US 20050134486 A1
An apparatus accepts randomly arriving blocks of parallel digital data of varying bit lengths termed datum segments that may have been generated by stripping leading zeros from bytes of a fixed size, each having associated therewith a bit count code that expresses the bit length of each datum segment in the form nnnnndddd . . . , the n being the bits of the bit count code in such number as to encompass the memory capacity of a receiving device to which the datum segments are to be sent, and the d representing the actual datum segment bits. The apparatus concatenates the nnnnndddd . . . expressions to form a continuous bit sequence that is saved so that each nnnnndddd . . . expression is accessible thereafter through the computer address therefor, such use preferably being by a circuit of matching bit length, the format, however, allowing the original form of the data to be recovered if desired.
1. A concatenator adapted to receive and process datum segments of varying bit lengths, comprising:
an input register having a predetermined bit length and having defined therein an initial input register address;
means for receiving within said input register one or more datum segments that may have varying bit lengths;
means for determining the full bit length of one or more received datum segments;
an output register having a predetermined bit length and having defined therein an initial output register address;
transfer means for transferring a first datum segment from said input register to said initial output register address as a first step of a concatenation series;
address calculation means for calculating respective addresses for one or more datum segments other than said first datum segment;
concatenation means for transferring, as further steps of a concatenation series, one or more additional datum segments from said input register into locations within said output register that are in immediate juxtaposition with respective datum segments just previously transferred thereto;
means for determining whether or not a received datum segment can be fit within said predetermined bit length of said output register at locations not occupied by any prior transferred datum segments; and
concatenation initiation means for initiating a new concatenation series upon one of said additional datum segments exceeding in length an amount of said predetermined length of said output register not occupied by one or more prior transferred datum segments;
whereby said output register can be filled to near said predetermined bit length by concatenating together a series of datum segments.
2. The concatenator of
3. The concatenator of
4. The concatenator of
5. The concatenator of
6. The concatenator of
7. The concatenator of
8. A datum segment positioner adapted to accept variable length datum segments each having associated therewith a specific register address to which said datum segment is to be sent, and to place individual ones of said datum segments into respective ones of said specific register addresses, comprising:
address identification means, wherein as to each said datum segment said datum segment positioner identifies said specific register address that is associated with each said individual datum segment; and
datum segment transfer means wherein each said individual datum segment is transferred to corresponding ones of said specific register addresses that are associated with each said individual datum segment.
9. A method of concatenating a series of datum segments comprising the following steps:
a) in a first transfer, transferring from a first register having a predetermined length, into an initial address within a second register having a predetermined length, a first datum segment having a predetermined length as defined by an end bit location that is less than said predetermined length of said first register, together with the full remaining content of said first register;
b) in a second transfer, transferring from said first register a second datum segment, together with the full remaining content of said first register, into a second address that begins within said second register at a bit location immediately following said end bit location of said first datum segment, whereby a portion of said first full remaining content as had been transferred into said second register in said first transfer is over-written by said second transfer.
c) repeating steps a) and b) until said second register is essentially filled with consecutively concatenated datum segments.
10. The method of
This application claims the priority of co-pending application Ser. No. 10/462,868 filed on Jun. 16, 2003.
1. Field of the Invention
This invention relates to methods and apparatus for improving the efficiency in the use of and data transmission to, from and within the circuitry of computers, ASICs, optical devices, and other devices that are adapted to accept and operate on serial and parallel binary digital data, and specifically to such data in the form of datum segments that may vary in length and, among other forms, can be expressed by the formula nnnnnddd . . . where the n's represent the binary code that expresses the bit length of the datum segment ddd . . . , the data so formed also providing security against the compromising of data.
2. Description of the Related Art
In the development of the computer art, data transfer has long been a critical issue. The speed of what would appear to be the fastest computer at present, as noted by Katie Greene in Simulators Face Real Problems, Science, Vol. 301, No. 5631, pp. 301-302 (18 Jul. 2003), is reported to be 35,860 gigaflops, but advances beyond that speed are said to be limited by the need to wait for data on which to operate. Two of the patents previously issued to this inventor, i.e., U.S. Pat. Nos. 6,208,275 and 6,580,378, as well as application Ser. No. 10/462,868 filed on Jun. 16, 2003, have addressed that data transfer process. Although that work was addressed generally to the transfer of data to and from a computer, the methods and apparatus described are equally applicable to processes occurring entirely within a computer. The present invention continues that development and sets out, in one aspect of seeking faster computer operations, another step in the creation of systems by which data can be provided more efficiently, more rapidly, and with less waste of bit space both to/from and within a computer or like device. In this description, a datum segment is simply an amount of sequential binary code that represents an item of information.
U.S. Pat. No. 6,208,275, issued to this inventor on Mar. 27, 2001, shows a concatenator that accepts a series of datum segments of a fixed size, such as 8-bit bytes, and then concatenates them together to form larger words of lengths that would be an integral multiple of that fixed size, thereby to yield words of 16, 24, 32 bits, etc. The reason for so doing is that in a computer having, e.g., 32-bit data busses, it is wasteful of space to use such a bus to transmit a datum segment only 8 bits long24 bit spaces are unused. The '275 concatenator serves to string together four 8-bit bytes and then transmit the resultant 32-bit data string, thus to use the full capacity of the bus so that one transmission does what would otherwise have taken four. No data transmission occurs during the actual concatenation process, but that gate-based process will ordinarily be much faster than the data transfer, hence there is a net saving of time. Of course, the same principles will apply to other fixed byte or bus sizes such as 64-bit, 128-bit, etc.
U.S. Pat. No. 6,580,378, issued to this inventor on Jun. 17, 2003, sets out a simple data enumerator that counts the bytes on which the concatenation process just noted is carried out, and tags each such byte with an index number. Those numbers will identify which positions within a destination register will contain which datum segments. When those datum segments are transmitted to a computer, those numbers also aid in recovering the original datum segments from the longer, concatenated data segment in the computer. The nnnnn bit count code is used in the present invention to aid in specifying the starting addresses of successive datum segments in the output register of the apparatus and in the computer or other device to which these data are to be sent, since, when treating variable length datum segments, those addresses will no longer be fixed in advance by a fixed datum segment size.
Application Ser. No. 10/462,868, filed Jun. 16, 2003, and issued to this inventor in part on ______, as U.S. Pat. No. ______, describes another method by which bit space can be saved in transmitting and using bit strings, in part by stripping therefrom any leading zeroes that take up space but do not convey any information. A second aspect of the ______ apparatus is that it can treat datum segments of varying length, that might have come about either as a result of that zero-stripping process, or such datum segments may have been provided to the apparatus originally. Included is a general method of forming and using datum segments of varying length, by way of a variable length shift register that can yield versions of all of the basic gates of digital electronics, e.g., AND, OR, XNOR, etc., that can also be of varying length. By a variable length gate is meant that only those bit spaces in a register that correspond in number to the bit length of a datum segment at hand need to be utilized, thus leaving other adjacent bit spaces for other uses. It is one specific purpose of the present invention, that would not otherwise be available, to provide means by which datum segments, fixed-length bytes, identifiable bit sequences or data in any other form under whatever name, that could have been zero-stripped to include only meaningful data bits, can then be processed as such, as a matter of routine, thereby to maximize the data handling capabilities of such a system relative to any other system.
The processes of the ______ patent lead to a form of encoding a datum segment as nnnnnddd . . . , where nnnnn, the bit count code, expresses the number of bits in the datum segment that immediately follows, and the ddd . . . represent the actual bits of the datum segment. That method of expressing the nature of the datum segment comes about firstly by independently establishing the number of bits in the datum segment, which the apparatus of the ______ patent carries out. The actual code of the datum segment itself is then established and concatenated onto the bit count code nnnnn, using 5 bits as an example. Concatenation of this type could be carried out in the apparatus of the ______ patent since there, as here, the number of bits n used to express the number of bits in the datum segment was fixed, and since the datum segment ddd . . . is concatenated onto the Least Significant Bit (LSB) end of the nnnnn code, the length of the datum segment itself does not affect the concatenation process.
An initial address is used for placement of the bit count code nnnnn, itself again having 4 bits, and the first bit of the datum segment is then placed at the address that immediately follows the code nnnnn, i.e., in the example using the 5-bit nnnnn code at the 5th register position. It is then necessary only to place that entire bit string into a register large enough to accommodate that entire nnnnnddd . . . code. The reason that the bit count code nnnnn is retained is to permit structuring of the circuitry within the computer to which these data are to be sent.
However, it is not immediately possible in the apparatus of the ______ patent to concatenate onto a first nnnnnddd . . . code a second such code, since the end point of that first nnnnnddd . . . datum segment will not be known, except indirectly through knowledge of the nnnnn value that is associated with each datum segment. What is needed is thus a means by which the position of the last bit of a variable length datum segment can be established within the hardware itself, so as not only to carry out concatenations of the form nnnnn+dddddd=nnnnndddddd as is done by the apparatus of the '275 patent, where here + means a concatenation and = means that the result of that concatenation then follows, but to concatenate those resultant terms, i.e., in concatenations of the form nnnnnddd . . . +nnnnnddd . . . +nnnnnddd . . . =nnnnnddd . . . nnnnnddd . . . nnnnnddd . . . where the number of data bits d and hence the length of the datum segment as a whole can vary, and of course the ellipses would no longer be present, but the numbers of d's that represent the actual number of bits in each datum segment would be shown instead.
For example, and for brevity using here a bit count code of 4 bits, a resultant code for three datum segments of varying size could come out to be as
A concatenator has as input at least one register that is adapted to receive datum segments of varying length, from 1 bit on up to the full size of the register. The data must either arrive in, or be placed into, the form nnnnnddd . . . , wherein the n's express the number of bits in the datum segment dsi=ddd . . . that follows that nnnnn sequence, and the ellipsis represents a continuing sequence of bits d, with the bit count code nnnnn now selected as an example containing 5 bits, so as to express any integer from 1 to 32. The concatenator accepts a series of distinguishable datum segments from within a data stream and transfers those datum segments to a series of pre-determined memory locations within a computer or other data processing system. The efficiency of that transfer is improved by concatenating together as many datum segments as may be, given the sizes of the datum segments that happen to have arrived and the size of the bus on which the data are to be transferred, and hence to transfer as much information as possible in each transmission. These datum segments may or may not have been zero stripped, thus to be shorter in length than would otherwise have been the case.
It would be of particular advantage as to data comprising numbers having widely varying magnitudes, including smaller numbers for which the binary code therefor would contain many leading zeroes, to zero strip such numbers and then concatenate together the results, thus saving bit space both by the zero stripping and by the concatenation. Such a procedure as to both zero stripping and concatenation also has a security aspect, in that if the data were put into the nnnnddd . . . varying length form before transmission, and those data were intercepted in transit or had otherwise fallen into unauthorized hands, any effort to interpret those data (perhaps as 8-bit bytes) in a computer not equipped with the apparatus described herein would yield meaningless results, and the data would remain secure in spite of having been intercepted.
The individual datum segments can arrive sequentially, or two or more such datum segments may arrive simultaneously, from two or more data streams into a corresponding number of variable length registers. (A reason for treating more than one data stream at a time is that the circuitry of the invention is expected to operate much faster than data can be transmitted thereto, so the introduction of a second bit stream would make better use of that circuitry if it did not have to wait for data, and means could be provided to cycle through the several bit streams in the old manner of time sharing on main frames.) In the case of the arrival of two or more data streams, the datum segments would still be treated in a pre-determined order, e.g., cycling from left to right in an array of buffers into which the datum segments had been stored on arrival. One of two datum segments, i.e., the first-arriving datum segment if the two datum segments arrived sequentially within a single bit stream, is placed to the left in a variable length output register, and then the second-arriving datum segment, starting with the bit count code that expresses the length thereof, is placed to the right of the last bit of the first datum segment, commencing at the position that immediately follows that last or Least Significant Bit (LSB) of that first datum segment, as determined by the value of nnnnn for that first datum segment. What is involved here is that since that first datum segment might have any length within some particular range, it will not be known in advance where that LSB will arrive and hence where the first bit of the bit count code for the second datum segment is to be placed, and to accomplish that determination so as to place that second and of course all later datum segments at proper places in a final destination for the data as a whole is what this variable length concatenator is to do.
In general, where ADDRESS1 is a starting address given in such number of bits as is needed to express the full size of the registers into which the datum segments are to be placed, the address of the first bit of each later datum segment is given by the sums of the lengths of the preceding full datum segments plus 1, where by a full datum segment of length Li is meant the length of the datum segment dsi itself plus the length of the code by which that length is expressed. That is, the 5-bit length of the bit count code nnnnn that expresses the length of each datum segment, that is then ((nnnnn)2)i, i.e., a code to the base 2 of the length of the ith datum segment, has added thereto the actual value of the datum segment length so expressed, and the address for the start of that bit count code for each such data segment dsj, with respect to a series of concatenations, becomes ADDRESSi=ADDRESS1+Σ Li+1, where the i refer to all of the datum segments of the series up to but not including that last dsj. The length of the bit count code can be fixed with respect to a given instance of the invention, or that code length can be made adjustable so as to be pre-set to a desired value based on the maximum length anticipated for a given body of data to be treated, whereby that bit code length could then be changed to accommodate another particular series of datum segments. By relevant memory may mean a block of memory selected out of the full memory capacity of the computer that has been set aside for this data input function, or it may be essentially all of that memory capacity.
Look-ahead means are provided for determining whether or not sufficient bit space remains available in the output register of the apparatus for each next datum segment, and if that space turns out to be insufficient, the register content as accumulated to that point will be transferred out therefrom at that time. After that transfer, a new concatenation series is started with that new datum segment then becoming the first datum segment of that new concatenation series, to be placed at the leftward end of that output register.
Selection of the number of bits to be used in designating the lengths of the datum segments to be treated depends only on the lengths of the datum segments one expects to receive, and has no relationship with the length of the output register into which the incoming datum segments are to be accumulated. That is, one could use an nnnnn code that would express datum segment lengths only up to 32 bits, as will be done herein, but the output register into which those datum segments are to be accumulated may be 128 bits, 256, or whatever size may be required for the intended uses. Even so, in order to illustrate the case in which concatenation of a next datum segment would exceed the size of the output register, in the example employed herein the size of the output register is also set at 32 bits. The purpose in carrying out the concatenation, after all, is to permit the transfer of as much information in one transmission as possible, hence the register from which each such transfer is to be carried out, and of course the associated transmission means, would be made as large as may be feasible. The register of the datum segment positioner to be seen later in this specification might well be 128 bits, the output registers could be 1 Kb, and the relevant memory in the computer or the like could be measured in Gb, or other examples could be cited as to the needs of other particular cases.
The addresses to which these datum segments are to be sent are calculated from the addresses of preceding datum segments, but if desired, as for example when it is known that a sequence of datum segments to be received will relate to different subjects of interest (e.g., after some known or measurable number of datum segments the data might change from the names of company personnel to their phone numbers, from accounts receivable to accounts payable, from the genes within one chromosome to those in another, etc.), means are provided for introducing a new starting address to form a separate memory block when the datum segments that treat a new subject classification are to begin.
Since the circuitry of the invention may need to treat bit streams arriving from a separate source as through a modem, or data arising from within the same computer, and in either event this circuitry is to transmit its output to some relevant computer memory, the circuitry of the invention is preferably placed physically at a location within the computer that is conveniently near to both such a modem or other input/output device and to the relevant computer memory, and preferably, if possible, on the same printed circuit board (PCB) as that relevant memory, with the modem or like device generally being on its own PCB, each such PCB to be installed in the mother board of the computer in the usual manner. In other words, this present concatenator might best be utilized if fabricated as an enclosure within a large array of memory locations, thereby to take maximum advantage of the concentration of data within that memory that this concatenator provides.
The broad aspects of the invention as to function are shown in
Concatenator 10 of the present invention thus adds the ability, by way of datum segment router 12 and datum segment distributor 14 as shown in
While concatenator 10 of the '275 patent shows registers of fixed size into which are fed a series of bytes b1, b2, . . . , each of which are also of that same fixed size, as just stated concatenator 10 of this invention employs a variable length input register 12 and a variable length output register 32. Full datum segments containing the bit count code nnnnni and also the datum segment dsi itself are placed into input register 12 in the first position thereof; the address of each additional datum segment thereafter is established by calculation from the address of the last of the previous full datum segments, and the datum segments as initially placed in each case into the first positions of input register 12 will ultimately be transferred to the addresses in output register 32 as had just been determined.
For example, for a first datum segment i shown in
The result of concatenating just those first two datum segments i and j is shown in the output register at the bottom of
With concatenator 10 other purposes can be pursued besides the orderly transfer of data, e.g., the data enumerator of the '378 patent can be incorporated with the circuitry of the invention to count the datum segments as they enter. If those datum segments were words of text, for example, that count could provide a word count of a document, or if those data were the contents of individual sales invoices, the count would yield the number of such sales. Also, in the interest of designing self-testing circuitry, if that data enumerator were accompanied by a clock signal, as would not ordinarily be the case with an asynchronous gate circuit such as concatenator 10, the circuitry could also provide means for measuring the speed of operation of concatenator 10.
What the invention provides as an adjunct, in its ability to place data into the most convenient locations within a computer, is an acceptance and subsequent distribution of data in forms, and having logical structures, that would have been established before those data had even reached the computer, that would then save time in executing READ and WRITE commands or other operations within the computer, rather than having such data be entered into an amorphous memory structure for which the actual addresses that happened to become used had no relationship to the data content. Put another way, instead of allowing incoming data just to stack up in one huge pile as in ordinary, non-discriminatory memory systems, this apparatus is adaptable, through the use of a ttt . . . address prefix, to carry out a pre-sorting of data such that the incoming data can be pre-directed as to destination. In such case the data are tagged for storage in a manner that will be meaningful, with each type of data then being placed in structured stacks or hard drive addresses in a way such as to be physically juxtaposed as to type, and thus be accessible more rapidly for further operations.
Since concatenator 10 accepts datum segments of varying length, the location in the output register into which is to be placed the first bit of each new full datum segment, i.e., the first bit of the bit count code nnnnni of that datum segment, must be defined. That is done by pre-selection as to a first datum segment, but can be done thereafter either by pre-selection or, through one major aspect of the invention, by calculation as to later datum segments, that calculation being based on the sizes of successive datum segments as they arrive and are accumulated within an output register. As noted earlier, datum segments of varying length also require a determination of whether or not an arriving datum segment can actually be accommodated by the bit locations left unused in the output register, and both processes are carried out by concatenator 10.
An initial ADDRESS1, that may be the same as an address in some relevant block of computer memory or other circuitry to which the datum segment is to be sent, establishes the placement of the first bit of the bit count code nnnnn1, which then, following the 5 bits of that bit count code nnnnn1 itself, defines the position for placement of the corresponding datum segment ds1. Of course, given that the length of the datum segment is known, as is the length of the bit count code nnnnn1 by which that length is expressed, placement of the first bit of the bit count code automatically fixes the locations of the rest of the bits in the full datum segment, as well as the starting point for the next datum segment.
ADDRESS1 will then arbitrarily be taken to be 00001, while for the corresponding computer address any accessible value, including those that may have been assigned to various nodes within the computer, can be used. That address might be held either as a permanent entry for a particular instance of the invention, or so as to be editable for data classification purposes as noted above, but in either case the value of ADDRESS1 is used to route that first datum segment and then, by adding to ADDRESS1 the full length Li of that first datum segment, to have defined the address for the bit count code and then the content of the next datum segment, and so on. Through use of that method, each datum segment will be placed in immediate contiguity with datum segments on either side thereof, except, of course, for the first and last datum segments, although even then by rare circumstance the first and last datum segments may be placed in contiguity one with the other, so that all datum segments would be in immediate contiguity with others, and in such case not even a single bit location in the computer would be left unused. In what follows, it is assumed that the number of bits to be used for the bit count code, the value for ADDRESS, the sizes of the registers, and other parameters that characterize any particular instance of the invention are held in non-volatile memory in the concatenator 10 circuitry, and that on startup those values will be loaded into the places required, subject to any further editing thereof by the user.
First adder 20 will carry out two additions with respect to each new incoming datum segment. For each of those additions, there must also be two routings carried out by router 18, which are (1) the selection of the appropriate inputs for first adder 20 for each addition; and (2) a routing of the result of each addition. (As will be discussed below, two subtractions will also be carried out in subtractor 24 with respect to each datum segment.) The operations that then take place are described in terms of 0 and 1 areas in first router 18 (and in the subsequent circuit blocks), those areas not designating distinct physical structures but rather what the operations will be when one or the other of the 0 and 1 designations, which ar in fact addition selection codes, is in effect. Thus, the four inputs that are shown as being connected to first router 18 relate to the 0 and 1 notations within first router 18 by way of their locations, with the single external input A shown entering first router 18 within a 0 area, along with the B entry 00101 that expresses the size of the nnnnn1 bit count code, those two quantities participating in the first addition under a 0 addition selection code; and then the quantities L1 and ADDRESS1 for this first datum segment ds1 similarly being shown at the C and D terminals in the bottom half of first router 18 that contains the 1 addition selection code that brings about the second addition. In the top half of first router 18 labeled 0, the two inputs shown entering that upper half will be those two inputs that will be sent on to first adder 20 on a 0 addition selection code for a first addition. Similarly, when the addition selection code is 1, the two inputs coming from the lower half of first router 18 and connecting to the lower half of first adder 20 will be those that are sent to first adder 20 for a second addition. As will be explained below, which addition selection code is in effect at a particular time is determined by first OR gate 28 and first toggle switch 30 that are also shown in
The continued course of that process, i.e., as to what is done with the value obtained for ADDRESS2, lies in identifying that particular datum segment of the sequence dsi, dsj, dsk, dsl, . . . for which the corresponding full datum segment lengths Li, Lj, Lk, Ll . . . total more than the 32-bit size of output register 32. Concatenator 10 thus tests each newly arriving datum segment in that regard. If a newly arrived datum segment will fit into output register 32, it will so be placed as part of the concatenation process then under way, while if it will not so fit, a new concatenation series will be started. With Ln representing each of the full datum segment lengths, if Σ Ln≦32, the last-arriving datum segment can be concatenated onto the pre-existing content of output register 32, but otherwise not, and hence a new concatenation must be started. (For reasons to be given below, the test is not actually carried out on the basis of that sum, but rather on the calculated address for each new datum segment. Either test could be used, and both are deemed to fall within the spirit and scope of the invention.)
A second task of concatenator 10 is to determine the address in the computer, for which ADDRESSi in concatenator 10 (a register address) will be a temporary corollary, to which each datum segment will be sent. That computer address will be the same whether or not a datum segment in question can be accommodated within the series of concatenations then being carried out or must be used to start a new series of concatenations. That is, although the location within output register 32 into which a datum segment will be placed for transfer will change when a new concatenation is begun, and will always be the first position on a new concatenation rather than a calculated register address, the computer addresses are cumulative, and the computer address for the next-arriving datum segment must remain the same even though the correlated register address may changethe datum segment will have a register address either of ADDRESS1 or some other value, depending upon whether or not it was used to initiate a new concatenation series, but the computer address to which it will be sent will be the same in either case. Successive computer addresses will continue to increase through whatever number of series of concatenations as may be necessary in order that all of the datum segments will be placed in juxtaposition within the computer as intended. Each datum segment will arrive at a pre-determined computer location, whether into memory or as immediate input to a circuit, regardless of how the concatenations that brought about those transfers were carried out.
To clarify how that result is brought about, there will exist a definable relationship between a particular location in input register 16 and output register 32, with both of these also being related to addresses in the computer. The relationship between output register 32 addresses ADDRESSi and the computer addresses will change, however, since while input register 16 and output register 32 are used over and over again, as just indicated the ultimate locations in the computer to which data will be sent are cumulative in nature and will be fixed (as to a particular course of operations). In being used over and over again, input register 16 conveys a single datum segment at a time, and starts each new concatenation series at ADDRESS1=00001. Output register 32, on the other hand, although it starts its own set of concatenated datum segments at its own leftmost address, accumulates a number of such datum segments in each concatenation series, and then transmits some number of those accumulated datum segments all at once.
As previously noted, in the examples to be used below datum segment lengths were selected such that the first two datum segments ds1 and ds2 (respectively parts of the full datum segments i and j shown in
The formalism used is that the first bit location of input register 16, i.e., ADDRESS1, is correlated with the computer address, and for ease of discussion as to the very first series of concatenations both the register address ADDRESS1 and corresponding computer address are given the address 00001. The second datum segment ds2 will have the register address ADDRESS2, which as it turns out will be 01001 for which (01001)2=9, hence the first bit of bit count code nnnnn2 thereof will ultimately be located at that address in output register 32, i.e., at the 9th bit location therein that correlates with another specific computer address. Any pre-sorting of these data segments as was noted above would be shown by the fact that if it were desired to store data into, say, a 2000 block in the computer, the register address of 00001 would correlate with a computer address of 2001, and likewise the register address 9 would correlate with the computer address 2009. More exactly, since the computer addresses will increase continuously, those computer addresses may be of a form even as large (in binary code) as 00000000000000000000000000000001 (4 Gb) or more, etc., i.e., being expressed by a rather larger number of bits so as to accommodate that larger capacity of the computer. That process is not a part of the concatenation process, however, and so is not discussed further here except to note that such process is more naturally carried out by the circuitry to which concatenator 10 connects rather than by concatenator 10 itself. Upon completion of as many concatenations as possible within a first set of 32 bits, in a new concatenation series there will again be a register address ADDRESS1, ADDRESS2, and so on, but the computer addresses to which these addresses correlate will commence where the previous series left off, i.e., at an address that follows the last bit of the last datum segment already treated.
To determine whether or not output register 32 can accept a next-arriving datum segment, it will be necessary to evaluate a sum effectively of the form Σ Ln as to each new datum segment, to determine whether or not 32Σ Ln<0, i.e., whether th bit lengths of the data segments add up to more than the bit length of output register 32, and several different ways of so doing might be used. One way would be to keep a running count of the values Li, Li+Lj, Li+Lj+Lk, etc., and then subtract each such value from the 32-bit size of output register 32. If that calculation yielded a result of <0, the capacity of output register 32 would have been exceeded, and the concatenation of that latest datum segment being considered could not be carried out. (A sum result of Σ Ln=0 would show that such capacity had been exceeded insofar as a next ADDRESSi is concerned, even though that for the datum segments alone had not, since it is the bit address next following the bit space taken up by Σ Ln that the address test actually to be used seeks to evaluate.)
It should be made clear, however, that a determination that some ADDRESSx can be accommodated within output register 32 says nothing about whether a corresponding datum segment dsx that commences at that address could be so accommodated. That ADDRESSx can be accommodated in output register 32 shows only that the preceding datum segment dsx−1 can be accommodated, since dsx−1 precedes ADDRESSx in order. To determine whether or not dsx can be fitted into output register 32 requires testing of the address that derives from the next datum segment, i.e., by testing ADDRESSx+1, which is established both as actual fact and by the test being carried out by the length of the datum segment dsx+1.
A second method of determining whether or not a new datum segment could be accommodated in output register 32 could be based on defining LM=32 as the maximum number of bits that output register 32 can accommodate, as in this instance of concatenator 10, wherein the i datum segment of
Turning back now to
The addition selection code designation for the first addition has been arbitrarily selected to be 0, and to start the process for the i datum segment of
As noted earlier, that nnnnn1=00011 value is placed on input line A of first router 18. The length of the bit count code nnnnni, selected for this instance of the invention to be 5 bits and hence constituting a permanent input, is placed on line B of first router 18. The sum of those two values, i.e., 3+5=8, is the length L1 of this first full datum segment in the general format nnnnnddd . . . , with the full datum segment i itself being 00011010. The above summation is the first addition to be carried out in first adder 20, and comes about by (a) having the addition selection code of 0 on first router 18; (b) the connections resulting therefrom that put 00011 (the length of dsi) on input line A of first adder 20 and 00101 (the length of the bit count code nnnnn1 by which that length of dsi is expressed) on input line B of first adder 20; and finally (c) the addition selection code 0 itself on first adder 20 so as to bring about the desired addition. Again, the addition of those two numbers yields the full datum segment length L1. Mor generally, if Ln is the bit length of nnnnni, here set at 5-bits, then Li=Ln+(nnnnni)2, where the subscript 2 merely means that the preceding binary code is to the base 2.
Those 8 bits are to be placed into output register 32, but it must first be determined whether or not output register 32 has sufficient bit space to accommodate them. For this first datum segment there must necessarily be sufficient space, but the determination thereof is carried through here even so, since it is indicative of the process as to later datum segments, and in any event will so proceed on its own. For that purpose, as seen in
For purposes of that second addition, ADDRESS1 for the first datum segment ds1 is seen in
To determine whether or not output register 32 can accommodate a particular datum segment, a choice of procedure is available. Two subtractions are carried out in subtractor 24, that could be based either on the address of the next-arriving datum segment or on the sum of all of the previous datum segment lengths. If a subtraction test relative to the first datum segment based on the value of ADDRESS2 obtained for the second datum segment ds2 were used, a positive result would indicate not only that the concatenation (or in the case now being discussed of a first datum segment of a series, simply a data transfer) could be carried out, but also that there would remain in output register 32 at least one more bit unused, that would be ADDRESS2 itself. A calculation based on the total length of all preceding datum segments (at the moment only on L1) would indicate with a 32L1=0 result (of course, that is not the present case) that the last bit position of ds1 coincided with bit location 32 of output register 32, i.e., there was just enough space in output register 32 for ds1 but not for ADDRESS2. That would be the most useful result, but the procedure to be adopted here is the former, since the ADDRESS2 value that in any event will be needed otherwise had just been calculated, is readily available, and can also be correlated immediately with the computer address to which the corresponding datum segment will be sent. Even so, which procedure is used, so long as it is internally consistent, is again a matter of design choice, so the use of either procedure would be deemed to fall within the spirit and scope of the appended claims.
The reasons for using the particular values in the subtractions now to be described will be set out below, but for the present it is simply noted that subtractor 24 is configured to carry out two subtractions, the first yielding the value of the quantity ADDRESS2-2 as the subtrahend for the second subtraction, and then the second subtraction LM-(ADDRESS2-2), where LM is the number of bit positions (32) in output register 32. The result of that second subtraction determines whether or not the datum segment being tested will fit into output register 32. By the initial premise that no incoming datum segments would exceed in length the 32-bit size of the concatenator 10 circuitry, there will necessarily be space for datum segment i in output register 32, and with the 3-bit length of the ds1 selected here as an example that is clearly the case.
As shown by the terminal designations A, B, C, and D in
The 00111 value on the H terminal of subtractor 24 shown in
Since ds1 would precede ADDRESS2 in output register 32, if ADDRESS2 can be fit into output register 32 then so can ds1, but it is also desired that ds1 be transferred into output register 32 whenever possible, even if there were no space for ADDRESS2. To define the test in that fashion, there are two adjustments that need to be made. The first of these derives from the fact just stated, namely, that it is the last bit of the datum segment that must be fit into output register 32, and not the next following bit that would be the address of the next datum segment. If that were the only consideration, the test to determine whether or not output register 32 could accommodate the datum segment under test would be to subtract from LM=32 the quantity (ADDRESS2-1). However, as will now be shown, it happens that yet another bit must be subtracted from that subtrahend because of the nature of the test that subtractor 24 carries out.
It is convenient here to use a 1's complement subtractor because of one feature that such subtractor type exhibits, the operation of which will be known to those of ordinary skill in the art. The circuitry of this type of subtractor is such that besides the actual subtraction result, it also provides a single bit code, here called a test bit, that indicates whether or not, in any given subtraction, the minuend was larger than the subtrahend. The test to be applied is whether or not minuend>subtrahend, not minuend≧subtrahend, and subtractor 24 is configured so as to yield a 0 routing bit if the minuend is larger than the subtrahend, i.e., the first expression just stated is satisfied, and a 1 routing bit if it is not, i.e., that expression is not satisfied. The former 0 result continues the concatenation series then being carried out, while a 1 routing code causes the concatenation series to terminate.
To account for the additional 1 value noted above that will yield the number 2 in the first subtraction, in a cas in which factually minuend=subtrahend, then using a minuend≧subtrahend test together with a subtrahend (ADDRESS2-1), the extreme of that test (in which the calculation was also minuend−subtrahend=0) would identify the last bit position of the preceding datum segment rather than the next-following position, which is ADDRESS2. The result would be affirmative since the result minuend−subtrahend=0 satisfies the minuend≧subtrahend test, thus to indicate that the datum segment under test would fit exactly within the remaining bit locations of output register 32. However, as just stated that is not the test that the 1's complement subtractor carries out, which is instead minuend>subtrahend. In order to have that test that the subtractor actually carries out yield that same result as just noted, by the artifice of defining the subtrahend for the second subtraction as (ADDRESS2-2), or more generally as (ADDRESSi-2), the resultant affirmative test will be that desired, i.e., indicating that a datum segment that has the same number of bits as the number of spaces remaining in output register 32 will indeed fit therein. With that test result the transfer would be carried out as a continuation of the concatenation series then being conducted.
So as to enable that ADDRESS2-2 subtraction to be carried out as the first of the two subtractions, with ADDRESS2=01001 on terminal E of subtractor 24 as shown in
The result of the present G−H subtraction is LM-(ADDRESS2-2)=32−7=25=(11001)2, which is a positive number that would clearly (and obviously) allow the first datum segment ds1 to be concatenated onto the content of, or in the case of this first datum segment merely entered into, both input register 16 and output register 32. As noted above, in this embodiment of the invention subtractor 24 is constructed so that if the second subtraction therein shows that minuend>subtrahend, i.e., LM>(ADDRESS2-2) or equivalently LM-(ADDRESS2-2)>0, subtractor 24 will yield a 0 routing code and the concatenation series will be continued, while a result LM-(ADDRESS2-2)≦0 will yield a 1 routing code and the concatenation series then in process will be terminated, that last-received datum segment being used instead to start a new concatenation series.
To clarify further the basis for these calculations, i.e., how concatenator 10 might operate under different circumstances, and specifically to illustrate the case in which a new and different third datum segment ds3′ would exactly fit into the remaining bit positions in output register 32, since L1+L2=8+12=20, at that stage there remain 32−20=12 bit spaces available in output register 32, wherein ADDRESS3′ for that third datum segment ds3′ would still be 21. A full 12-bit datum segment k of 5 bits for the bit count code nnnnn3′ and 7 bits for the datum segment ds3′ itself should then fit into output register 32, i.e., in addresses 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32. An entry into terminal C of first router 18 of L3′=01100, wherein (01100)2=12, would yield an ADDRESS4′=C+D=12+21=33, which lies one bit beyond the length of output register 32 and hence could not be used. However, when the procedure using the (ADDRESSi-2) subtrahend in the second subtraction is used, whereby the first subtraction in subtractor 24 is E−F=ADDRESS4′-2=31 and that result is placed into terminal H of subtractor 24, the second subtraction thereby becomes G−H=LM-(ADDRESS4′-2)=32−31=1, which is a positive number to produce a 0 routing bit, thus to indicate that while this new, hypothetical ADDRESS4′ will not fit into output register 32, the new, hypothetical full third datum segment ds3′ having 12 bits that immediately precedes ADDRESS4′ will indeed so fit. Concatenation of that 12-bit full datum segment onto the two full datum segments ds1 and ds2 that were already present in output register 32 would then proceed, which is the desired result (in this present illustration) and confirms that the quantity (ADDRESSi-2) evaluated in the first subtraction in subtractor 24 is indeed the proper subtrahend for general use relative to the second subtraction in subtractor 24.
Those subtractions as to the 3-bit first datum segment ds1=010 are shown in detail in
With one of the subtraction selection codes 0 or 1 being in effect in subtractor 24, the subtractions are self-executing once the required inputs are provided. Consequently, with a 0 subtraction selection code in place, as best seen in
The second subtraction LM-(ADDRESS2-2) yields a positive result and hence a 0 routing code, since 32−7>0. In accordance with step 3 of the second subtraction, that 0 routing code is sent to third router 26 to establish that the concatenation is to proceed, as shown in
Turning back now to the general process, the manner by which ADDRESS2 is obtained, and by extension an ADDRESS3, ADDRESS4, etc., has been explained, but what must still be shown specifically is how ADDRESS1 is acquired. As noted earlier that value is automatically loaded into concatenator 10 at startup, but what must be shown is how ADDRESS1 is acquired when a new concatenation series is started. As shown in
In transferring the first datum segment 1 from input register 16 to output register 32, it might seem necessary to ensure that only the bit string nnnnn1ds1 should be transferred to output register 32, i.e., only the 8 bits that make up i=00011010. To transfer any more of the content of input register 16 to output register 32 of
To see further how that occurs, it may be recalled that although the ADDRESSi values in concatenator 10 recycle, i.e., each new concatenation series starts with ADDRESS1=00001 (in this example), the corresponding addresses in the computer are cumulative and continue to increase as more datum segments are transferred in, without regard to what may have been the particular concatenation series in which a given datum segment had been contained. The computer address for a new datum segment is determined from the last calculated ADDRESSi value, and the last-arriving datum segment will be sent precisely to that specific computer address, whether as a part of an ongoing concatenation series or as the first datum segment of a new concatenation series. Thus, although particular transmissions may contain some small number of unused bit spaces, there will be no such gaps in circuit usage within the computer, but only perhaps a small number of bit locations that are left unused after completion of all of the concatenations and subsequent data transfers would have sought to fill the memory of the computer completely. As noted earlier, upon the occurrence of a 1 routing bit, as shown in both
Further as to treating the data, it was noted earlier that the present circuit provides means both for accepting pre-classified data and for counting the datum segments, and the means for so doing will now be discussed. The distribution of data as to subject matter requires that the desired classifications already be present in the datum segments as received, and that option is shown in
As to the counting of received datum segments, as shown in
That yynnnnnddd . . . sequence can also be used together with the type code ttt noted earlier that can be used to identify the type or subject matter of each particular type of data being treated and, if reset for each ttt type of data, this index number yy could be us d to count out the number of entries within each type. In the continuous addressing used in the computer itself, the addresses for each data type could also be compartmentalized, whereby tttADDRESS1's would be the first computer address within each memory block as designated to accept data of the particular type ttt, ultimately to yield a complete yytttnnnnnddd . . . . code. The particular order in which these various elements are assembled is of course quite arbitrary, but the foregoing order is adopted since the assignment of type codes ttt must already have been established at the time that the datum segments are received.
The purpose in adopting these practices, given the time and commitment of the ALU and other computer circuitry often required to carry out sorting and similar such operations, is to provide data to that computer that merely in the course of being transmitted thereto will already have been organized into logical structures that will be most convenient for use. All of the aforesaid coding variations can be used in the present apparatus with the kinds of minor circuit adaptations (for appending one code or the other, etc.) just discussed, as will be known to persons of ordinary skill in the art, based on the concepts of the '275 and '378 patents, application Ser. No. 10/462,868 and the present disclosure.
So as to track the course of these events, in the steps to be set out below in Table II for the first datum segment i it is assumed that what would have occurred just prior to the first event of this concatenation series would have been either an initial startup of the circuit or the transfer of a 1 bit from third router 26. Recognizing that ADDRESS1=00001 is already present at input D of first router 18, as a result either of a startup or that 1 bit, the circumstances that exist prior to the full treatment of a new datum segment in its path to output register 32, and an initial sketch of what will then transpire, are as follows:
The foregoing course of events can be summarized in the following Table II in which the values referred to in the preceding list have been entered in, and in particular the point at which the value for ADDRESS2 is obtained is marked:
Datum segments beyond the first one, although they will likewise arrive in the leftmost positions of input register 16, must be repositioned before they are placed into output register 32. Description of that process will b deferred, however, until after it is shown how a second and then a third datum segment are treated in accordance with the same procedure as was just described relative to the first datum segment. Table III below is thus just like Table II except in pertaining to the second datum segment j, and although not discuss d in as full detail as was Table II, the locations therein where differences exist between the data in the two tables will be briefly noted, the first of these differences being that Table III has no initial row in it since it follows after Table II, Table III then being as follows:
The following Table IV shows the same process as to the third datum segment, which is k of
The bit count code nnnnn3 for the k datum segment is seen in
As shown in
The explicit consequences of having acquired a 1 routing bit can be seen in
Besides determining whether or not each new datum segment will fit into the space remaining in output register 32, the second major task noted earlier is that of positioning the datum segments so as to achieve proper concatenation, wherein each datum segment has its beginning immediately after the end of a preceding datum segment. Datum segments will all arrive initially in the leftmost positions of input register 16, and the first datum segment will be sent to the leftmost positions of output register 32, but then each subsequent datum segment must be placed at a different address, whereby the second datum segment has the address ADDRESS2, the third must arrive in output register 32 at ADDRESS3, and so on. It must now be explained is how that shift in position of a datum segment is brought about.
The method of so doing is to put the datum segment into a shift register and then bring about the number of shifts necessary to place each datum segment as received into the numerical address that follows immediately after each preceding datum segment. Datum segment distributor 14 accepts each new datum segment, that would also be passing through the aforesaid routing process, and based on whether a 0 or a 1 routing bit had resulted therefrom, establish that each particular datum segment is to be in an ADDRESSi position as a continuation of a concatenation series on a 0 routing bit, or in the first ADDRESS1 position of output register 32 as the first datum segment of a new concatenation series on a 1 routing bit. Datum segment distributor 14 includes circuitry to accomplish that bit shifting, shown in
To accomplish that bit shifting will first require determining the number of bit shifts to be carried out. As can be seen from a review of Tables II-IV, and also by comparing the data shown to be present at the D line of first router 18 in
Also, the datum segment in question must have been placed into a location in which it can be bit shifted in accordance with that determination. For the latter purpose,
Although it was noted above that every datum segment will be used to calculate a new ADDRESSi for each next-arriving datum segment, from which the number of bit shifts for that next datum segment can be determined, that information actually comes more directly when the length Li of the datum segment dsi and its bit count code nnnnni at hand have been determined by the first addition in first adder 20. It is thus that Li value that is used to count out the number of bit shifts required for a datum segment that is to be concatenated as part of a current concatenation series. Consequently, besides being used to determine the address of the next datum segment and the routing code for the datum segment at hand, that Li value is also copied into datum segment distributor 14 so as to determine the number of bit shifts required.
There are two aspects of this bit shifting process that must be carefully noted. The first of these is that the length Li of a particular datum segment that has just arrived pertains not to the bit shifting of that datum segment itself, but rather to that of the next following datum segment. Secondly, it is only with respect to the bit shifting of a third datum segment in a series that the length Li of the preceding datum segment alone will yield the proper number of bit shifts. That is, the location of the third datum segment depends only on the bit shifting of the second datum segment, there having been no bit shifting of the first datum segment. The fourth datum segment, however, must be placed at a location just after the sum of the bit shifts of the second and third datum segments, and so on. Thus, in general the number of bit shifts applicable to the ith datum segment, i.e., the bit shift count bsci, can be expressed as bsci=Σ Li-1, i.e., the sum of the lengths of all of the preceding datum segments and their bit count codes (which of course is 5 in every case). Or put another way, bsci=bsci-2+Li-1, i.e., to obtain the bit shift for the ith datum segment, the length of the immediately preceding datum segment and its bit count code is added on to the number of bit shifts that had previously been carried out for all of the preceding datum segment(s). The latter expression above describes directly and exactly how the final bit shift value is actually obtained.
Specifically, it can be seen in bit shifter 54 as shown in
As noted in the earlier discussion, however, the length of the third datum segment k precludes it from being included in the concatenation series that includes datum segments i and j. With datum segment k having a length of 17 bits, it will fit into input register 16 and both holding register 58 and shifting register 60 upon its arrival, but upon being bit shifted 8+12=20 times in shifting register 60, the bits at the LSB end of datum segment k will be pushed off the LSB end of shifting register 60. That will not matter, however, since the fact that datum segment k will not fit into the space remaining in output register 32 will be signaled by a 1 routing bit from third router 26, and hence it will not be that bit-shifted version of datum segment k in shifting register 60 that is transferred to output register 32 in any event, but rather the unshifted datum segment k in holding register 58, and indeed, as will be described below in the discussion pertaining to datum segment positioner 56, into the ADDRESS1 positions of output register 32 to start a new concatenation series.
However, that step alone will not resolve the matter of proper treatment of the datum segment that will be next to follow. The several additions and subtractions that occur in the circuit path from first router 18 to third router 26 will already have taken place when that 1 routing code is acquired, and the various values deriving from those calculations, which are appropriate only to the case of a continuing concatenation, will nevertheless have been entered into the various terminals involved, and particularly into bit shift accumulator 62. Those values derive from calculations based on what would have been a third datum segment of that series, with the number of bit shifts required then pertaining, of course, to what would have been a fourth datum segment of that series, when the situation actually existing is that th putative third datum segment has instead become a first datum segment, and the putative fourth datum segment has become a second datum segment, with the former situation having present a number of accumulated bit shifts that do not apply to what has become the real situation.
Specifically, upon the entry of datum segment k into input register 16, the bit count code nnnnni=01100=12 thereof would have been entered into line A of first router 18, the length of the bit count code itself, or 00110=5 would have been added thereto in the first addition in first adder 20, and an Li value=10001=17 would have been entered both into line C of first router 18 and bit shift accumulator 62 to establish in part the number of bit shifts to be applied to the next datum segment. Moreover, the cumulative Li values for the first and second datum segments, or 8+12=20, would have been added thereto in second adder 64 so as to yield a total bit shift of 37 for the next (or fourth) datum segment. However, when that third k datum segment comes to be treated as it mustnot as the third datum segment of a continuing series but rather as the first of a new seriesand is thus placed into ADDRESS1, the number of bit shifts appropriate to the next-received datum segment, i.e., what will now have become the second datum segment of that new series, will be only the full length of that k datum segment so as to be concatenated onto the end thereof, i.e., only by those 17 bits and not by any prior accumulation of earlier bit shifts. The appearance of that 1 routing bit must then be applied also to the entry only of that 17-bit figure into both shift count register 66, for purposes of counting out the correct number of bit shifts, and into datum segment positioner 56 wherein only those 17 bit shifts will correctly position that next-arriving datum segment.
The means for accomplishing that step is provided as a part of the bit shifting process itself. It has been noted that the correct bit shift value of 17 shifts for the next-following datum segment has already been placed, as usual, at terminal C of first router 18 and at the input to bit shift accumulator 62. To avoid including in the bit shift count the number of bit shifts that had already taken place, it is only necessary to prevent the occurrence of that Σ bsci addition in second adder 64, so that the correct 17 bit value will pass directly into shift count register 66. For that purpose, it would be possible to add another router to the circuit such that second adder 64 was simply routed around on a 1 routing bit, and that procedure would of course fall within the spirit and scope of the invention. However, as indicated above a more simple method of accomplishing that purpose can be found within the bit shifting process itself, as will now be described.
As shown in
In more detail, at startup or on the completion of a concatenation series, both shift count register 66 and second counter 70 will have default values of 00000. That is brought about by the fact that the end of a bit shifting process is signaled by a 0 bit from XOR gate 68, that will cause a reset of second counter 70 back to its 00000 reset value, and the line extending from the r box on second counter 70 shows that the same 00000 value is sent to shift count register 66. That is, a 0 bit will be produced from XOR gate 68 both in having 00000 values in both shift count register 66 and second counter 70 and (in this particular case) in having 01000 values in both shift count register 66 and second counter 70. Or, more generally, a 0 bit from XOR gate 68 will appear at the end of any bit shifting process as to any datum segment, at such time as second counter 70 has counted out that number of bit shifts as shift count register 66 had indicated was required. Bit shifting is started by the entry of a bit shift count bsci into shift count register 66, and more specifically in the present case, as shown in
Besides the fact that a new concatenation series begins with a first datum segment to which no bit shifting is applied, it is also true that as to that concatenation series there would also have been no bit shifting of any preceding datum segments. The way in which that fact is reflected, i.e., wherein the various data entries previously mentioned that would have b en developed in the course of deriving a 1 routing bit from third datum segment k are not included in setting the bit shifts of the new concatenation series, is not by routing around the Σ bsci addition in second adder 64, as was noted above as being one possibility, but rather by resetting the input to second adder 64 to reflect the actual number of prior bit shifts that had already been carried out, which of course is zero when treating the second datum segment of a new concatenation series. The line from the r portion of second counter 70 that resets the content of shift count register 66 to 00000 is thus sent also to second adder 64.
The value of 17 bits as the length Li of datum segment k, that has now become the first datum segment of the new concatenation series, still remains as an input to bit shift accumulator 62, but when passed through second adder 64 to have the value of zero added thereto retains that same, correct value as to the number of bit shifts to be applied to a next-arriving datum segment, e.g., as to an l datum segment following after that k datum segment. As to additional datum segments that can be included within the concatenation series then being carried out, the number of bit shifts for the third and later datum segments would derive from the addition of the real bit shifts that would have been carried out, e.g., as (L1)+L2; (L1+L2)+L3+ . . . , etc., where the term in parentheses would be the value brought back around into second adder 64 by the Σ bsci line, to which would be added the bit count comprising the length of each immediately prior-received datum segment, thus to place each successively new datum segment immediately adjacent the preceding one, i.e., at places further and further down along output register 32 until as many datum segments as output register 32 could then accommodate had been included.
While the bit shifting of a datum segment is proceeding, that datum segment would also have been getting processed through the circuitry of
Besides needing to determine the number of bit shifts required to place a datum segment into the desired end-to-end relationship with adjacent datum segments and then to bring about that bit shifting, both of which ar done by bit shifter 54 as just described, it is also necessary to effect that actual placement of the datum segment in output register 32, and that is done by datum segment positioner 56 shown in detail in
In the datum segment transfers discussed so far, it would have been the full content of a register that was moved into some other register. What must be accomplished by datum segment positioner 56, however, is the transfer of only a portion of the content of a register, specifically transfer register 74, into output register 32, and in such a way as not to overwrite whatever datum segment content might already have been placed into output register 32. The method of so doing is based on the ADDRESSi of the particular datum segment. As will be seen further below, on a 0 routing code datum segment positioner 56 transfers out the data in transfer register 74 that is located in positions starting at a particular ADDRESSi and then on to the right to the LSB end of the register, but data lying on the MSB or left side of ADDRESSi are untouched. The data present in those last-mentioned locations will in fact be a series of 1 bits that would have arisen from the bit shifting that had placed the datum segment of interest to commence at ADDRESSi. The data already contained in output register 32 on the MSB side of ADDRESSi are necessarily also left untouched, which is precisely the desired result since those data would be made up of one or more datum segments that had already been transferred into output register 32, and the data transfer that actually does take place concatenates the new datum segment onto those already present, which is the basic purpose of concatenator 10.
For reasons of space in
The structure of datum segment positioner 56 centers firstly on an array of 32 datum release latches 78-140 through selected ones of which are to pass the respective contents of the 32 bit locations in transfer register 74 that connect to respective D terminals of datum release latches 78-140, with the 32 Q output terminals of datum release latches 78-140 then connecting respectively to the 32 bit locations of output register 32. Secondly, the selection of the bits actually to be transferred is accomplished by an array of XNOR gates 142-204, the outputs of which connect respectively to the G terminals of datum release latches 78-140. The inputs to XNOR gates 142-204 consist of a common 5-bit address bus 206 as one input thereto, and a second input comprising an array of 5-bit address buffers 208-270 that contain in order the respective 5-bit binary codes for the integers 1-32 and connect individually and respectively to each of the XNOR gates 142-204. A 1 bit will be produced directly at the output of any XNOR gate 142-204 into which address bus 206 provides a 5-bit code input that matches th value in an address buffer 208-270 that connects to that same XNOR gate.
Additional connections between the output sides of XNOR gates 142-204 and the G terminals of datum release latches 78-140 are such that upon entry onto address bus 206 of the binary code for any one of the numbers 1-32, a 1 bit will appear on th output side not only of the XNOR gate to which was connected the same binary code as was present on a corresponding one of the address buffers 208-270, but also on the output sides of every other XNOR gate for which the binary numbers serving as one input thereto are larger than the binary number that had been entered. Thus, it can be seen in
To see more specifically how that result is achieved, attention may first be drawn to the horizontal line in
That procedure clearly accomplishes transfer of the 5-bit nnnnn bit count code for the first datum segment, but it may be wondered why connection is also made to the 6th position. That is answered by considering what is the lowest possible address that could be used for the second datum segment. In order to have a second datum segment, there must first have been a first datum segment, which must contain therein at least one bit. That one bit would occupy the 6th position, which then establishes the limitation that ADDRESS2≧7. (Since the first datum segment in this example has 3 bits, it turns out actually that ADDRESS2=01001=9.) Moreover, because of the right-ward pointing diodes 272 that interconnect respective positions 6-7; 7-8; 8-9; etc., a 1 bit at the 1-6 positions will also appear at all of the rest of the positions to the right therefrom, i.e., at positions 7-32 (the outputs of XNOR gates 144-204). As to the first datum segment i in particular,
Proper placement of later datum segments, however, will require not only having both rightward reaching and electrically conductive lines leading from each XNOR gate output to the next adjacent outputs of the XNOR gates located at higher numbered positions, but also means for ensuring that whatever may be written into an ADDRESSx in output register 32 does not write over what might previously have been written therein for the previous datum segment starting at ADDRESSx−1, and likewise for any earlier datum segments. Both of those tasks are accomplished by that array of diodes 272 that connect between all adjacent XNOR gate outputs beyond the first six, since diodes 272 have a directional orientation that will permit the transfer of a 1 bit to the right in
The manner just described in which the locations for the bit count code nnnnn1 and the consequent placement of the datum segment are treated is thus unique to the first datum segment of a concatenation series. Again, in this example the number of spaces required for that bit count code will always be five, since the instance of concatenator 10 selected here as an example was given that size, and of course a corresponding size for input, output registers 16, 32 of 32 bits, given that a 5-bit code can express the integers from 1 to 32. The actual implementation of that 5-bit limitation is precisely at this point, i.e., in the fixed connections to those datum release latches 78-88 that control the inputs to the 6 left-most positions of output register 32.&
In the transfer of later datum segments of a concatenation series for which the routing code will be 0, each such datum segment will have its own ADDRESSi that must be correctly placed within datum segment positioner 56 in order to yield the proper positioning within output register 32. The required ADDRESSi value to be applied to transfer register 74 for the datum segment at hand is obtained in the same way as was the number of bit shifts that were required to place that same datum segment in proper condition for such transfer into transfer register 74. That is, there will be no bit shifting of a first datum segment, but then the bit shifting of the second datum segment is based on the length Li of the first datum segment, as described with reference to
To illustrate how a first datum segment that had previously been transferred into output register 32 will be left undisturbed by later transmissions thereto, an input on address bus 206 of the address 00111, wherein (00111)2=7, would cause a 1 bit to appear at the output of the XNOR gate that connects to that 7th position address bus 206, i.e., XNOR gate 144. That could occur, of course, only if the first datum segment had but one bit, and hence with its 5-bit bit count code (00001)2=1 had filled only the first 6 positions of input register 16. Appearance of the aforesaid 1 bit on the G terminal of datum release latch 90, that is directly connected to that output of XNOR gate 144, would then cause the content of the 7th position on transfer register 74, which connects to the D terminal of datum release latch 90, to appear on the Q output of datum release latch 90 that connects to the 7th position of output register 32. Because of the rightward connection from the output of XNOR gate 90 to the right-oriented diode 272 and thence to the output of the adjacent XNOR gate 92, that 1 bit will also appear on the G terminal of datum release latch 92, thereby allowing the content of position 8 on transfer register 74 to pass through datum release latch 92 to the 8th position on output register 32. Similarly, the continuing rightward connection of diodes 272 to successively next adjacent XNOR gate outputs then causes all of the remaining rightward content of transfer register 74 to be transferred to output register 32.
However, because of the rightward orientation of diodes 272, and especially the diode 272 that lies between positions 6 and 7, that 1 bit at position 7 is not felt on the G terminal of datum release latch 88 that connects to position 6 (and thence to positions 1-5) of transfer register 74, so there will be no change in the content of the 6th position of output register 32, nor of the content of the 1st-5th positions. Whatever may be the content of a new datum segment, given the bit shifts that would have occurred on the content of transfer register 74 for all datum segments other than the first, the second datum segment will be transferred into those positions in output register 32 that commence with the 7th position. The same will apply to the entry of any ADDRESSi larger than 7 (either in lieu of or following thereafter), i.e., if transfer is to occur at all, those data bits that are held at or in higher numbered positions in transfer register 74 than any particular ADDRESSi will be transferred therefrom to output register 32, and none will be transferred to any lower-numbered positions, whose previously transferred content in output register 32 will remain intact.
With specific reference now to the second datum segment j, the first 8 positions in transfer register 74 would have contained the first datum segment 00011010, but after bit shifting and transfer, second datum segment j will instead appear in transfer register 74 as 11111111001111011101, having been bit shifted past those first eight positions into which would have been placed instead the eight 1 bits from the bit shifting process. That is, the full content of output register 32 will appear as 00011010001111011101xxxxxxxxxxxx, where the x's may either be left over from a previous concatenation series or may be a default entry of a zero-string 000000000000 upon startup. The content of the first datum segment will remain, and will not be replaced by those 81 bits arising from the bit shifting process, because the transfer of the second datum segment only begins at ADDRESS2=9.
Alternatively, an ADDRESS2 and later such ADDRESS, values can be obtained for use in datum segment positioner 56, for the aforesaid datum segment positioning purposes, from the output of the second addition in first adder 20, as these are established as each new datum segment arrives as shown in
Although the invention has been described in terms of ordinary electronic digital gates embodied in semiconductor technology (but substituting mechanical switches for purposes of illustration when that seemed to make a description more clear), it should be recognized that the circuit structures and overall architecture of the invention do not depend upon any specific technology. Implementation of the invention could as well be based on other kinds of hardware to form the necessary binary logic, even including, for example, carbon nanotube transistors as described by Ali Javey, Jing Guo, Qian Wang, Mark Lundstrom & Hongjie Dai, Ballistic Carbon Nanotube Field-Effect Transistors, NATURE, Vol. 424, Issue No. 6949, pp. 654-657 (7 Aug. 2003); the kind of gates formed of magnetoresistive elements as described in A. Ney, C. Pampuch, R. Koch & K. H. Ploog, Programmable Computing With a Single Magnetoresistive element, NATURE, Vol. 425, Issue No. 6957, pp. 485-487 (2 Oct. 2003); semiconductor light sources and fiber optics using photon transmissions, or, in a quantum computing context, controllable quantum dot interchanges using electron spin for the 0 and 1 bits. Implementation of the invention using these or other technologies, whether presently known or yet to be conceived and demonstrated, are deemed also to fall within the spirit and scope of the present invention and of the claims appended hereto. More particularly, in lieu of datum segment positioner 56 as described and shown herein for purposes of placing datum segments into desired positions in an output register there might be used the data selector 40 and data release latch 42 of U.S. application Ser. No. 10/462,868, filed by the present inventor on Jun. 16, 2003, having the title Gate-Based Zero-Stripping and Varying Datum Segment Length and Arithmetic Method and Apparatus, and which is incorporated herein by this reference and serves the purpose, as does datum segment positioner 56, of placing datum segments that have undergone some previous processing into some set of previously identified register addresses. (In lieu of the system in datum segment positioner 6 that employs diodes 272 to separate those datum release latches that are to pass data therethrough from those that are not, data selector 40 and data release latch 42 of U.S. application Ser. No. 10/462,868 uses selectable direct connections through an array of datum release latches between input and output registers.) That datum segment distribution procedure, and other variations of other particular methods and apparatus described herein that in such cases would be known to a person of ordinary skill in the art, are similarly deemed to fall within the spirit and scope of the present invention, which must be interpreted only by the claims appended hereto.
It should also be noted that the data produced by concatenator 10 will be unreadable by any computer or other data processing system that has not been adapted to treat variable length datum segments, e.g., by transforming such data as derive from concatenator 10 into some predetermined fixed length as may be used by a particular computer or data processing system. The reason is that standard data processing systems use Central Processing Units (CPUs) or at least processors, that are designed to accommodate bytes of some fixed size. From concatenator 10, however, instead of receiving bit strings of a corresponding length such processors could receive within, say, one 16- or 32-bit address, a number of datum segments other than one, one datum segment and part of a second, just a part of one datum segment, and so on, and any attempt to interpret those data by fixed size means would generally yield only gibberish. Concatenator 10 can thus function also a hardware-based computer security device.
The design and construction of other variations in the forms of the electronic, light or other components than those already mentioned herein, and in the particular selection of components, could easily be carried out by a person of ordinary skill in the art, based on the present description of the manner of so doing and the functions being accomplished, hence all such variations are deemed to fall within the spirit and scope of the invention and of the claims appended hereto. Other arrangements and dispositions of the aforesaid or like components, the descriptions of which are intended to be illustrative only and not limiting, may also be made without departing from the spirit and scope of the invention, which must be identified and determined only from the following claims and the equivalents thereof.