partition size, reach operation system
Unite and required using big page, the method so as to lift big page utilization rate, all in this patent protection domain.
Claims (4)
1. a kind of method for improving operating system big page utilization rate, its step is：
1) system increases by a variable a in the virtual address space data structure of each process, has divided for being recorded as process
Heap top position with virtual address；And change heap starting and function is set so as to the big page alignment of heap initial address of return；
2) during process initiation, variable a is initialized as 0 by system；When the process calls heap top setting function to carry out internal memory application,
To incoming a pile top parameter b of system, b is the new heap top position that request is arranged；
3) according to the heap top position of the last request of the process and the heap top position b of current request, calculating meets the process to system
The heap top position of current memory demand value c upwards after big page alignment；
4) currency of process variable a is assigned to the record of process storage allocation highest heap address variable by system, and
It is compared with c：If less than c, then system increases the stack space of the process according to the two difference；If greater than c, then
Internal memory release is carried out according to the two difference, reduces the stack space of the process, if equal, do not carried out internal memory and call；
Wherein, if the heap top position of the last request of the process is asked more than the heap top position b of current request from last
Heap top position and c between choose a smaller value, whether systems inspection b to the virtual address between the smaller value has been assigned
Physical memory, if be allocated, by this section of physical memory clear 0.
2. the method for claim 1, it is characterised in that if the value of variable a is 0, i.e. system also not to the heap of the process
Distributed internal memory, then system is by the initial address of heap that the heap top position value of the last request of the process is the process.
3. the method for claim 1, it is characterised in that the heap starting arrange the heap original position that function returns be to
A setting value is deducted again after upper big page alignment.
4. the method as described in claim 1 or 2 or 3, it is characterised in that the variable a is long shaping address variable.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410146873.3A CN103984599B (en) | 2014-04-14 | 2014-04-14 | Method for improving utilization rate of large pages of operating system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410146873.3A CN103984599B (en) | 2014-04-14 | 2014-04-14 | Method for improving utilization rate of large pages of operating system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103984599A CN103984599A (en) | 2014-08-13 |
| CN103984599B true CN103984599B (en) | 2017-05-17 |
Family
ID=51276590
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410146873.3A Active CN103984599B (en) | 2014-04-14 | 2014-04-14 | Method for improving utilization rate of large pages of operating system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103984599B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106326144B (en) * | 2015-06-24 | 2019-08-06 | 龙芯中科技术有限公司 | Method for reading data and device based on big page mapping |
| CN105893269B (en) * | 2016-03-31 | 2018-08-21 | 武汉虹信技术服务有限责任公司 | EMS memory management process under a kind of linux system |
| CN106843906B (en) * | 2017-02-22 | 2021-02-02 | 苏州浪潮智能科技有限公司 | Method and server for adjusting system page size |
| CN108664419A (en) * | 2018-04-03 | 2018-10-16 | 郑州云海信息技术有限公司 | A kind of method and its device of determining memory big page number |
| CN110109761B (en) * | 2019-05-11 | 2021-06-04 | 广东财经大学 | Method and system for managing kernel memory of operating system in user mode |
| CN112035379B (en) * | 2020-09-09 | 2022-06-14 | 浙江大华技术股份有限公司 | Method and device for using storage space, storage medium and electronic device |
| CN112596913B (en) * | 2020-12-29 | 2022-08-02 | 海光信息技术股份有限公司 | Method and device for improving performance of transparent large page of memory, user equipment and storage medium |
| CN112905497B (en) * | 2021-02-20 | 2022-04-22 | 迈普通信技术股份有限公司 | Memory management method and device, electronic equipment and storage medium |
| CN113687873B (en) * | 2021-07-30 | 2024-02-23 | 济南浪潮数据技术有限公司 | Large page memory configuration method, system and related device in cloud service page table |
| CN115061954B (en) * | 2022-08-18 | 2022-11-29 | 统信软件技术有限公司 | Missing page interrupt processing method, computing device and storage medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8417913B2 (en) * | 2003-11-13 | 2013-04-09 | International Business Machines Corporation | Superpage coalescing which supports read/write access to a new virtual superpage mapping during copying of physical pages |
| EP2449469B1 (en) * | 2009-06-29 | 2019-04-03 | Hewlett-Packard Enterprise Development LP | Hypervisor-based management of local and remote virtual memory pages |
| CN102446136B (en) * | 2010-10-14 | 2014-09-03 | 无锡江南计算技术研究所 | Self-adaptive large-page allocation method and device |
| CN103019949B (en) * | 2012-12-27 | 2015-08-19 | 华为技术有限公司 | A kind of distribution method and device writing merging Attribute Memory space |
| CN103257929B (en) * | 2013-04-18 | 2016-03-16 | 中国科学院计算技术研究所 | A kind of virutal machine memory mapping method and system |
-
2014
- 2014-04-14 CN CN201410146873.3A patent/CN103984599B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN103984599A (en) | 2014-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103984599B (en) | Method for improving utilization rate of large pages of operating system | |
| CN105893269B (en) | EMS memory management process under a kind of linux system | |
| US8866831B2 (en) | Shared virtual memory between a host and discrete graphics device in a computing system | |
| US8176282B2 (en) | Multi-domain management of a cache in a processor system | |
| EP2266040B1 (en) | Methods and systems for dynamic cache partitioning for distributed applications operating on multiprocessor architectures | |
| US8661181B2 (en) | Memory protection unit in a virtual processing environment | |
| US9183157B2 (en) | Method for creating virtual machine, a virtual machine monitor, and a virtual machine system | |
| US9280486B2 (en) | Managing memory pages based on free page hints | |
| EP2581828B1 (en) | Method for creating virtual machine, virtual machine monitor and virtual machine system | |
| CN103116556B (en) | Internal storage static state partition and virtualization method | |
| KR20120106696A (en) | Extended page size using aggregated small pages | |
| Haldar et al. | Operating systems | |
| US9208088B2 (en) | Shared virtual memory management apparatus for providing cache-coherence | |
| DE112005003736T5 (en) | Virtual Translation Buffer | |
| US20160259689A1 (en) | Managing reuse information in caches | |
| US20160259732A1 (en) | Managing reuse information for memory pages | |
| US9471509B2 (en) | Managing address-independent page attributes | |
| US10268592B2 (en) | System, method and computer-readable medium for dynamically mapping a non-volatile memory store | |
| US10013360B2 (en) | Managing reuse information with multiple translation stages | |
| CN112445767A (en) | Memory management method and device, electronic equipment and storage medium | |
| US7562204B1 (en) | Identifying and relocating relocatable kernel memory allocations in kernel non-relocatable memory | |
| CN112596913A (en) | Method and device for improving performance of transparent large page of memory, user equipment and storage medium | |
| KR20220024206A (en) | Hardware-Based Memory Compression | |
| Wu et al. | DWARM: A wear-aware memory management scheme for in-memory file systems | |
| CN116681578B (en) | Memory management method, graphic processing unit, storage medium and terminal equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |