1. Field of the Invention
The present invention relates to a saturating multiplier. In one example, the present invention relates to methods and apparatus for using the saturating multiplier to implement a multiply-accumulate (MAC) unit.
2. Description of Related Art
Hardware devices such as programmable logic chips use multipliers and associated logic to perform various operations. In typical instances, multipliers handle data having a fixed width. For example, a multiplier may be a multiplier on a 32-bit processor. In this case, the multiplier is configured to perform an operation on two 32-bit operands. In some case, the product of the two 32-bit operands can be represented as a 32-bit product. However, in other instances, a 32-bit data width may not be able to represent the 32-bit product. For example, multiplication of two 32-bit numbers may yield a 64-bit product. Because the datapath of a processor may not support the 64-bit product, saturation is provided as a solution.
When saturation occurs, the 64-bit product is simply represented as the closest value presentable by a 32-bit data sequence. The saturated value can then be added to or subtracted from other values. The adding or subtracting may involve other conditions such as carry overflow.
However, mechanisms for implementing multipliers and handling saturation associated with a multiplier are limited. In some instances, the saturation condition is explicitly detected. However, explicitly detecting saturation conditions can be inefficient, particularly on programmable devices. Consequently, techniques and mechanisms are provided for more efficiently handling saturation associated with a multiplier and multiply accumulate blocks.
Methods and apparatus are provided for implementing an efficient saturating multiplier associated with addition and subtraction logic. The result of the multiplier is saturated before accumulating. The result of the multiplier can be stored in a result register in unsaturated form. The output of the result register can then be saturated and provided to addition and subtraction logic to allow efficient implementation of a saturating multiplier.
In one embodiment, a method for performing multiplication on a programmable chip is provided. A multiply operation is performed using a multiplier associated with addition/subtraction (add/sub) logic. The result of the multiplier is saturated before accumulating by using add/sub logic. Saturating the result of the multiplier before accumulating allows implementation of multiply accumulate using a reduced number of logic elements.
In another embodiment, a programmable chip is provided. The programmable chip includes a signed multiplier, an extractor, an adder, and a saturation block. The signed multiplier is configured to receive a first input A and a second input B. The signed multiplier is configured to provide an output C. The extractor is configured to take the output C and truncate the output to provide a truncated value M. The adder is configured to take truncated value M as a first input and provide a value D. The saturation block is operable to saturate the value D to provide a value Z as the second input to the adder.
These and other features and advantages of the present invention will be presented in more detail in the following specification of the invention and the accompanying figures, which illustrate by way of example the principles of the invention.
The invention may best be understood by reference to the following description taken in conjunction with the accompanying drawings, which illustrate specific embodiments of the present invention.
Reference will now be made in detail to some specific embodiments of the invention including the best modes contemplated by the inventors for carrying out the invention. Examples of these specific embodiments are illustrated in the accompanying drawings. While the invention is described in conjunction with these specific embodiments, it will be understood that it is not intended to limit the invention to the described embodiments. On the contrary, it is intended to cover alternatives, modifications, and equivalents as may be included within the spirit and scope of the invention as defined by the appended claims. For example, the techniques of the present invention will be described in the context of particular processors and memory.
For example, the techniques of the present invention will be described in the context of particular components and devices. However, it should be noted that the techniques of the present invention can be applied to a variety of types of components and devices. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. The present invention may be practiced without some or all of these specific details. In other instances, well known process operations have not been described in detail in order not to unnecessarily obscure the present invention. Furthermore, techniques and mechanisms of the present invention will sometimes be described in singular form for clarity. However, it should noted that some embodiments can include multiple iterations of a technique or multiple instantiations of a mechanism unless noted otherwise. For example, a processor is used in a variety of contexts. However, it will be appreciated that multiple processors can also be used while remaining within the scope of the present invention.
A variety of devices including Digital Signal Processing (DSP) processors make extensive use of multiplication and addition/subtraction. In many instances, it is not possible to represent the result of a multiplication (or addition) in the number of bits allowed, so many standards including DSP standards specify that the result be saturated to the largest or smallest number representable.
In one example, the 1q15 format is used to represent floating point numbers. That is, the 1q15 representation is used for representing numbers in the range: −1 to +0.999.
Performing operations on fixed bit-length numbers can yield results that can not be represented in the same bit-length. For example, multiplying two 1q15 numbers gives a 2q30 result. However, this result is often truncated back to a 1q15 format, resulting in a loss of precision (from truncating the 30 bits after the decimal point to 15 bits). In most cases, the truncation leads to an accurate result. However, multiplying (−1)*(−1) results in an incorrect result after truncation: (giving (−1) instead of (+1)). In this case, the result is saturated instead.
According to various embodiments, multiplication is often associated with an addition or subtraction. In some examples, a multiply-accumulate unit provided is capable of the following operations:
performing a multiplication, storing the result (MUL);
performing a multiplication, adding the result to the previously stored value (MAC);
performing a multiplication, subtracting the result to the previously stored value (MSC).
In many implementations, saturation of the multiplication result occurs before any further addition/subtraction.
According to various embodiments, entries 701 and 707 show no saturation condition. Consequently, the output provided is the same as the M[15:0] input provided. However, entries 703 and 705 show saturation conditions. That is, M[16] is not the same as M[15]. In one example, when M[16] is 0 and M[15] is 1 at 703, the untruncated value provided should be +1 or greater. Consequently, the saturated output provided is the value closest to +1, or 0111111111111111 or +0.999. In another example, when M[16] is 1 and M[15] is 0 at 705, the untruncated value provided should be −1 or less. Consequently, the saturated output provided is the value closest to −1, or 1000000000000000 or −1.
Invert-M=˜DoAddnSub XOR M[16] XOR M[15]; and
Carry-In=˜DoAddnSub.
Although, inversion and carry-in can be implemented in a wide variety of manners, the following description outlines a particularly convenient mechanism for implementing inversion and carry-in. The mechanisms and techniques of the present invention can also be implemented using a typical logic element in arithmetic mode.
Although the techniques and mechanisms of the present invention are applicable to a variety of different devices including general purpose processors, video accelerators, cryptography accelerators, digital signal processors, microcontrollers, etc., the techniques and mechanisms of the present invention are particularly applicable to programmable chips. Implementing mechanisms such as MAC blocks for performing multiply accumulate operations is typically expensive on a programmable chip.
The programmable chip uses different types of resources that can be interchangeably used in different allocations to implement the system on a programmable chip. In one example, the programmable chip uses logic elements for implementing each of the various components on the programmable chip.
Logic elements typically can be implemented using components such as antifuses, static RAM, and EPROMS. Any mechanism on a programmable chip that performs an operation on a given number of input lines to provide one or more outputs based on information programmed is herein referred to as a logic element. Some logic elements are implemented as combinations of lookup tables and switches for performing Boolean operations on input lines. In one example, a logic element includes a 16-bit SRAM lookup table (LUT) that can implement an arbitrary 4-input logic function, circuitry that forms a fast carry chain and a fast cascade chain, a register and preset/reset logic for the register. In other examples, a logic element is an adaptive logic module (ALM) that provides 8 inputs, allowing it to efficiently support various configurations such as 6-input or 7-input lookup tables or two independent 4-input lookup tables.
The techniques and mechanisms of the present invention allow the implementation of a system on a programmable chip from a high-level language program. In one example, variable latency and fixed latency can be supported on a system using a conventional bus architecture.
A system bus typically has a set width (e.g. 64 bits, 128 bits) and allows only one primary component to actively use the bus at any one time. In conventional systems, only one primary component can access any one of the given secondary components at any given time. Multiple primary components accessing secondary components in a manner that would cause data bit collisions if performed on the same data lines is referred to herein as accessing secondary components at the same time.
In one example, an Ethernet component is accessing a secondary PIO. While the Ethernet component is accessing the secondary PIO, a processor can not access an SDRAM through a peripheral interface even if both the primary streaming output device and the peripheral interface are available.
According to various embodiments of the present invention, it is recognized that a bus is no longer required in certain applications such as system on a chip, system on a programmable chip, and other computer system implementations. A device such as a programmable logic device (PLD) or a field programmable gate array (FPGA) using a hardware descriptor language (HDL) is herein referred to as a programmable chip or a programmable device. Instead of implementing complicated bus sharing schemes using mechanisms such as splitting, the bus itself can be eliminated to improve system performance.
According to specific embodiments, it is recognized that primary components and secondary components need not be routed through a construct such as a bus. By not routing signals through a bus, a streaming output device can be implemented in a much more efficient manner. The conventional data and address lines that make up the bus are no longer the resource in contention. Instead, secondary components are resources in contention, as the number of physical lines connected to each secondary component in a system can be fixed. Furthermore, by not using a bus, interconnection flexibility is enhanced. For example, an Ethernet component can be allocated a variety of ports for directly accessing a memory and the only resource in contention would be the memory.
Consequently, a system bus arbitrator associated with all the secondary components in a computer system is no longer needed. Instead, secondary components themselves that may be accessed by more than one primary component are assigned individual secondary side arbitrators. An arbitrator that corresponds to a specific secondary component accessible by more than one primary component is referred to herein as a secondary side arbitrator. In one embodiment, there is a secondary side arbitrator for each secondary component in a computer system. In other embodiments, there is a secondary side arbitrator for selected secondary components in a system. The techniques of the present invention recognize that Ethernet component support for high-level language programs can be more efficiently and effectively provided in a system by using secondary side arbitration.
According to various embodiments, a secondary side arbitrator 1651 allows a first secondary component in a system to be accessed by a first primary component at the same time a second secondary component in the system is accessed by a second primary component. For example, peripheral interface 1625 can be accessed by primary Ethernet 1615 through secondary side arbitrator 1651 at the same time, secondary UART 1621 is accessed by primary CPU 1613.
By allowing a CPU to access a secondary component at the same time another primary component such as a streaming output device or an Ethernet component is accessing memory, bus bottlenecks can be reduced. By using the simultaneous multiple primary component architecture, more direct connections between components can also be supported.
In one example, an input stage 1701 often allows selection and parameterization of components to be used on an electronic device. The input stage 1701 also allows configuration of variable or fixed latency support. In some examples, components provided to an input stage include intellectual property functions, megafunctions, and intellectual property cores. The input stage 1701 may be a graphical user interface using wizards for allowing efficient or convenient entry of information. The input stage may also be a text interface or a program reading a data file such as a spreadsheet, database table, or schematic to acquire selection information. The input stage 1701 produces an output containing information about the various modules selected.
In typical implementations, the generator program 1705 can identify the selections and generate a logic description with information for implementing the various modules. The generator program 1705 can be a Perl script creating HDL files such as Verilog, Abel, VHDL, and AHDL files from the module information entered by a user. In one example, the generator program identifies a portion of a high-level language program to accelerate. The other code is left for execution on a processor core. According to various embodiments, the generator program 1705 identifies pointers and provides ports for each pointer. One tool with generator program capabilities is System on a Programmable Chip (SOPC) Builder available from Altera Corporation of San Jose, Calif. The generator program 1705 also provides information to a synthesis tool 1707 to allow HDL files to be automatically synthesized. In some examples, a logic description is provided directly by a designer. Hookups between various components selected by a user are also interconnected by a generator program. Some of the available synthesis tools are Leonardo Spectrum, available from Mentor Graphics Corporation of Wilsonville, Oreg. and Synplify available from Synplicity Corporation of Sunnyvale, Calif. The HDL files may contain technology specific code readable only by a synthesis tool. The HDL files at this point may also be passed to a simulation tool 1709.
As will be appreciated by one of skill in the art, the input stage 1701, generator program 1705, and synthesis tool 1707 can be separate programs. The interface between the separate programs can be a database file, a log, or simply messages transmitted between the programs. For example, instead of writing a file to storage, the input stage 1701 can send messages directly to the generator program 1705 to allow the generator program to create a logic description. Similarly, the generator program can provide information directly to the synthesis tool instead of writing HDL files. Similarly, input stage 1701, generator program 1705, and synthesis tool 1707 can be integrated into a single program.
A user may select various modules and an integrated program can then take the user selections and output a logic description in the form of a synthesized netlist without intermediate files. Any mechanism for depicting the logic to be implemented on an electronic device is referred to herein as a logic description. According to various embodiments, a logic description is an HDL file such as a VHDL, Abel, AHDL, or Verilog file. A logic description may be in various stages of processing between the user selection of components and parameters to the final configuration of the device. According to other embodiments, a logic description is a synthesized netlist such as an Electronic Design Interchange Format Input File (EDF file). An EDF file is one example of a synthesized netlist file that can be output by the synthesis tool 1707.
A synthesis tool 1707 can take HDL files and output EDF files. Tools for synthesis allow the implementation of the logic design on an electronic device. Some of the available synthesis tools are Leonardo Spectrum, available from Mentor Graphics Corporation of Wilsonville, Oreg. and Synplify available from Synplicity Corporation of Sunnyvale, Calif. Various synthesized netlist formats will be appreciated by one of skill in the art.
A verification stage 1713 typically follows the synthesis stage 1707. The verification stage checks the accuracy of the design to ensure that an intermediate or final design realizes the expected requirements. A verification stage typically includes simulation tools and timing analysis tools. Tools for simulation allow the application of inputs and the observation of outputs without having to implement a physical device. Simulation tools provide designers with cost effective and efficient mechanisms for both functional and timing verification of a design. Functional verification involves the circuit's logical operation independent of timing considerations. Parameters such as gate delays are disregarded.
Timing verification involves the analysis of the design's operation with timing delays. Setup, hold, and other timing requirements for sequential devices such as flip-flops are confirmed. Some available simulation tools include Synopsys VCS, VSS, and Scirocco, available from Synopsys Corporation of Sunnyvale, Calif. and Cadence NC-Verilog and NC-VHDL available from Cadence Design Systems of San Jose, Calif. After the verification stage 1713, the synthesized netlist file can be provided to physical design tools 1719 including place and route and configuration tools. A place and route tool typically locates logic cells on specific logic elements of a target hardware device and connects wires between the inputs and outputs of the various logic elements in accordance with logic required to implement an electronic design. The device can also be physically tested at 1723.
For programmable logic devices, a programmable logic configuration stage can take the output of the place and route tool to program the logic device with the user selected and parameterized modules. According to various embodiments, the place and route tool and the logic configuration stage are provided in the Quartus Development Tool, available from Altera Corporation of San Jose, Calif. As will be appreciated by one of skill in the art, a variety of synthesis, place and route, and programmable logic configuration tools can be tested using various techniques of the present invention.
As noted above, different stages and programs can be integrated in a variety of manners. According to one embodiment, the input stage 1701, the generator program 1705, the synthesis tool 1707, the verification tools 1713, and physical design tools 1719 are integrated into a single program. The various stages are automatically run and transparent to a user. The program can receive the user selected modules, generate a logic description depicting logic for implementing the various selected modules, and implement the electronic device. As will be appreciated by one of skill in the art, HDL files and EDF files are mere examples of a logic description. Other file formats as well as internal program representations are other examples of a logic description.
Both of these memory devices may include any suitable type of the computer-readable media described above. A mass storage device 1808 is also coupled bi-directionally to CPU 1802 and provides additional data storage capacity and may include any of the computer-readable media described above. The mass storage device 1808 may be used to store programs, data and the like and is typically a secondary storage medium such as a hard disk that is slower than memory. The mass storage device 1808 can be used to hold a library or database of prepackaged logic or intellectual property functions, as well as information on generating particular configurations. It will be appreciated that the information retained within the mass storage device 1808, may, in appropriate cases, be incorporated in standard fashion as part of memory 1806 as virtual memory. A specific mass storage device such as a CD-ROM 1814 may also pass data uni-directionally to the CPU.
CPU 1802 is also coupled to an interface 1810 that includes one or more input/output devices such as such as video monitors, track balls, mice, keyboards, microphones, touch-sensitive displays, transducer card readers, magnetic or paper tape readers, tablets, styluses, voice or handwriting recognizers, or other well-known input devices such as, of course, other computers. Finally, CPU 1802 optionally may be coupled to a computer or telecommunications network using a network connection as shown generally at 1812. With such a network connection, it is contemplated that the CPU might receive information from the network, or might output information to the network in the course of performing the above-described method steps. It should be noted that the system 1800 may also be associated with devices for transferring completed designs onto a programmable chip. The above-described devices and materials will be familiar to those of skill in the computer hardware and software arts.
The hardware elements described above may be configured (usually temporarily) to act as multiple software modules for performing the operations of this invention. For example, instructions for running a generator program, input stage (e.g., a wizard), and/or compiler may be stored on mass storage device 1808 or 1814 and executed on CPU 1808 in conjunction with primary memory 1806.
Although many of the components and processes are described above in the singular for convenience, it will be appreciated by one of skill in the art that multiple components and repeated processes can also be used to practice the techniques of the present invention.
While the invention has been particularly shown and described with reference to specific embodiments thereof, it will be understood by those skilled in the art that changes in the form and details of the disclosed embodiments may be made without departing from the spirit or scope of the invention. For example, embodiments of the present invention may be employed with a variety of processors and instruction lengths and should not be restricted to the ones mentioned above. It is therefore intended that the invention be interpreted to include all variations and equivalents that fall within the true spirit and scope of the present invention.
This application is a continuation of and claims benefit of priority under 35 U.S.C. 120 of U.S. patent application Ser. No. 10/918,117 filed on Aug. 12, 2004 now U.S. Pat. No. 7,487,196, which claims benefit of priority under 35 U.S.C. 119(e) of U.S. Provisional Application No. 60/530,445 filed Dec. 16, 2003 and titled “METHODS AND APPARATUS FOR IMPLEMENTING A SATURATING MULTIPLIER”, the entireties of both applications are incorporated by reference for all purposes.
Number | Name | Date | Kind |
---|---|---|---|
6209017 | Lim et al. | Mar 2001 | B1 |
20050027773 | Machnicki et al. | Feb 2005 | A1 |
Number | Date | Country | |
---|---|---|---|
20090172067 A1 | Jul 2009 | US |
Number | Date | Country | |
---|---|---|---|
60530445 | Dec 2003 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 10918117 | Aug 2004 | US |
Child | 12346692 | US |