byucc.jhdl.Logic.Modules.FloatingPoint
Class FPDivide

java.lang.Object
  byucc.jhdl.base.Nameable
      byucc.jhdl.base.Node
          byucc.jhdl.base.Cell
              byucc.jhdl.base.Structural
                  byucc.jhdl.Logic.LogicGates
                      byucc.jhdl.Logic.LogicStatic
                          byucc.jhdl.Logic.Logic
                              byucc.jhdl.Logic.Modules.FloatingPoint.FPDivide

All Implemented Interfaces:

BooleanFlags, Clockable, Pipelineable, byucc.jhdl.base.Propagateable, TreeListable

public class FPDivide

extends Logic

implements Pipelineable

Floating-point divider.

For all available constructors, see Constructor Summary or Constructor Detail.

General Description

FPDivide is a floating-point divider with generic exponent and significand sizes.

FPDivide takes two floating-point inputs of the same width. It divides them and outputs the quotient. The quotient has the same width as the inputs.

FPDivide rounds the result using the "round to nearest even" rounding mode. This is the standard rounding mode on computers.

To learn about the floating-point format, see the FloatingPoint package summary.

Implementation

The floating-point divider subtracts the exponents, divides the significands, and handles special cases The significand division is done using an integer divider.

The divider may be pipelined. Pipelining is recommended, since it greatly improves the maximum clock rate.

In addition, the floating-point divider handles several special cases.

Special Cases

Dividend /	Divisor =	Quotient	Exceptions4
Normal1	Normal	Normal
		Infinity3	Overflow
		0.0	Underflow
NaN	anything	NaN
anything	NaN	NaN
0.0	0.0	NaN	Invalid Operation
Infinity	Infinity	NaN
Normal	0.0	Infinity	Division by Zero
Infinity	0.0	Infinity	Invalid Operation
Infinity	Normal	Infinity
0.0	Normal	0.0
0.0	Infinity	0.0
Normal	Infinity	0.0

Notes
¹ Signs are not included in this table. The sign of the result depends on the signs of the operands.
² In this table, "Normal" means a nonzero finite floating-point representation of a value.
³ In the IEEE floating-point standard, the result of overflow actually depends on the rounding mode. If the rounding mode is "round to nearest / even," then an overflow results in an infinity. If the rounding mode is "round to zero," then an overflow results in the largest representable finite number. Other modes behave differently. (See the IEEE specification.) The value shown in this table is the value actually returned by this module.
⁴ This module does not currently support these exceptions. They are listed as a guide for possible future expansion. The IEEE exceptions are:

1. Invalid Operation
2. Division by Zero
3. Overflow
4. Underflow
5. Inexact (this should be implemented assuming no overflow trap)

The conditions for these exceptions are given in the IEEE specification.

Block Diagram

Version:

$Revision: 1.22 $

Author:

Aaron Stewart

Field Summary

static CellInterface[]	cell_interface
static java.lang.String	cellname
static int	ROUND_TO_NEAREST A constant for the default rounding mode: Round to nearest.
static int	ROUND_TOWARD_ZERO A constant for truncation mode (no rounding).

Fields inherited from class byucc.jhdl.Logic.Logic

ABOVE, ALIGN_BOTTOM, ALIGN_CENTER, ALIGN_LEFT, ALIGN_LSB, ALIGN_MSB, ALIGN_RIGHT, ALIGN_TOP, BELOW, DOWN, EAST_OF, LEFT_OF, MAX_PACK, NORTH_OF, ON, ONTOP, ONTOP_OF, RIGHT_OF, SOUTH_OF, TOLEFT, TORIGHT, UNCONSTRAINED, UP, WEST_OF

Fields inherited from class byucc.jhdl.Logic.LogicGates

tech_mapper

Fields inherited from class byucc.jhdl.base.Cell

BOOLEAN, CELL_NAME_DECLARATION, CellInterfaceDeterminesUniqueNetlistStructure, DETERMINE_FROM_STRING, GENERICS_DECLARATION, implicit_interface, IMPLICIT_INTERFACE_DECLARATION, INTEGER, INTERFACE_DECLARATION, LONG, PORT_IOS_DECLARATION, PORT_NAMES_DECLARATION, PORT_NET_NAMES_DECLARATION, PORT_PROPERTIES, PORT_WIDTHS_DECLARATION, SIGN_EXT, STRING, ZERO_PAD

Fields inherited from interface byucc.jhdl.base.BooleanFlags

ANTECEDANT_IS_BEHAVIORALLY_MODELED, ASYNC_PORT, ASYNCHRONOUS_RESOLVED, ATOMICALLY_PLACEABLE, ATOMICALLY_UNMAPPABLE, BEHAVIORALLY_MODELED_BRANCH, CLK_PORT, CLOCK_METHOD_IMPLEMENTED_BY_USER, CLOCK_METHOD_IS_DISABLED, CLOCKABLE_IS_SCHEDULED, DANGLING_IS_OK, DELETE_MARK, FATAL_BUILD_ERROR_OCCURED, HAS_BEEN_TRACED, HAS_USER_SPECIFIED_NAME, HWUPDATE, IMPLICIT_PORT, IN_CLK_PORT, IN_PORT, INOUT_PORT, IO_TYPE_FLAGS, IS_BEHAVIORALLY_MODELED, IS_ON_BUILD_STACK, IS_ON_PROP_LIST, IS_PLACED, METHODS_IMPLEMENTED_BY_USER, NETLISTABLE, ORIG_WIRE_IS_ATOMIC, OUT_PORT, PLACEMENT_IS_LOCKED, PROPAGATE_METHOD_IMPLEMENTED_BY_USER, PROPAGATE_METHOD_IS_DISABLED, RECURSION_FLAG, RESET_METHOD_IMPLEMENTED_BY_USER, SIMULATEABLE, SOURCELESS_IS_OK, SYNC_PORT, VISIBLE

Constructor Summary

FPDivide(Node parent, Wire dividend_in, Wire divisor_in, Wire quotient, Wire overflow_out, Wire underflow_out, int exponentWidth, int pipelining)
Builds a floating-point divider without exception flags.

FPDivide(Node parent, Wire dividend_in, Wire divisor_in, Wire quotient, Wire overflow_out, Wire underflow_out, int exponentWidth, int pipelining, java.lang.String instanceName)
Builds a floating-point divider.

Method Summary

void	clock() Used by the simulator.
BV[]	compute(BV dividend, BV divisor, int width, int exponentWidth, int mantSize) Used in behavioral model to compute the output
protected boolean	defaultSimulationModelIsBehavioral() The default simulation model is behavioral
java.lang.String	getCellName() Returns the cellname.
int	getInputLatency()
int	getIntDivideLatency()
int	getLatency() Returns the latency of the design.
void	propagate() Used by the simulator.
void	reset() Used by the simulator.

Methods inherited from class byucc.jhdl.Logic.Logic

clockDriver, clockDriver, connect_implicit_ports, connectImplicitPorts, constructSubCell, constructSubCellNoImplicitPorts, enableNewPlacement, enableNewPlacement, extend, extend, getDefaultClock, getDefaultTechMapper, getGlobalClock, getSinkCell, getSourceCell, getSourceCell, getSourceLeaf, getSourcePlaceable, getSourcePlaceableLeaf, getSubCellClass, getTechMapHint, getTechMapHint, getTechMapper, growAndShiftl, lockChildPlacement, lsb, lsb, map, map, map, map, map, map, map, map, map, map, map, map, map, map, map, map, map, map, map, msb, msb, msbIndx, netlist, netlist, netlist, netlist, netlist, netlist, netlist, netlist, padClock_o, padClock_o, padClock_o, padClock, padClock, padClock, padIn_o, padIn_o, padIn_o, padIn, padIn, padIn, padInout_o, padInout_o, padInout_o, padInout, padInout, padInout, padOut_o, padOut_o, padOut_o, padOut, padOut, padOut, padOutT_o, padOutT_o, padOutT_o, padOutT, padOutT, padOutT, place, place, place, place, place, place, place, place, place, place, place, place, place, place, place, place, place, printTechMapHints, range, rotate, rotate, scale, scale, setBBox, setDefaultTechMapper, setFloorPlannerIsMaster, setTechMappingEnabled, setWandH, signExtend_o, signExtend, signExtend, sink, source, takeBot_o, takeBot, takeBot, takeBotSigned_o, takeBotSigned, takeTop_o, takeTop, takeTop, techmap, techMappingEnabled, translate, translate, zeroExtend_o, zeroExtend, zeroExtend, zeroExtendRight_o, zeroExtendRight

Methods inherited from class byucc.jhdl.Logic.LogicStatic

add_o, add_o, and_o, and_o, and, and, buf_o, buf_o, buf, buf, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, constant_o, constant_o, constant_o, constant_o, constant_o, constant_o, constant_o, constant_o, constant, constant, constant, constant, constant, constant, constant, constant, constant, constant, gnd_o, gnd_o, gnd, gnd, gnd, gnd, mux_o, mux, nc, nc, nc, nc, nc, nc, not_o, not_o, not, not, or_o, or_o, or, or, reg_o, reg, vcc_o, vcc_o, vcc, vcc, vcc, vcc, wire, wire, wire, wire, xnor_o, xnor_o, xnor, xor_o, xor

Methods inherited from class byucc.jhdl.Logic.LogicGates

add_o, add_o, add_o, add_o, add_o, add_o, add, add, add, add, addsub_o, addsub_o, addsub_o, addsub_o, addsub_o, addsub_o, addsub, addsub, addsub, addsub, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and_o, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, and, ashiftr_o, ashiftr, ashiftr, buf_o, buf_o, buf, buf, checkValueRepresentableInWidth, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat_o, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, concat, constant_o, constant_o, constant_o, constant_o, constant_o, constant_o, constant_o, constant_o, constant, constant, constant, constant, constant, constant, constant, constant, constant, constant, gnd_o, gnd_o, gnd, gnd, gnd, gnd, mux_o, mux_o, mux_o, mux_o, mux_o, mux_o, mux, mux, mux, mux, mux, mux, name, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand_o, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nand, nc, nc, nc, nc, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor_o, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, nor, not_o, not_o, not, not, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or_o, or, or, or, or, or, or, or, or, or, or, or, or, or, or, or, or, or, or, pulldown, pulldown, pullup, pullup, reg_o, reg_o, reg_o, reg_o, reg, reg, reg, reg, regc_o, regc_o, regc_o, regc_o, regc, regc, regc, regc, regce_o, regce_o, regce_o, regce_o, regce, regce, regce, regce, regp_o, regp_o, regp_o, regp_o, regp, regp, regp, regp, regpe_o, regpe_o, regpe_o, regpe_o, regpe, regpe, regpe, regpe, regr_o, regr_o, regr_o, regr_o, regr, regr, regr, regr, regre_o, regre_o, regre_o, regre_o, regre, regre, regre, regre, regs_o, regs_o, regs_o, regs_o, regs, regs, regs, regs, regse_o, regse_o, regse_o, regse_o, regse, regse, regse, regse, shiftl_o, shiftl, shiftl, shiftr_o, shiftr, shiftr, sub_o, sub_o, sub_o, sub_o, sub_o, sub_o, sub, sub, sub, sub, tbuf_o, tbuf_o, tbuf, tbuf, vcc_o, vcc_o, vcc, vcc, vcc, vcc, wire, wire, wire, wire, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor_o, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xnor, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor_o, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor, xor

Methods inherited from class byucc.jhdl.base.Structural

behavioralModelIsAvailable, clockMethodIsDisabled, clockMethodIsDisabled, clockMethodIsEnabled, clockMethodIsEnabled, hasBeenTraced, hasBeenTraced, hasBehaviorInClockMethod, hasBehaviorInPropagateMethod, isAsynchronouslyScheduled, isAsynchronouslyScheduled, isFallingEdgeTriggered, isReadyToBeAsynchronouslyScheduled, isRisingEdgeTriggered, needsToBeAsynchronouslyScheduled, needsToBeClocked, propagateMethodIsDisabled, propagateMethodIsDisabled, propagateMethodIsEnabled, propagateMethodIsEnabled, willUseHWUpdate, willUseHWUpdate

Methods inherited from class byucc.jhdl.base.Cell

addPort, addPorts, addProperties, addProperties, addProperty, addProperty, addProperty, addProperty, antecedantIsBehaviorallyModeled, antecedantIsBehaviorallyModeled, bind, bind, bind, bind, cellInterfaceDeterminesUniqueNetlistStructure, clk, connect, connectAllWires, connectOptional, disableAllBehavioralModels, disableBehavioralModel, enableBehavioralModel, getArgument, getAttachedPort, getAttachedWire, getAttachedWireNoException, getCellNetlist, getCellNetList, getCellNetlist, getCellNetlist, getDescendents, getFlatNetlist, getFlatNetlistableChildren, getGeneric, getHeight, getNetlistableChildren, getPlacementInfo, getPortProperties, getPortRecord, getPortRecords, getProperties, getProperty, getPropertyValue, getSinkWires, getSourceWires, getUniqueCellName, getWidth, getX, getY, hasPort, hasPorts, in, in, inout, inout, isAsynchronousSourceSinkResolved, isBehaviorallyModeled, isBehaviorallyModeledBranch, isInput, isLeafCell, isNetlistable, isNetlistable, isNetlistablePort, isNetlistLeaf, isNotNetlistable, isNotNetlistablePort, isNotVisible, isOutput, isPlaceable, isPlaceable, isPlaced, isPlaced, isPlacementLocked, isRoot, isSimulateable, isSimulateable, isSink, isSource, isVisible, isVisible, join, lockPlacement, nc, out, out, param, popHierarchy, port, port, port, postorderCheck, preorderCheck, pushHierarchy, pushHierarchy, pushHierarchy, pushHierarchy, pushHierarchyNoImplicitPorts, pushHierarchyNoImplicitPorts, removeAllUnconnectedPorts, removePort, removeProperty, replaceProperty, replaceProperty, resetBehavioralModelsToDefaults, setAsynchronousSourceSinkResolved, setGeneric, setHeight, setNotNetlistable, setNotNetlistable, setNotVisible, setNotVisible, setPlacementInfo, setPortNotNetlistable, setPortNotNetlistable, setProperty, setWidth, subClassDelete, toString, uniquifyCell, userDefinedClockCount, verifyAndCleanup

Methods inherited from class byucc.jhdl.base.Node

addObservable, addSimulatorCallback, checkAll, delete, getBuildingFlag, getChildren, getChildrenEnumeration, getInstanceName, getParent, getParentCell, getRelatives, getSystem, getWires, optimize, orphanAllowed, printAllChildren, printTree, removeSimulatorCallback, setDefaultClock

Methods inherited from class byucc.jhdl.base.Nameable

caseSensitivity, caseSensitivity, disableNameClashChecking, getFullName, getFullNameNoTestBench, getHierNameNoTestBench, getInstanceNo, getInstanceNumber, getLeafName, getLeafName, getRelativeName, getUserName, getUserName, hasUserSpecifiedName, isDescendantOf, setInstanceNumber

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

cell_interface

public static CellInterface[] cell_interface

cellname

public static final java.lang.String cellname

See Also:

Constant Field Values

ROUND_TO_NEAREST

public static final int ROUND_TO_NEAREST

A constant for the default rounding mode: Round to nearest. If the value is exactly at the halfway point, it is rounded up (incremented) if odd, or down (truncated) if even, so that the resulting bit pattern is even. This "tie breaker" is designed to prevent the roundoff error from being biased in a particular direction. This may make a difference in long or repeated computations.

Examples (binary):

• 01.100001 is rounded up to 10.
• 01.100000 is rounded up to 10.
• 10.100000 is rounded down to 10.
• 01.0xxxxx is rounded down to 01.

See Also:

Constant Field Values

ROUND_TOWARD_ZERO

public static final int ROUND_TOWARD_ZERO

A constant for truncation mode (no rounding). This is also called "round toward zero".

Example (binary):

• 01.xxxxx is always rounded to 01.

See Also:

Constant Field Values

Constructor Detail

FPDivide

public FPDivide(Node parent,
                Wire dividend_in,
                Wire divisor_in,
                Wire quotient,
                Wire overflow_out,
                Wire underflow_out,
                int exponentWidth,
                int pipelining)

Builds a floating-point divider without exception flags. Use the getLatency() method of the cell to determine the number of pipeline stages.

Example:

int latency = new Divider (this, dividend, divisor, quotient, pipelining, exponentWidth, "MyDivider") . getLatency();

FPDivide

public FPDivide(Node parent,
                Wire dividend_in,
                Wire divisor_in,
                Wire quotient,
                Wire overflow_out,
                Wire underflow_out,
                int exponentWidth,
                int pipelining,
                java.lang.String instanceName)

Builds a floating-point divider. Use the getLatency() method of the cell to determine the number of pipeline stages.

Example:

int latency = new Divider (this, dividend, divisor, quotient, overflow_out, underflow_out, pipelining, exponentWidth, "MyDivider") . getLatency();

Method Detail

defaultSimulationModelIsBehavioral

protected boolean defaultSimulationModelIsBehavioral()

The default simulation model is behavioral

Overrides:

defaultSimulationModelIsBehavioral in class Structural

Returns:

true if TestBench or leafCell, false otherwise.

getCellName

public java.lang.String getCellName()

Returns the cellname.

Overrides:

getCellName in class Cell

Returns:

the cell name associated with a derived class, null if not declared.

getLatency

public int getLatency()

Returns the latency of the design.

Specified by:

getLatency in interface Pipelineable

Returns:

The number of clock cycles from the time valid inputs appear until the outputs appear. If the circuit is fully combinational, zero is returned.

getIntDivideLatency

public int getIntDivideLatency()

getInputLatency

public int getInputLatency()

reset

public void reset()

Used by the simulator. This is part of the behavioral model, for clocked logic.

Specified by:

reset in interface Clockable

Overrides:

reset in class Structural

clock

public void clock()

Used by the simulator. This is part of the behavioral model, for clocked logic.

Specified by:

clock in interface Clockable

Overrides:

clock in class Structural

propagate

public void propagate()

Used by the simulator. This is part of the behavioral model, for combinational logic.

Specified by:

propagate in interface byucc.jhdl.base.Propagateable

Overrides:

propagate in class Structural

compute

public BV[] compute(BV dividend,
                    BV divisor,
                    int width,
                    int exponentWidth,
                    int mantSize)

Used in behavioral model to compute the output

Returns:

An array containing [0]: the quotient, [1]: the status bits: OVERFLOW_MASK, UNDERFLOW_MASK, DIVIDE_BY_ZERO_MASK