Search Images Maps Play YouTube News Gmail Drive More »
Sign in
Screen reader users: click this link for accessible mode. Accessible mode has the same essential features but works better with your reader.

Patents

  1. Advanced Patent Search
Publication numberUS20010011342 A1
Publication typeApplication
Application numberUS 09/796,037
Publication dateAug 2, 2001
Filing dateFeb 28, 2001
Priority dateJul 9, 1998
Also published asUS6343356, US6430677, USRE40883
Publication number09796037, 796037, US 2001/0011342 A1, US 2001/011342 A1, US 20010011342 A1, US 20010011342A1, US 2001011342 A1, US 2001011342A1, US-A1-20010011342, US-A1-2001011342, US2001/0011342A1, US2001/011342A1, US20010011342 A1, US20010011342A1, US2001011342 A1, US2001011342A1
InventorsGerald Pechanek, Edwin Barry
Original AssigneePechanek Gerald G., Barry Edwin F.
Export CitationBiBTeX, EndNote, RefMan
External Links: USPTO, USPTO Assignment, Espacenet
Methods and apparatus for dynamic instruction controlled reconfigurable register file with extended precision
US 20010011342 A1
Abstract
A reconfigurable register file integrated in an instruction set architecture capable of extended precision operations, and also capable of parallel operation on lower precision data is described. A register file is composed of two separate files with each half containing half as many registers as the original. The halves are designated even or odd by virtue of the register addresses which they contain. Single width and double width operands are optimally supported without increasing the register file size and without increasing the number of register file ports. Separate extended registers are also employed to provide extended precision for operations such as multiply-accumulate operations.
Images(11)
Previous page
Next page
Claims(24)
We claim:
1. A reconfigurable register file capable of storing and retrieving double and single width data types on a cycle by cycle basis comprising:
two register files of equivalent capacity and equivalent number of store (write) and retrieve (read) ports;
single and dual width data type instructions;
a plurality of multiplexers to select one of the two register files or both for a single width or double width read operation; and
combinatorial logic to select one of the two register files or both for a single width or double width write operation.
2. The reconfigurable register file of
claim 1
wherein the most significant bits for a double width access are consistently stored in either the odd register file or the even register file, and the least significant bits for the double width access are consistently stored in either the even register file or the odd register file.
3. The reconfigurable register file of
claim 1
wherein the odd register file has an associated extended register file.
4. The reconfigurable register file of
claim 1
wherein the even register file has an associated extended register file.
5. A ManArray processing machine comprising:
a processing element having Store, Load, MAU, ALU and DSU execution units;
a 32 bit odd register file having eight read ports and 4 write ports;
a 32 bit even register file having eight read ports and 4 write ports; and
a plurality of multiplexers to select the odd or even register file or both for a read operation.
6. The ManArray processing machine of
claim 5
wherein the 32 bit odd register file has an associated 8 bit extended register.
7. The ManArray processing machine of
claim 5
wherein the 32 bit even register file has an associated 8 bit extended register.
8. The ManArray processing machine of
claim 5
further comprising combinatorial logic to enable the 32 bit odd register file or the 32 bit even register file or both for a write operation.
9. A method of operating on double width and single width files utilizing a single reconfigurable register file, the method comprising the following steps:
configuring a double width file into a most significant bits half (MSBH) and least significant bits half (LSBH);
storing the MSBH in a first register file;
storing the LSBH in a second register file; and
utilizing a multiplexer to select the first register file or the second register file or both to perform a read operation.
10. The method of
claim 9
further comprising the step of utilizing combinatorial logic to perform a write operation and store the MSBH or the LSBH or both.
11. The method of
claim 9
further comprising the step of: extending the precision of the reconfigurable register file by employing an extended register file in conjunction with the first register file and employing an extended register file in conjunction with the second register file.
12. A reconfigurable register file capable of storing and retrieving double and single width data types on a cycle by cycle basis comprising:
an odd register file having an associated extended register file;
an even register file having an associated extended register file;
a multiplexer to select the odd or the even register file or both for a single width or double width read operation; and
combinatorial logic to select the odd or the even register file or both for a single width or double width write operation.
13. The reconfigurable split register file of
claim 12
wherein the most significant bits for a double width access are consistently stored in either the odd register file or the even register file, and the least significant bits for the double width access are consistently stored in either the even register file or the odd register file.
14. The reconfigurable register file of
claim 12
wherein the odd and even register files have associated extended register files.
15. A ManArray processing machine comprising:
a processing element having Store, Load, MAU, ALU and DSU execution units;
a p-bit odd register file having at least 8 read ports and at least 4 write ports;
a p-bit even register file having at least 8 read ports and at least 4 write ports; and
a plurality of multiplexers to select the odd or even register file for a read operation.
16. The processing machine of
claim 15
wherein p equals 32.
17. The processing machine of
claim 15
wherein the p-bit odd and even register files have k-bit associated extended register files.
18. The ManArray processing machine of
claim 16
wherein the 32 bit odd register file has an associated 8 bit extended register file.
19. The ManArray processing machine of
claim 16
wherein the 32 bit even register file has an associated 8 bit extended register file.
20. A VLIW ManArray processing machine comprising:
a processing element having multiple execution units;
a p-bit odd register file having q-read ports and r-write ports;
a p-bit even register file having q-read ports and r-write ports; and
a plurality of multiplexers to select the odd or even register file for a read operation.
21. The VLIW ManArray processing machine of
claim 20
further comprising:
k-bit extended precision register files associated with the p-bit odd and even register files; and
said processing element further comprises logic control for enabling selection of the k-bit extended precision register files.
22. A reconfigurable register file apparatus having extended precision comprising:
at least two x/2-bit registers;
at least two extended precision registers;
a multiplier for performing x/2 times x/2 multiply operations on x/2-bit numbers from two of said at least two x/2-bit registers to produce an x-bit result;
an extended accumulator for extending the x-bit result; and
a multiplexer for selecting one of the at least two extended precision registers for storage of a partial sum of product value for a next multiply accumulate operation.
23. The reconfigurable register file apparatus of
claim 22
wherein x equals 16, 32 or 64.
24. The reconfigurable register file apparatus of
claim 22
wherein said extended precision registers area y bit registers, where y equals 4, 8, 16 or 32.
Description
FIELD OF THE INVENTION

[0001] The present invention relates generally to improvements to processing, and more particularly to advantageous techniques for providing a scalable building block register file which in a first application of the register file provides a low cost lower capacity register file, while in a second application, a higher capacity register file with dynamic reconfiguration support for flexible data type operations is provided. The present invention also relates to advantageous techniques for providing a dynamically reconfigurable register file of variable size width for different levels of data precision operations when executing algorithms demanding variable data types of variable precision requirements and for conducting multiple parallel operations on lower precision data in 32 bit and 64 bit forms.

BACKGROUND OF THE INVENTION

[0002] When executing algorithms it is desirable to have a register file that can be organized to more advantageously support processing of the varying data types and formats that dynamically occur in a programming application. For example, a register file of large width for high precision operations can be required in one part of an application while single and multiple parallel operations on lower precision data can be required in a different part of the same application. This desire is offset by the hardware cost to implement a wider register file or the hardware cost to implement additional read and write ports. The problem is how to achieve a dynamically configurable register file with extended precision at a reduced hardware cost without affecting general capabilities including performance.

SUMMARY OF THE INVENTION

[0003] The present invention advantageously addresses these problems while achieving a variety of advantages as addressed in further detail below. In one aspect of the present invention, to achieve the effect of a doublewide register file, two single wide register files, each with the same number of registers, are used in combination to provide a single register model that uses less read and write ports individually than a single register file of twice the capacity would require. Due to the reduced size of the register files and reduced number of read and write ports, higher performance implementations can be achieved as compared to a single register file of equivalent combined capacity of data width and read and write ports. The architecture designates one reduced register file to contain even register addresses and the other to contain odd register addresses. In a second aspect of this invention, the architecture designates one register file configured as two banks of registers wherein the even and odd registers are selectable by means of the read/write port address lines. In a third aspect of this invention, an additional register set of at least one register can be dynamically associated with any register in the register file to flexibly provide extended precision data width to any selected file register.

[0004] By appropriate multiplexing and control logic, single width, double width, and extended precision accessing are made available. By architecture definition, double width accesses are constrained to only work on even-odd register pairs thereby treating the two separate register files as a single addressable file of twice the width of an individual register. By convention and as dictated by the architecture, either the even or odd register file is designated as containing the upper half of the bits in a double width access. Double width accesses may occur on the read, write operations, or both depending on the operation to be performed. In this way, the access width of the register file is doubled without the addition of costly read/write ports or more bits per each register and the number of required read and write ports per half is reduced. The double width register file achieved by this invention provides the single width accesses for a simpler programming model when dealing with data types of single width. Additionally, since the same number of read and write ports exist on both halves, single width accesses across the full even plus odd register address space are possible.

[0005] These and other features, aspects and advantages of the invention will be apparent to those skilled in the art from the following detailed description taken together with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

[0006]FIG. 1A illustrates a first prior art register file arrangement;

[0007]FIG. 1B illustrates a second prior art register file arrangement;

[0008]FIG. 1C illustrates a first reconfigurable register file in accordance with the present invention;

[0009]FIG. 1D illustrates an exemplary add instruction for use in conjunction with a reconfigurable register file;

[0010]FIG. 2 illustrates a ManArray indirect very long instruction word (iVLIW) processor in conjunction with a reconfigurable register file in accordance with the present invention;

[0011]FIG. 3A illustrates two x/2 extended precision registers used with the reconfigurable register file for extended precision;

[0012]FIG. 3B illustrates four x/4 extended precision registers used with the reconfigurable register file for extended precision;

[0013]FIG. 3C illustrates an exemplary MPXYA instruction for use with a reconfigurable register file; and

[0014]FIG. 4 illustrates two x/4 extended precision registers used with a building block register file that is a subset of the reconfigurable register file.

DETAILED DESCRIPTION

[0015] Further details of a presently preferred ManArray architecture are found in U.S. patent application Ser. Nos. 08/885,310 and 08/949,122 filed Jun. 30, 1997 and Oct. 10, 1997, respectively, Provisional Application Serial No. 60/064,619 entitled Methods and Apparatus for Efficient Synchronous MIMD VLIW Communication” filed Nov. 7, 1997, Provisional Application Serial No. 60/067,511 entitled “Method and Apparatus for Dynamically Modifying Instructions in a Very Long Instruction Word Processor” filed Dec. 4, 1997, Provisional Application Serial No. 60/068,021 entitled “Methods and Apparatus for Scalable Instruction Set Architecture” filed Dec. 18, 1997, Provisional Application Serial No. 60/071,248 entitled “Methods and Apparatus to Dynamically Expand the Instruction Pipeline of a Very Long Instruction Word Processor” filed Jan. 12, 1998, Provisional Application Serial No. 60/072,915 entitled “Methods and Apparatus to Support Conditional Execution in a VLIW-Based Array Processor with Subword Execution filed Jan. 28, 1988, Provisional Application Serial No. 60/088,148 entitled “Methods and Apparatus for ManArray PE-PE Switch Control” filed on Jun. 5, 1998, Provisional Application Serial No. 60/092,148 entitled “Methods and Apparatus for Dynamic Instruction Controlled Reconfigurable Register File with Extended Precision” filed Jul. 9, 1998, and Provisional Application Serial No. 60/092,130 entitled “Methods and Apparatus for Instruction Addressing in Indirect VLIW Processors” filed Jul. 9, 1998, all of which are assigned to the assignee of the present invention and incorporated herein by reference in their entirety.

[0016]FIG. 1A depicts a first prior art register file arrangement 100 (Prior Art) 1 consisting of “n” registers R0 . . . R(n−1) 110 with four read data output ports, Rx0 112, Rx1 114, Rx2 116, and Rx3 118, each of x-bits width. In addition, there are two write ports, Rt0 124 and Rt1 126, each of x-bits width. A total of six x-bit width ports are required to provide double width accesses. The data bit width “x” is typically 8-bit, 16-bit, 32-bit, 64-bit and other sizes such as 9-bit, 18-bit etc. The register file read data output ports connect to an execution unit, for example a Multiply Accumulate Unit 120 consisting of a multiplier 121 and an accumulator 123. The equation of operation for the MAU 120 shown in FIG. 1A is Rt1.w∥Rt0.w=Rx1.w*Rx0.w+Rx3.w∥Rx2.w where the notation Ra#.w has a=x for an input source operand and a=t for a target result operand, the # indicative of the register number from the register file 110, and the .w indicates the data type which in this case is a single width x-bit word size. The “∥” symbol indicates a concatenation of input or output bus widths due to the granularity of the read and write ports of the register file 110. Other execution unit types include Arithmetic Logic Units, specialized functional units, etc. as dictated by a particular processor architecture.

[0017]FIG. 1B depicts a second prior art register file arrangement 200 (Prior Art 2) consisting of “n” registers R0 . . . R (n−1) 210 with three 2x-bit wide read ports, Rx0 212, Rx1 214, and Rx2 216. In addition, there is a single 2x-bit wide write port Rt 224. A total of three 2x-bit width ports are required to provide double width accesses. The data bit width “2x” is typically 16-bit, 32-bit, 64-bit, 128-bit and other sizes such as 18-bit, 36-bit, etc. The register file read data output ports typically connect to an execution unit, for example a Multiply Accumulate Unit 220 consisting of a multiplier 221 and an accumulator 223. The equation of operation for the MAU 220 shown in FIG. 1B is Rt#0.2w=Rx1.w*Rx0.w+Rx2.2w where the notation Ra#.w has a=x for an input source operand and a=t for a target result operand, the # indicative of the register number from the register file 110, the 0.2w indicates the data type which in this case is a double width 2x-bit word size, and the single width “.w” word size uses only half of the read ports Rx0 212 and Rx1 214. In the example indicated, the other halves of the Rx0 and Rx1 data ports are unused. Other execution unit types include Arithmetic Logic Units, specialized functional units, and the like as dictated by a particular processor architecture.

[0018] There are problems associated with each of these prior art designs which are solved by the present invention. In the Prior Art 1 approach of FIG. 1A, a large number of read and write single width ports are required to achieve the input data width granularity of operand accesses and 2x-bit result storage requirements of the indicated execution unit. In a VLIW machine supporting multiple execution units, the number of ports grows for each supported execution unit. The number of read and write ports limits the width of the VLIW and affects the performance that can be achieved potentially to an unacceptable level. Increasing the width of the VLIW can be counterproductive, since adding ports to a single “n” register file increases the critical path slowing read and write access.

[0019] In FIG. 1B the Prior Art 2 approach, double width ports are used reducing the number of ports but at the expense of granularity of operand accesses and results storage. This means that if a single x-bit width operation is required, the additional x-bits accessed in a 2x-bit read access could be wasted causing a loss of storage, potentially increased power, quite possibly additional programmer overhead to conserve register usage depending upon the whether an execution unit has a single width x-bit write execute action, and potentially slower performance due to programming overhead to conserve register storage space. In addition, if the same number of “n” registers is to be maintained as compared to the Prior Art 1 approach, the “n” registers must be of 2x width doubling the size of the register file which, in general, does not necessarily provide a significant performance advantage that would justify the added expense.

[0020]FIG. 1C depicts a reconfigurable file and execution unit 300 in accordance with the present invention. The reconfigurable register file consists of a first portion or file 330 and a second portion or file 340, each containing three single x-bit width read access ports, and a single x-bit write port. File 330 consists of read ports Ryo 332, Rxo 334, Rso 336, and write port Rto 326. File 340 consists of read ports Rye 342, Rxe 344, Rse 346 and write port Rte 324. Multiplexers 301 and 302 allow single width accesses 352 and 354, respectively, from either half of the composite register file. The MAU execution unit 320, consists of a multiplier 321 and an accumulator 323. In FIG. 1C each register file portion 330 and 340 is required to have only three x-bit read ports and a single x-bit write port. In the present invention, a small number of ports for each register file portion are utilized to achieve x-bit width granularity of storage. Consequently, this design more readily supports a VLIW architecture containing multiple execution units. In addition, by reducing the register size in each portion of the composite register file of “n/2” registers along with a minimum number of ports, the critical path in the register file is reduced thereby improving its read and write access performance. It is noted that an n/4 register arrangement is also a feasible approach for low cost applications.

[0021] In addition, the present invention does not sacrifice granularity of accesses, single width and double width accesses are optimally supported without increasing the register file size. This is important since all applications contain a control portion that typically has sequential execution with little or no parallelism and an algorithm execution portion that typically contains operations that can be executed in parallel. The parallel code portion can be operated upon by packed data operations and VLIW operations while the sequential control section usually requires single width data type operation support. It is consequently of great importance to efficiently support the sequential code data types as well as the parallel code data types. A reconfigurable register file provides this support.

[0022] A presently preferred add instruction is shown in FIG. 1D showing support for single x=32 bit width operations 430, 431, and 432 as well as double x=64 bit width operations 440, 441, 442, and 443. With this novel register file design integrated into the instruction set architecture, single width and double width instructions can be mixed on a cycle by cycle basis. For example with x=32-bits, the present invention allows the reconfigurable register file to be treated as a 32×32-bit register file in one cycle and a 16×64-bit register file in the very next cycle.

[0023] ManArray Reconfigurable Register File

[0024] In a ManArray indirect Very Long Instruction Word (iVLIW) processor's execution units 250 as shown in FIG. 2, there are 8 read ports and 4 write ports for each half of the reconfigurable register file 200. These ports support single width 32-bit and in combination double width 64-bit accesses for any of the 5 execution units. Address and control logic are not shown in FIG. 2 to improve the clarity of illustration. It will be recognized that registers having numbers of bits (p), and different numbers of read ports (q) and write ports (r) may be employed.

[0025] An exemplary instruction that takes advantage of this configuration of the register file is the 32-bit multiply-accumulate. The operation performed by this instruction is:

Rto∥Rte←(Rx * Ry)+Rto∥Rte

[0026] Rx and Ry are 32-bit quantities and Rto∥Rte is a 64-bit quantity. In a traditional non-split 32-bit wide register file implementation, it would take 1(Rx)+1(Ry)+2(Rto∥Rte)=4 32-bit read ports and 2(Rto∥Rte←) 32-bit write ports to accommodate this instruction. However, using the two register file blocks described above, this same function can be implemented with 3 read ports and 1 write port per block by using even/odd pairs for the 64-bit quantities.

[0027] For operations that do not need 64-bit quantities, the mux on the input to the functional unit is controlled to select the proper register file. As an example, consider the add instruction executing on the ALU that performs the function:

Rt←Rx+Ry

[0028] where Rx, Ry, and Rt are 32-bit quantities. If Rx is R1, Ry is R3, and Rt is R5 then the mux on the lower 32-bit inputs selects the odd register file for both inputs. Since the ALU has two read ports on the odd register file this operation is accomplished without any problems. The 32-bit write to R5 is also easily accomplished by only enabling the write for the odd register file. Any combination of even or odd registers can be selected without restrictions.

[0029] Extended Precision

[0030] An approach to increasing the width of the register file at a reduced hardware cost comes from taking into consideration where the extra precision gained from a wider register file is really needed. For example, in multiply-accumulate operations, extra precision is needed for the accumulation in some applications to increase the number of times accumulation can occur without overflow. In addition, even though providing extended precision support to all register files is a general case, in specific applications this is usually not required and would be considered unnecessarily expensive to implement. It is also not desirable to explicitly specify which registers are specially enabled to support extended precision operations. Further, it is not desirable to have additional architecturally defined extended precision accumulator registers in addition to an existing register file. Consequently, for low cost implementations, as well as, for a flexible programming model for extended precision support, the present reconfigurable register file with extended precision invention advantageously addresses such concerns.

[0031] To accommodate such specific needs without increasing the number of ports or the width of the entire register file, the reconfigurable register file concept is extended by adding, in the simplest case, a single additional register known as the extended precision register. FIG. 3A illustrates a system 500 employing two (x/2)-bit registers 553 and 555 labeled XH1 and XH0 which are used to extend the precision of the accumulation operation that occurs in the Extended accumulator unit 523. The Multiply with Extended Accumulate operation is defined in FIG. 3C which defines the MPYXA instruction. The apparatus of FIG. 3A is adapted for an 80-bit extended accumulate operation where a 32×32-bit multiply is carried out by multiplier 521 which produces a 64-bit result that is extended to 80-bits in the accumulate operation of extended accumulator 523. This can be seen in FIG. 3A where depending upon the least significant bit (LSB) of the target register field in the MPYXA instruction, bit 17 of FIG. 3C, one of two extended precision registers XH1 553 or XH0 555 is selected via multiplexer 563. The least significant bit of the Register Target field allows the extended precision register to be arbitrarily used with any pair of registers in the register file. This powerful but simple feature allows a programmer to utilize any pair of registers for an extended precision operation without any mode control or specialized accumulator hardware added to the architecture. The inputs of multiplexer 563 are the (x/2)-bit length extended precision input operands XH0 552 and XH1 554. The multiplexer 563 selects XH0 552 when its input control line 556 is a “0”. The multiplexer 563 selects XH1 554 when its input control line 556 is a “1”. The output of multiplexer 563 is signal line 564 which is (x/2)-bits and is an input to the extended accumulator 523. The extended output 566 is a partial sum of product value that is stored in the extended precision registers in preparation for the next multiply accumulate operation. The output 566 is written to either XH1 553 or XH0 555 under control of a Write (Wr) signal 562. The pipeline stored LSB of the Rte field 551 is used to control the Wr signal via logical AND type function where the Wr 562 is passed onto the register depending on the state of the LSB. The AND gates 557 and 559 control this function, where the LSB input to AND 559 is an inverted version 561 of whatever bit appears on line 556. The output of the AND gates 558 and 560 control the writing of the output extended precision data 566 to their extended precision registers. The extended precision registers XH1 553 and XH0 555 are part of the special purpose or miscellaneous registers that are used in the processor and consequently are load-able and read-able by the programmer. The read and write buses that accomplish this task for the programmer are not shown in FIG. 3A for reasons of clarity.

[0032]FIG. 3B depicts a quad extended precision apparatus 600 supporting the MPYXA multiply with extended accumulate instruction of FIG. 3C which shows dual 40 bit accumulation 702 and double width 80 bit accumulation 703. In FIG. 3B, four (x/4)-bit registers are provided as partitions of two (x/2)-bit registers 653 and 655 labeled XB3 and XB2 in register 653 and XB1 and XB0 in register 655. The four (x/4)-bit registers are used to extend the precision of the accumulation operation that occurs in the Extended accumulator units 621 and 625. The Multiply with Extended Accumulate operation is defined in FIG. 3C which defines the MPYXA instruction for dual 40-bit extended accumulates 702. The apparatus of FIG. 3B supports the dual 40-bit extended accumulate operation where two 16×16-bit multiplies 619 and 623 each produce a 32-bit result that are each extended to 40-bits in the accumulate operations performed by accumulators 621 and 625, respectively. This operation can be seen in FIG. 3B where depending upon the least significant bit (LSB) of the target register field in the MPYXA instruction, bit 17 of FIG. 3C, one of two extended precision registers XB3 and XB2 653 or XB1 and XB0 655 are selected via multiplexers 663 and 665. The least significant bit of the Register Target field allows the extended precision register to be arbitrarily used with any pair of registers in the register file. This powerful but simple feature allows a programmer to utilize any pair of registers for an extended precision operation without any mode control or specialized accumulator hardware added to the architecture. The input of multiplexers 663 and 665 are the (x/2)-bit length extended precision input operands XB0 622 and XB2 626 for multiplexer 663, and XB1 624 and XB3 628 for multiplexer 665. The multiplexer 663 selects XB0 622 when its input control line 630 is a “0”. The multiplexer 665 selects XB1 624 when its input control line 630 is a “0”. The multiplexer 663 selects XB2 626 when its input control line 630 is a “1”. The multiplexer 665 selects XB3 628 when its input control line 630 is a “1”. The output 670 of multiplexer 663 is (x/4)-bits and serves an input to the extended accumulator 621. The extended output 636 is a partial sum of product value that is stored in the extended precision registers in preparation for the next multiply accumulate operation. The output 672 of multiplexer 665 is (x/4)-bits and serves as an input to the extended accumulator 625. The extended output 638 is a partial sum of product value that is stored in the extended precision registers in preparation for the next multiply accumulate operation. The output 636 is written to either XB2 or XB0 and the output 638 is written to either XB3 or XB1 all under control of a Write (Wr) signal 648. The pipeline stored LSB of the Rte field 651 is used to control the Wr signal via a logical AND type function where the Wr 648 is passed onto the register depending on the state of the LSB. The AND gates 657 and 659 control this function, where the LSB input to AND 659 is an inverted 661 version of 630. The output of the AND gates 632 and 634 control the writing of the output extended precision data 636 and 638 to their extended precision registers. The partitioned extended precision registers 653 and 655 are part of the special purpose or miscellaneous registers that are used in the processor and consequently are load-able and read-able by the programmer. The read and write buses that accomplish this task for the programmer are not shown in FIG. 3B for reasons of clarity.

[0033] In a typical application, x is 32-bits, with (x/2)=16-bits and (x/4)=8-bits though different extended precision bit widths are not precluded. The present approach allows dual accumulations of 40-bits of precision for dual 16×16 multiply-accumulates, as specified in the MPYXA instruction FIG. 3C and for the exemplary apparatus shown in FIG. 3B. For 32×32 multiply-accumulate operations, 80-bits of precision are available for the accumulation. The extended precision concept can be further extended to support quad 20 bit accumulations where x is 16-bits and there are 4 extended precision bits. The concept can be further generalized by using more than one x-bit extended precision register and basing the selection of the register extended precision portions on more than the single LSB of the Instruction Pte field. Since a single 32-bit extended precision register provides support for up to two 80-bit extended accumulate operations and up to four 40-bit extended accumulate operations, further extensions, even though feasible, for practical reasons presently appear to be of limited use.

[0034] Due to the nature of many applications, a processor can be designed utilizing a subset of the ManArray architecture that is based upon a single 16×32 register file, i.e. one of the building blocks for a reconfigurable register file. Dual 8×32 register files can be also used to create a reconfigurable 16×32 register file. An important aspect is that a low cost register file design point can be reached by subsetting the ManArray architecture that allows future growth into higher performance processors that remain code compatible with the lower cost subset design. An exemplary apparatus 700 implementing this use of the extended precision concept with a single register file design is shown in FIG. 4.

[0035] While the present invention has been described in the context of a number of presently preferred embodiments, it will be recognized that the teachings of the present invention may be advantageously applied to a variety of processing arrays and variously adopted consistent with the claims which follow.

Referenced by
Citing PatentFiling datePublication dateApplicantTitle
US8423745Nov 16, 2009Apr 16, 2013Convey ComputerSystems and methods for mapping a neighborhood of data to general registers of a processing element
US8443147Dec 5, 2011May 14, 2013Convey ComputerMemory interleave for heterogeneous computing
US8561037Aug 29, 2007Oct 15, 2013Convey ComputerCompiler for generating an executable comprising instructions for a plurality of different instruction sets
US8583897 *Feb 2, 2009Nov 12, 2013Arm LimitedRegister file with circuitry for setting register entries to a predetermined value
CN101930355BAug 24, 2010Jul 24, 2013中国航天科技集团公司第九研究院第七七一研究所Register circuit realizing grouping addressing and read write control method for register files
CN101930356BAug 24, 2010Mar 20, 2013中国航天科技集团公司第九研究院第七七一研究所Method for group addressing and read-write controlling of register file for floating-point coprocessor
Classifications
U.S. Classification712/16, 712/E09.025, 712/225, 712/E09.027
International ClassificationG06F9/30
Cooperative ClassificationG06F9/3012, G06F9/30112, G06F9/30105, G06F9/30036, G06F9/3016
European ClassificationG06F9/30R5, G06F9/30A1P, G06F9/30T4, G06F9/30R4C, G06F9/30R4
Legal Events
DateCodeEventDescription
Aug 29, 2006ASAssignment
Owner name: ALTERA CORPORATION, CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:PTS CORPORATION;REEL/FRAME:018184/0423
Effective date: 20060824
Owner name: ALTERA CORPORATION,CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:PTS CORPORATION;US-ASSIGNMENT DATABASE UPDATED:20100316;REEL/FRAME:18184/423
Dec 28, 2005FPAYFee payment
Year of fee payment: 4
Aug 3, 2004RFReissue application filed
Effective date: 20040419
Nov 24, 2003ASAssignment
Owner name: PTS CORPORATION, CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:ALTERA CORPORATION;REEL/FRAME:014683/0914
Effective date: 20030407
Owner name: PTS CORPORATION 101 INNOVATION DRIVESAN JOSE, CALI
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:ALTERA CORPORATION /AR;REEL/FRAME:014683/0914
Nov 20, 2003ASAssignment
Owner name: ALTERA CORPORATION, CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BOPS, INC.;REEL/FRAME:014683/0894
Effective date: 20030407
Owner name: ALTERA CORPORATION 101 INNOVATION DRIVESAN JOSE, C
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:BOPS, INC. /AR;REEL/FRAME:014683/0894