WO2001001265A1

WO2001001265A1 - Device for managing data exchanges between data processing equipment

Info

Publication number: WO2001001265A1
Application number: PCT/FR2000/001767
Authority: WO
Inventors: Fabrice Rouillier; Jean-Charles Faugere
Original assignee: Inria Institut National De Recherche En Informatique Et En Automatique; Universite Pierre Et Marie Curie
Priority date: 1999-06-25
Filing date: 2000-06-23
Publication date: 2001-01-04
Also published as: FR2795573A1; FR2795573B1; EP1190332A1

Abstract

The invention concerns a data converting device comprising storage means (7) for storing first and second symbol sets, all different, representing a first word arrangement, individually coding primary elementary data, and a second set of symbols, all different, representing a second word arrangement. It further comprises an operating unit (8) which receives the first and second sets of symbols and a primary elementary information to operate on said primary elementary information transformations of the words solely by the first and second sets of symbols so as to supply an equivalent secondary information.

Description

Device for managing data exchanges between computer hardware

The invention relates to the field of data exchange between computer hardware of identical or different types, such as software, microprocessors, databases, and the like.

In the IT field, the use of increasingly powerful microprocessors makes it possible to reduce ever more computation or processing times. However, this leads to an manipulation, in real time, of an ever-increasing amount of data, for example greater than 1 GB (Gigabytes) in the case of systems such as FGB and RS, or in the case of communication systems. database management (DBMS) and image processing.

In a microprocessor, or more generally in computer hardware, the data are generally stored in the form of "packets" of k bits (binary information) in registers with a capacity of (n * k) bits. By register is meant here, floating registers, certain memory components of microprocessors, graphics cards, MMX and the like. A microprocessor therefore does not know how to process (whole) data the size of which is greater than the capacity of its registers, ie (n * k) bits.

In general, each "packet" includes k = 8 bits. We then speak of a byte. For example, the largest integer that a 32-bit microprocessor can process includes 4 bytes (n = 4, k = 8).

The order in which the k-bit packets (for example bytes) are stored often varies from machine to machine. It is then said that the hardware has different internal encodings, or in other words arrangements of packets of k different bits. However, whatever the machine, the binary representation of a given object (for example an integer) of size less than k bits is invariant (all the bits are given in the same order). Such a representation is therefore common to everyone.

A real problem therefore arises when two pieces of equipment operating according to different internal codings wish to exchange data objects (scalars) whose dimensions are greater than or equal to k bits. This problem is further reinforced when the dimensions (k * n) of the material registers differ.

For example, the integer 33,751,553 which breaks down in the base {2 ⁸ } in the form 1 + 2 * 2 ⁸ + 3 * 2 ¹⁶ + 2 * 2 ²⁴ , is coded in a microprocessor of the ALPHA type or PC thereafter of n = 4 coefficients [1,2,3,2]. Now, in a microprocessor of the SPARC type, this integer is coded thereafter by n = 4 coefficients [2,3,2,1] which for a microprocessor designates the integer 16,909,058 (2 + 3 * 2 ⁸ + 2 * 2 ¹⁶ + 2 ²⁴ ).

In this example, we see that a permutation of the coefficients differentiates the two internal codings.

To allow such materials to exchange their integers, it is therefore essential that they know their respective internal codings, or in other words the permutations which will allow them to transform their respective codings.

However, currently, the permutations are given as a function of couples (k, n) fixed once and for all, generally using software such as XDR (Trademark registered by the SUN Company).

Such software in fact provides transcription of machine and floating-point integers codable on 8, 16 and 32 bits (a convention is proposed for integers of 64, but not beyond) in an external coding (or transmission coding) which himself is found to be identical to the internal coding of microprocessors of the SPARC type. External coding can be described as "common or universal language". In this software, the arrangement of the k bits of each packet (byte for k = 8) is always invariant.

This type of transcription requires, for each data exchange, a first conversion (or encoding) of the first internal code of the "transmitting" material to the external encoding, then a second conversion (or decoding) of the external code to the second internal code of the receiving equipment.

Double conversion is also carried out when the materials are compatible with each other (same internal coding and incompatible with external XDR type coding). To avoid this, it is of course possible to reconfigure the XDR software, but this requires manipulation by an operator.

Furthermore, as it stands, XDR software is difficult to use in 64-bit environments, and unusable in 128-bit environments. More generally, as soon as scalar data exceeds 32 bits, the XDR software leaves the direction of operations ("the hand") to the user to process data greater than 32 bits. On the other hand, it is not intended to work with different 8-bit packets.

The known solutions therefore do not allow dynamic parameterization and / or exchanges between materials independently of their respective architectures.

In summary, no known solution brings complete satisfaction in terms of speed, efficiency and adaptability.

The invention therefore aims to improve the situation in terms of data exchange. To this end, it proposes a device intended to work on primary elementary data coded individually according to a first arrangement of words (or internal coding), and comprising: * storage means where the first and second sets of symbols are stored, all distinct, respectively forming a representation of the first arrangement and of a second arrangement of words (or external coding), a priori distinct from the first, and * an operator capable of receiving as input the first and second sets of symbols and a primary elementary datum , such as an integer, in order to effect on it transformations of words defined only by the first and second sets of symbols, so as to provide as output a corresponding secondary datum and equivalent to the primary elementary datum.

This provides a fully configurable and dynamic converter.

The invention finds a particularly interesting application when first and second hardware wish to exchange primary elementary data. In this case, the first material delivers primary elementary data coded according to the first arrangement (or first internal coding), while primary elementary data coded according to a fourth arrangement (second internal coding) by a second material are converted by a means of conversion in the form of secondary data coded according to a third arrangement (second external coding).

As indicated in the introduction, the word arrangement must be considered here as an arrangement of groups of bits in a register (in practice, each group generally being formed of 8 bits (or byte)).

According to the invention, the operator of the device comprises interrogation means which carry out the following operations: * first of all, they provide the second hardware with a message which contains the second set of symbols and requires the sending, in return, of a primary elementary datum, transformed from the second set of symbols by the coding according to the fourth arrangement;

* then, they deduce from this primary elementary datum and from the first and second sets of symbols a third set of symbols forming a representation of the fourth arrangement;

* then, they replace the second set of symbols with the third set of symbols, both in the operator and in the conversion means, so that:

- in the event of transmission of a primary elementary data coded according to the first arrangement and intended for the second material, the operator delivers to it, directly, a primary elementary data coded according to the fourth arrangement, and in the event of transmission of a primary elementary data coded according to the fourth arrangement and intended for the first material, the operator delivers to it, directly, a primary elementary data coded according to the first arrangement.

In this way, in particular when the materials are of radically different types, and consequently have different internal (first and fourth arrangements) and external (second and third arrangements) codings, the device according to the invention can be configured independently of the architectures of the equipment wishing to exchange data.

Once the device has established a direct link (that is to say a "direct conversion operator" between the first and second internal codings), the time necessary for the exchange of data between the materials is very significantly reduced .

The invention also relates to the methods which will be described below and which allow the device to ensure their conversions. Other characteristics and advantages of the invention will appear on examining the detailed description below, and the appended drawings, in which:

- Figure 1 is a diagram illustrating an embodiment of one invention in an application to the exchange of data between two computers (or more generally two computer hardware) of different types; and

- Figure 2 is an algorithm describing an embodiment of the method according to the invention.

The attached drawings are, for the most part, certain. Consequently, they may not only serve to supplement it, but also contribute to the definition of the invention if necessary.

We first refer to FIG. 1 to draw up an inventory of what was done before the present invention in terms of data exchange between two computer hardware, such as computers (or workstation) Ml and M2.

Of course, it could more simply be microprocessors, or even different software or databases, possibly installed in the same machine (or computer), but operating according to different internal and / or external codings.

It is important to note that FIG. 1 does not represent the prior art as such, but that it makes it possible to present the elements

In the example illustrated in FIG. 1, the computer M1 comprises a microprocessor 1, for example of the 32-bit SPARC type. This microprocessor 1 is installed on an electronic card in order to be able to cooperate with a hard disk 2. The microprocessor 1, under the control of the operating system 3 of the computer, stored on the hard disk 2, performs operations on data, and delivers 32 bit data on an output 4 (when it is of the type SPARC) according to a first arrangement (or first internal coding).

Each 32-bit datum is then delivered at the output 4 of the microprocessor 1 in the form of an ordered sequence of four bytes (8 bits). The internal code Ml _kn (E) is called the ordered sequence of four (n) bytes (k = 8) representing an integer E in a 32-bit SPARC microprocessor. With such a SPARC microprocessor, the internal coding of a 32-bit integer is the result of the coefficients of its base decomposition {2 ^k }, ordered according to the decreasing powers.

For example, in a PC or ALPHA microprocessor, the internal coding of a 32-bit integer is the result of the coefficients of its base decomposition {2 ⁸ }, ordered according to increasing powers.

To allow the transmission of such integers E, from the computer Ml to the computer M2, it is necessary that the integers have the same format, or external coding. This is not always the case, as we will see below.

Consequently, a conversion module 5 is conventionally provided, generally installed in the form of software (or program) on the hard disk of each computer. Of course, the conversion module can be produced in the form of an electronic circuit.

Anyway, this conversion module 5 is connected to an interface 6 coupled, for example by a wired link, to the interface 6 'of the computer M2 with which it wishes to exchange data.

To define the external coding (or second arrangement), use is made of a base {1, 2 ^k , 2 ^2k , ... 2 ^nk }, where k and n denote respectively the number of bits of each word of the arrangement and the number of elements of the base, and where n * k is equal to the number of bits of the data.

For example, in the aforementioned case, the first arrangement is defined by n = 4 words of k = 8 bits, that is n * k ≈ 32 bits.

In this example, the conversion module 5 therefore has the function of converting a primary elementary datum delivered by the output 4 of the microprocessor 1 (that is to say provided according to the first internal coding or first arrangement M _xkn ) into a datum secondary coded according to external coding, or second arrangement, O _{λ kn} .

Generally, what differentiates internal coding from external coding, in the same hardware, is a permutation type operation. In this case, we have the relation: D _l , _kf n (E) = Φι, _k , _n <Mι _fk , _n (B))

It is quite obvious, that in certain cases, the permutation ^φ _k n ( ⁼ l'2) can be the identity. In this case, M and D are made up of the same ordered sequence of words defining the integer. They are said to have the same arrangement, or format.

What has just been said for the first computer Ml also applies to the second computer M2. Only the internal coding M _{2 kn} , as well as possibly the external coding ^D _{2 k} n 'are different. Here we mean by different, either arrangements (or sequences) whose elements (or words) are ordered differently, or arrangements which do not have the same number of elements (kl and k2 different and / or ni and n2 different).

Generally, especially when one is in a client-server type environment, the hardware is substantially homogeneous, so that the external encodings they use are identical. In this case, we have the following relation: Φ2,, n (M ₂ , _kf n (E)) = D _k , _n (E) = • ι _fk , _π (M _{lf kf ll} (E))

In conventional machines, for each primary elementary data item delivered by the microprocessor 1, on its output 4, and coded according to the first arrangement (or first internal coding), the conversion module 5 performs an encoding (or first conversion) intended to provide at the interface 6 a secondary datum (generally an integer E) coded according to the second arrangement (or first external coding). In other words, at the output of the conversion means 5, the following data is available:

D _k , n, ⁽ E ⁾ = • ι _f k, n <M _lfkf n <= >>

This secondary data is then sent to the second computer M2, which will decode it (second conversion) using its 5 'conversion module (M _{2 n} (E) = Φ ^_1 _{2 kn} ( ^D _k n, ()) ) 'to provide the microprocessor with primary elementary data coded according to its second internal coding (or fourth arrangement), so that it can process this data.

The same applies when the second computer M2 wishes to transmit primary elementary data to the first computer M1, and in particular to its microprocessor 1. The encoding then consists in forming the secondary data:

^D 2, k, n (E) - Φ ₂ , _k , n ( ^M 2, k, n (E))

and the decoding of the secondary data supplied by the second computer M2 is carried out in the conversion module 5 of the first computer M1. Which provides a primary elementary data according to the first arrangement (or first internal coding): M _{1 k} , _n (E) = - ¹ ₁ , _k , _n (D _{kfri /} (E))

The conversion module 5 of each material Mi (i = 1.2) can only carry out a single coding from the internal coding to the external coding, and vice versa. As a result, this type of hardware can only operate with hardware having at least one common external coding D _kn . Those skilled in the art have proposed to install data exchange software in certain hardware, for example on their hard disk. These include, for example, the XDR software (Trademark) from the manufacturer of SUN workstations. This software offers a library of encoding / decoding functions which allow a first device of a first type to exchange data with a second device of a second type, once the respective types of this device have been declared. It is, therefore, a purely static type conversion operation, since it requires the intervention of an operator knowing the respective types of the two materials.

In addition, this type of software systematically performs, for each data item to be exchanged, a double conversion. The first conversion consists of the encoding of the primary elementary data according to the first internal coding into a secondary data according to the external coding. The second conversion consists in a decoding of the secondary data according to the external coding into a primary elementary data according to the second internal coding.

In addition, unless reconfigured by hand, by a specialist, the XDR software continues to perform its double conversion when the two materials are identical.

This double conversion greatly slows (at least a factor of two (2)) the data processing speeds.

The invention provides a solution to this drawback.

In what follows, the invention will be described, with reference to FIGS. 1 and 2, in an application to the exchange of data between two computer hardware, such as computers (or workstation) M1 and M2.

Of course, it could be, more simply, microprocessors, or even different software, possibly- They are all installed in the same machine (or computer), but operating according to different internal codings.

The invention provides a device comprising a first part which replaces the conversion module 5 in the first material M1, and a second part which completes the first and is implanted at least partially in the first material Ml (the rest then being implanted in the second M2 material).

Preferably, the device is produced in the form of software modules installed on the hard disk (s). However, it can also be produced in the form of electronic circuits. A combination of the two (software and circuit) can also be considered.

In the first computer M1, there are installed memory means capable of storing a first set (or series) of symbols, all distinct, forming a representation of the first arrangement (or first internal coding). Such storage means are, for example, produced in the form of program lines which refer to addresses of registers or memories of the hard disk 2 of the computer M1.

The storage means 7 also store a second set (or series) of symbols, all distinct, forming a representation of the second arrangement of words (or first external coding), generally distinct from the first arrangement. It can indeed be identical in certain cases.

Preferably, the symbol sets consist of an ordered sequence of n components which characterize, as indicated above, the first and second arrangements.

The device further comprises an operator 8 coupled to the storage means 7, as well as to the output 4 of the microprocessor 1. It can thus receive on an input the first and second sets of symbols, as well as each elementary data. primary shutter coded according to the first arrangement by the microprocessor 1.

This operator 8 is preferably produced in the form of a software module (or program) using a library of mathematical calculations. Its function is to perform on the primary elementary data, received in the form of an ordered sequence of words, transformations of defined words, only, from the first and second sets of symbols. At the output of this operator 8, secondary data equivalent to the primary elementary data received is delivered. The word transformation must be understood here in its mathematical definition, that is, as a function or application.

In this way, an improved conversion module is produced which is fully configurable and adaptable to any type of material.

The device according to the invention also makes it possible to accelerate very significantly the speed of data exchange between two materials Ml and M2 of different types, in particular. In what follows, it will be considered that not only the internal codings of the two materials M1 and M2 are different, but that their external codings are also different. For example, the microprocessor 1 of the computer M1 is of the 32-bit SPARC type, while the microprocessor 1 ′ of the second computer M2 is of the 64-bit ALPHA type.

To allow these two materials M1 and M2 to communicate, the device according to the invention implements an interrogation protocol. This protocol is initiated by the first computer M1 in order to determine the internal coding (or fourth arrangement) of the microprocessor l 'of the second computer M2.

The interrogation is carried out by an interrogation module 9, which is constituted by a software (or program) module. This interrogation software module is preferably installed for a part 9-1 on the hard disk of the computer Ml, and for another complementary part 9-2 on the hard disk 2 'of the computer M2. Of course, as indicated above, each part of the interrogation module 9-1 and 9-2 can be produced in the form of electronic circuits.

Furthermore, it is possible to envisage a variant in which the entire interrogation module 9 is located on the first computer M1. In this case, the interrogation software module 9 is designed so as to spontaneously implant, during a first interrogation, a few selected program lines (substantially equivalent to 9-2) on the hard disk 2 ′ of the second computer M2, and preferably in its 5 ′ conversion module, when this is produced in the form of a software module.

We will now describe a mode (or method) of operation of the device according to the invention in which the external coding ^D _k , _n ^is assumed to be fixed in advance and identical for the two computers. This mode emulates the operating mode of the XDR software.

According to the invention, the set of second symbols (representative of the second arrangement) is an ordered sequence composed of elements distinct from each other: [1, 2, 3, ..., n] and equal to D _kn (E). This ordered sequence is associated with the integer E = 1 + 2 * 2 ^k + 3 * 2 ^2k + ... + n * 2 < ⁿ " ^{1) *} , which will be sent to the second computer M2 to determine its fourth arrangement .

For reasons of convenience, it is assumed here that n is less than 2 ^k , the general case can be easily treated by composing the methods (or operating modes).

In other words, by addressing the integer E indicated above, or more exactly by addressing the ordered sequence [1, 2, 3, ..., n] comprising numbers all distinct from each other, we will , in return, deduce the second internal code (or fourth arrangement) of the second machine M2. As an example, we take k = 8 and k * n ≥ 32.

To completely determine the permutation _{2 kn} implemented by the conversion module 5 'of the second computer M2, it suffices to take the integer E = 1 + 2 * 2 ⁸ + 3 * 2 ¹⁶ + 4 * 2 ²⁴ and the data secondary D _kn (E) formed by the sequence of symbols

In practice, the determination of the permutation Φ _{2 kn} amounts to constructing a noted table, for k fixed, permut, such that permut [n] is the ordered list representing Φ _mkn .

The program for detecting the internal coding and the construction of the external coding is given as an example, in C language, in the Mod2 module of the appendix.

The functions construct ordered lists external_codingfi] and internal_coding [i], for values of i taking the sizes of the usual integers of the C language (char, short, int, and long). It is recalled that in C language (which is only an example of a usable programming language) the long variable indicates the maximum size of the registers that the microprocessor uses to store scalar data). The variables external__coding and internal_coding respectively represent the external coding D _kn and the internal coding M _kn . These variables are given by way of example in the Modl module of the appendix, where "B8" is the invariant type which makes it possible to represent an integer from the value 0 (zero) to the value 255.

For a fixed value i, we construct the integer E = 1 + 2 * 2 ^k + ... + n * 2 * ⁽ⁱ -D.

The aforementioned ordered lists form tables. The array external_coding [i] is then equal, for k = 8, to the external coding D _{8 fi} (E) = [1, 2, 3, -, i]. The internal_coding [i] table is equal to M- ^ g ^ E) which varies depending on the computer (or machine considered). We then define a set (or multiplicity) of several permutation functions Φ _{m 8 ±} for values of i varying from 1 to sizeof (long) (or in some cases from 1 to sizeof (long). the integer scalar having the largest admissible size in the C language; in general it is the largest integer understandable by machine or hardware. These permutation functions take into account integers codable on 8, 16, 32 and 64 bits, they can be easily extended to larger values, especially 128 bits.

The permutation functions are defined, for example, in the Mod3 module of the appendix. It will be noted that the formulation of this program module makes it possible to detect the values of i for which no permutation is necessary.

It is then necessary to describe the encoding / decoding functions which will be used when the parameters of the problem, namely the internal coding, external coding and permutation function have been determined.

For example, the function allowing the external coding of an integer (or array of m 8-bit elements) of given size (sz) to be transformed into its internal format can be defined by the Mod4 module of the appendix.

As will be appreciated by those skilled in the art, the Mod4 module describing the encoding / decoding functions offers generic functions, insofar as, on the one hand, they only depend on the size of the integers to be processed, and that on the other hand, care is taken to make permutations only if this proves necessary.

In the same way, the inverse functions are defined which make it possible to transform the internal coding of an integer (array m of element of 8 bits) of given size (sz) into another internal coding (array p of elements of 8 bits). These inverse functions are defined, for example, by the Mod5 module in the appendix. They therefore make it possible to directly convert a primary elementary datum coded according to the first, respectively fourth, arrangement into a primary elementary datum coded according to the fourth, respectively first, arrangement.

Using Modl to Mod5 modules, a protocol (or process) for exchanging binary data is implemented, the implementation of which does not depend in any way on the respective architectures of the computers (or machines) considered. Of course, this implementation is linked to the language used, here the C language. However, a transposition to another computer language can be easily obtained.

To operate the exchange protocol, the first computer M1 opens the communication channel which connects it to the second computer M2, then activates this second computer M2 and sends it the whole E defined by the second set of symbols, here [1 , 2, 3, 4], on the chosen communication channel. The second computer M2 reads the whole, transforms it by coding it according to its fourth arrangement (or second internal coding) and returns this transform on the communication channel.

An example of a program module making it possible to carry out the operations mentioned above is given in the Mod6 module (as regards part 9-1 installed in the first computer Ml) and in the Mod7 module (as regards the part 9-2 located in the second M2 computer).

Of course, and as indicated above, the part 9-2 intended to activate the second computer M2 can be installed remotely by the first computer Ml. To do this, it suffices that the interrogation module 9 of MI addresses an adaptation of the program module proposed in Mod7.

In the Mod6 module, two arguments are called: a first machine name (first computer) and a second machine name (second computer) which must be activated by the first machine. By the way, in Mod7, two arguments are also called: a machine (or computer) name and a port.

In these two modules, the functions send_n (nb, bus, buf, n), respectively read_n (nb, bus, buf, n), write, respectively read, an array buf of n * k bits (k is in practice fixed at the value 8) consisting of integer machines of nb * k bits on a channel denoted bus.

It is clear that these functions will depend on the chosen communication channel (shared memory, files, databases (DBMS), graphics cards, microprocessor, image and / or sound storage formats (multimedia), sockets and the like) . Such functions call (in the form of loops) the functions conv_machine_2_prot_UI and reorder_UI, and manage the input / output buffers if necessary.

In the above, the external coding D _kn was assumed to be fixed in advance. It therefore corresponded to an embodiment of the device according to the invention which is particularly well suited to equipment which can communicate with one another due to the same external coding D _kn .

However, the device according to the invention can also be adapted to the exchange of data between hardware having different external codings. In this case, the device dynamically calculates a common external coding, which amounts to not a priori fixing a data exchange format.

A preferential solution consists in fixing as internal exchange coding the internal coding of one of the two machines (or computers) so that one of the two permutations _kn and Φ _{2 k} , _n , used by these two machines is equal to identity.

The main steps of dynamic determination of an exchange format (or external coding) according to the invention will now be described with reference to FIG. 2. First of all, in a step 10, the microprocessor 1 of the first computer M1 addresses to the implanted device, here on the hard disk 2, a data item coded according to the first arrangement (or first internal coding).

This data being intended for the second computer M2, for which neither the external coding (or third arrangement) nor the internal coding (or fourth arrangement) is known, an initialization of the part 9-1 of the protocol is carried out in a step 20 contained in the first computer Ml, and in particular in the operator 8. This consists in fixing a default value for D _kn . In fact, this default value for D _kn is the ordered sequence described previously [1, 2, 3, ..., n]. Then follows, in a step 30, an initialization of the part 9-2 of the protocol which is in the second computer M2.

It is clear, as indicated above, that when the entire interrogation program is installed in the first computer M1, the program lines 9-2 ensuring the initialization of the second computer M2 must be transferred to it ci, for example in its 5 'conversion module. This step 30 therefore consists in replacing the external coding D _{2 k} , _n , (or third arrangement) of the second computer M2 by the default value D _kn fixed in step 20.

In fact, Ml sends an interrogation message to M2 containing the second set of symbols which represents the second arrangement, or one or more variants thereof, comprising a different number (s) of symbols ( s) (higher or lower), so that in the case of different exchange formats, at least one of these symbol sets can be processed by the 5 'conversion module of M2. In this case, it is the memory means 7 of M1 which store the different variants of symbol sets comprising numbers n of different words and / or words of number k ± of different bits. The microprocessor 1 'of M2 returns in return to the conversion module 5', which now codes in the exchange format D _kn imposed by default, a primary elementary datum coded according to its fourth arrangement (or second external coding) M ₂ , _n , . This data is converted (encoded) according to the external coding D _kn and addressed to Ml on the communication channel.

In a step 40, the interrogation module 9-1 installed in the first computer M1 will read on the communication channel the data M _{2 k} , _n , (transformed from the second set of symbol D _kn ) which represents the internal coding of the second M2 computer.

In a step 50, the interrogation module 9-1 of the first computer M1 then replaces the default value of the external coding D _kn with the data received from the second computer M2, namely M _2fk , _fn ..

Then, in a step 60, the default value of D _kn supplied during step 30 is replaced in the second computer M2, and in particular in its conversion module 5 ′.

We can note that, due to the replacement in M2 of D _kn (external coding imposed by default) by M _{2 k} , _n , (internal coding of the microprocessor l '), we find ourselves in the 5' conversion module of M2 with a permutation Φ ₂ < _n <equal to the identity. Similarly, we find ourselves in the "conversion module" 5 of Ml with an external coding identical to the internal coding of M2, so that any primary elementary data of a microprocessor 1 or the, coded according to the first or the fourth arrangement, can be converted directly into primary elementary data of the other microprocessor 1 ′ or 1, coded according to the fourth or the first arrangement.

The two machines M1 and M2 are now ready to exchange primary elementary data directly in a step 70. Primary elementary data delivered by the microprocessor l 'of M2, or by the conversion module 5 of Ml, will therefore not undergo conversion in the conversion module 5'. This 5 'conversion module is, in a way, short-circuited. Furthermore, data intended for the microprocessor 1 of Ml is converted, directly, from the internal format (fourth arrangement) of M2 to the internal format (first arrangement) of Ml.

It is clear that the process which has just been described with reference to the algorithm illustrated in FIG. 2 adapts automatically, through a negotiation phase, to the case described above, in which the two computers (or machines) have the same external coding as original. In this case, the initialization steps 20 and 30 are unnecessary.

An example program, making it possible to implement the algorithm illustrated in FIG. 2, is given in module Mod8 of the appendix. To this program are associated, as indicated previously, two programs, of the type of those given in the Mod6 and Mod7 modules of the appendix. These are the programs allowing respectively to activate the conversion module 5 of the first computer (Mod9 module of the appendix) and the module allowing to activate the second computer M2 (ModlO module of the appendix).

In this invention, the definition of scalar data must be taken in its broad sense, that is to say as basic units. Under these conditions, the notion of coding must also be taken in its broadest sense, that is to say as a method (or mode) intended to order scalar units. Coding can therefore be forced or initiated by a user, as needed.

The invention is not limited to the embodiments of the device and method described above, only by way of example, but it encompasses all the variants that a person skilled in the art may envisage within the framework of the claims below. -after. Thus, a device and the associated method have been described in which the second set of symbols (representative of the second arrangement) consisted of a series of n elements, all distinct and with values increasing from 1 to n. However, it is clear that any other series of n distinct elements can be used.

Annex

* Modl:

Typedef longlong Ilonglong

Typedef unsigned int UI32

# dβf ine _max_sizθ_typβ 255

B8 _θ itβrnal_coding [_ ax_sizβ_typβ] [_πιax_sizβ_typβ]; B8 intθrnal_coding [_ ax_3izβ_typβ] [_max_sizθ_typθ]; UI32 pβπButL_œax_sizβ_typβ] Lιnax_sizθ_typθD;

* Mod2:

τoid inlt_peπnαt_c ar (String tab.int) {tab_int [0] -l; >

“Definβ hinder θ_perαut (typ) \ τόîd inlt.perαα _ #t typ t # (String tab.int) \ {\ typ rββ * l; \

pov ** inc; \ res + «(top) * pow; \ tπp ++; \> \ memcpyCCvoid *) tab_int, (void *) (very), sizeof (typ)); \

generθ.peπαut (short) genβre.pθrααt (int) geaere_peraut (long) genere.permut (Ilonglong) void set _permutation_0 (String int_coding, const UI32 sz)

{if (sz≈≈sizeof (char) H init_permut_char (int_coding);

> if (sz≈sizeof (short)) {init_permut_short (int_coding);

> if (sz * "sizeof (int)) i init_peπmιt_int (int_coding); y if (sz≈≈sizeof (long)) - C init_pβπnut_long (iαt_coding);

> if (sz "sizeof (Ilonglong)) {init_permut_Ilonglong (int_coding);

>

void sβt_pem-utations_00 C

UI32 i-1; vhile (i <= maxsize_prot) {set .permutât ion.0 (fc (int ernal_coding [i] [0]), i); i * 2;

>

/ * nazsize_prot init init the maximum size of the integers to be considered by the protocol * /

void set_exchangβ_0 () <

UI32 il.j; vhile (i <= maxsize_prot) -C for (j = 0; j <i; j ++ H external_coding [i] [j] = j + l;> i * »2; * Mθd3:

UI32 find_elem (B8 elt, String tab, const UI32 sz) {

UI32 i * 0; vhile ((i <sz) ftft ((tab [i])! - (elt))) i ++; return (i); >

void define_permutation (String int .coding, String ezt.coding.

UI32 ** peπα_ coding, const UI32 sz)

-VS

UI32 i; if (strncmp (int.codin, ext .codin, sz)). permnt.type [sz] * NEED_PERlfuT; for (i "0; i <sz; i ++) {

(* peπa_coding ⁾ [i] = f ind_elem (int_coding [i3, ezt.coding, sz); >> else pertπut_type [sz] "HO_PERMϋT; ^}

void C initProtocolO

{

UI32 i-1; UI32 * ptr; set .exchange.O ()

C initProtocol_0 (); vhile

{ptr = k (permut [i] [0]); def ine_perπιαtation (fc (internal_coding [i] [0] ⁾ , external_coding [i3, fcptr, i ⁾ ; i * 2;

> * Mod4: void reordβr.UI (const UI32 sz, String m) i UI32 i; static B8 tmp jnax.Bize.type]; UI32 * permut_tmp ~ NULL; if (permut .type [sz] ≈HEED.PEBMUT) {. permut _tmp "* (permut [sz] [0]); for (i = 0; i <sz; i ++ H tmp [i] * m [permut _tmp [i]];

> for (i "0; i <sz; i ++) {m [i] = tmp [i]; >}>

* Mod5: void conv_nachine_2.prot_UI (const UI32 sz, String p.Cste.String m) i.

U132 i;

permut _tmp * fc (permut [sz] [0]); f or (i "0; i <sz; i ++) -C p [permut _tmp [i]] * m [i];

>> else {for (i-0; i <sz; i ++) {p [i] ^ [i]; >>>

* Mod6:

* / int jnain (int argc, char ** argv)

UI32 te-123, re * 0;

/ * open the communication channel and start the server * /

BUS GB * creat eBuβ (argv [1], argv [2]);

C initProtocolO;

Bβndjα (sizeof (UI32), GB, (char *) (fcte), 1); read_n (sizeof (UI32), GB, (char *) (axis), 1); fprintf (stderr, »'Xu *', te); fprintf (stderr, »'Xu *', re);

/ * * Mod7:

* / int main (int argc, char ** argv) i

UI32 te "0;

/ * open the communication channel and launch the Bβrveur * /

C initProtocolO; rβad_n (sizeof (UI32), GB, (char *) (ftte), 1); sβnd_n (sizeof (UI32), GB, (char *) (ftte), 1);

* Mod8:

void C rβinitProtocolO

UI32 i-1; UI32 * ptr; set .θxchange.O () C__initProtόcoi «.d (); vhile (i <= maxsize_prot) ptr≈fc (permut [i] [0]);

* Mod9:

* / int main (int argc, char ** argv) i.

UI32 te * 1234, re-0;

/ * maximum size of integers supported by this processor * /

B8 maxsizθ_loc "sizeof (long);

/ • maximum size of integers supported by the client * /

B8 maxsize_ext ~ 0;

/ * open the communication channel and launch the server * / BUS GB * ereateBus (argv [l], argv [2]);

/ * initialization via protocol ut / * initProtocolO;

read_n (l, GB, (char *) C & Cmaxsize.ext)), 1); send_n (l, GB, (char *) (fc (maxsize.loc)), 1); / * calculation of the maximum size of integers supported by the protocol * / maxsize_prot = min (maxsize.loc, maxsizθ_ext);

/ * reception of the internal coding for the machine on which the integer server running is eligible for the protocol, allocation to the exchange coding

* / f or (i≈l; i <= ((UI32) maxsize_prot); i ** 2H read_n (l, GB, (char *) (fc (external_coding [i] [0])), i);

>

/ * re-initialization of the protocol with the new exchange format * / C reinitProtocolO; / * sending orders * /

8βnd_n (sizeof (UI32), GB, (char *) (fcte), 1);

/ * reading the result • / read_n (sizeof (UI32), GB, (char *) (axis), 1); fprintf (stderr, '' ta ", te); fprintf (stderr,"% u ", re);

* Mod 10:

• int main (int argc, char ** argv) i

/ * maximum size of integers supported by this processor * /

B8 maxsize_loc ^s 8izeof (long);

/ * maximum size of integers supported by the .client * / B8 maxsize_ext * 0;

/ * open the communication channel and start the server * /

BUS GB * createBus (argv [l], arg [2]);

/ * initialization of the protocol by default * / C initProtocolO; send_n (l, GB, (char *) (c (maxsize_loc)), 1); read_n (l, GB, (char *) (sdαaxsize.ext)), 1); / * calculation of the maximum size of integers supported by the protocol

naxsiz _θ _prot = min (max8izθ_loc, maxsize_ext>;

* sending of the internal coding for this machine of the integers admissible for the protocol * / f or (i ^" l; i <= ((UI32) maxsize.prot); i * -2) {sendjι ⁽ l, GB, ⁽ char * ) (ft (internal_coding [i] [0])), i);

> f * the internal format of integers ^β admissible for the protocol is assigned to the exchange format

* / forCi-l; i <»CCUT32) maxsizβ_prot); i ** 2) {forCj-0; j <i; j ++) {external.codingCi] [j] ≈internal_coding [i] [j]; >}

/ * re-initialization of the protocol with the new exchange format * / C reinitProtocolC); / * read commands * / read_n (l, GB, (char *) (ft (te)), l); you ++;

/ * returns the result * / send_n (sizeof (UI32), GB, (char *) (fctβ), 1);

Claims

claims

1. Data conversion device, intended to work on primary elementary data coded individually according to a first arrangement of words, characterized in that it comprises:

memorization means (7) for storing a first set of symbols, all distinct, forming a representation of said first arrangement and a second set of symbols all distinct, forming a representation of a second arrangement of words, and

* an operator (8) arranged to receive as input a primary elementary data item, as well as said first and second sets of symbols, and to effect on this primary elementary data word transformations defined only by said first and second sets of symbols so as to output a corresponding secondary data equivalent to said primary elementary data.

2. Device according to claim 1, wherein a first hardware (1) delivers said primary elementary data coded according to said first arrangement, and a conversion means (5 ') delivers secondary data coded according to a third arrangement, after data conversion primary elementary coded according to a fourth arrangement by a second material (1 '), said material (1,1') wishing to exchange primary elementary data, characterized in that said operator (8) comprises interrogation means (9) arranged for: * providing said second hardware (1 ') with a message containing said second set of symbols and requesting the sending back to the operator (8) of primary elementary data, transformed from said second set of symbols by coding according to said fourth arrangement , * deduce from this primary elementary data as well as from the first and second sets of symbols a third set of symbols forming a representation of said fourth arrangement, * replace said second set of symbols with said third set of symbols, in said operator (8) and in said conversion means (5 '), so that in the event of transmission of a primary elementary datum coded according to the first , respectively fourth, arrangement and intended for said second, respectively first, material, said operator (8) delivers to the latter, directly, a primary elementary data item coded according to the fourth, respectively first arrangement.

3. Device according to claim 2, characterized in that said first, second and third sets of symbols are ordered sequences of numbers.

4. Device according to claim 3, characterized in that said second set of symbols is the sequence [1, 2, 3, ..., n-1, n], n being the number of components of a base on which the secondary data item is broken down into words of k bits, k being greater than or equal to 1, and in particular equal to 8.

5. Device according to one of claims 2 to 4, characterized in that said second and third arrangements are identical.

6. Device according to one of claims 2 to 4, characterized in that said second and third arrangements are different and are associated with sets of symbols comprising different numbers of words and / or words of different number of bits, and in that said interrogation means are arranged to address to said conversion means (5 ') at least a second set of symbols so that it is substituted for said third arrangement.

7. Device according to claim 6, characterized in that said storage means (7) store several second sets of symbols comprising numbers n _A of different words and / or words of number k _± of different bits, and in that said interrogation means (9) are arranged to provide said second material (1 ′), by message, a chosen number of first sets of different symbols.

8. Device according to one of claims 2 to 7, wherein said first equipment (1) is located in a first machine (Ml), in particular a computer, characterized in that said storage means (7) and a part at least said operator (8,9-1) are located in said first machine.

9. Device according to claim 8, characterized in that said storage means (7) and a part (8, 9-1) at least of said operator are installed in the form of a program in said first machine (Ml).

10. Device according to one of claims 8 and 9, in which said second material (1 ') and said conversion means (5) are installed in a second machine (M2), in particular a computer, characterized in that a first part (9-1) of said interrogation means is located in said first machine, while a second complementary part (9-2) is located in said second machine (M2).

11. Device according to claim 10, characterized in that said interrogation means (9-2) are installed in the form of a program.

12. Device according to claim 11, characterized in that said operator (8) is arranged to implant said second part (9-2) of the interrogation means in said second machine (M2) when said first material attempts, for the first times, to exchange primary elementary data with said second material (1 ′).

13. Method for converting primary elementary data coded individually according to a first arrangement of words, characterized in that it comprises the following steps: a) providing a first set of symbols, all distinct, forming a representation of said first arrangement and a second set of symbols all distinct, forming a representation of a second arrangement of words, and b) receiving primary elementary data, as well as said first and second sets of symbols, and c) performing on this primary elementary data transformations of words defined only by said first and second sets of symbols so as to provide as output a secondary data corresponding and equivalent to said primary elementary data.

14. The method of claim 13, wherein there is received from a first material a primary elementary data item coded according to said first arrangement, and from a conversion means a secondary item data encoded according to a third arrangement, resulting from the conversion of primary elementary data coded according to a fourth arrangement by a second piece of equipment, characterized in that in step b): * said second piece of equipment is provided with a message containing said second set of symbols and requiring the sending of a piece of data in return elementary elementary, transformed from said first set of symbols by coding according to said fourth arrangement, then

* from this primary elementary data as well as from the first and second sets of symbols is deduced a third set of symbols forming a representation of said fourth arrangement, and

* said second set of symbols is replaced everywhere by said third set of symbols, so that in the event of transmission of a primary elementary data item coded according to the first, respectively fourth, arrangement and intended for said second, respectively first, material, a primary elementary data item coded according to the fourth, first arrangement respectively, is delivered to the latter.