CN105045724A - Space allocation based wear leveling method for nonvolatile memory - Google Patents

Abstract

The present invention discloses a space allocation based wear leveling method for a nonvolatile memory. The method comprises the following steps of: 1. making statistics on the number of write operations of each variable and the dimension of the variable; 2. allocating an appropriate space dimension for each variable according to the unoccupied space dimension to enable the maximum number of the write operations in a nonvolatile memory to be reduced as far as possible; and 3. executing a recompiled program in an embedded system to realize wear leveling of the nonvolatile memory. The space allocation based wear leveling method for the nonvolatile memory has the following advantages: 1. the method has very short execution time and very low storage overhead; 2. no hardware overhead is required; 3. for a variable level, the method has a better fine grain property and can realize better wear leveling; and 4. the method does not need to depend on any hardware and has very high practicality.

Description

A kind of nonvolatile memory abrasion equilibrium method based on allocation of space

Technical field

The invention belongs to computer memory technical field, be specifically related to a kind of abrasion equilibrium method of nonvolatile memory.

Background technology

Phase transition storage (Phasechangememory, PCM) a kind of novel nonvolatile memory is belonged to, it have non-volatile, low in energy consumption, read at a high speed, store many remarkable characteristics such as intensive, extensively thought that most probable substitutes the storer of FLASH and DRAM, and become the main flow internal memory in following embedded system.But PCM is limited to its limited write durability.A typical PCM cell only can tolerate 10 ⁷to 10 ⁹secondary write operation, otherwise it will lose the ability of change state.The write operation of typical embedded program, usually writes on internal memory in the extremely unbalanced mode of one, such as: most of write operation concentrates on a small amount of variable.Write operation is variable frequently, large, erasable to this position continually to the position wearing and tearing of depositing this variable, accelerates the wearing and tearing of PCM.If do not have suitable management, PCM chip may be damaged in tens seconds.

In order to extend the life-span of PCM, prior art adopts the method reducing and write number of times or obtain abrasion equilibrium.For the write operation number of times reduced on PCM, because write operation number of times total on PCM does not directly determine the life-span of PCM, the life-span of PCM is depended on PCM is write position the most frequently, on PCM any unit write number of times exceed it write durability restriction, the reliability of PCM becomes even worse, and may make a mistake, therefore when the write operation of program is inconsistent, this reduces write operation number of times is unable.About abrasion equilibrium method, namely write operation is distributed on PCM uniformly, the core concept of abrasion equilibrium method is the optimization of system architecture level, such as periodically the region frequently write and the mapping of field infrequently write, or randomness ground exchange address spaces, makes each position be write number of times consistent.But due to very high exchange frequency, huge writing expense and run expense may occur, and mostly all needs the support of hardware and operating system, thus limits the practical application of phase transition storage PCM in embedded system.

Summary of the invention

For the deficiencies in the prior art, technical matters to be solved by this invention is the aspect from software compiler, according to access module, a kind of nonvolatile memory abrasion equilibrium method based on allocation of space is provided, it effectively can improve phase transition storage serviceable life in embedded systems, and there is very low time and storage overhead, without any need for hardware supported.

Technical matters to be solved by this invention is realized by such technical scheme, and it comprises the following steps:

Step 1, count the write operation number of times of each variable and the size of variable;

Step 2, be that each variable distributes suitable space size according to unappropriated space size, the maximum write operation number of times on nonvolatile memory is reduced as far as possible;

Step 3, compiled program, embedded system to perform, to obtain the abrasion equilibrium on nonvolatile memory.

Tool of the present invention has the following advantages: above-mentioned treatment step completes in compilation time, has very low execution time and storage overhead; Because do not need the support of operating system or hardware, there is no hardware spending; Prior art the present invention is directed to variable level for page level, has better fine granularity, so can obtain better abrasion equilibrium; The present invention does not need to rely on any hardware, has very high practicality.

Accompanying drawing explanation

Accompanying drawing of the present invention is described as follows:

Fig. 1 is the schematic diagram in the present invention, variable being extended to array;

Fig. 2 is allocation of space process flow diagram of the present invention;

Fig. 3 decision function judge of the present invention decision flowchart.

Embodiment

Be that the invention will be further described for embodiment below in conjunction with accompanying drawing with phase transition storage:

The present invention proposes a kind of space allocation method, namely space allocation method is that each variable is expanded to the array manifold varied in size, make the write operation amount on a position can expand to multiple position equably, thus obtain a write operation distribution balanced on phase transition storage.

As shown in Figure 1, (a) is extended to the schematic diagram of an array for canonical variable, and the array that (b) is one dimension is expanded into the schematic diagram of two-dimensional array, and write operation amount larger position is like this diffused into many times of spaces.C () to have the schematic diagram of n*w write operation for variable in storage unit 1, (d) write operation is diffused into n unit uniformly for n*w time, and each unit is only write w time.So first this method invention is concentrated and is made each unit in each canonical variable or data structure have a sizeable space, the namely value of middle h and Fig. 1 (b) of Fig. 1 (a) middle i, j, k.Then, we recompile program, make the write operation of variable be distributed to these array manifold uniformly.

The present invention includes following steps:

Step 1, count the write operation number of times of each variable and the size of variable;

Amendment compiler, compiles program source code, calculates the size of variable, and record the write operation number of times of each variable.

Step 2, be that each variable distributes suitable space size according to unappropriated space size, the maximum write operation number of times on phase transition storage is reduced as far as possible;

The variable frequently write is extended to the array of different size, for each array set threshold values, whenever the number of times write a position exceedes this threshold values, Data Migration to next position, and next for the index point of array position, make the write operation amount on a position can expand to multiple position equably, thus obtain one balance write distribution;

Step 3, compiled program, embedded system to perform, to obtain the abrasion equilibrium on phase transition storage.

Fig. 2 is space size allocation algorithm process flow diagram of the present invention.This flow process starts from step 301, then:

In step 302, input will leave write operation number of times and the variable size of the variable on phase transition storage in, and the size of phase transition storage,

In step 303, determine the next T of threshold values _minbe exactly the write operation number of times of next time all variablees of threshold values is distributed on phase transition storage uniformly, total the write operation number of times equaling variable divides exactly the size in phase transition storage;

In step 304, determine the previous T of threshold values _max, the previous of threshold values is exactly the maximum write operation number of times by original mode variable left on phase transition storage, namely the write operation number of times of write operation variable the most frequently;

Next, we make use of the basic thought of binary chop algorithm.Binary chop algorithm, also known as binary search, takes full advantage of the order relation between element, adopts divide-and-conquer strategy.Thus, we improve binary chop algorithm, search a qualified minimum threshold values, make any storage unit of phase transition storage all be less than minimum threshold values, thus prolongation phase transition storage is in the serviceable life of embedded system.Basic thought is, in the valued space of threshold value, gets the next T of threshold value _minas tested object, if T _minsatisfy condition, then search successfully; If T _mindo not satisfy condition, then the intermediate value of test threshold.If intermediate value satisfies condition, continue to search in the left half-court of intermediate value; If intermediate value does not satisfy condition, then continue to search in the right half-court of intermediate value.Continuous repetition said process, until find.

In step 305, by the next T of decision function judge decision threshold _minwhether satisfy condition, this decision function judge introduces in figure 3; If satisfy condition, then perform step 306; Otherwise, perform step 307;

In step 306, the next T of threshold values _minsatisfy condition, then T _minfor best threshold values, threshold values T=T namely undetermined _min;

In step 307, give the space size of each variable according to best threshold values dispensed, the principle of distributing array manifold has just held the write operation of variable, and a few space then can not hold, and vmi is the integral multiple of si, can try to achieve the array manifold size of each variable

In formula, T is threshold values undetermined, and Wi is the write operation number of times of variable, and si is variable size; Then, step 313 is performed

In step 308, at the next T of threshold values _mindo not satisfy condition down, then by the previous T of decision function judge decision threshold _maxwhether satisfy condition; If met, then perform step 310; Otherwise, perform step 309;

In step 309, report an error low memory, performs step 313;

In step 310, judge intermediate value ((T by decision function judge _max+ T _min)/2) whether satisfy condition, if satisfy condition, then perform step 311; Otherwise, perform step 312;

In step 311, reduce hunting zone, intermediate value is assigned to the previous T of threshold values _max, T _max=(T _max+ T _min)/2), increase the next T of threshold values _min, T _min=t _min+ 1; Turn and perform step 305;

In step 312, the previous of threshold values gets T _max; The next T of intermediate value threshold values _min=(T _max+ T _min)/2)+1, turn and perform step 305;

In step 313, EOP (end of program).

As shown in Figure 3, this flow process starts from step 401 to above-mentioned decision function judge determination flow, then:

In step 402, input the write operation number of times W of threshold values T undetermined, variable _i, variable size S _iwith the space S of phase transition storage _pCM;

In step 403, according to the information of threshold values T undetermined and variable, calculate array manifold size needed for each variable

In step 404, need array manifold size according to each variable, calculate whole program requisite space total amount

In formula, n is total number of variable, and i is the sequence number of variable, vm _iit is the space size of i-th variable;

In step 405, judge that whether variable requisite space total amount RM is than actual transformation storage space S _pCMgreatly; If so, then threshold values T undetermined is too small, performs step 406; If not, then continue judgement and write expense, perform step 407;

In step 406, export false; Perform step 411

In step 407, variable write operation sum equals variable and writes the summation that number of times is multiplied by variable size, $WT=\underset{i=1}{\overset{n}{\Σ}}{S}_i\×W_i;$

In step 408, calculate write operation expense: because each movement can cause reposition that once extra write operation occurs, so the vacuum space size that write operation expense equals to distribute deducts variable space size, $WO=\underset{i=1}{\overset{n}{\Σ}}({\mathrm{vm}}_i-S_i);$

In step 409, limit the ratio value of write operation expense and variable write operation sum, make to obtain abrasion equilibrium in restriction write operation expense: in this method invention, the ratio value of setting write operation expense is 1%; If exceed this ratio value 1%, then threshold values T is too small, performs step 306; If do not exceed this ratio 1%, then perform step 410

In step 410, export true;

In step 411, EOP (end of program).

Claims (3)

1., based on a nonvolatile memory abrasion equilibrium method for allocation of space, it is characterized in that, comprise the following steps:

Step 1, count the write operation number of times of each variable and the size of variable;

Step 2, be that each variable distributes suitable space size according to unappropriated space size, the maximum write operation number of times on nonvolatile memory is reduced as far as possible;

Step 3, compiled program, embedded system to perform, to obtain the abrasion equilibrium on nonvolatile memory.

2. the nonvolatile memory abrasion equilibrium method based on allocation of space according to claim 1, is characterized in that, described variable space size allocation algorithm comprises the following steps:

Step 1) determine the next T of threshold values _minwith the previous T of threshold values _max;

Step 2) by the next T of decision function judge decision threshold _minwhether satisfy condition, in this way, then threshold values T=T undetermined _min; The array manifold size of variable

In formula, T is threshold values undetermined, and Wi is the write operation number of times of variable, and si is variable size;

Step 3) by the previous T of decision function judge decision threshold _maxwhether satisfy condition, as no, report an error low memory;

Step 4) the previous T of threshold values _maxsatisfy condition, judge intermediate value ((T by decision function judge _max+ T _min)/2) whether satisfy condition, in this way, intermediate value is assigned to the previous T of threshold values _max, T _max=(T _max+ T _min)/2), increase the next T of threshold values _min, T _min=T _min+ 1; Proceed to step 2); Step 5) intermediate value ((T _max+ T _min)/2) do not satisfy condition, then the previous of threshold values gets T _max; The next T of intermediate value threshold values _min=(T _max+ T _min)/2)+1, proceed to step 2).

3. the nonvolatile memory abrasion equilibrium method based on allocation of space according to claim 2, is characterized in that, described decision function judge judges to comprise the following steps:

Step (1) according to threshold values T to be determined, the write operation number of times W of variable _i, variable size S _iwith the size S of phase transition storage _pCM; Calculate array manifold size needed for each variable calculate whole program requisite space total amount

In formula, n is total number of variable, and i is the sequence number of variable, vm _iit is the space size of i-th variable;

Step (2) judges that whether variable requisite space total amount RM is than actual transformation storage space S _pCMgreatly; If so, then threshold values T undetermined is too small, exports false;

Step (3) at variable requisite space total amount RM than actual transformation storage space S _pCMwhen little, calculate variable write operation sum calculate extra write operation expense $WO=\underset{i=1}{\overset{n}{\Σ}}({\mathrm{vm}}_i-S_i);$

Step (4) limits the ratio value of write operation expense and variable write operation sum, if exceed this ratio value, then threshold values T is too small, exports false; If do not exceed this ratio, then export true.