MICROCHIP -logo

MICROCHIP CoreFPU Core Floating Point Unit

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit-produxt

 

Introduction 

  • The Core Floating Point Unit (CoreFPU) is designed for floating-point arithmetic and conversion operations, for single and double precision floating-point numbers. CoreFPU supports fixed-point to floating-point and floating-point to fixed-point conversions and floating-point addition, subtraction, and multiplication operations. The IEEE® Standard for Floating-Point Arithmetic (IEEE 754) is a technical standard for floating-point computation.
  • Important: CoreFPU supports calculations with normalized numbers only, and only the Verilog language is supported; VHDL is not supported.

Summary
The following table provides a summary of the CoreFPU characteristics.

Table 1. CoreFPU Characteristics 

Core Version This document applies to CoreFPU v3.0.
Supported Device Families
  • PolarFire® SoC
  • PolarFire
  • RTG4™
Supported Tool Flow Requires Libero® SoC v12.6 or later releases.
Licensing CoreFPU is not license locked.
Installation Instructions CoreFPU must be installed to the IP Catalog of Libero SoC automatically through the IP Catalog update function. Alternatively, CoreFPU could be manually downloaded from the catalog. Once the IP core is

installed, it is configured, generated and instantiated within SmartDesign for inclusion in the project.

Device Utilization and Performance A summary of utilization and performance information for CoreFPU is listed in Device Resource Utilization and Performance.

CoreFPU Change Log Information
This section provides a comprehensive overview of the newly incorporated features, beginning with the most recent release. For more information about the problems resolved, see the Resolved Issues section.

Version What’s New
v3.0 Implemented additional output flags to enhance the accuracy of the IP
v2.1 Added the double precision feature
v2.0 Updated the timing waveforms
v1.0 First production release of CoreFPU

1. Features

CoreFPU has the following key features:

  • Supports Single and Double Precision Floating Numbers as per IEEE-754 Standard
  • Supports Conversions as listed:
    • Fixed-point to Floating-point conversion
    • Floating-point to Fixed-point conversion
  • Supports Arithmetic Operations as listed:
    • Floating-point addition
    • Floating-point subtraction
    • Floating-point multiplication
  • Provides the Rounding Scheme (Round to nearest even) for the Arithmetic Operations only
  • Provides Flags for Overflow, Underflow, Infinity (Positive Infinity, Negative Infinity), Quiet NaN (QNaN) and Signalling NaN (SNaN) for Floating-Point Numbers.
  • Supports Fully pipelined implementation of Arithmetic Operations
  • Provides Provision to configure the Core for Design Requirements

Functional Description

  • The IEEE Standard for Floating-Point Arithmetic (IEEE 754) is a technical standard for floating-point computation. The term floating-point refers to the radix point of the number (decimal point or binary point), which is placed anywhere with respect to the significant digits of the number.
    A floating-point number is typically expressed in the scientific notation, with a fraction (F), and an exponent (E) of a certain radix (r), in the form of F × r^E. Decimal numbers use radix of 10 (F × 10^E); while binary numbers use radix of 2 (F × 2^E).
  • The representation of the floating-point number is not unique. For example, the number 55.66 is represented as 5.566 × 10^1, 0.5566 × 10^2, 0.05566 × 10^3, and so on. The fractional part is normalized. In the normalized form, there is only a single non-zero digit before the radix point. For example, decimal number 123.4567 is normalized as 1.234567 × 10^2; binary number 1010.1011B is normalized as 1.0101011B × 2^3.
  • It is important to note that floating-point numbers suffer from loss of precision when represented with a fixed number of bits (for example, 32-bit or 64-bit). This is because there are an infinite number of real numbers (even within a small range from 0.0 to 0.1). On the other hand, an
    n- bit binary pattern represents a finite 2^n distinct numbers. Hence, not all the real numbers are represented. The nearest approximation is used instead, which results in the loss of accuracy.

The single precision floating-point number is represented as follows:

  • Sign bit: 1-bit
  • Exponent width: 8 bits
  • Significand precision: 24 bits (23 bits are explicitly stored)

Figure 2-1. 32-bit Frame

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (2)The double precision floating-point number is represented as follows:

  • Sign bit: 1-bit
  • Exponent width: 11 bits
  • Significand precision: 53 bits (52 bits are explicitly stored)

Figure 2-2. 64-bit Frame MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (3)The CoreFPU is the top-level integration of the two conversion modules (Fixed to Float point and Float to Fixed point) and three arithmetic operations (FP ADD, FP SUB, and FP MULT). The user can configure any one of the operations based on the requirement so that the resources are utilized for the selected operation.
The following figure shows the top level CoreFPU block diagram with ports.

Figure 2-3. CoreFPU Ports Block Diagram

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (4)The following table lists the width of the Input and Output ports. Table 2-1. Input and Output Port Width

Signal Single Precision Width Double Precision Width
ain [31:0] [63:0]
bin [31:0] [63:0]
aout [31:0] [63:0]
pout [31:0] [63:0]

Fixed-Point to Floating-Point (Conversion)

CoreFPU configured as fixed to floating-point infers the fixed-point to floating-point conversion module. The input (ain) to CoreFPU is any fixed-point number containing the integer and fractional bits. The CoreFPU configurator has the options to select the input integer and fraction widths. The input is valid on di_valid signal and output is valid on do_valid. The output (aout) of the fixed to float operation is in single or double precision floating-point format.
Example for fixed-point to floating-point conversion operation is listed in the following table.
Table 2-2. Example for Fixed-Point to Floating-Point Conversion

Fixed-Point Number Floating-Point Number
ain Integer Fraction aout Sign Exponent Mantissa
0x12153524 (32-bit) 00010010000101010 011010100100100 0x4610a9a9 0 10001100 00100001010100110101001
0x0000000000008CCC

(64-bit)

0000000000000000000000000000000000000000000000001 000110011001100 0x3FF199999999999A 0 01111111111 0001100110011001100110011001100110011001100110011010

Floating-Point to Fixed-Point (Conversion) 
CoreFPU configured as floating to fixed-point infers the floating-point to fixed-point conversion module. The input (ain) to CoreFPU is any single or double precision floating-point number and produces an output (aout) in fixed-point format containing integer and fractional bits. The input is valid on di_valid signal and output is valid on do_valid. The CoreFPU configurator has the options to select the output integer and fraction widths.
Example for floating-point to fixed-point conversion operation is listed in the following table.

Table 2-3. Example for Floating-Point to Fixed-Point Conversion

Floating-Point Number Fixed-Point Number
ain Sign Exponent Mantissa aout Integer Fraction
0x41bd6783 (32-bit) 0 10000011 01111010110011110000011 0x000bd678 00000000000010111 101011001111000
0x4002094c447c30d3

(64-bit)

0 10000000000 0010000010010100110001000100011111000011000011010011 0x0000000000012095 0000000000000000000000000000000000000000000000010 010000010010101

Floating-Point Addition (Arithmetic Operation)
CoreFPU configured as FP ADD infers the floating-point addition module. It adds the two floating-point numbers (ain and bin) and provides the output (pout) in floating-point format. The input and output are single or double precision floating-point numbers. The input is valid on di_valid signal and output is valid on do_valid. The core produce ovfl_fg (Overflow), qnan_fg (Quiet Not a Number), snan_fg (Signalling Not a Number), pinf_fg(Positive Infinity), and ninf_fg (Negative Infinity) flags based on the addition operation.
Examples for floating-point addition operation are listed in the following tables.
Table 2-4. Example for Floating-Point Addition Operation (32-bit)

Floating-Point Value Sign Exponent Mantissa
Floating-point input 1 ain (0x4e989680) 0 10011101 00110001001011010000000
Floating-point input 2 bin (0x4f191b40) 0 10011110 00110010001101101000000
Floating-point addition output pout (0x4f656680) 0 10011110 11001010110011010000000

Table 2-5. Example for Floating-Point Addition Operation (64-bit)

Floating-Point Value Sign Exponent Mantissa
Floating-point input 1

ain (0x3ff4106ee30caa32)

0 01111111111 0100000100000110111011100011000011001010101000110010
Floating-point input 2

bin (0x40020b2a78798e61)

0 10000000000 0010000010110010101001111000011110011000111001100001
Floating-point addition output pout (0x400c1361e9ffe37a) 0 10000000000 1100000100110110000111101001111111111110001101111010

Floating-Point Subtraction (Arithmetic Operation) 
CoreFPU configured as FP SUB infers the floating-point subtraction module. It subtracts the two floating-point numbers (ain and bin) and provides the output (pout) in floating-point format. The input and output are single or double precision floating-point numbers. The input is valid on di_valid signal and output is valid on do_valid. The core produce ovfl_fg (Overflow), unfl_fg (underflow), qnan_fg (Quiet Not a Number), snan_fg (Signalling Not a Number), pinf_fg (Positive Infinity), and ninf_fg (Negative Infinity) flags based on the subtraction operation.
Examples for floating-point subtraction operation are listed in the following tables.
Table 2-6. Example for Floating-Point Subtraction Operation (32-bit)

Floating-Point Value Sign Exponent Mantissa
Floating-point input 1 ain (0xac85465f) 1 01011001 00001010100011001011111
Floating-point input 2 bin (0x2f516779) 0 01011110 10100010110011101111001
Floating-point subtraction output pout (0xaf5591ac) 1 01011110 10101011001000110101011
Floating-Point Value Sign Exponent Mantissa
Floating-point input 1

ain (0x405569764adff823)

0 10000000101 0101011010010111011001001010110111111111100000100011
Floating-point input 2

bin (0x4057d04e78dee3fc)

0 10000000101 0111110100000100111001111000110111101110001111111100
Floating-point subtraction output pout (0xc02336c16ff75ec8) 1 10000000010 0011001101101100000101101111111101110101111011001000

Floating-Point Multiplication (Arithmetic Operation)
CoreFPU configured as FP MULT infers the floating-point multiplication module. It multiplies the two floating-point numbers (ain and bin) and provides the output (pout) in floating-point format. The input and output are single or double precision floating-point numbers. The input is valid on di_valid signal and output is valid on do_valid. The core produce ovfl_fg (Overflow), unfl_fg (Underflow), qnan_fg (Quiet Not a Number), snan_fg (Signalling Not a Number), pinf_fg (Positive Infinity), and ninf_fg (Negative Infinity) flags based on the multiplication operation.
Examples for floating-point multiplication operation are listed in the following tables.
Table 2-8. Example for Floating-Point Multiplication Operation (32-bit)

Floating-Point Value Sign Exponent Mantissa
Floating-point input 1 ain (0x1ec7a735) 0 00111101 10001111010011100110101
Floating-point input 2 bin (0x6ecf15e8) 0 11011101 10011110001010111101000
Floating-point Multiplication output pout (0x4e21814a) 0 10011100 01000011000000101001010
Floating-Point Value Sign Exponent Mantissa
Floating-point input 1

ain (0x40c1f5a9930be0df)

0 10000001100 0001111101011010100110010011000010111110000011011111
Floating-point input 2

bin (0x400a0866c962b501)

0 10000000000 1010000010000110011011001001011000101011010100000001
Floating-point multiplication output pout (0x40dd38a1c3e2cae9) 0 10000001101 1101001110001010000111000011111000101100101011101001

 Truth Table for Addition and Subtraction 
The following truth tables list the values for addition and subtraction operation. Table 2-10. Truth Table for Addition

Data A Data B Sign Bit Result Overflow Underflow SNaN QNaN PINF NINF
QNaN/SNaN x 0 POSQNaN 0 0 0 1 0 0
x QNaN/SNaN 0 POSQNaN 0 0 0 1 0 0
zero zero 0 POSZERO 0 0 0 0 0 0
zero posfinite(y) 0 posfinite(y) 0 0 0 0 0 0
zero negfinite(y) 1 negfinite(y) 0 0 0 0 0 0
zero posinfinite 0 posinfinite 0 0 0 0 1 0
zero neginfinite 1 neginfinite 0 0 0 0 0 1
posfinite(y) zero 0 posfinite(y) 0 0 0 0 0 0
posfinite posinfinite 0 posinfinite 0 0 0 0 1 0
Table 2-10. Truth Table for Addition (continued)
Data A Data B Sign Bit Result Overflow Underflow SNaN QNaN PINF NINF
posfinite neginfinite 1 neginfinite 0 0 0 0 0 1
negfinite(y) zero 1 negfinite(y) 0 0 0 0 0 0
negfinite posinfinite 0 posinfinite 0 0 0 0 1 0
negfinite neginfinite 1 neginfinite 0 0 0 0 0 1
posinfinite zero 0 posinfinite 0 0 0 0 1 0
posinfinite posfinite 0 posinfinite 0 0 0 0 1 0
posinfinite negfinite 0 posinfinite 0 0 0 0 1 0
posinfinite posinfinite 0 posinfinite 0 0 0 0 1 0
posinfinite neginfinite 0 POSQNaN 0 0 0 1 0 0
neginfinite zero 1 neginfinite 0 0 0 0 0 1
neginfinite posfinite 1 neginfinite 0 0 0 0 0 1
neginfinite negfinite 1 neginfinite 0 0 0 0 0 1
neginfinite posinfinite 0 POSQNaN 0 0 0 1 0 0
neginfinite neginfinite 1 neginfinite 0 0 0 0 0 1
posfinite posfinite 0 posfinite 0 0 0 0 0 0
posfinite posfinite 0 posinfinite 0 0 0 0 1 0
posfinite posfinite 0/1 QNaN 0 0 0 1 0 0
posfinite posfinite 0/1 SNaN 0 0 1 0 0 0
posfinite posfinite 0 POSSNaN 1 0 1 0 0 0
posfinite negfinite 0 posfinite 0 0 0 0 0 0
posfinite negfinite 1 negfinite 0 0 0 0 0 0
posfinite negfinite 0 POSSNaN 0 1 1 0 0 0
negfinite posfinite 0 posfinite 0 0 0 0 0 0
negfinite posfinite 1 negfinite 0 0 0 0 0 0
negfinite posfinite 0 POSSNaN 0 1 1 0 0 0
negfinite negfinite 1 negfinite 0 0 0 0 0 0
negfinite negfinite 1 neginfinite 0 0 0 0 0 1
negfinite negfinite 0/1 QNaN 0 0 0 1 0 0
negfinite negfinite 0/1 SNaN 0 0 1 0 0 0
negfinite negfinite 0 POSSNaN 1 0 1 0 0 0
Data A Data B Sign Bit Result Overflow Underflow SNaN QNaN PINF NINF
QNaN/SNaN x 0 POSQNaN 0 0 0 1 0 0
x QNaN/SNaN 0 POSQNaN 0 0 0 1 0 0
zero zero 0 POSZERO 0 0 0 0 0 0
zero posfinite(y) 1 negfinite(y) 0 0 0 0 0 0
zero negfinite(y) 0 posfinite(y) 0 0 0 0 0 0
zero posinfinite 1 neginfinite 0 0 0 0 0 1
zero neginfinite 0 posinfinite 0 0 0 0 1 0
posfinite(y) zero 0 posfinite(y) 0 0 0 0 0 0
posfinite posinfinite 1 neginfinite 0 0 0 0 0 1
posfinite neginfinite 0 posinfinite 0 0 0 0 1 0
negfinite(y) zero 1 negfinite(y) 0 0 0 0 0 0
negfinite posinfinite 1 neginfinite 0 0 0 0 0 1
Table 2-11. Truth Table for Subtraction (continued)
Data A Data B Sign Bit Result Overflow Underflow SNaN QNaN PINF NINF
negfinite neginfinite 0 posinfinite 0 0 0 0 1 0
posinfinite zero 0 posinfinite 0 0 0 0 1 0
posinfinite posfinite 0 posinfinite 0 0 0 0 1 0
posinfinite negfinite 0 posinfinite 0 0 0 0 1 0
posinfinite posinfinite 0 POSQNaN 0 0 0 1 0 0
posinfinite neginfinite 0 posinfinite 0 0 0 0 1 0
neginfinite zero 1 neginfinite 0 0 0 0 0 1
neginfinite posfinite 1 neginfinite 0 0 0 0 0 1
neginfinite negfinite 1 neginfinite 0 0 0 0 0 1
neginfinite posinfinite 1 neginfinite 0 0 0 0 0 1
neginfinite neginfinite 0 POSQNaN 0 0 0 1 0 0
posfinite posfinite 0 posfinite 0 0 0 0 0 0
posfinite posfinite 1 negfinite 0 0 0 0 0 0
posfinite posfinite 0 POSSNaN 0 1 1 0 0 0
posfinite negfinite 0 posfinite 0 0 0 0 0 0
posfinite negfinite 0 posinfinite 0 0 0 0 1 0
posfinite negfinite 0/1 QNaN 0 0 0 1 0 0
posfinite negfinite 0/1 SNaN 0 0 1 0 0 0
posfinite negfinite 0 POSSNaN 1 0 1 0 0 0
negfinite posfinite 1 negfinite 0 0 0 0 0 0
negfinite posfinite 1 neginfinite 0 0 0 0 0 1
negfinite posfinite 0/1 QNaN 0 0 0 1 0 0
negfinite posfinite 0/1 SNaN 0 0 1 0 0 0
negfinite posfinite 0 POSSNaN 1 0 1 0 0 0
negfinite negfinite 0 posfinite 0 0 0 0 0 0
negfinite negfinite 1 negfinite 0 0 0 0 0 0
negfinite negfinite 0 POSSNaN 0 1 1 0 0 0

Important:

  • They in the preceding tables denotes any number.
  • The in the preceding tables denotes a don’t care condition.

Truth Table for Multiplication 
The following truth table lists the values for multiplication operation.

Table 2-12. Truth Table for Multiplication

Data A Data B Sign Bit Result Overflow Underflow SNaN QNaN PINF NINF
QNaN/SNaN x 0 POSQNaN 0 0 0 1 0 0
x QNaN/SNaN 0 POSQNaN 0 0 0 1 0 0
zero zero 0 POSZERO 0 0 0 0 0 0
zero posfinite 0 POSZERO 0 0 0 0 0 0
zero negfinite 0 POSZERO 0 0 0 0 0 0
zero posinfinite 0 POSQNaN 0 0 0 1 0 0
zero neginfinite 0 POSQNaN 0 0 0 1 0 0
Table 2-12. Truth Table for Multiplication (continued)
Data A Data B Sign Bit Result Overflow Underflow SNaN QNaN PINF NINF
posfinite zero 0 POSZERO 0 0 0 0 0 0
posfinite posinfinite 0 posinfinite 0 0 0 0 1 0
posfinite neginfinite 1 neginfinite 0 0 0 0 0 1
negfinite zero 0 POSZERO 0 0 0 0 0 0
negfinite posinfinite 1 neginfinite 0 0 0 0 0 1
negfinite neginfinite 0 posinfinite 0 0 0 0 1 0
posinfinite zero 0 POSQNaN 0 0 0 1 0 0
posinfinite posfinite 0 posinfinite 0 0 0 0 1 0
posinfinite negfinite 1 neginfinite 0 0 0 0 0 1
posinfinite posinfinite 0 posinfinite 0 0 0 0 1 0
posinfinite neginfinite 1 neginfinite 0 0 0 0 0 1
neginfinite zero 0 POSQNaN 0 0 0 1 0 0
neginfinite posfinite 1 neginfinite 0 0 0 0 0 1
neginfinite negfinite 0 posinfinite 0 0 0 0 1 0
neginfinite posinfinite 1 neginfinite 0 0 0 0 0 1
neginfinite neginfinite 0 posinfinite 0 0 0 0 1 0
posfinite posfinite 0 posfinite 0 0 0 0 0 0
posfinite posfinite 0 posinfinite 0 0 0 0 1 0
posfinite posfinite 0 POSQNaN 0 0 0 1 0 0
posfinite posfinite 0 POSSNaN 0 0 1 0 0 0
posfinite posfinite 0 POSSNaN 1 0 1 0 0 0
posfinite posfinite 0 POSSNaN 0 1 1 0 0 0
posfinite negfinite 1 negfinite 0 0 0 0 0 0
posfinite negfinite 1 neginfinite 0 0 0 0 0 1
posfinite negfinite 0 POSQNaN 0 0 0 1 0 0
posfinite negfinite 0 POSSNaN 0 0 1 0 0 0
posfinite negfinite 0 POSSNaN 1 0 1 0 0 0
posfinite negfinite 0 POSSNaN 0 1 1 0 0 0
negfinite posfinite 1 negfinite 0 0 0 0 0 0
negfinite posfinite 1 neginfinite 0 0 0 0 0 1
negfinite posfinite 0 POSQNaN 0 0 0 1 0 0
negfinite posfinite 0 POSSNaN 0 0 1 0 0 0
negfinite posfinite 0 POSSNaN 1 0 1 0 0 0
negfinite posfinite 0 POSSNaN 0 1 1 0 0 0
negfinite negfinite 0 posfinite 0 0 0 0 0 0
negfinite negfinite 0 posinfinite 0 0 0 0 1 0
negfinite negfinite 0 POSQNaN 0 0 0 1 0 0
negfinite negfinite 0 POSQNaN 0 0 1 0 0 0
negfinite negfinite 0 POSQNaN 1 0 1 0 0 0
negfinite negfinite 0 POSQNaN 0 1 1 0 0 0

Important:

Sign Bit ‘0’ defines positive output and ‘1’ defines negative output.
The x in the preceding table denotes don’t care condition.

CoreFPU Parameters and Interface Signals
This section discusses the parameters in the CoreFPU Configurator settings and I/O signals.

Configuration GUI Parameters 
There are number of configurable options that apply to the FPU unit as shown in the following table. If a configuration other than default is required, configuration dialog box is used to select appropriate values for the configurable option.

Table 3-1. CoreFPU Configuration GUI Parameters 

Parameter Name Default Description
Precision Single Select the operation as required:

Single Precision
Double Precision

Conversion Type Fixed-point to Floating-point conversion Select the operation as required:
  • Fixed-point to Floating-point conversion
  • Floating-point to Fixed-point conversion
  • Floating-point addition
  • Floating-point subtraction
  • Floating-point multiplication
Input Fraction Width1 15 Configures the fractional point in the Input ain and bin signals

Valid range is 31–1

Output Fraction Width2 15 Configures the fractional point in the Output aout signals

Valid range is 51–1

Important:

  1. This parameter is configurable only during fixed-point to floating-point conversion.
  2. This parameter is configurable only during floating-point to fixed-point conversion.

Input and Output Signals (Ask a Question)
The following table lists the input and output port signals of CoreFPU.

Table 3-2. Port Description 

Signal Name Width Type Description
clk 1 Input Main system clock
rstn 1 Input Active-low asynchronous reset
di_valid 1 Input Active-high input valid

This signal indicates that the data present on ain[31:0], ain[63:0] and bin[31:0], bin[63:0] is valid.

ain 32/64 Input A Input Bus (It is used for all operations)
bin1 32/64 Input B Input Bus (It is used for arithmetic operations only)
aout2 32/64 Output Output value when fixed to floating-point or floating to fixed-point conversion operations are selected.
pout1 32/64 Output Output value when addition, subtraction, or multiplication operations are selected.
Table 3-2. Port Description (continued)
Signal Name Width Type Description
do_valid 1 Output Active-high signal

This signal indicates that the data present on pout/aout data bus is valid.

ovfl_fg3 1 Output Active-high signal

This signal indicates the overflow during floating-point operations.

unfl_fg 1 Output Active-high signal

This Signal indicates the underflow during floating point operations.

qnan_fg3 1 Output Active-high signal

This signal indicates the Quiet Not a Number (QNaN) during floating-point operations.

snan_fg 1 Output Active-high signal

This signal indicates the Signalling Not-a-Number (SNaN) during floating point operations.

pinf_fg3 1 Output Active-high signal

This signal indicates the positive infinity during floating-point operations.

ninf_fg 1 Output Active-high signal

This signal indicates the negative infinity during floating-point operations.

Important:

  1. This port is available only for floating-point addition, subtraction, or multiplication operations.
  2. This port is available only for fixed-point to floating-point and floating-point to fixed-point conversion operations.
  3. This port is available for floating-point to fixed-point, floating-point addition, floating-point subtraction, and floating-point multiplication.

Implementation of CoreFPU in Libero Design Suite

This section describes the implementation of CoreFPU in the Libero Design Suite.

SmartDesign 

CoreFPU is available for download in the Libero IP catalog through the web repository. Once it is listed in the catalog, the core is instantiated using the SmartDesign flow. For information on using SmartDesign to configure, connect, and generate cores, see Libero SoC online help.
After configuring and generating the core instance, the basic functionality is simulated using the testbench supplied with the CoreFPU. The testbench parameters automatically adjust to the CoreFPU configuration. The CoreFPU is instantiated as a component of a larger design.
Figure 4-1. SmartDesign CoreFPU Instance for Arithmetic Operations

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (5)Figure 4-2. SmartDesign CoreFPU Instance for Conversion Operation MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (6)

 

Fixed-Point to Floating-Point Conversion
During fixed-point to floating-point conversion, the Input Fraction Width is configurable. The Output Width is set to 32-bit for single precision and 64-bit for double precision floating-point by default.
To convert from fixed-point to floating-point, select Fixed to floating point Conversion type, as shown in the following figure.

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (7)Floating-Point to Fixed-Point 
During floating-point to fixed-point conversion, the Output Fractional Width is configurable, and the Input Width is set to 32-bit for single precision and 64-bit for double precision floating-point by default.
To convert from floating-point to fixed-point, select Floating point to fixed Conversion type, as shown in the following figure.
Figure 4-4. CoreFPU Configurator for Floating Point to Fixed MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (8)Floating-Point Addition/Subtraction/Multiplication
During floating-point addition, subtraction, and multiplication operation, the Input Fraction Width and Output Fraction Width are not configurable as these are floating-point arithmetic operations, and the Input/Output Width is set to 32-bit single precision and 64-bit for double precision floating-point by default.
The following figure shows the CoreFPU configurator for floating point subtraction operation.

Figure 4-5. CoreFPU Configurator for Floating Point SubtractionMICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (9)Simulation (Ask a Question)
To run simulations, in the core configuration window, select User Testbench. After generating the CoreFPU, the pre-synthesis testbench Hardware Description Language (HDL) files are installed in Libero.

Simulation Waveforms (Ask a Question)
This section discusses the simulation waveforms for CoreFPU.
The following figures show the waveform of fixed-point to floating-point conversion for both 32-bit and 64-bit.MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (10)

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (11) MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (12) MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (13)

System Integration
The following figure shows an example of using the core. In this example, the design UART is used as a communication channel between the design and the host PC. The signals ain and bin (each of 32-bit or 64-bit width) are the inputs to the design from UART. After the CoreFPU receives the di_valid signal, it computes the result. After computing the result, the do_valid signal goes high and stores the result (aout/pout data) in the output buffer. This same procedure is applicable for conversion and arithmetic operations. For conversion operations, only input ain is sufficient whereas for arithmetic operations, both ain and bin inputs are required. Output aout is enabled for conversion operations and pout port is enabled for arithmetic operations.
Figure 4-16. Example of the CoreFPU System

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (14)

 

  1. Synthesis (Ask a Question)
    To run synthesis on the CoreFPU, set the design root to the IP component instance and from the Libero design flow pane, run the Synthesis tool.
    Place and Route (Ask a Question)
    After the design is synthesized, run the Place-and-Route tool. CoreFPU requires no special placeand- route settings.
  2. User Testbench (Ask a Question)
    A user testbench is provided with the CoreFPU IP release. Using this testbench, you can verify functional behavior of CoreFPU.

A simplified block diagram of the user testbench is shown in the following figure. The user testbench instantiates the Configured CoreFPU design (UUT), and includes behavioral test data generator, necessary clock, and reset signals.
Figure 4-17. CoreFPU User Testbench

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (15)Important: You have to monitor the output signals in ModelSim simulator, see Simulation section.

Additional References (Ask a Question)
This section provides a list for additional information.
For updates and additional information about the software, devices, and hardware, visit the

Intellectual Property pages on the Microchip FPGAs and PLDs website.

  1. Known Issues and Workarounds (Ask a Question)
    There are no known issues and workarounds for CoreFPU v3.0.
  2. Discontinued Features and Devices (Ask a Question)
    There are no discontinued features and devices with this IP release.

Glossary

The following are the list of terms and definitions used in the document.
Table 6-1. Terms and Definitions

Term Definition
FPU Floating Point Unit
FP ADD Floating-Point Addition
FP SUB Floating-Point Subtraction
FP MULT Floating-Point Multiplication

Resolved Issues 
The following table lists all the resolved issues for the various CoreFPU releases.

Table 7-1. Resolved Issues

Release Description
3.0 The following is the list of all resolved issues in the v3.0 release:

Case Number: 01420387 and 01422128

Added the rounding scheme logic (round to the nearest even number).

2.1 The following is the list of all resolved issues in the v2.1 release:
The design encounters issues due to the presence of duplicate modules when multiple cores are instantiated.
Renaming the CoreFPU IP instance results in an “Undefined module” error.
1.0 Initial Release

Device Resource Utilization and Performance

The CoreFPU macro is implemented in the families listed in the following table.
Table 8-1. FPU PolarFire Unit Device Utilization for 32-Bit

FPGA Resources Utilization
Family 4LUT DFF Total Math Block Device Percentage Performance Latency
Fixed-Point to Floating-Point
PolarFire® 260 104 364 0 MPF300T 0.12 310 MHz 3
Floating-Point to Fixed-Point
PolarFire 591 102 693 0 MPF300T 0.23 160 MHz 3
Floating-Point Addition
PolarFire 1575 1551 3126 0 MPF300T 1.06 340 MHz 16
Floating-Point Subtraction
PolarFire 1561 1549 3110 0 MPF300T 1.04 345 MHz 16
Floating-Point Multiplication
PolarFire 465 847 1312 4 MPF300T 0.44 385 MHz 14
FPGA Resources Utilization
Family 4LUT DFF Total Math Block Device Percentage Performance Latency
Fixed-Point to Floating-Point
RTG4™ 264 104 368 0 RT4G150 0.24 160 MHz 3
Floating-Point to Fixed-Point
RTG4 439 112 551 0 RT4G150 0.36 105 MHz 3
Floating-Point Addition
RTG4 1733 1551 3284 0 RT4G150 1.16 195 MHz 16
Floating-Point Subtraction
RTG4 1729 1549 3258 0 RT4G150 1.16 190 MHz 16
Floating-Point Multiplication
RTG4 468 847 1315 4 RT4G150 0.87 175 MHz 14
FPGA Resources Utilization
Family 4LUT DFF Total Math Block Device Percentage Performance Latency
Fixed-Point to Floating-Point
PolarFire® 638 201 849 0 MPF300T 0.28 305 MHz 3
Floating-Point to Fixed-Point
PolarFire 2442 203 2645 0 MPF300T 0.89 110 MHz 3
Floating-Point Addition
PolarFire 5144 4028 9172 0 MPF300T 3.06 240 MHz 16
Floating-Point Subtraction
PolarFire 5153 4026 9179 0 MPF300T 3.06 250 MHz 16
Floating-Point Multiplication
PolarFire 1161 3818 4979 16 MPF300T 1.66 340 MHz 27
FPGA Resources Utilization
Family 4LUT DFF Total Math Block Device Percentage Performance Latency
Fixed-Point to Floating-Point
RTG4™ 621 201 822 0 RT4G150 0.54 140 MHz 3
Floating-Point to Fixed-Point
RTG4 1114 203 1215 0 RT4G150 0.86 75 MHz 3
Floating-Point Addition
RTG4 4941 4028 8969 0 RT4G150 5.9 140 MHz 16
Floating-Point Subtraction
RTG4 5190 4026 9216 0 RT4G150 6.07 130 MHz 16
Floating-Point Multiplication
RTG4 1165 3818 4983 16 RT4G150 3.28 170 MHz 27

Important: To increase the frequency, select Enable retiming option in synthesis setting.

Revision History

The revision history describes the changes that were implemented in the document. The changes are listed by revision, starting with the most current publication.

MICROCHIP-CoreFPU-Cor-Floating-Point-Unit- (1)

Microchip FPGA Support

Microchip FPGA products group backs its products with various support services, including Customer Service, Customer Technical Support Center, a website, and worldwide sales offices. Customers are suggested to visit Microchip online resources prior to contacting support as it is very likely that their queries have been already answered.
Contact Technical Support Center through the website at www.microchip.com/support. Mention the FPGA Device Part number, select appropriate case category, and upload design files while creating a technical support case.
Contact Customer Service for non-technical product support, such as product pricing, product upgrades, update information, order status, and authorization.

  • From North America, call 800.262.1060
  • From the rest of the world, call 650.318.4460
  • Fax, from anywhere in the world, 650.318.8044

Microchip Information

Trademarks
The “Microchip” name and logo, the “M” logo, and other names, logos, and brands are registered and unregistered trademarks of Microchip Technology Incorporated or its affiliates and/or subsidiaries in the United States and/or other countries (“Microchip Trademarks”). Information regarding Microchip Trademarks can be found at https://www.microchip.com/en-us/about/legal-information/microchip-trademarks
ISBN: 979-8-3371-0947-3

Legal Notice
This publication and the information herein may be used only with Microchip products, including to design, test, and integrate Microchip products with your application. Use of this information in any other manner violates these terms. Information regarding device applications is provided only for your convenience and may be superseded by updates. It is your responsibility to ensure that your application meets with your specifications. Contact your local Microchip sales office for additional support or, obtain additional support at www.microchip.com/en-us/support/design-help/client-support-services

THIS INFORMATION IS PROVIDED BY MICROCHIP “AS IS”. MICROCHIP MAKES NO REPRESENTATIONS OR WARRANTIES OF ANY KIND WHETHER EXPRESS OR IMPLIED, WRITTEN OR ORAL, STATUTORY OR OTHERWISE, RELATED TO THE INFORMATION INCLUDING BUT NOT LIMITED TO ANY IMPLIED WARRANTIES OF NON-INFRINGEMENT, MERCHANTABILITY, AND FITNESS FOR A PARTICULAR PURPOSE, OR WARRANTIES RELATED TO ITS CONDITION, QUALITY, OR PERFORMANCE.

IN NO EVENT WILL MICROCHIP BE LIABLE FOR ANY INDIRECT, SPECIAL, PUNITIVE, INCIDENTAL, OR CONSEQUENTIAL LOSS, DAMAGE, COST, OR EXPENSE OF ANY KIND WHATSOEVER RELATED TO THE INFORMATION OR ITS USE, HOWEVER CAUSED, EVEN IF MICROCHIP HAS BEEN ADVISED OF THE POSSIBILITY OR THE DAMAGES ARE FORESEEABLE. TO THE FULLEST EXTENT ALLOWED BY LAW, MICROCHIP’S TOTAL LIABILITY ON ALL CLAIMS IN ANY WAY RELATED TO THE INFORMATION OR ITS USE WILL NOT EXCEED THE AMOUNT OF FEES, IF ANY, THAT YOU HAVE PAID DIRECTLY TO MICROCHIP FOR THE INFORMATION.

Use of Microchip devices in life support and/or safety applications is entirely at the buyer’s risk, and the buyer agrees to defend, indemnify and hold harmless Microchip from any and all damages, claims, suits, or expenses resulting from such use. No licenses are conveyed, implicitly or otherwise, under any Microchip intellectual property rights unless otherwise stated.

Microchip Devices Code Protection Feature
Note the following details of the code protection feature on Microchip products:

  • Microchip products meet the specifications contained in their particular Microchip Data Sheet.
  • Microchip believes that its family of products is secure when used in the intended manner, within operating specifications, and under normal conditions.
  • Microchip values and aggressively protects its intellectual property rights. Attempts to breach the code protection features of Microchip products are strictly prohibited and may violate the Digital Millennium Copyright Act.
  • Neither Microchip nor any other semiconductor manufacturer can guarantee the security of its code. Code protection does not mean that we are guaranteeing the product is “unbreakable”. Code protection is constantly evolving. Microchip is committed to continuously improving the code protection features of our products.

Documents / Resources

MICROCHIP CoreFPU Core Floating Point Unit [pdf] User Guide
v3.0, v2.1, v2.0, v1.0, CoreFPU Core Floating Point Unit, Core Floating Point Unit, Floating Point Unit, Point Unit

References

Leave a comment

Your email address will not be published. Required fields are marked *