CA1285075C - Dual byte order computer architecture - Google Patents
Dual byte order computer architectureInfo
- Publication number
- CA1285075C CA1285075C CA000529196A CA529196A CA1285075C CA 1285075 C CA1285075 C CA 1285075C CA 000529196 A CA000529196 A CA 000529196A CA 529196 A CA529196 A CA 529196A CA 1285075 C CA1285075 C CA 1285075C
- Authority
- CA
- Canada
- Prior art keywords
- data
- external
- unit
- bus
- ordering sequence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Lifetime
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F7/00—Methods or arrangements for processing data by operating upon the order or content of the data handled
- G06F7/76—Arrangements for rearranging, permuting or selecting data according to predetermined rules, independently of the content of the data
- G06F7/768—Data position reversal, e.g. bit reversal, byte swapping
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F12/00—Accessing, addressing or allocating within memory systems or architectures
- G06F12/02—Addressing or allocation; Relocation
- G06F12/04—Addressing variable-length words or parts of words
Abstract
ABSTRACT
A CPU or other function unit is disclosed which follows one data ordering scheme internally, and in which incoming and/or outgoing data pass through a data order conversion unit for adapting it to a selectable external data ordering scheme. The means for specifying the external data ordering scheme is accessible from outside the physical package(s) in which the functional unit is housed. The data order conversion unit may comprise a load aligner and/or a store aligner, one or both of which may comprise means for shifting informational unit of a smaller size within informational units of a larger size.
The shift amount may derive from the low order address bits and may be altered depending on the external data ordering means selected.
A CPU or other function unit is disclosed which follows one data ordering scheme internally, and in which incoming and/or outgoing data pass through a data order conversion unit for adapting it to a selectable external data ordering scheme. The means for specifying the external data ordering scheme is accessible from outside the physical package(s) in which the functional unit is housed. The data order conversion unit may comprise a load aligner and/or a store aligner, one or both of which may comprise means for shifting informational unit of a smaller size within informational units of a larger size.
The shift amount may derive from the low order address bits and may be altered depending on the external data ordering means selected.
Description
DUAL BYTE ORDE~ COMPUT~R ARCHITECTURE
BACKGROUND OE THE INVENTION
1. Field of the Invention S The invention relates to computer architectures, and more particularly to means for accommodating various ~ ordering schemes for the storage of information~
;
BACKGROUND OE THE INVENTION
1. Field of the Invention S The invention relates to computer architectures, and more particularly to means for accommodating various ~ ordering schemes for the storage of information~
;
2. Description of Related Art Data is organized in modern computers in either of two . 10 formats or in some combination of those formats. ~he formats are known as "big endian", in which the high order bit, byte or other unit of information is loeated in the lower numbered unit address, and "little endian,"
in which the high order unit of information is located in the higher numbered unit address. ~ee Cohent "On Holy Wars and a Plea for Peace," Com~_er, 10/81, pp.
48-54, which contains a good discussion of the data organization problem. Thus, in a true big endian computer architecture, bits of data are thought of as being lined up from left to right, the lowest numbered and most significant bit being on the left. When this string of bits is divided lnto, for example, 8-bit bytes, 16-bi~ halfwords and/or 32-bit word~, the lowest numbered and most significant bytef halfword or word ~, . . ~
~2~ 7~
continues to he located on the left. In a true little endian architecture, the scheme is exactly the opposite.
Bits, bytes, halfwords and words are numbered right to left, the least significant bit, byte, halfword or word being located on the right~ Within a given computer, or within a given functional unit ~such as a CPU) within a computer, it makes no difference which scheme is chosen because all the hardware and software can be designed around the chosen scheme. The difference becomes important when one computer or functional unit communicates with another, the d~ta ordering scheme of which is not known. This is because information which is transmitted from the lvwest numbered unit to the highest numbered unit will be most significant to least significant under the big endian scheme, but least significant to most significant under the little endian scheme.
In many modern computers, the smallest addressable unit of information (i.e., the smallest group of bits which one computer or functional unit may receive from, transmit to or refer to in another) is a byte. This eases the data ordering conflict because all the bits in the byte may be transmitted in parallel. Whereas the bits in the byte may be numbered differently for different data ordering schemes, the least significant bit is always on the right. The different data ordering schemes for such computers are therefore functionally identical for informational units smaller than a byte.
Howeverl it is still necessary to be concerned about byte, halfword and word ordering.
Also, it should be noted that some computers implement a mixed data ordering scheme. In these computers, words and larger units, for example, may have a big endian ordering, whereas halfwords, bytes and bits may have a li~tle endian ordering. This adds complexity because on multiword transmissions of data it becomes necessary to know both that ~he first word is the high order word, and also that the first byte of that word is the low order byte.
S The data ordering problem is particularly important to manufacturers vf chip- and board-level microcomputer based products which are specific to a chosen data ordering scheme, such as CPU's. The conflict is not as important for other products, such as memory, because such products are usable without modification with any ordering scheme.
Manufactures of chip- and board-le~el products typically intend their products to be incorporated into a computer with other chip- or board-level products of other manufacturers which may use a different data ordering scheme. Manufacturers therefore have, in the past, designed their products according to one data ordering scheme and marketed them only to customers with a compatible architecture. This effectively halves the manufacturer's available market. As an alternative, some manufacturers make and sell two versions of every product which needs a predefined data ordering scheme, one f or each type of architecture. This solution is also unsatisfactory, however, because it requires the manufacturer to design, manufacture and maintain inventory on twice as many products. It also prevents the manufacturer from taking full advanta~e of the economies of scale, because only half the production volume of each version can be expected.
It is also possible to design products according to one ordering scheme, and rely on software to realign the informational units if the external hardware uses a different scheme. The oftware could move around informational units of whatever size is necessary. ~his avoids the necessity of designing, manufacturing and stocking different versions of each product, but it merely moves the difficulty to the software.
Different versions of each compiler, operating system or other program must be written and supyorted, or a single version must be written which is capable of adapting itself for use with more than one ordering scheme. The latter solution can be extremely complicated to implement and can, under certain circumstances, result in intolerable overhead for software-managed data realignment.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide a functional unit such as a CPU
which is not subject to the above difficulties.
It is another object of the present invention to provide a CPU which avoids the necessity of designing, manufacturing, and/or maintaining inventory of two or more versions of the same product.
It is another object of the present invention to provide an effective division between hardware and software of the complexity required to adapt a single product to different data ordering schemes.
It is another object of the present invention to provide a CPU which is adaptable to one of two or more predefined external schemes for ordering informational units, such as bytes, which are smaller than the data bus width.
In accordance with the present invention there is provided a data processing unit, which comprises:
~.
. .
~ 2~7~
- ~a -an execution unit for processing data;
an internal data bus coupled to said execution unit for input of internal bus data to said execution unit, the internal bus data being organized into information units, each comprising a series of individual data groups, the individual data groups of each respective information unit beiny arranged within the respective information unit and relative to one another, in a preselected internal data ordering sequence for processing by said execution unit;
means for aligning data groups within a respective information unit coupled to said internal data bus;
an external data bus coupled to said means for aligning data groups, for input of external bus data to said means for aligning data groups, said external bus data being organized into the informa-tion units, each comprising a series of the indi-vidual data groups, the individual data groups of each information unit of the external bus data being arranged within the respective information unit and relative to one another, in an external data ordering ; sequence;
: 25 a shift amount converter unit coupled to said means for aligning data groups within a respective information unit;
means for storing information indicative of the preselected internal data ordering sequence and the external data ordering sequence and for generating a control signal indicative of a relationship between -the preselected internal data ordering sequence and the external data ordering sequence;
.
7~;
- 4b -said means for generating a control signal being coupled to said shift amount converter unit for input of said con-trol signal to said shift amount converter;
said shif-t amount converter being responsive to said control signal to control sald means for aligning to realign the data groups of each information unit of the external bus data, input to said means for aligning data groups, from the external data ordering sequence into the preselected internal data ordering sequence to provide the internal bus data for input to said execution unit;
and means to selectively input information indicative of the external data ordering sequence to said means for storing information.
According to the present invention, the functional unit ls designed -to follow any desired data ordering scheme c~ 7~i internally, and to have both incoming and outgoing data pass through a data order conversion unit for adapting it to a selectable external data ordering scheme. The data order conversion unit should be capable of realigning informational units of a smaller size within informational units of a larger size, where the smaller size is the size below which all data may be considered in parallel, usually a byte, and the larger size is equal to the width of the largest informational unit to which the software may make external reference, usually equal to the data bus width. Realignment of informational units smaller than the smaller size is unnecessary because there is no ambiguity if bits are considered in parallel. Realignment of data across the boundaries of informational units of the larger size, a type of work which is performed frequently by software anyway, may be left to the software.
The means for specifying the external ordering scheme should be accessible from outside the physical package in which the functional unit is housed, so that only one of such packages need be designed, manufactured and maintained in inventory. If the functional unit is implemented (perhaps with other units) as a single chip, one OE more pins may be dedicated to or usable for this purpose. If the functional unit is implemented as a module, the means for specifying the data ordering scheme may comprise a DIP switch located on the module, perhaps under a cover which is sealed by the manufacturer after being set.
In one embodiment of the invention the functional unit is a CPU chip or chip set in which data is transmitted to and from memory in the form of 32-bit words, but in which the software may address individual 8-bit bytes.
The data order conversion unit realigns bytes within incoming or outgoing word in accordance with the ;~$7~
specified external byte and halfword ordering scheme.
Realignment of multiword strings of data is left to the software.
The data order conversion unit may c~mprise a load aligner and a store aligner, one or both of which comprises means for shif~ing (as distinguished from randomly realigning) informational units of the smaller size within an informational unit of the larger size by an amount dependent on the low order address bits generated within the CPU, wherein the shift amount is altered depending on the data order selection means.
One or both of the means for shifting may also double as a general purpose shifter in the CPU.
RIEF DSCRIPTION OF THE DRAWlNGS
The invention will be described with respect to particular embodiments thereof. Reference will be made to the drawings, in which like members are given like designations, and in which:
FIGURES 1 and 2 are block diagrams of embodiments of the present invention.
DETAILED DESCRIPTIQN
In Fig. 1 there is shown a block diagram of a CPU 10 constructed in accordance with the invention. It co~prises an execution unit 12 which performs, among other things, instruction decode and data manipulation designated by software. The execution unit 12 has full 32-bit data paths and, wherever a choice must be made, is little endian in configuration. The execution unit 12 is connected to a 32-bit internal address bus 14 and a 32-bit internal data bus 16. The internal address bus 14 is connected to a 32-bit external address bus 18, which is connected in turn ~o main memory and various other external de~ices ~not shown). The internal data bus 16 is connected to a load aligner 20 and a store aligner 22, whieh are in turn connected to a 32-bit S external data bus 24. The load aligner realigns the position of the four bytes of incoming 32-bit data words before transmitting them to the execution unit 12 over the internal data bus 16. The load aligner takes ints account such software-selectable factors as the 10 reference type ~byte, halfword or full word) and whether the incoming data is ~o be zero- or sign~extended.
These characteristics are sen~ to the load aligner 20 by the execution unit 12 over alignment control leads 26.
The load aligner 20 is also eonnected to the low order two bits of the internal address bus 14 so that the particular byte or bytes desired by ~he software may be selected and shifted to their rightmost positions in the incoming word. The two low order address bits are connected to the load aligner via a shift amount converter 28 (the function of which will be described below) and two shift amount leads 30.
The store aligner 22 is also connected to the alignment control leads 26 and the shift amount leads 30. It functions similarly to the load aligner 20, except that its operations are reversed. Since it realigns bytes on the internal data bus 16 before sending them out over the external data bus 24, its operations mirror those of the load aligner 20. Thus, for example, on a halfword load from the left half of a 32-bi~ memory location, the load aligner 20 would shift the incoming data to the right by 16 bits; whereas on a halfw~rd store to the left half of the same memory location, the same shift would have to be made to the left. Additionally, the store aligner 22 generates write control signals which are transmitted to the main memory over write control leads 32 to designate which if any of the bytes on the external data bus 24 are to be written.
The purpose of the shift amount converter 2B is to adapt the CPU to whatever byte ordering scheme is used in the external hardware. If the external scheme i5 little endian, then no adaptation is necessary and the tw~ low order address bits are transmitted to the shift amount leads 30 unaltered. Thus, on a load, the high order 30 bits of the external address bus 18 are used to select the word from main memory in which the desired byte or bytes are located. That entire word is transmitted to the load aligner 20, which shifts the information to the r1ght depending on the reference type and the shift amount leads 30. If the reference type is "byte", then the number of bytes by which the information is shifted is given directly by the two shift amount leads 30 (i.e., the low order two address bits). If the reference type is "halfword", then the number of bytes by which the information is shifted is given by twice the high order shift amoun~ lead (i.e., 0 if the 21 address bit is 0, 2 if the 21 address bit is 1). If the reference type is "word", then no shifting takes place.
Similar shifting takes place on a store through the store aligner 22, except that all shifting is to the left.
On the other hand, if the external hardware is big endian, then byte alignment is different. Bytes and halfwords are numbered from left to right instead of right to left, and the high order byte or halfword in a word is to the right of the remaining bytes or halfwords. One way to accommodate ~his difference would be to specify the byte order of the external hardware as an input to combinational logic in the load aligner 20 for selecting one byte of the external word to be put in each byte position of the internal word. However, since an entire 32-bit word is transmitted along the external ;
; .
data bus 24 regardless of the reference type, it is preferred that the difference be accommodated in the shift amount converter 2B merely by altering the two low order address bits before they enter the load and store aligners ~0 and 22. Thus, if the reference type is "byte", both low order address bits are inverted before being passed to the shift amount leads 30. If the reference type is "halfword", then only the 21 address bit is inverted. If the reference type is "word", then neither low order address bit is inverted. This logic effectively realigns bytes as necessary to adapt an internal little endian architecture to an external big-endian architecture.
In order for the shift amount converter 23 to perform its function, it is connected to the alignmen~ control leads 26 and, via a data srder select inpu~ 33, to a data order storage unit 34. The data order storage unit 34 may comprise a D flip flop the D input of which is hardwired externally to a selected voltage level to indicate the data ordering scheme of the external architecture, and the clock input of which is connected to the power-up and system reset lines ~not shown~.
Alternatively, ~he data order storage unit 34 may be omitted and the data order select input to the shift amount converter 28 connected directly to an external package pin which may be hardwired to the selected voltage level. It should be noted that no significant advantage would be gained by having ~he data ordering scheme selectable by the software, because the ordering scheme of the external hardware is by definition fixed.
It will not change from power-up to power-down. Having the ordering scheme software selectable could also add significant complexity, since the software must accommodate such flexibility. There may, however, be circumstances in which software selectability is desired, and the benefits of providing it would have to 5~375 be weighed against the detriments of increased software complexity.
In the above description, nothing has been said about the physical implementation of the embodiment. It should be apparent to a person of ordinary skill that for the most part, the physical implementation is not relevant to the invention. All the circuitry shown as constituting part of the CPU 10 in Fig. 1 can be implemented on one chip, or it can be divided into several. Additional circuitry not shown in Fig. 1 may also be integrated onto one or more of such chips. In order to obtain the benefits of the invention as described above, however, the means to select the external data ordering scheme should be accessible from outside the bounds of the physical unit of which it is desired to avoid manufacturing more than one version.
Thus, if a CPU such as that designated 10 in Fig. 1 is to be manufactured as a single chip, perhaps with a memory management unit and cache control logic integrated thereon, then the data order storage unit 34 should be accessible from ou~side the chip, for example via a dedicated pin. If the CPU is to be implemented on a printed oircuit board module, and the module is the physical unit which is to be usable in both big and little endian machines, then the data order storage unit 34 should be accessible from outside the module, for example by manual switching of a DIP switch either on or off the printed circuit board.
The advantages of an architecture such as that o the above embodiment are apparent. First, only the one ! version of the CPU need be designed, manufactured and stocked, since it is usable in both big and little endian computers. Contrary to previous techniques in which the da~a ordering scheme of the device had to be specified in the design process, the ordering scheme of ~ t7~
this CPU need not be specified until it is installed in a larger package whose data ordering scheme is fixed.
Second, no s~ftware complexity is dedicated to the realignment of bytes within words for conversion to a different external ordering scheme, and no time is wasted performing such conversions. All such conversions are performed in combinational logic which may be integrated on one chip with the execution unit and thereby made extremely fast. Multiword data storage and retrieval is not accommodated in this embodiment, and to the extent multiword accesses are dependent upon the data ordering scheme, the software must be able to re-order the words as necessary. The data order select input 33 may be made readable by the software for that purpose. ~o~ever, the inability to automatically realign data on multiword accesses is not a significant drawback. Since the software usually moves around ful~
words anyway, it does not add significant overhead to incorporate a selectahle word-ordering scheme into program designs.
It should be noted that the architecture of a functional unit implementing the present invention does not necessarily prevent specialized software from directly accessing selected bytes in memory without regard to byte ordering ~chemes. This can be ~ccomplished in the above embodiment by providing an alignment control lead which "turns of" the shift amoun~ converter 28, causing it to pass the two low order address bits directly to the shift amount leads 30, regardless of the signal on the data order select input 33. Such an alignment control lead may be activated on certain addressing modes, and may be accompanied by additional signals to the load and store aligners 20 and 22.
In Fig. 2 there is shown another embodiment of the present invention, in which a general purpose shi~ter s doubles as the store aligner. The embodiment comprises a CPU 50, containin~ a load aligner 20 similar to that shown in Fig. 1. The load aligner 20 in the CPU 50 is connected on its input to the 32-bit external data bus 24, and on its output to a register file 52 and to an instruction decoder and control unit 54. The output of the load aligner 20 is also connected t~ an arithmetic logic unit (ALU3 56, as is the output of the register file 52. The ALU 56 is in turn connected to a general purpose shifter 58, which is capable of shifting data right or left by any number of bits from zero to 31. It is sufficient for the purposes of the invention, however, if the general purpose shifter 58 is capable of shifting data only by multiples of 8 bits. The output of the shifter 58 may be fed back to the register file 52 or enabled onto the external data bus 24. The CPU 50 also includes an address unit 60 which gener~tes addresses and puts them out onto the 32-bit external address bus 18. As with the embodiment of Pig. 1, the two low order address bits in the embodiment of Fig. Z
are also transmitted to the shift amount converter 28, and the shift amount leads 30 on the output of the shift amount converter 28 are connected to the load aligner 20. The shift amount converter 2B and the load aligner 20 in Pig. 2 are identical to ~heir counterparts in P'g.
1. Instead of also being connected to a separate store aligner as in Fig. 1, however, the shift amount leads 30 in Fig. 2 are connected to the ~eneral purpose shifter 5B through a multiplexer 62, ~he multiplexer 62 selects either the in~vrmation ~n the shift amount leads 30 or shift amount information from another souree (not shown) for presentation to the shifter 58.
The operation of the CPU 50 on a load is identical to the operation of the CPU 10 in Fig. 1 on a load and will not be repeated here. On an internal register to register operation, data from the register file 52 5~
and/or the load aligner 20 passes through the ALU 56 and is transmitted to the shifter 58. The multiplexer 62 selects a shift amount and direction from sources in the CPU 50 not shown, and transmits that as control information to the shifter 58. The shif~er 58 performs the desired operation and the result is fed back into the register file 52. The operation of the CPU 50 on a store is the same as its operation register to register, except that the output of the shifter 58 is enabled onto the external data bus 24, and the multiplexer 62 selects shift amount and direction information from the shift amount leads 30 instead of from the sources not shown.
This ensures that data bytes are aligned appropriately for the byte order of ~he external hardware before the da~a i5 placed out on the external data bus 24. The embodiment of Fig. 2 therefore provides all the advantayes of that of ~ig. 1, and additionally saves space by combining the function of a s~ore aligner shifter with a general purpose shifter already present in the CPU.
~he invention has been described with respect to particular embodiments thereof, and it will be understood that numerous variations on these embodiments are possible within the scope of the invention. ~or example, a CPU may be designed according to the invention which is capable of realigning data but not instructions. In the embodiment of Fig. 2, this may be accomplished by connecting the input of the instruction decoder and control uni~ 54 to the external da~a bus 24, instead of to the output of load aligner 20. As another example, it was mentioned previously that the invention is not limited to CPUs. Thus, the invention may be applied to other functional units which are specific to a chosen data ordering scheme. It may be applied to functional units which do not have an address generating means, but instead merely pass along addresses generated 7~i elsewhere. It may also be applied to functional units which do not provide addresses at all, such as certain coprocessors. These and other variations on the described embodiments are all within the scope of the invention.
in which the high order unit of information is located in the higher numbered unit address. ~ee Cohent "On Holy Wars and a Plea for Peace," Com~_er, 10/81, pp.
48-54, which contains a good discussion of the data organization problem. Thus, in a true big endian computer architecture, bits of data are thought of as being lined up from left to right, the lowest numbered and most significant bit being on the left. When this string of bits is divided lnto, for example, 8-bit bytes, 16-bi~ halfwords and/or 32-bit word~, the lowest numbered and most significant bytef halfword or word ~, . . ~
~2~ 7~
continues to he located on the left. In a true little endian architecture, the scheme is exactly the opposite.
Bits, bytes, halfwords and words are numbered right to left, the least significant bit, byte, halfword or word being located on the right~ Within a given computer, or within a given functional unit ~such as a CPU) within a computer, it makes no difference which scheme is chosen because all the hardware and software can be designed around the chosen scheme. The difference becomes important when one computer or functional unit communicates with another, the d~ta ordering scheme of which is not known. This is because information which is transmitted from the lvwest numbered unit to the highest numbered unit will be most significant to least significant under the big endian scheme, but least significant to most significant under the little endian scheme.
In many modern computers, the smallest addressable unit of information (i.e., the smallest group of bits which one computer or functional unit may receive from, transmit to or refer to in another) is a byte. This eases the data ordering conflict because all the bits in the byte may be transmitted in parallel. Whereas the bits in the byte may be numbered differently for different data ordering schemes, the least significant bit is always on the right. The different data ordering schemes for such computers are therefore functionally identical for informational units smaller than a byte.
Howeverl it is still necessary to be concerned about byte, halfword and word ordering.
Also, it should be noted that some computers implement a mixed data ordering scheme. In these computers, words and larger units, for example, may have a big endian ordering, whereas halfwords, bytes and bits may have a li~tle endian ordering. This adds complexity because on multiword transmissions of data it becomes necessary to know both that ~he first word is the high order word, and also that the first byte of that word is the low order byte.
S The data ordering problem is particularly important to manufacturers vf chip- and board-level microcomputer based products which are specific to a chosen data ordering scheme, such as CPU's. The conflict is not as important for other products, such as memory, because such products are usable without modification with any ordering scheme.
Manufactures of chip- and board-le~el products typically intend their products to be incorporated into a computer with other chip- or board-level products of other manufacturers which may use a different data ordering scheme. Manufacturers therefore have, in the past, designed their products according to one data ordering scheme and marketed them only to customers with a compatible architecture. This effectively halves the manufacturer's available market. As an alternative, some manufacturers make and sell two versions of every product which needs a predefined data ordering scheme, one f or each type of architecture. This solution is also unsatisfactory, however, because it requires the manufacturer to design, manufacture and maintain inventory on twice as many products. It also prevents the manufacturer from taking full advanta~e of the economies of scale, because only half the production volume of each version can be expected.
It is also possible to design products according to one ordering scheme, and rely on software to realign the informational units if the external hardware uses a different scheme. The oftware could move around informational units of whatever size is necessary. ~his avoids the necessity of designing, manufacturing and stocking different versions of each product, but it merely moves the difficulty to the software.
Different versions of each compiler, operating system or other program must be written and supyorted, or a single version must be written which is capable of adapting itself for use with more than one ordering scheme. The latter solution can be extremely complicated to implement and can, under certain circumstances, result in intolerable overhead for software-managed data realignment.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide a functional unit such as a CPU
which is not subject to the above difficulties.
It is another object of the present invention to provide a CPU which avoids the necessity of designing, manufacturing, and/or maintaining inventory of two or more versions of the same product.
It is another object of the present invention to provide an effective division between hardware and software of the complexity required to adapt a single product to different data ordering schemes.
It is another object of the present invention to provide a CPU which is adaptable to one of two or more predefined external schemes for ordering informational units, such as bytes, which are smaller than the data bus width.
In accordance with the present invention there is provided a data processing unit, which comprises:
~.
. .
~ 2~7~
- ~a -an execution unit for processing data;
an internal data bus coupled to said execution unit for input of internal bus data to said execution unit, the internal bus data being organized into information units, each comprising a series of individual data groups, the individual data groups of each respective information unit beiny arranged within the respective information unit and relative to one another, in a preselected internal data ordering sequence for processing by said execution unit;
means for aligning data groups within a respective information unit coupled to said internal data bus;
an external data bus coupled to said means for aligning data groups, for input of external bus data to said means for aligning data groups, said external bus data being organized into the informa-tion units, each comprising a series of the indi-vidual data groups, the individual data groups of each information unit of the external bus data being arranged within the respective information unit and relative to one another, in an external data ordering ; sequence;
: 25 a shift amount converter unit coupled to said means for aligning data groups within a respective information unit;
means for storing information indicative of the preselected internal data ordering sequence and the external data ordering sequence and for generating a control signal indicative of a relationship between -the preselected internal data ordering sequence and the external data ordering sequence;
.
7~;
- 4b -said means for generating a control signal being coupled to said shift amount converter unit for input of said con-trol signal to said shift amount converter;
said shif-t amount converter being responsive to said control signal to control sald means for aligning to realign the data groups of each information unit of the external bus data, input to said means for aligning data groups, from the external data ordering sequence into the preselected internal data ordering sequence to provide the internal bus data for input to said execution unit;
and means to selectively input information indicative of the external data ordering sequence to said means for storing information.
According to the present invention, the functional unit ls designed -to follow any desired data ordering scheme c~ 7~i internally, and to have both incoming and outgoing data pass through a data order conversion unit for adapting it to a selectable external data ordering scheme. The data order conversion unit should be capable of realigning informational units of a smaller size within informational units of a larger size, where the smaller size is the size below which all data may be considered in parallel, usually a byte, and the larger size is equal to the width of the largest informational unit to which the software may make external reference, usually equal to the data bus width. Realignment of informational units smaller than the smaller size is unnecessary because there is no ambiguity if bits are considered in parallel. Realignment of data across the boundaries of informational units of the larger size, a type of work which is performed frequently by software anyway, may be left to the software.
The means for specifying the external ordering scheme should be accessible from outside the physical package in which the functional unit is housed, so that only one of such packages need be designed, manufactured and maintained in inventory. If the functional unit is implemented (perhaps with other units) as a single chip, one OE more pins may be dedicated to or usable for this purpose. If the functional unit is implemented as a module, the means for specifying the data ordering scheme may comprise a DIP switch located on the module, perhaps under a cover which is sealed by the manufacturer after being set.
In one embodiment of the invention the functional unit is a CPU chip or chip set in which data is transmitted to and from memory in the form of 32-bit words, but in which the software may address individual 8-bit bytes.
The data order conversion unit realigns bytes within incoming or outgoing word in accordance with the ;~$7~
specified external byte and halfword ordering scheme.
Realignment of multiword strings of data is left to the software.
The data order conversion unit may c~mprise a load aligner and a store aligner, one or both of which comprises means for shif~ing (as distinguished from randomly realigning) informational units of the smaller size within an informational unit of the larger size by an amount dependent on the low order address bits generated within the CPU, wherein the shift amount is altered depending on the data order selection means.
One or both of the means for shifting may also double as a general purpose shifter in the CPU.
RIEF DSCRIPTION OF THE DRAWlNGS
The invention will be described with respect to particular embodiments thereof. Reference will be made to the drawings, in which like members are given like designations, and in which:
FIGURES 1 and 2 are block diagrams of embodiments of the present invention.
DETAILED DESCRIPTIQN
In Fig. 1 there is shown a block diagram of a CPU 10 constructed in accordance with the invention. It co~prises an execution unit 12 which performs, among other things, instruction decode and data manipulation designated by software. The execution unit 12 has full 32-bit data paths and, wherever a choice must be made, is little endian in configuration. The execution unit 12 is connected to a 32-bit internal address bus 14 and a 32-bit internal data bus 16. The internal address bus 14 is connected to a 32-bit external address bus 18, which is connected in turn ~o main memory and various other external de~ices ~not shown). The internal data bus 16 is connected to a load aligner 20 and a store aligner 22, whieh are in turn connected to a 32-bit S external data bus 24. The load aligner realigns the position of the four bytes of incoming 32-bit data words before transmitting them to the execution unit 12 over the internal data bus 16. The load aligner takes ints account such software-selectable factors as the 10 reference type ~byte, halfword or full word) and whether the incoming data is ~o be zero- or sign~extended.
These characteristics are sen~ to the load aligner 20 by the execution unit 12 over alignment control leads 26.
The load aligner 20 is also eonnected to the low order two bits of the internal address bus 14 so that the particular byte or bytes desired by ~he software may be selected and shifted to their rightmost positions in the incoming word. The two low order address bits are connected to the load aligner via a shift amount converter 28 (the function of which will be described below) and two shift amount leads 30.
The store aligner 22 is also connected to the alignment control leads 26 and the shift amount leads 30. It functions similarly to the load aligner 20, except that its operations are reversed. Since it realigns bytes on the internal data bus 16 before sending them out over the external data bus 24, its operations mirror those of the load aligner 20. Thus, for example, on a halfword load from the left half of a 32-bi~ memory location, the load aligner 20 would shift the incoming data to the right by 16 bits; whereas on a halfw~rd store to the left half of the same memory location, the same shift would have to be made to the left. Additionally, the store aligner 22 generates write control signals which are transmitted to the main memory over write control leads 32 to designate which if any of the bytes on the external data bus 24 are to be written.
The purpose of the shift amount converter 2B is to adapt the CPU to whatever byte ordering scheme is used in the external hardware. If the external scheme i5 little endian, then no adaptation is necessary and the tw~ low order address bits are transmitted to the shift amount leads 30 unaltered. Thus, on a load, the high order 30 bits of the external address bus 18 are used to select the word from main memory in which the desired byte or bytes are located. That entire word is transmitted to the load aligner 20, which shifts the information to the r1ght depending on the reference type and the shift amount leads 30. If the reference type is "byte", then the number of bytes by which the information is shifted is given directly by the two shift amount leads 30 (i.e., the low order two address bits). If the reference type is "halfword", then the number of bytes by which the information is shifted is given by twice the high order shift amoun~ lead (i.e., 0 if the 21 address bit is 0, 2 if the 21 address bit is 1). If the reference type is "word", then no shifting takes place.
Similar shifting takes place on a store through the store aligner 22, except that all shifting is to the left.
On the other hand, if the external hardware is big endian, then byte alignment is different. Bytes and halfwords are numbered from left to right instead of right to left, and the high order byte or halfword in a word is to the right of the remaining bytes or halfwords. One way to accommodate ~his difference would be to specify the byte order of the external hardware as an input to combinational logic in the load aligner 20 for selecting one byte of the external word to be put in each byte position of the internal word. However, since an entire 32-bit word is transmitted along the external ;
; .
data bus 24 regardless of the reference type, it is preferred that the difference be accommodated in the shift amount converter 2B merely by altering the two low order address bits before they enter the load and store aligners ~0 and 22. Thus, if the reference type is "byte", both low order address bits are inverted before being passed to the shift amount leads 30. If the reference type is "halfword", then only the 21 address bit is inverted. If the reference type is "word", then neither low order address bit is inverted. This logic effectively realigns bytes as necessary to adapt an internal little endian architecture to an external big-endian architecture.
In order for the shift amount converter 23 to perform its function, it is connected to the alignmen~ control leads 26 and, via a data srder select inpu~ 33, to a data order storage unit 34. The data order storage unit 34 may comprise a D flip flop the D input of which is hardwired externally to a selected voltage level to indicate the data ordering scheme of the external architecture, and the clock input of which is connected to the power-up and system reset lines ~not shown~.
Alternatively, ~he data order storage unit 34 may be omitted and the data order select input to the shift amount converter 28 connected directly to an external package pin which may be hardwired to the selected voltage level. It should be noted that no significant advantage would be gained by having ~he data ordering scheme selectable by the software, because the ordering scheme of the external hardware is by definition fixed.
It will not change from power-up to power-down. Having the ordering scheme software selectable could also add significant complexity, since the software must accommodate such flexibility. There may, however, be circumstances in which software selectability is desired, and the benefits of providing it would have to 5~375 be weighed against the detriments of increased software complexity.
In the above description, nothing has been said about the physical implementation of the embodiment. It should be apparent to a person of ordinary skill that for the most part, the physical implementation is not relevant to the invention. All the circuitry shown as constituting part of the CPU 10 in Fig. 1 can be implemented on one chip, or it can be divided into several. Additional circuitry not shown in Fig. 1 may also be integrated onto one or more of such chips. In order to obtain the benefits of the invention as described above, however, the means to select the external data ordering scheme should be accessible from outside the bounds of the physical unit of which it is desired to avoid manufacturing more than one version.
Thus, if a CPU such as that designated 10 in Fig. 1 is to be manufactured as a single chip, perhaps with a memory management unit and cache control logic integrated thereon, then the data order storage unit 34 should be accessible from ou~side the chip, for example via a dedicated pin. If the CPU is to be implemented on a printed oircuit board module, and the module is the physical unit which is to be usable in both big and little endian machines, then the data order storage unit 34 should be accessible from outside the module, for example by manual switching of a DIP switch either on or off the printed circuit board.
The advantages of an architecture such as that o the above embodiment are apparent. First, only the one ! version of the CPU need be designed, manufactured and stocked, since it is usable in both big and little endian computers. Contrary to previous techniques in which the da~a ordering scheme of the device had to be specified in the design process, the ordering scheme of ~ t7~
this CPU need not be specified until it is installed in a larger package whose data ordering scheme is fixed.
Second, no s~ftware complexity is dedicated to the realignment of bytes within words for conversion to a different external ordering scheme, and no time is wasted performing such conversions. All such conversions are performed in combinational logic which may be integrated on one chip with the execution unit and thereby made extremely fast. Multiword data storage and retrieval is not accommodated in this embodiment, and to the extent multiword accesses are dependent upon the data ordering scheme, the software must be able to re-order the words as necessary. The data order select input 33 may be made readable by the software for that purpose. ~o~ever, the inability to automatically realign data on multiword accesses is not a significant drawback. Since the software usually moves around ful~
words anyway, it does not add significant overhead to incorporate a selectahle word-ordering scheme into program designs.
It should be noted that the architecture of a functional unit implementing the present invention does not necessarily prevent specialized software from directly accessing selected bytes in memory without regard to byte ordering ~chemes. This can be ~ccomplished in the above embodiment by providing an alignment control lead which "turns of" the shift amoun~ converter 28, causing it to pass the two low order address bits directly to the shift amount leads 30, regardless of the signal on the data order select input 33. Such an alignment control lead may be activated on certain addressing modes, and may be accompanied by additional signals to the load and store aligners 20 and 22.
In Fig. 2 there is shown another embodiment of the present invention, in which a general purpose shi~ter s doubles as the store aligner. The embodiment comprises a CPU 50, containin~ a load aligner 20 similar to that shown in Fig. 1. The load aligner 20 in the CPU 50 is connected on its input to the 32-bit external data bus 24, and on its output to a register file 52 and to an instruction decoder and control unit 54. The output of the load aligner 20 is also connected t~ an arithmetic logic unit (ALU3 56, as is the output of the register file 52. The ALU 56 is in turn connected to a general purpose shifter 58, which is capable of shifting data right or left by any number of bits from zero to 31. It is sufficient for the purposes of the invention, however, if the general purpose shifter 58 is capable of shifting data only by multiples of 8 bits. The output of the shifter 58 may be fed back to the register file 52 or enabled onto the external data bus 24. The CPU 50 also includes an address unit 60 which gener~tes addresses and puts them out onto the 32-bit external address bus 18. As with the embodiment of Pig. 1, the two low order address bits in the embodiment of Fig. Z
are also transmitted to the shift amount converter 28, and the shift amount leads 30 on the output of the shift amount converter 28 are connected to the load aligner 20. The shift amount converter 2B and the load aligner 20 in Pig. 2 are identical to ~heir counterparts in P'g.
1. Instead of also being connected to a separate store aligner as in Fig. 1, however, the shift amount leads 30 in Fig. 2 are connected to the ~eneral purpose shifter 5B through a multiplexer 62, ~he multiplexer 62 selects either the in~vrmation ~n the shift amount leads 30 or shift amount information from another souree (not shown) for presentation to the shifter 58.
The operation of the CPU 50 on a load is identical to the operation of the CPU 10 in Fig. 1 on a load and will not be repeated here. On an internal register to register operation, data from the register file 52 5~
and/or the load aligner 20 passes through the ALU 56 and is transmitted to the shifter 58. The multiplexer 62 selects a shift amount and direction from sources in the CPU 50 not shown, and transmits that as control information to the shifter 58. The shif~er 58 performs the desired operation and the result is fed back into the register file 52. The operation of the CPU 50 on a store is the same as its operation register to register, except that the output of the shifter 58 is enabled onto the external data bus 24, and the multiplexer 62 selects shift amount and direction information from the shift amount leads 30 instead of from the sources not shown.
This ensures that data bytes are aligned appropriately for the byte order of ~he external hardware before the da~a i5 placed out on the external data bus 24. The embodiment of Fig. 2 therefore provides all the advantayes of that of ~ig. 1, and additionally saves space by combining the function of a s~ore aligner shifter with a general purpose shifter already present in the CPU.
~he invention has been described with respect to particular embodiments thereof, and it will be understood that numerous variations on these embodiments are possible within the scope of the invention. ~or example, a CPU may be designed according to the invention which is capable of realigning data but not instructions. In the embodiment of Fig. 2, this may be accomplished by connecting the input of the instruction decoder and control uni~ 54 to the external da~a bus 24, instead of to the output of load aligner 20. As another example, it was mentioned previously that the invention is not limited to CPUs. Thus, the invention may be applied to other functional units which are specific to a chosen data ordering scheme. It may be applied to functional units which do not have an address generating means, but instead merely pass along addresses generated 7~i elsewhere. It may also be applied to functional units which do not provide addresses at all, such as certain coprocessors. These and other variations on the described embodiments are all within the scope of the invention.
Claims (4)
1. A data processing unit, which comprises:
an execution unit for processing data;
an internal data bus coupled to said execution unit for input of internal bus data to said execution unit, the internal bus data being organized into information units, each comprising a series of individual data groups, the individual data groups of each respective information unit being arranged within the respective information unit and relative to one another, in a preselected internal data ordering sequence for processing by said execution unit;
means for aligning data groups within a respective information unit coupled to said internal data bus;
an external data bus coupled to said means for aligning data groups, for input of external bus data to said means for aligning data groups, said external bus data being organized into the informa-tion units, each comprising a series of the indi-vidual data groups, the individual data groups of each information unit of the external bus data being arranged within the respective information unit and relative to one another, in an external data ordering sequence;
a shift amount converter unit coupled to said means for aligning data groups within a respective information unit;
means for storing information indicative of the preselected internal data ordering sequence and the external data ordering sequence and for generating a control signal indicative of a relationship between the preselected internal data ordering sequence and the external data ordering sequence;
said means for generating a control signal being coupled to said shift amount converter unit for input of said control signal to said shift amount converter;
said shift amount converter being responsive to said control signal to control said means for aligning to realign the data groups of each information unit of the external bus data, input to said means for aligning data groups, from the external data ordering sequence into the preselected internal data ordering sequence to provide the internal bus data for input to said execution unit;
and means to selectively input information indicative of the external data ordering sequence to said means for storing information.
an execution unit for processing data;
an internal data bus coupled to said execution unit for input of internal bus data to said execution unit, the internal bus data being organized into information units, each comprising a series of individual data groups, the individual data groups of each respective information unit being arranged within the respective information unit and relative to one another, in a preselected internal data ordering sequence for processing by said execution unit;
means for aligning data groups within a respective information unit coupled to said internal data bus;
an external data bus coupled to said means for aligning data groups, for input of external bus data to said means for aligning data groups, said external bus data being organized into the informa-tion units, each comprising a series of the indi-vidual data groups, the individual data groups of each information unit of the external bus data being arranged within the respective information unit and relative to one another, in an external data ordering sequence;
a shift amount converter unit coupled to said means for aligning data groups within a respective information unit;
means for storing information indicative of the preselected internal data ordering sequence and the external data ordering sequence and for generating a control signal indicative of a relationship between the preselected internal data ordering sequence and the external data ordering sequence;
said means for generating a control signal being coupled to said shift amount converter unit for input of said control signal to said shift amount converter;
said shift amount converter being responsive to said control signal to control said means for aligning to realign the data groups of each information unit of the external bus data, input to said means for aligning data groups, from the external data ordering sequence into the preselected internal data ordering sequence to provide the internal bus data for input to said execution unit;
and means to selectively input information indicative of the external data ordering sequence to said means for storing information.
2. The data processing unit of Claim 1 wherein said means to selectively input information indicative of the external data ordering sequence includes a data order select input line for selective input of the information indicative of the external data ordering sequence from a source external to said data processing unit.
3. The data processing unit of Claim 2 further comprising a D flip flop coupled to said data order select line for storing the information indicative of the external data ordering sequence input on said data order select line, said D flip flop being coupled to said shift amount converter.
- 16 a -
- 16 a -
4. The data processing unit of Claim 1 further comprising:
an address bus coupled to said execution unit and said shift amount converter;
said execution unit operating to generate an address for each information unit of the external data on said external data bus, for output to said address bus;
each address including a first segment uniquely identifying a respective information unit of the external bus data and a second segment uniquely identifying a data group within the respective information unit;
said shift amount converter receiving the second segment of each address on said address bus and operating to invert a preselected portion of the second segment of each address when the control signal input by said means for generating a control signal is indicative of the preselected internal data ordering sequence corresponding to a first data ordering sequence and the external data ordering sequence corresponding to a second data ordering sequence, the inverted second address segment being used to control said means for aligning data groups within a respective information unit.
an address bus coupled to said execution unit and said shift amount converter;
said execution unit operating to generate an address for each information unit of the external data on said external data bus, for output to said address bus;
each address including a first segment uniquely identifying a respective information unit of the external bus data and a second segment uniquely identifying a data group within the respective information unit;
said shift amount converter receiving the second segment of each address on said address bus and operating to invert a preselected portion of the second segment of each address when the control signal input by said means for generating a control signal is indicative of the preselected internal data ordering sequence corresponding to a first data ordering sequence and the external data ordering sequence corresponding to a second data ordering sequence, the inverted second address segment being used to control said means for aligning data groups within a respective information unit.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US82728286A | 1986-02-06 | 1986-02-06 | |
| US827,282 | 1986-02-07 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA1285075C true CA1285075C (en) | 1991-06-18 |
Family
ID=25248802
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA000529196A Expired - Lifetime CA1285075C (en) | 1986-02-06 | 1987-02-06 | Dual byte order computer architecture |
Country Status (13)
| Country | Link |
|---|---|
| US (1) | US4959779A (en) |
| EP (1) | EP0236615B1 (en) |
| JP (1) | JPH0734174B2 (en) |
| KR (1) | KR870008252A (en) |
| AT (1) | ATE82412T1 (en) |
| CA (1) | CA1285075C (en) |
| DE (1) | DE3687124T2 (en) |
| DK (1) | DK60987A (en) |
| ES (1) | ES2007031A6 (en) |
| GR (1) | GR870205B (en) |
| IE (1) | IE870310L (en) |
| IL (1) | IL81239A (en) |
| NO (1) | NO870416L (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8683182B2 (en) | 1995-08-16 | 2014-03-25 | Microunity Systems Engineering, Inc. | System and apparatus for group floating-point inflate and deflate operations |
Families Citing this family (52)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS6428752A (en) * | 1987-07-24 | 1989-01-31 | Toshiba Corp | Data processor |
| JP3041308B2 (en) * | 1987-09-30 | 2000-05-15 | 健 坂村 | Data processing device |
| JP2633331B2 (en) * | 1988-10-24 | 1997-07-23 | 三菱電機株式会社 | Microprocessor |
| US5265237A (en) * | 1988-12-07 | 1993-11-23 | Xycom, Inc. | Byte swapping apparatus for selectively reordering bytes of an N-bit word communicated between an at computer and VME bus |
| GB2228348A (en) * | 1989-01-13 | 1990-08-22 | Texas Instruments Ltd | Memory interface integrated circuit |
| GB2229832B (en) * | 1989-03-30 | 1993-04-07 | Intel Corp | Byte swap instruction for memory format conversion within a microprocessor |
| US6038584A (en) * | 1989-11-17 | 2000-03-14 | Texas Instruments Incorporated | Synchronized MIMD multi-processing system and method of operation |
| US5168561A (en) * | 1990-02-16 | 1992-12-01 | Ncr Corporation | Pipe-line method and apparatus for byte alignment of data words during direct memory access transfers |
| DE69124437T2 (en) * | 1990-08-09 | 1997-07-03 | Silicon Graphics Inc | Method and device for reversing byte order in a computer |
| JP2763207B2 (en) * | 1991-04-25 | 1998-06-11 | 株式会社東芝 | Information processing device |
| US5524250A (en) * | 1991-08-23 | 1996-06-04 | Silicon Graphics, Inc. | Central processing unit for processing a plurality of threads using dedicated general purpose registers and masque register for providing access to the registers |
| US5313231A (en) * | 1992-03-24 | 1994-05-17 | Texas Instruments Incorporated | Color palette device having big/little endian interfacing, systems and methods |
| US5408664A (en) * | 1992-06-19 | 1995-04-18 | Silicon Graphics, Incorporated | System and Method for booting computer for operation in either of two byte-order modes |
| US5471628A (en) * | 1992-06-30 | 1995-11-28 | International Business Machines Corporation | Multi-function permutation switch for rotating and manipulating an order of bits of an input data byte in either cyclic or non-cyclic mode |
| US5412782A (en) | 1992-07-02 | 1995-05-02 | 3Com Corporation | Programmed I/O ethernet adapter with early interrupts for accelerating data transfer |
| JP3187539B2 (en) * | 1992-07-28 | 2001-07-11 | 株式会社東芝 | Data transfer device |
| US5517627A (en) * | 1992-09-18 | 1996-05-14 | 3Com Corporation | Read and write data aligner and method |
| US5392406A (en) * | 1992-09-18 | 1995-02-21 | 3Com Corporation | DMA data path aligner and network adaptor utilizing same |
| US5572235A (en) * | 1992-11-02 | 1996-11-05 | The 3Do Company | Method and apparatus for processing image data |
| US5838389A (en) * | 1992-11-02 | 1998-11-17 | The 3Do Company | Apparatus and method for updating a CLUT during horizontal blanking |
| US5596693A (en) * | 1992-11-02 | 1997-01-21 | The 3Do Company | Method for controlling a spryte rendering processor |
| US5481275A (en) | 1992-11-02 | 1996-01-02 | The 3Do Company | Resolution enhancement for video display using multi-line interpolation |
| US5752073A (en) * | 1993-01-06 | 1998-05-12 | Cagent Technologies, Inc. | Digital signal processor architecture |
| US5519842A (en) * | 1993-02-26 | 1996-05-21 | Intel Corporation | Method and apparatus for performing unaligned little endian and big endian data accesses in a processing system |
| US5574923A (en) * | 1993-05-10 | 1996-11-12 | Intel Corporation | Method and apparatus for performing bi-endian byte and short accesses in a single-endian microprocessor |
| US5564056A (en) * | 1994-03-01 | 1996-10-08 | Intel Corporation | Method and apparatus for zero extension and bit shifting to preserve register parameters in a microprocessor utilizing register renaming |
| US5781763A (en) * | 1995-04-07 | 1998-07-14 | International Business Machines Corporation | Independent control of DMA and I/O resources for mixed-endian computing systems |
| US5742840A (en) * | 1995-08-16 | 1998-04-21 | Microunity Systems Engineering, Inc. | General purpose, multiple precision parallel operation, programmable media processor |
| US7301541B2 (en) * | 1995-08-16 | 2007-11-27 | Microunity Systems Engineering, Inc. | Programmable processor and method with wide operations |
| US7483935B2 (en) * | 1995-08-16 | 2009-01-27 | Microunity Systems Engineering, Inc. | System and method to implement a matrix multiply unit of a broadband processor |
| US6295599B1 (en) * | 1995-08-16 | 2001-09-25 | Microunity Systems Engineering | System and method for providing a wide operand architecture |
| US5953241A (en) * | 1995-08-16 | 1999-09-14 | Microunity Engeering Systems, Inc. | Multiplier array processing system with enhanced utilization at lower precision for group multiply and sum instruction |
| US5819117A (en) * | 1995-10-10 | 1998-10-06 | Microunity Systems Engineering, Inc. | Method and system for facilitating byte ordering interfacing of a computer system |
| US5854914A (en) * | 1996-02-13 | 1998-12-29 | Intel Corporation | Mechanism to improved execution of misaligned loads |
| US5848436A (en) * | 1996-03-06 | 1998-12-08 | International Business Machines Corporation | Method and apparatus for efficiently providing data from a data storage medium to a processing entity |
| US5781923A (en) * | 1996-05-28 | 1998-07-14 | Hewlett-Packard Company | Adding a field to the cache tag in a computer system to indicate byte ordering |
| US6148302A (en) * | 1998-02-26 | 2000-11-14 | Sun Microsystems, Inc. | Method, apparatus, system and computer program product for initializing a data structure at its first active use |
| US6457121B1 (en) * | 1999-03-17 | 2002-09-24 | Intel Corporation | Method and apparatus for reordering data in X86 ordering |
| US6819325B2 (en) | 2000-03-07 | 2004-11-16 | Microsoft Corporation | API communications for vertex and pixel shaders |
| US7159041B2 (en) * | 2000-03-07 | 2007-01-02 | Microsoft Corporation | Method and system for defining and controlling algorithmic elements in a graphics display system |
| FR2818145B1 (en) * | 2000-12-18 | 2003-11-28 | Oreal | ANTISOLAR COSMETIC COMPOSITIONS BASED ON A SYNERGETIC MIXTURE OF FILTERS AND USES |
| US6828975B2 (en) * | 2001-03-01 | 2004-12-07 | Microsoft Corporation | Method and system for managing graphics objects in a graphics display system |
| US6874150B2 (en) | 2001-03-01 | 2005-03-29 | Microsoft Corporation | Method and system for maintaining connections between surfaces and objects in a graphics display system |
| US6812923B2 (en) | 2001-03-01 | 2004-11-02 | Microsoft Corporation | Method and system for efficiently transferring data objects within a graphics display system |
| US7023431B2 (en) * | 2001-03-01 | 2006-04-04 | Microsoft Corporation | Method and system for providing data to a graphics chip in a graphics display system |
| US6831635B2 (en) * | 2001-03-01 | 2004-12-14 | Microsoft Corporation | Method and system for providing a unified API for both 2D and 3D graphics objects |
| WO2003021423A2 (en) | 2001-09-04 | 2003-03-13 | Microunity Systems Engineering, Inc. | System and method for performing multiplication |
| US7251752B2 (en) | 2001-10-01 | 2007-07-31 | Adams Phillip M | Computerized product improvement apparatus and method |
| US7634762B1 (en) | 2005-04-19 | 2009-12-15 | Paravirtual Corp. | Selective post-compile conversion |
| US7685405B1 (en) * | 2005-10-14 | 2010-03-23 | Marvell International Ltd. | Programmable architecture for digital communication systems that support vector processing and the associated methodology |
| US8099448B2 (en) * | 2005-11-02 | 2012-01-17 | Qualcomm Incorporated | Arithmetic logic and shifting device for use in a processor |
| US20170123792A1 (en) * | 2015-11-03 | 2017-05-04 | Imagination Technologies Limited | Processors Supporting Endian Agnostic SIMD Instructions and Methods |
Family Cites Families (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3573744A (en) * | 1968-11-01 | 1971-04-06 | Bell Telephone Labor Inc | Data buffer system for transferring information from a first to a second storage medium |
| US3699525A (en) * | 1970-11-27 | 1972-10-17 | Honeywell Inf Systems | Use of control words to change configuration and operating mode of a data communication system |
| US3846763A (en) * | 1974-01-04 | 1974-11-05 | Honeywell Inf Systems | Method and apparatus for automatic selection of translators in a data processing system |
| US3996566A (en) * | 1974-12-16 | 1976-12-07 | Bell Telephone Laboratories, Incorporated | Shift and rotate circuit for a data processor |
| US4494186A (en) * | 1976-11-11 | 1985-01-15 | Honeywell Information Systems Inc. | Automatic data steering and data formatting mechanism |
| GB1563165A (en) * | 1977-11-16 | 1980-03-19 | Ibm | Character display system |
| US4281315A (en) * | 1979-08-27 | 1981-07-28 | Bell Telephone Laboratories, Incorporated | Collection of messages from data terminals using different protocols and formats |
| JPS5714957A (en) * | 1980-06-30 | 1982-01-26 | Toshiba Corp | Memory device |
| JPS5945545A (en) * | 1982-09-06 | 1984-03-14 | Mitsubishi Electric Corp | 8-bit data converting circuit |
| US4512018A (en) * | 1983-03-08 | 1985-04-16 | Burroughs Corporation | Shifter circuit |
| US4559614A (en) * | 1983-07-05 | 1985-12-17 | International Business Machines Corporation | Interactive code format transform for communicating data between incompatible information processing systems |
| US4595911A (en) * | 1983-07-14 | 1986-06-17 | Sperry Corporation | Programmable data reformat system |
| US4595980A (en) * | 1983-07-27 | 1986-06-17 | International Business Machines Corp. | Interactive data processing system having concurrent multi-lingual inputs |
-
1986
- 1986-07-11 DE DE8686305355T patent/DE3687124T2/en not_active Expired - Lifetime
- 1986-07-11 AT AT86305355T patent/ATE82412T1/en not_active IP Right Cessation
- 1986-07-11 EP EP86305355A patent/EP0236615B1/en not_active Expired - Lifetime
- 1986-07-24 JP JP61174692A patent/JPH0734174B2/en not_active Expired - Lifetime
-
1987
- 1987-01-13 IL IL81239A patent/IL81239A/en unknown
- 1987-02-03 NO NO870416A patent/NO870416L/en unknown
- 1987-02-05 IE IE870310A patent/IE870310L/en unknown
- 1987-02-05 ES ES878700270A patent/ES2007031A6/en not_active Expired
- 1987-02-06 CA CA000529196A patent/CA1285075C/en not_active Expired - Lifetime
- 1987-02-06 KR KR870000944A patent/KR870008252A/en not_active Application Discontinuation
- 1987-02-06 DK DK060987A patent/DK60987A/en not_active Application Discontinuation
- 1987-02-06 GR GR870205A patent/GR870205B/en unknown
-
1988
- 1988-11-28 US US07/277,406 patent/US4959779A/en not_active Expired - Lifetime
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8683182B2 (en) | 1995-08-16 | 2014-03-25 | Microunity Systems Engineering, Inc. | System and apparatus for group floating-point inflate and deflate operations |
| US8769248B2 (en) | 1995-08-16 | 2014-07-01 | Microunity Systems Engineering, Inc. | System and apparatus for group floating-point inflate and deflate operations |
Also Published As
| Publication number | Publication date |
|---|---|
| EP0236615A2 (en) | 1987-09-16 |
| JPS62184530A (en) | 1987-08-12 |
| NO870416L (en) | 1987-08-10 |
| DK60987A (en) | 1987-08-07 |
| US4959779A (en) | 1990-09-25 |
| EP0236615A3 (en) | 1988-10-05 |
| DE3687124T2 (en) | 1993-03-18 |
| JPH0734174B2 (en) | 1995-04-12 |
| GR870205B (en) | 1987-08-19 |
| NO870416D0 (en) | 1987-02-03 |
| KR870008252A (en) | 1987-09-25 |
| IL81239A0 (en) | 1987-08-31 |
| EP0236615B1 (en) | 1992-11-11 |
| DE3687124D1 (en) | 1992-12-17 |
| ES2007031A6 (en) | 1989-06-01 |
| IL81239A (en) | 1991-03-10 |
| ATE82412T1 (en) | 1992-11-15 |
| DK60987D0 (en) | 1987-02-06 |
| IE870310L (en) | 1987-08-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA1285075C (en) | Dual byte order computer architecture | |
| US3949379A (en) | Pipeline data processing apparatus with high speed slave store | |
| US4878174A (en) | Flexible ASIC microcomputer permitting the modular modification of dedicated functions and macroinstructions | |
| US3983539A (en) | Polymorphic programmable units employing plural levels of sub-instruction sets | |
| US4395758A (en) | Accelerator processor for a data processing system | |
| US4016545A (en) | Plural memory controller apparatus | |
| US3781812A (en) | Addressing system responsive to a transfer vector for accessing a memory | |
| US5280595A (en) | State machine for executing commands within a minimum number of cycles by accomodating unforseen time dependency according to status signals received from different functional sections | |
| US4558411A (en) | Polymorphic programmable units employing plural levels of sub-instruction sets | |
| US3629857A (en) | Computer input buffer memory including first in-first out and first in-last out modes | |
| JPS6259822B2 (en) | ||
| US4204252A (en) | Writeable control store for use in a data processing system | |
| JPS598846B2 (en) | Microprogrammable peripheral controller | |
| US4339793A (en) | Function integrated, shared ALU processor apparatus and method | |
| US5854939A (en) | Eight-bit microcontroller having a risc architecture | |
| US4179738A (en) | Programmable control latch mechanism for a data processing system | |
| EP0126247B1 (en) | Computer system | |
| EP1271305B1 (en) | Data processing apparatus | |
| US4430711A (en) | Central processing unit | |
| US5179671A (en) | Apparatus for generating first and second selection signals for aligning words of an operand and bytes within these words respectively | |
| US5680632A (en) | Method for providing an extensible register in the first and second data processing systems | |
| EP0338564B1 (en) | Microprogram branching method and microsequencer employing the method | |
| US4811266A (en) | Multifunction arithmetic indicator | |
| JPS605979B2 (en) | electronic control device | |
| JP2703884B2 (en) | Data write control method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| MKEX | Expiry |