Summary of the invention
In view of this, the application provides a kind of memorizer compiler joining method, is used for solving prior art
The problem of the area loss of middle memorizer.
To achieve these goals, it is proposed that scheme as follows:
A kind of memorizer compiler joining method, including:
Determine that signal is driven into the delay time of distalmost end memory element and the columns n of memory element by wordline
And the number n of the Fin of the driving MOS of wordline drivingWLDBetween functional relationship, wherein said nWLD
It it is a dispersion number;
Described n array storage unit is divided into default section, at nWLDSpan in, according to described letter
Number relation calculates the minimum delay time of every section of memory element, when described minimum delay time is pre-less than first
If during value, obtain the n that described minimum delay time is correspondingWLD, it is designated as preferred value, described preferred value
It is wordline in described memorizer and drives the driving MOS's of the buffer between every two sections of memory element
Fin number;
Assembled Fin is that the wordline of preferred value drives and the domain of buffer.
Preferably, above-mentioned memorizer compiler joining method, also include:
When described minimum delay time is not less than the first preset value, the equal segmentation of described n array storage unit
Number adds 1, until at nWLDSpan in, the minimum delay time of every section of memory element be less than first
Preset value.
Preferably, in above-mentioned memorizer compiler joining method, described signal is driven into by memory word line
The delay time of distalmost end memory element and the columns n of memory element and wordline drive the Fin's of MOS
Number nWLDBetween functional relationship, for:
Wherein, t0 is that signal is driven into the delay time of distalmost end memory element, R by memory word lineC_WL
For the resistance value of memory element bit line, CC_WLFor the capacitance of memory element bit line, a and b is constant.
Preferably, in above-mentioned memorizer compiler joining method, also include:
Line number m of described memorizer compiler is tracked, particularly as follows:
Calculate the resistance R of the pull-up PMOS of memorizerC_PUWith the resistance R writing drivingWRDAnd m
The resistance R of line storage unitC_BLThe ratio T of sum, wherein, described;Described nWRD
For writing the number of the Fin of the NMOS of driving, c is fixed value;
At nWRDSpan in, calculate the driving of writing making described ratio T more than the second preset value
The number of the minimum Fin of NMOS, is designated as nWRD1;
Assembled Fin is nWRD1The domain writing driving.
Preferably, in above-mentioned memorizer compiler joining method, also include:
Use sequential control circuit that line number m and the columns n of described memorizer compiler are tracked, tool
Body is:
Determine the time delay t1 of sequential control circuit of memorizer and functional module number p of inverter drive with
And the number n of the Fin of phase inverterDriverBetween functional relationship, wherein said nDriverIt it is a dispersion number;
Ask at described nDriverSpan in so that Fin's when described time delay t1 meets timing requirements
Minimal amount, is designated as nDriver1;
Assembled Fin is nDriver1Sequential control circuit domain.
Preferably, in above-mentioned memorizer compiler joining method, the time delay t1 of described sequential control circuit with
Functional module number p of inverter drive and the number n of the Fin of phase inverterDriverBetween functional relationship be Wherein, described d, e are fixed value, described Cleafcell
By the electric capacity of the functional module that phase inverter is driven, described RleafcellThe functional module driven by phase inverter
Connection resistances.
A kind of memorizer, described memorizer is spelled for using above-mentioned any one memorizer compiler joining method
The memorizer of dress.
From above-mentioned technical scheme it can be seen that memorizer compiler joining method disclosed in the present application,
By the columns n of described memorizer compiler is tracked so that the WL Driver of memory element and
The number of the Fin of Buffer is optimal value so that the size of described WL Driver and Buffer is optimum,
Therefore enhance the area utilization of memorizer, reduce area loss.
Detailed description of the invention
The circuit size being directed in memory function module of the prior art is according to the worst case of array
Depending on, thus cause the problem of the area loss of memorizer, this application discloses a kind of memorizer compiler
Joining method and memorizer.
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out
Clearly and completely describe, it is clear that described embodiment is only a part of embodiment of the present invention, and
It is not all, of embodiment.Based on the embodiment in the present invention, those of ordinary skill in the art are not doing
Go out the every other embodiment obtained under creative work premise, broadly fall into the scope of protection of the invention.
Under fin field-effect transistor (Fin Field-Effect Transistor, FinFET) technique, MOS
The channel length of device is fixed, and width can only the value of quantification.Therefore user can customize MOS device
Size, MOS device can be carried out computer control.The most original memorizer compiler skill
Art can be optimized reform under FinFET technique.
As it is shown in figure 1, the FinFET in Fig. 1 is 3D transistor (metal-oxide-semiconductor), see in Fig. 1
The channel length of 3D transistor is fixed value, width can only quantized value, the most described metal-oxide-semiconductor
Size is determined by total number of Fin: FinSum=Fin1Finger* Finger, metal-oxide-semiconductor be sized at one discrete
Function, thus can facilitate and process with computer, control domain arrangement, carry out assembled MOS
Pipe.Wordline in memorizer compiler drive (WL Driver), write driving (Write Driver), time
Sequence is followed the trail of in circuit (Tracking Path) and is provided with large-size device, and described WL Driver is used for driving
The wordline of memory element in storage array, described Write Driver is used for driving memory element in storage array
Bit line, described Tracking Path is for controlling the sequential of memorizer, and applicant finds to utilize by research
The characteristic of FinFET size discrete, can make the size of large-size device carry out line number and columns
Follow the trail of, so that the area loss reducing memorizer is possibly realized.
Fig. 2 is the long line model of the WL Driver in memorizer compiler disclosed in the embodiment of the present application.
Wherein, R shown in Fig. 2WLDThe resistance of MOS, C is driven for WL DriverWLDMOS is driven for WL Driver
Electric capacity, RC_WLFor the resistance of memory element bit line, CC_WLFor the electric capacity of memory element bit line, WL Driver
Driving altogether n array storage unit, described t0 is the signal time delay by the memory element of WL Driver to distalmost end
Time, described t0=f (nWLD, n), wherein, described nWLDThe catalogue number of the Fin of MOS is driven for WL Driver,
And described nWLDIt it is a dispersion number.
Based on described signal by the delay time t0 of the memory element of WL Driver to distalmost end and described WL
Driver drives the catalogue number n of the Fin of MOSWLDAnd the functional relationship between the columns n of memory element, this Shen
Please disclose a kind of memorizer compiler joining method, by the columns of memorizer compiler is tracked,
Reach to optimize the purpose of the PPA of memorizer compiler.
Fig. 3 is the disclosed method being tracked the columns of memorizer compiler of the embodiment of the present application.
Seeing Fig. 1, the method includes:
The columns n of described memorizer compiler is tracked, particularly as follows:
Step S301: determine that signal is driven into delay time t0 and the storage of distalmost end memory element by wordline
The number n of the Fin of the driving MOS that the columns n of unit and wordline driveWLDBetween functional relationship;
Step S302: described n array storage unit is divided into i section, i=1;
Step S303: at nWLDSpan in, be calculated the minimum time delay of n/i array storage unit
Time tn/i;
Step S304: judge that whether described minimum delay time tn/i is less than the first preset value ttarget, if
Perform step S306, if it does not, perform step S305, described first preset value ttargetSize according to use
Depending on the demand of family:
Step S305: make i=i+1, continues executing with step S303;
At nWLDSpan in, continue to be calculated the minimum delay time tn/i of n/i array storage unit,
Until judging that described minimum delay time tn/i is less than described first preset value ttarget;
Step S306: if described minimum delay time tn/i is less than the first preset value ttarget, obtain and make institute
State the number (n that wordline corresponding when minimum delay time tn/i is minima drives the Fin of MOSWLD),
It is designated as nmeet;
Step S307: insert a buffer Buffer between every two sections of memory element, wordline drives and every
The Fin number of the driving of individual described buffer Buffer is nmeet;
Step S308: assembled Fin is nmeetWordline drive and the domain of buffer.
Shown in Fig. 3, first nWLDIn the number of Fin permits span, obtain n memory element
Minimum time delay, if less than target (the first preset value) ttarget, then ask in span and meet constraint
The minimum Fin value of condition;If greater than the first preset value ttarget, then be averaged segmentation, asks every section
Minimum time delay, and by every section of minimum time delay and the first preset value ttargetCompare, if less than the first preset value
ttarget, then seek the minimum Fin value meeting constraints, otherwise continue segmentation and go down until meeting goal constraint
(every section of minimum time delay is less than the first preset value t for conditiontargetTill).Choose the most always and meet condition
Segments and the value of Fin, select the result of optimum always, and therefore every section of time delay overall altogether then exists
Meeting also is minimum under constraints.
Wherein, said method can be referred to as greedy algorithm, permissible for this iteration selecting greed
Solving with analog simulation, under limited, the discrete value of Fin, Dynamic simulation obtains prolonging of correspondence
Time, the time delay obtained is compared with the first preset value, if being less than first and arranging, then terminates imitative
Very, then time delay is carried out computing;If more than the first preset value, then described n array storage unit being carried out
Segmentation, continues under discrete value, and emulation obtains every section of time delay corresponding to memory element, and follow-up changes
Generation by that analogy, when time delay corresponding to every section of memory element is less than the first preset value till.The most every
Insert between two sections of memory element and determine WL Driver and Buffer with the value of buffer Buffer, Fin
Size.Different columns (n) obtains segments and value n of Fin of correspondencemeet, pass through computer
Generating the domain of corresponding WL Driver and Buffer, such memorizer compiler just can follow the trail of columns
Change so that the PPA of the WL Driver of each memorizer can be optimum.
See disclosed in the above embodiments of the present application, the columns n of described memorizer compiler is tracked
Method, the number of the Fin of WL Driver and Buffer of memory element is optimal value so that
The size of described WL Driver and Buffer is optimum, thus enhances the area utilization of memorizer, subtracts
Little area loss.
It is understood that see Fig. 2, described it is driven into prolonging of distalmost end memory element by memory word line
Time the columns n of time t0 and memory element and wordline drive the number n of Fin of MOSWLDBetween
Functional relationship, for:
If whereinCWLD=bnWLD, described a, b are constant, then:
Wherein RC_WLFor the resistance value of memory element bit line, CC_WLCapacitance for memory element bit line;
Therefore, described t0 Yu n and n can be tried to achieveWLDBetween functional relation be:
Fig. 4 is the long wire model figure writing driving Write Driver worst case of memorizer compiler.Wherein,
Shown RC_PUFor the resistance of the pull-up PMOS of memory element, RC_BLFor the resistance of memory element BL,
CC_BLFor the electric capacity of memory element bit line BL (bit line), RWRDFor writing driving (Write Driver)
Resistance, m is the line number of memory element.When needs write upset memory element, the ratio of resistance needs
Meet condition:
Wherein said λflipIt it is the second preset value.
Wherein, if the number of the Fin of the NMOS of described Write Driver is nWRD, then, describedDescribed c is constant, and the most above-mentioned formula can be deformed into:
Fig. 5 is the disclosed method stream being tracked line number m of memorizer compiler of the embodiment of the present application
Cheng Tu.
It is understood that for the PPA optimizing memorizer further, described memorizer is compiled above-mentioned
Translate on the basis of the columns n of device is tracked, it is also possible to line number m of memorizer compiler is tracked,
Seeing Fig. 4 and Fig. 5, its detailed process is:
Step S501: calculate the resistance R of the pull-up PMOS of memorizerC_PUHinder with the resistance writing driving
Value RWRDResistance R with m line storage unitC_BLThe ratio T of sum;
Step S502: at nWRDSpan in, calculate and make described ratio T more than the second preset value
The number of minimum Fin of the NMOS writing driving, be designated as nWRD1;
Step S503: assembled Fin is nWRD1The domain writing driving.
Visible, it is analyzed by arranging the concrete structure that described memorizer editing machine is write driving,
Resistance R to the pull-up PMOS of described memorizerC_PUWith the resistance R writing drivingWRDWith m row
The resistance R of memory elementC_BLThe ratio T of sum, meanwhile, analyzes the resistance writing driving described in obtaining
Resistance RWRDNumber n with the Fin of the NMOS of described Write DriverWRDBetween relation, thus logical
Cross to bring into and obtain, number n of the Fin of the NMOS of described Write DriverWRDAnd between described ratio T
Relation, i.e. can be obtained so that minimum n when described ratio T meets pre-conditioned by analytical calculationWRD's
Value, thus reach the purpose that line number m of reservoir compiler is tracked so that described Write Driver
Size be optimum, thus further enhancing the area utilization of memorizer, reduce area loss.
Fig. 6 is the structure chart of the sequential tracking circuit of memorizer.Wherein, Cell memory function mould in figure
Block, described WL FMS is the timing triggers of WL signal, and described SA FSM is SAEN signal
Timing triggers.
Seeing Fig. 6, described sequential follows the trail of circuit operationally, when shown CK end gets clock signal
During rising edge, trigger WLEN and be set to ' 1 ' so that shown DWL is also set to ' 1 ' unlatching phase inverter,
RBL starts electric discharge, finally makes WL_END become ' 1 ' replacement WL FSM, corresponding WLEN
Putting ' 0 ', DWL puts ' 0 ', and RBL reverts to ' 1 ', and WL_END puts ' 0 ', terminates WL
Sequential follow the trail of;When WL_END is set to ' 1 ', the sequential opening SA is followed the trail of, and SAEN puts ' 1 ',
SA_END puts ' 1 ' replacement SA FSM and makes SAEN be ' 0 ', and then SA_END also becomes ' 0 ',
The sequential terminating SA is followed the trail of.Driving one long line of inverter drive of WLEN, its signal lag is subject to
The impact of the line number (m) of storage array in memorizer;Drive one long line of inverter drive of RBL,
Time delay is affected by the line number (m) of storage array in memorizer equally;The inverter drive driving SAEN is long
Line, time delay is affected by the columns (n) of storage array in memorizer;Drive the phase inverter of SA_END
Driving long line, time delay is affected by storing columns (n) in memorizer equally.And above-mentioned these phase inverters
It is large-size device, affects the overall sequential of memorizer.The sequential of existing memorizer compiler is followed the trail of
The worst case of array is only designed by circuit, so causes array sequential in the case of other to be followed the trail of also
It not optimal so that it is not optimum that sequential follows the trail of the PPA of circuit, adds the area of memorizer
Waste.
Fig. 7 is the long line model that sequential follows the trail of the large scale phase inverter in circuit.
Being directed to sequential in memorizer and following the trail of the PPA of circuit is not optimum problem, applicant establish as
Long line model shown in Fig. 7, wherein, shown RDriverFor driving the MOS resistance of phase inverter, shown
CDriverFor driving the mos capacitance of phase inverter, described RleafcellFor the connection resistances of memory function module,
Described CleafcellFor the electric capacity of memory function module, the memorizer that described p is connected by driving phase inverter
The number of functional module, described time delay t1 is by by driving phase inverter to the farthest merit driving phase inverter to be connected
The delay time of energy module.
Can be obtained by computational analysis, described time delay:
If driving the number of the Fin of MOS in above-mentioned long line is nDriver, then describedCDriver=e*nDriver, described d, e are constant.The most above-mentioned formula can be deformed into:
By above-mentioned formula, described at nDriverSpan in, try to achieve the minimum meeting delay requirement
The size of driving MOS, and be assembled into corresponding domain, so that what above-mentioned long line drove
PPA optimizes, and this solution procedure can solve with analog simulation: at discrete nDriverSize
In span, the time delay t value of simulated measurement correspondingly-sized, compare timing requirements, the condition of being met obtains
The n of minimum dimensionDriver, and assembled corresponding domain.Therefore follow the trail of sequential and follow the trail of line number and the columns of circuit
Change, the sequential of each memorizer of generation is followed the trail of circuit from being in harmony, thus is reduced memorizer
Area loss.
Concrete, see Fig. 8, the above embodiments of the present application use sequential control circuit to described memorizer
The method that line number m of compiler and columns n are tracked, concretely:
Step S801: determine the time delay t1 of the sequential control circuit of memorizer and the function mould of inverter drive
The number n of the Fin of block number p and phase inverterDriverBetween functional relationship, wherein said nDriverIt is one
Dispersion number;
Step S802: ask at described nDriverSpan in so that described time delay t1 for meet preset bar
The minimal amount of Fin during part, is designated as nDriver1, i.e. make described t1 less than the minimum of Fin during predetermined value
Number;
Step S803: assembled Fin is nDriver1Sequential control circuit domain.
It is understood that corresponding to method disclosed in the above embodiments of the present application, disclosed herein as well is
A kind of memorizer using any one method above-mentioned to make.
A kind of memorizer, it is characterised in that described memorizer is for using the claims 1-5 any one
The memorizer that memorizer compiler joining method is assembled.
It is understood that memorizer disclosed in the embodiment of the present application can use the WL in memorizer
The columns of memorizer is tracked and/or uses Writer Driver to chase after the line number of memorizer by Driver
Line number and columns are all tracked by track and/or employing sequential control circuit, therefore, it is possible to in memorizer
Large-size device is tracked customization, so that the overall PPA of each memorizer is optimum.
In this specification, each embodiment uses the mode gone forward one by one to describe, and each embodiment stresses
Being the difference with other embodiments, between each embodiment, identical similar portion sees mutually.
Described above to the disclosed embodiments, makes professional and technical personnel in the field be capable of or uses
The present invention.Multiple amendment to these embodiments will be aobvious and easy for those skilled in the art
See, generic principles defined herein can without departing from the spirit or scope of the present invention,
Realize in other embodiments.Therefore, the present invention is not intended to be limited to the embodiments shown herein,
And it is to fit to the widest scope consistent with principles disclosed herein and features of novelty.