The TrueType Instruction Set (OpenType 1.5)

TrueType provides instructions for each of the following tasks and a set of general purpose instructions. This chapter describes the TrueType instruction set. Instruction descriptions are organized by category based on their function.

• Pushing data onto the interpreter stack
• Managing the Storage Area
• Managing the Control Value Table
• Modifying Graphics State settings
• Managing outlines
• General purpose instructions

Anatomy of a TrueType Instruction

TrueType instructions are uniquely specified by their opcodes. For convenience, this book will refer to instructions by their names. Each instruction name is a mnemonic intended to aid in remembering that instruction’s function. For example, the MDAP instruction stands for Move Direct Absolute Point. Similarly, RUTG is short for Round Up To Grid. A brief description of each instruction clarifying the mnemonic marks the start of a new instruction.

One name may actually refer to several different but closely related instructions. A bracketed list of Boolean values follows each name to uniquely specify a particular variant of a given instruction. The Boolean list can be converted to a binary number and that number added to the base opcode for the instruction to obtain the opcode for any instruction variant.

To obtain the opcode for any instruction, take the lower of the two opcode values given in the code range and add the unsigned binary number represented by the list of binary digits. The left most bit is the most significant. For example, given an instruction with the opcode range 0xCO–0xDF and five Boolean flags (a through e) the opcode for a given instruction base can be computed as shown:

Opcode = 0xC0 + a · 2⁴ + b · 2³ + c · 2² + d · 2¹ + e · 2⁰

If these flags were set to 11101 the code would be computed as follows:

0xC0 + 1 · 2⁴ + 1 · 2³ + 1 · 2² + 0 · 2¹ + 1 · 2⁰

• = 0xC0 + 0x10 + 0x8 + 0x4 + 0x1
• = 0xC0 + 0x1D = 0xDD

Instruction opcodes are part of the instruction stream, a sequence of opcodes and data. The instruction stream is not a stack. While the stream of opcodes and data on the instruction stream is gradually used up, no new data is added to the instruction stream by the execution of another instruction (i.e. there is no equivalent of a push instruction that adds data to the instruction stream). It is possible to alter the flow of control through the instruction stream using one of the jump instructions described in a later section.

The instruction stream is shown as a sequence of opcodes and data. Since the instruction stream is 1-byte wide, words will be broken up into high bytes and low bytes with high bytes appearing first in the stream. For added readability, instruction names are used in illustrations instead of opcodes. An arrow will point to the next instruction awaiting execution.

A few instructions known collectively as push instructions move data from the instruction stream to the interpreter stack. These instructions are unique in taking their arguments from the instruction stream. All other TrueType instructions take any data needed from the stack at the time they are executed. Any results produced by a TrueType instruction are pushed onto the interpreter stack.

An instruction that expects two arguments and pushes a third would expect the two arguments to be at the top of the stack. Any result pushed by that instruction appears at the top of the stack.

To easily remember the order in which stack values are handled during arithmetic or logical operations, imagine writing the stack values from left to right, starting with the bottom value. Then insert the operator between the two furthest right elements. For example, subtract a,b would be interpreted as (b-a):

c b - a

GT a,b would be interpreted as (b>a):

c b > a

The statement push d, e means push d then push e adding two elements to the stack as shown.

To indicate that the top two stack elements are to be removed the statement would be pop e, d.

It has already been noted that the bracketed list of binary digits that follows the instruction name uniquely identifies an instruction variant. This is done by having the bits represent a list of Boolean flags that can be set to TRUE with a value of 1 or to FALSE with a value of 0. Binary digits that follow the name can also be grouped to form a larger binary number. In such cases, the documentation specifies the meaning associated with each possible numerical combination.

An instruction specification consists of the instruction name followed by its bracketed Boolean flags. Additional information describing the flags and explaining the stack interaction and any Graphics State dependencies is provided in tabular form:

Code Range	the range of hexadecimal codes identifying this instruction and its variants
Flags	an explanation of the meaning of a bracketed binary number
From IS	any arguments taken from the instruction stream by push instructions
Pops	any arguments popped from the stack
Pushes	any arguments pushed onto the stack
Uses	any state variables whose value this instruction depends upon
Sets	any state variables set by this instruction

Instruction descriptions include illustrations intended to clarify stack interactions, Graphics State effects, and changes to interpreter tables.

In the case of instructions that move points, an illustration will be provided to clarify the direction and magnitude of the movement. In these illustrations, shades of gray will be used to indicate the sequence in which points have been moved. The darker the fill, the more recently a point has been moved.

Data types

The instruction stream

Instruction opcodes are always bytes. Values in the instruction stream are bytes.

The stack

Values pushed onto the stack or popped from the stack are always 32 bit quantities (LONG or ULONG). When values that are less than 32 bits are pushed onto the stack, bytes are expanded to 32 bit quantities by padding the upper bits with zeroes and words are sign extended to 32 bits. In cases where two instruction stream bytes are combined to form a word, the high order bits appear first in the instruction stream.

Note: On a 16-bit system, such as Windows, all stack operations are on 16-bit values (SHORT or USHORT). Care must be taken to avoid overflow. It is also important to note that F26dot6 values (used for internal scalar math) are represented instead as 10 dot 6 values (i.e. the upper 16 bits are not supported).

All values on the stack are signed. Instructions, however, interpret these 32-bit quantities in a variety of ways. The interpreter variously understands quantities as integers and as fixed point numbers.

Values such as pixel coordinates are represented as 32-bit quantities consisting of 26 bits of whole number and 6 bits of fraction. These are fixed point numbers with the data type name F26Dot6.

The setting of the freedom_vector and projection_vector are represented as 2.14 fixed point numbers. The upper 16 bits of the 32 bit quantity are ignored.

A given set of 32 bits will have a different value depending upon how it is interpreted. The following 32 bit value interpreted as an integer has the value 264.

The same 32 bit quantity interpreted as an F26Dot6 fixed point number has the value 4.125.

The figure below gives several examples of expressing pixel values as 26.6 words.

Pushing data onto the interpreter stack

Most TrueType instructions take their arguments from the interpreter stack. A few instructions, however, take their arguments from the instruction stream. Their purpose is to move data from the instruction stream to the interpreter stack. Collectively these instructions are known as the push instructions.

PUSH N Bytes

NPUSHB[ ]
Code Range	0x40
From IS	n: number of bytes to push (1 byte interpreted as an integer) b1, b2,...bn: sequence of n bytes
Pushes	b1, b2, bn: sequence of n bytes each padded to 32 bits (ULONG)

Takes n unsigned bytes from the instruction stream, where n is an unsigned integer in the range (0..255), and pushes them onto the stack. n itself is not pushed onto the stack.

PUSH N Words

NPUSHW[ ]
Code Range	0x41
From IS	n: number of words to push (one byte interpreted as an integer) w1, w2,...wn: sequence of n words formed from pairs of bytes, the high byte appearing first
Pushes	w1, w2,...wn: sequence of n words each sign extended to 32 bits (LONG)

Takes n 16-bit signed words from the instruction stream, where n is an unsigned integer in the range (0..255), and pushes them onto the stack. n itself is not pushed onto the stack.

PUSH Bytes

PUSHB[abc]
Code Range	0xB0 – 0xB7
abc	number of bytes to be pushed - 1
From IS	b0, b1,..bn: sequence of n + 1 bytes
Pushes	b0, b1, ...,bn: sequence of n + 1 bytes each padded to 32 bits (ULONG)

Takes the specified number of bytes from the instruction stream and pushes them onto the interpreter stack.

The variables a, b, and c are binary digits representing numbers from 000 to 111 (0-7 in binary). Because the actual number of bytes (n) is from 1 to 8, 1 is automatically added to the ABC figure to obtain the actual number of bytes pushed.

Example:

PUSH Words

PUSHW[abc]
Code Range	0xB8 - 0xBF
abc	number of words to be pushed - 1.
From IS	w0,w1,..wn: sequence of n+1 words formed from pairs of bytes, the high byte appearing first
Pushes	w0, w1,...wn: sequence of n+1 words each sign extended to 32 bits (LONG)

Takes the specified number of words from the instruction stream and pushes them onto the interpreter stack.

The variables a, b, and c are binary digits representing numbers from 000 to 111 (0-7 binary). Because the actual number of bytes (n) is from 1 to 8, 1 is automatically added to the abc figure to obtain the actual number of bytes pushed.

Example:

Managing the Storage Area

The interpreter Storage Area is a block of memory that can be used to store and later access 32 bit values. Instructions exist for writing values to the Storage Area and retrieving values from the Storage Area. Attempting to read a value from a storage location that has not previously had a value written to it will yield unpredictable results.

Read Store

RS[ ]
Code Range	0x43
Pops	location: Storage Area location (ULONG)
Pushes	value: Storage Area value (ULONG)
Gets	Storage Area value

This instruction reads a 32 bit value from the Storage Area location popped from the stack and pushes the value read onto the stack. It pops an address from the stack and pushes the value found in that Storage Area location to the top of the stack. The number of available storage locations is specified in the maxProfile table in the font file.

Example:

Write Store

WS[ ]
Code Range	0x42
Pops	value: Storage Area value (ULONG)
	location: Storage Area location (ULONG)
Pushes	–
Gets	Storage Area value

This instruction writes a 32 bit value into the storage location indexed by locations. It works by popping a value and then a location from the stack. The value is placed in the Storage Area location specified by that address. The number of storage locations is specified in the maxProfile table in the font file.

Example:

Managing the Control Value Table

The Control Value Table stores information that is accessed by the indirect instructions. Values can be written to the CVT in FUnits or pixel units as proves convenient. Values read from the CVT are always in pixels (F26Dot6). This table, unlike the Storage Area, is initialized by the font and is automatically scaled.

Write Control Value Table in Pixel units

WCVTP[ ]
Code Range	0x44
Pops	value: number in pixels (F26Dot6 fixed point number)
	location: Control Value Table location (ULONG)
Pushes	–
Sets	Control Value Table entry

Pops a location and a value from the stack and puts that value in the specified location in the Control Value Table. This instruction assumes the value is in pixels and not in FUnits.

Write Control Value Table in FUnits

WCVTF[ ]
Code Range	0x70
Pops	value: number in FUnits (ULONG)
	location: Control Value Table location (ULONG)
Pushes	–
Sets	Control Value Table entry

Pops a location and a value from the stack and puts the specified value in the specified address in the Control Value Table. This instruction assumes the value is expressed in FUnits and not pixels. The value is scaled before being written to the table.

Read Control Value Table

RCVT[ ]
Code Range	0x45
Pops	location: CVT entry number (ULONG)
Pushes	value: CVT value (F26Dot6)
Gets	Control Value Table entry

Pops a location from the stack and pushes the value in the location specified in the Control Value Table onto the stack.

Managing the Graphics State

Instructions can be used to set the value of Graphics State variables and, in some cases, to retrieve their current value.

Getting a value

Instructions that retrieve the value of a state variable have names that begin with the word get. Get instructions will return the value of the state variable in question by placing that value on the top of the stack.

The illustration shows the effect of a GPV or get projection_vector instruction. It takes the x and y components of the projection_vector from the Graphics State and places them on the stack.

Setting a value

Instructions that change the value of a Graphics State variable have a name that begins with the word set. Set instructions expect their arguments to be at the top of the interpreter stack

In addition to simple sets and gets, some instructions exist to simplify management of the values of state variables. For example, a number of instructions exist to set the direction of the freedom_vector and the projection_vector. In setting a vector, it is possible to set it to either of the coordinate axes, to the direction specified by a line, or to a direction specified by values taken from the stack. An instruction exists that directly sets the freedom_vector to the same value as the projection_vector.

Set freedom and projection Vectors To Coordinate Axis

SVTCA[a]
Code Range	0x00 - 0x01
a	0: set vectors to the y-axis
	1: set vectors to the x-axis
Pops	-
Pushes	-
Sets	projection_vector
	freedom_vector

Sets both the projection_vector and freedom_vector to the same one of the coordinate axes.

The SVTCA is a shortcut for using both the SFVTCA and SPVTCA instructions. SVTCA[1] is equivalent to SFVTCA[1] followed by SPVTCA[1]. This instruction ensures that both movement and measurement are along the same coordinate axis.

Example:

SVTCA[1]

SVTCA[0]

Set Projection_Vector To Coordinate Axis

SPVTCA[a]
Code Range	0x02 - 0x03
a	0: set the projection_vector to the y-axis
	1: set the projection_vector to the x-axis
Pops	-
Pushes	-
Sets	projection_vector

Sets the projection_vector to one of the coordinate axes depending on the value of the flag a.

Example:

SPVTCA[0]

Set Freedom_Vector to Coordinate Axis

SFVTCA[a]
Code Range	0x04 - 0x05
a	0: set the freedom_vector to the y-axis
	1: set the freedom_vector to the x-axis
Pops	-
Pushes	-
Sets	freedom_vector

Sets the freedom_vector to one of the coordinate axes depending upon the value of the flag a.

Example:

SFVTCA[0]

Set Projection_Vector To Line

SPVTL[a]
Code Range	0x06 - 0x07
a	0: sets projection_vector to be parallel to line segment from p1 to p2
	1: sets projection_vector to be perpendicular to line segment from p1 to p2; the vector is rotated counter clockwise 90 degrees
Pops	p1: point number (ULONG)
	p2: point number (ULONG)
Pushes	-
Uses	point p1 in the zone pointed at by zp2
	point p2 in the zone pointed at by zp1
Sets	projection_vector

Sets the projection_vector to a unit vector parallel or perpendicular to the line segment from point p1 to point p2.

Example:

Case 1:

SPVTL[1]

Case 2:

SPVTL[1]

Case 3:

SPVTL[1]

Set Freedom_Vector To Line

SFVTL[a]
Code Range	0x08 - 0x09
a	0: set freedom_vector to be parallel to the line segment defined by points p1 and p2
	1: set freedom_vector perpendicular to the line segment defined by points p1 and p2; the vector is rotated counter clockwise 90 degrees
Pops	p1: point number (ULONG)
	p2: point number (ULONG)
Pushes	-
Sets	freedom_vector
Uses	point p1 in the zone pointed at by zp2
	point p2 in the zone pointed at by zp1

Set Freedom_Vector To Projection Vector

SFVTPV[ ]
Code	0x0E
Pops	-
Pushes	-
Sets	freedom_vector

Sets the freedom_vector to be the same as the projection_vector.

Set Dual Projection_Vector To Line

SDPVTL[a]
Code Range	0x86 - 0x87
a	0: Vectors are parallel to line
	1: Vectors are perpendicular to line
Pops	p1: first point number (ULONG)
	p2: p2: second point number (ULONG)
Pushes	-
Sets	dual_projection_vector and projection_vector
Uses	point p1 in the zone pointed at by zp2
	point p2 in the zone pointed at by zp1

Pops two point numbers from the stack and uses them to specify a line that defines a second, dual_projection_vector. This dual_projection_vector uses coordinates from the scaled outline before any grid-fitting took place. It is used only with the IP, GC, MD, MDRP and MIRP instructions. Those instructions will use the dual_projection_vector when they measure distances between ungrid-fitted points. The dual_projection_vector will disappear when any other instruction that sets the projection_vector is used.

Set Projection_Vector From Stack

SPVFS[ ]
Code Range	0x0A
Pops	y: y component of projection_vector (2.14 fixed point number padded with zeroes)
	x: x component of projection_vector (2.14 fixed point number padded with zeroes)
Pushes	-
Sets	projection_vector

Sets the direction of the projection_vector, using values x and y taken from the stack, so that its projections onto the x and y-axes are x and y, which are specified as signed (two’s complement) fixed-point (2.14) numbers. The square root of (x2 + y2) must be equal to 0x4000 (hex).

If values are to be saved and used by a glyph program, font program or preprogram across different resolutions, extreme care must be used. The values taken from or put on the stack are 2.14 fixed-point values for the x and y components of the vector in question. The values are based on the normalized vector lengths. More simply, the values must always be set such that (X**2 + Y**2) is 1.

If a TrueType program uses specific values for X and Y to set the vectors to certain angles, these values will not produce identical results across different aspect ratios. Values that work correctly at 1:1 aspect ratios (such as VGA and 8514) will not necessarily yield the desired results at a ratio of 1.33:1 (e.g. the EGA).

By the same token, if a TrueType program is making use of the values returned by GPV and GFV, the values returned for a specific angle will vary with the aspect ratio in use at the time.

Example:

SPVFS[ ]

Set Freedom_Vector From Stack

SFVFS[ ]
Code	0x0B
Pops	y: y component of freedom_vector (2.14 fixed point number padded with zeroes)
	x: x component of freedom_vector (2.14 fixed point number padded with zeroes)
Pushes	-
Sets	freedom_vector

Sets the direction of the freedom_vector using the values x and y taken from the stack. The vector is set so that its projections onto the x and y -axes are x and y, which are specified as signed (two’s complement) fixed-point (2.14) numbers. The square root of (x2 + y2) must be equal to 0x4000 (hex).

If values are to be saved and used by a glyph program, font program or preprogram across different resolutions, extreme care must be used. The values taken from or put on the stack are 2.14 fixed-point values for the x and y components of the vector in question. The values are based on the normalized vector lengths. More simply, the values must always be set such that (X**2 + Y**2) is 1.

If a TrueType program uses specific values for X and Y to set the vectors to certain angles, these values will not produce identical results across different aspect ratios. Values that work correctly at 1:1 aspect ratios (such as VGA and 8514) will not necessarily yield the desired results at a ratio of 1.33:1 (e.g. the EGA).

By the same token, if a TrueType program is making use of the values returned by GPV and GFV, the values returned for a specific angle will vary with the aspect ratio in use at the time.

Example:

Get Projection_Vector

GPV[ ]
Code Range	0x0C
Pops	-
Pushes	x: x component of projection_vector (2.14 fixed point number padded with zeroes)
	y : y component of projection_vector (2.14 fixed point number padded with zeroes)
Gets	projection_vector

Pushes the x and y components of the projection_vector onto the stack as two 2.14 numbers.

If values are to be saved and used by a glyph program, font program or preprogram across different resolutions, extreme care must be used. The values taken from or put on the stack are 2.14 fixed-point values for the x and y components of the vector in question. The values are based on the normalized vector lengths. More simply, the values must always be set such that (X**2 + Y**2) is 1.

If a TrueType program uses specific values for X and Y to set the vectors to certain angles, these values will not produce identical results across different aspect ratios. Values that work correctly at 1:1 aspect ratios (such as VGA and 8514) will not necessarily yield the desired results at a ratio of 1.33:1 (e.g. the EGA).

By the same token, if a TrueType program is making use of the values returned by GPV and GFV, the values returned for a specific angle will vary with the aspect ratio in use at the time.

Example:

Case 1:

Case 2:

Case 3:

Get Freedom_Vector

GFV[ ]
Code Range	0x0D
Pops	-
Pushes	x: x-component of freedom_vector (2.14 number padded with zeroes)
	y: y component of freedom_vector (2.14 number padded with zeroes)
Gets	freedom_vector

Puts the x and y components of the freedom_vector on the stack. The freedom_vector is put onto the stack as two 2.14 coordinates.

If values are to be saved and used by a glyph program, font program or preprogram across different resolutions, extreme care must be used. The values taken from or put on the stack are 2.14 fixed-point values for the x and y components of the vector in question. The values are based on the normalized vector lengths. More simply, the values must always be set such that (X**2 + Y**2) is 1.

If a TrueType program uses specific values for X and Y to set the vectors to certain angles, these values will not produce identical results across different aspect ratios. Values that work correctly at 1:1 aspect ratios (such as VGA and 8514) will not necessarily yield the desired results at a ratio of 1.33:1 (e.g. the EGA).

By the same token, if a TrueType program is making use of the values returned by GPV and GFV, the values returned for a specific angle will vary with the aspect ratio in use at the time.

Example:

GFV[ ]

Set Reference Point 0

SRP0[ ]
Code Range	0x10
Pops	p: point number (ULONG)
Pushes	-
Sets	rp0
Affects	IP, MDAP, MIAP, MIRP, MSIRP, SHC, SHE, SHP

Pops a point number from the stack and sets rp0 to that point number.

Set Reference Point 1

SRP1[ ]
Code Range	0x11
Pops	p: point number (ULONG)
Pushes	-
Sets	rp1
Affects	IP, MDAP, MDRP, MIAP, MSIRP, SHC, SHE, SHP

Pops a point number from the stack and sets rp1 to that point number.

Set Reference Point 2

SRP2[ ]
Code Range	0x12
Pops	p: point number (ULONG)
Pushes	-
Sets	rp2

Pops a point number from the stack and sets rp2 to that point number.

Set Zone Pointer 0

SZP0[ ]
Code Range	0x13
Pops	n: zone number (ULONG)
Pushes	-
Sets	zp0
Affects	ALIGNPTS, ALIGNRP, DELTAP1, DELTAP2, DELTAP3, IP, ISECT, MD, MDAP, MIAP, MIRP, MSIRP, SHC, SHE, SHP, UTP

Pops a zone number, n, from the stack and sets zp0 to the zone with that number. If n is 0, zp0 points to zone 0. If n is 1, zp0 points to zone 1. Any other value for n is an error.

Example:

Set Zone Pointer 1

SZP1[ ]
Code Range	0x14
Pops	n: zone number (ULONG)
Pushes	-
Sets	zp1
Affects	ALIGNRPTS, ALIGNRP, IP, MD, MDRP, MSIRP, SHC, SHE, SHP, SFVTL, SPVTL

Pops a zone number, n, from the stack and sets zp1 to the zone with that number. If n is 0, zp1 points to zone 0. If n is 1, zp1 points to zone 1. Any other value for n is an error.

Example:

Set Zone Pointer 2

SZP2[ ]
Code Range	0x15
Pops	n: zone number (ULONG)
Pushes	-
Sets	zp2
Affects	ISECT, IUP, GC, SHC, SHP, SFVTL, SHPIX, SPVTL, SC

Pops a zone number, n, from the stack and sets zp2 to the zone with that number. If n is 0, zp2 points to zone 0. If n is 1, zp2 points to zone 1. Any other value for n is an error.

Set Zone PointerS

SZPS[ ]
Code Range	0x16
Pops	n: zone number (ULONG)
Pushes	-
Sets	zp0, zp1, zp2
Affects	ALIGNPTS, ALIGNRP, DELTAP1, DELTAP2, DELTAP3, GC, IP, ISECT, IUP, MD, MDAP, MDRP, MIAP, MIRP, MSIRP, SC, SFVTL, SHPIX, SPVTL, SHC, SHE, SHP, SPVTL, UTP

Pops a zone number from the stack and sets all of the zone pointers to point to the zone with that number. If n is 0, all three zone pointers will point to zone 0. If n is 1, all three zone pointers will point to zone 1. Any other value for n is an error.

Round To Half Grid

RTHG[ ]
Code Range	0x19
Pops	-
Pushes	-
Sets	round_state
Affects	MDAP, MDRP, MIAP, MIRP, ROUND
Uses	freedom_vector, projection_vector

Sets the round_state variable to state 0 (hg). In this state, the coordinates of a point are rounded to the nearest half grid line.

Example

Round To Grid

RTG[ ]
Code Range	0x18
Pops	-
Pushes	-
Sets	round_state
Affects	MDAP, MDRP, MIAP, MIRP, ROUND
Uses	freedom_vector, projection_vector

Sets the round_state variable to state 1 (g). In this state, distances are rounded to the closest grid line.

Example

Round To Double Grid

RTDG[ ]
Code Range	0x3D
Pops	-
Pushes	-
Sets	round_state
Affects	MDAP, MDRP, MIAP, MIRP, ROUND
Uses	freedom_vector, projection_vector

Sets the round_state variable to state 2 (dg). In this state, distances are rounded to the closest half or integer pixel.

Example

Round Down To Grid

RDTG[ ]
Code Range	0x7D
Pops	-
Pushes	-
Sets	round_state
Affects	MDAP, MDRP, MIAP, MIRP, ROUND
Uses	freedom_vector, projection_vector

Sets the round_state variable to state 3 (dtg). In this state, distances are rounded down to the closest integer grid line.

Example

Round Up To Grid

RUTG[ ]
Code Range	0x7C
Pops	-
Pushes	-
Sets	round_state
Affects	MDAP, MDRP, MIAP, MIRP, ROUND
Uses	freedom_vector, projection_vector

Sets the round_state variable to state 4 (utg). In this state distances are rounded up to the closest integer pixel boundary

Example

Round OFF

ROFF[ ]
Code Range	0x7A
Pops	-
Pushes	-
Sets	round_state
Affects	MDAP, MDRP, MIAP, MIRP, ROUND
Uses	freedom_vector, projection_vector

Sets the round_state variable to state 5 (off). In this state rounding is turned off.

Example

Super ROUND

SROUND[ ]
Code Range	0x76
Pops	n: number decomposed to obtain period, phase, threshold
Pushes	-
Sets	round_state
Affects	MDAP, MDRP, MIAP, MIRP, ROUND

SROUND allows you fine control over the effects of the round_state variable by allowing you to set the values of three components of the round_state: period, phase, and threshold.

More formally, SROUND maps the domain of 26.6 fixed point numbers into a set of discrete values that are separated by equal distances. SROUND takes one argument from the stack, n, which is decomposed into a period, phase and threshold.

The period specifies the length of the separation or space between rounded values in terms of grid spacing.

The threshold specifies the part of the domain that is mapped onto each value. More intuitively, the threshold tells a value when to “fall forward” to the next largest integer.

Only the lower 8 bits of the argument n are used. For SROUND gridPeriod is equal to 1.0 pixel. The byte is encoded as follows: bits 7 and 6 encode the period, bits 5 and 4 encode the phase and bits 3, 2, 1 and 0 encode the threshold as shown here.

period

0	period = gridPeriod/2
1	period = gridPeriod
2	period = gridPeriod * 2
3	Reserved

phase

0	phase = 0
1	phase = period/4
2	phase = period/2
3	phase = gridPeriod * 3/4

threshold

0	threshold = period -1
1	threshold = -3/8 * period
2	threshold = -2/8 * period
3	threshold = -1/8 * period
4	threshold = 0/8 * period
5	threshold = 1/8 * period
6	threshold = 2/8 * period
7	threshold =3/8 * period
8	threshold = 4/8 * period
9	threshold = 5/8 * period
10	threshold = 6/8 * period
11	threshold = 7/8 * period
12	threshold = 8/8 * period
13	threshold = 9/8 * period
14	threshold = 10/8 * period
15	threshold = 11/8 * period

For example, SROUND(01:01:1000) maps numbers into the values 0.25, 1.25, 2.25, The numbers from -0.25 to 0.75 are mapped into 0.25. The range of numbers [0.75, 1.75) map into 1.25. Similarly, the numbers from [1.75, 2.75) map into the number 2.25 and so on.

Rounding occurs after compensation for engine characteristics, so the steps in the rounding of a number n are:

• add engine compensation to n.
• subtract the phase from n.
• add the threshold to n.
• truncate n to the next lowest periodic value (ignore the phase).
• add the phase back to n.
• if rounding caused a positive number to become negative, set n to the positive round value closest to 0.
• if rounding caused a negative number of become positive, set n to the negative round value closest to 0.
• the period parameters can have values of 1/2 pixel, 1 pixel, or 2 pixels.
• the phase parameters can have values of 0 pixels, 1/4 pixel, 1/2 pixel, or 3/4 pixel.
• the threshold parameters can have values of -3/8 period, -2/8 period, 11/8 period. It can also have the special value largest-number-smaller-than-period which causes rounding equivalent to CEILING.

Super ROUND 45 degrees

S45ROUND[ ]
Code Range	0x77
Pops	n: ULONG decomposed to obtain period, phase, threshold (ULONG)
Pushes	-
Sets	round_state
Affects	MDAP, MDRP, MIAP, MIRP, ROUND

S45ROUND is analogous to SROUND. The gridPeriod is SQRT(2)/2 pixels rather than 1 pixel. It is useful for measuring at a 45 degree angle with the coordinate axes.

Set LOOP variable

SLOOP[ ]
Code Range	0x17
Pops	n: value for loop Graphics State variable (integer)
Pushes	-
Sets	loop
Affects	ALIGNRP, FLIPPT, IP, SHP, SHPIX

Pops a value, n, from the stack and sets the loop variable count to that value. The loop variable works with the SHP[a], SHPIX[a], IP[ ], FLIPPT[ ], and ALIGNRP[ ]. The value n indicates the number of times the instruction is to be repeated. After the instruction executes, the loop variable is reset to 1.

Set Minimum_Distance

SMD[ ]
Code Range	0x1A
Pops	distance: value for minimum_distance (F26Dot6)
Pushes	-
Sets	minimum_distance

Pops a value from the stack and sets the minimum_distance variable to that value. The distance is assumed to be expressed in sixty-fourths of a pixel.

INSTRuction execution ConTRoL

INSTCTRL[ ]
Code Range	0x8E
Pops	s: selector flag (int32)
	value: USHORT (padded to 32 bits) used to set value of instruction_control.
Pushes	-
Sets	instruction_control

Sets the instruction control state variable making it possible to turn on or off the execution of instructions and to regulate use of parameters set in the CVT program. INSTCTRL[ ] can only be executed in the CVT program.

This instruction clears and sets various control flags in the rasterizer. The selector flag determines valid values for the value argument. The value determines the new setting of the raterizer control flag. In this version there are three flags in use:

• Selector flag 1 is used to inhibit grid-fitting. If s=1, valid values for the value argument are 0 (FALSE) and 1 (TRUE). If the value argument is set to TRUE (v=1), any instructions associated with glyphs will not be executed. For example, to inhibit grid-fitting when a glyph is being rotated or stretched, use the following sequence on the preprogram:

  PUSHB[000] 6    /* ask GETINFO to check for stretching or rotation */
  GETINFO[]       /* will push TRUE if glyph is stretched or rotated */
  IF[]            /* tests value at top of stack */
  PUSHB[000] 1    /* value for INSTCTRL */
  PUSHB[000] 1    /* selector for INSTCTRL */
  INSTRCTRL[]     /* based on selector and value will turn grid-fitting off */
  EIF[]
  

• Selector flag 2 is used to establish that any parameters set in the CVT program should be ignored when instructions associated with glyphs are executed. These include, for example, the values for scantype and the CVT cut-in. If s=1, valid values for the value argument are 0 (FALSE) and 2 (TRUE). If the value argument is set to TRUE (v=2), the default values of those parameters will be used regardless of any changes that may have been made in those values by the preprogram. If the value argument is set to FALSE (v=0), parameter values changed by the CVT program will be used in glyph instructions.

• Selector flag 3 is used to control some aspects of how some specific instructions are interpreted when ClearType™ is being used. If the selector is enabled, a font will be rendered in native ClearType mode, as opposed to backwards compatibility mode. If the value argument is set to TRUE (v=4) then the interpreter will be in native ClearType mode. If the value argument is set to FALSE (v=0) then the interpreter will be in backwards compatibility mode.

SCAN conversion ConTRoL

SCANCTRL[ ]
Code Range	0x85
Pops	n: flags indicating when to turn on dropout control mode (16 bit word padded to 32 bits)
Pushes	-
Sets	scan_control

SCANCTRL is used to set the value of the Graphics State variable scan_control which in turn determines whether the scan converter will activate dropout control for this glyph. Use of the dropout control mode is determined by three conditions:

1. Is the glyph rotated?
2. Is the glyph stretched?
3. Is the current setting for ppem less than a specified threshold?

The interpreter pops a word from the stack and looks at the lower 16 bits.

Bits 0-7 represent the threshold value for ppem. A value of FF in bits 0-7 means invoke dropout_control for all sizes. A value of 0 in bits 0-7 means never invoke dropout_control.

Bits 8-13 are used to turn on dropout_control in cases where the specified conditions are met. Bits 8, 9 and 10 are used to turn on the dropout_control mode (assuming other conditions do not block it). Bits 11, 12, and 13 are used to turn off the dropout mode unless other conditions force it. Bits 14 and 15 are reserved for future use.

Bit	Meaning if set
8	Set dropout_control to TRUE if other conditions do not block and ppem is less than or equal to the threshold value.
9	Set dropout_control to TRUE if other conditions do not block and the glyph is rotated
10	Set dropout_control to TRUE if other conditions do not block and the glyph is stretched.
11	Set dropout_control to FALSE unless ppem is less than or equal to the threshold value.
12	Set dropout_control to FALSE unless the glyph is rotated.
13	Set dropout_control to FALSE unless the glyph is stretched.
14	Reserved for future use.
15	Reserved for future use.

For example

0x0000	No dropout control is invoked
0x01FF	Always do dropout control
0x0A10	Do dropout control if the glyph is rotated and has less than 16 pixels per-em

The scan converter can operate in either a “normal” mode or in a “fix dropout” mode depending on the value of a set of enabling and disabling flags.

SCANTYPE

SCANTYPE[ ]
Code Range	0x8D
Pops	n: 16 bit integer
Pushes	-
Sets	scan_control

Pops a 16-bit integer whose value is used to determine which rules the scan converter will use. If the value of the argument is 0, the fast scan converter will be used. If the value of the integer is 1 or 2, simple dropout control will be used. If the value of the integer is 4 or 5, smart dropout control will be used. More specifically,

• if n=0 rules 1, 2, and 3 are invoked (simple dropout control scan conversion including stubs)
• if n=1 rules 1, 2, and 4 are invoked (simple dropout control scan conversion excluding stubs)
• if n=2 rules 1 and 2 only are invoked (fast scan conversion; dropout control turned off)
• if n=3 same as n = 2
• if n = 4 rules 1, 2, and 5 are invoked (smart dropout control scan conversion including stubs)
• if n = 5 rules 1, 2, and 6 are invoked (smart dropout control scan conversion excluding stubs)
• if n = 6 same as n = 2
• if n = 7 same as n = 2

The scan conversion rules are shown here:

Rule 1	If a pixel’s center falls within the glyph outline, that pixel is turned on.
Rule 2	If a contour falls exactly on a pixel’s center, that pixel is turned on.
Rule 3	If a scan line between two adjacent pixel centers (either vertical or horizontal) is intersected by both an on-Transition contour and an off-Transition contour and neither of the pixels was already turned on by rules 1 and 2, turn on the left-most pixel (horizontal scan line) or the bottom-most pixel (vertical scan line). This is “Simple” dropout control.
Rule 4	Apply Rule 3 only if the two contours continue to intersect other scan lines in both directions. That is, do not turn on pixels for ‘stubs.’ The scanline segments that form a square with the intersected scan line segment are examined to verify that they are intersected by two contours. It is possible that these could be different contours than the ones intersecting the dropout scan line segment. This is very unlikely but may have to be controlled with grid-fitting in some exotic glyphs.
Rule 5	If a scan line between two adjacent pixel centers (either vertical or horizontal) is intersected by both an on-Transition contour and an off-Transition contour and neither of the pixels was already turned on by rules 1 and 2, turn on the pixel which is closer to the midpoint between the on-Transition contour and off-Transition contour. This is “Smart” dropout control.
Rule 6	Apply Rule 5 only if the two contours continue to intersect other scan lines in both directions. That is, do not turn on pixels for ‘stubs.’

New fonts wishing to use the new modes of the ScanType instruction, but still wishing to work correctly on old rasterizers that don’t recognize the new modes should:

1. First execute a ScanType instruction using an old mode which will give the best approximation to the desired new mode (e.g. Simple Stubs for Smart Stubs), and then
2. Immediately execute another ScanType instruction with the desired new mode.

Set Control Value Table Cut In

SCVTCI[ ]
Code Range	0x1D
Pops	n: value for cut_in (F26Dot6)
Pushes	-
Sets	control_value_cut_in
Affects	MIAP, MIRP

Sets the control_value_cut_in in the Graphics State. The value n is expressed in sixty-fourths of a pixel.

Increasing the value of the cut_in will increase the range of sizes for which CVT values will be used instead of the original outline value.

Set Single_Width_Cut_In

SSWCI[ ]
Code Range	0x1E
Pops	n: value for single_width_cut_in (F26dot6)
Pushes	-
Sets	single_width_cut_in
Affects	MIAP, MIRP

Sets the single_width_cut_in in the Graphics State. The value n is expressed in sixty-fourths of a pixel.

Set Single-width

SSW[ ]
Code Range	0x1F
Pops	n: value for single_width_value (FUnits)
Pushes	-
Sets	single_width_value

Sets the single_width_value in the Graphics State. The single_width_value is expressed in FUnits.

Set the auto_flip Boolean to ON

FLIPON[ ]
Code Range	0x4D
Pops	-
Pushes	-
Sets	auto_flip
Affects	MIRP

Sets the auto_flip Boolean in the Graphics State to TRUE causing the MIRP instructions to ignore the sign of Control Value Table entries. The default auto_flip Boolean value is TRUE.

Set the auto_flip Boolean to OFF

FLIPOFF[ ]
Code Range	0x4E
Pops	-
Pushes	-
Sets	auto_flip
Affects	MIRP

Set the auto_flip Boolean in the Graphics State to FALSE causing the MIRP instructions to use the sign of Control Value Table entries. The default auto_flip Boolean value is TRUE.

Set Angle_Weight

SANGW[ ]
Code Range	0x7E
Pops	weight: value for angle_weight
Pushes	-
Sets	angle_weight

SANGW is no longer needed because of dropped support to the AA (Adjust Angle) instruction. AA was the only instruction that used angle_weight in the global graphics state.

Pops a weight value from the stack and sets the value of the angle_weight state variable accordingly

Set Delta_Base in the graphics state

SDB[ ]
Code Range	0x5E
Pops	-
Pushes	-
Sets	delta_base
Affects	DELTAP1, DELTAP2, DELTAP3, DELTAC1, DELTAC2, DELTAC3

Pops a number, n, and sets delta_base to the value n. The default for delta_base is 9.

Set Delta_Shift in the graphics state

SDS[ ]
Code Range	0x5F
Pops	n: value for the delta_shift (ULONG)
Pushes	-
Sets	delta_shift
Affects	DELTAP1, DELTAP2, DELTAP3, DELTAC1, DELTAC2, DELTAC3

Sets delta_shift to the value n. The default for delta_shift is 3.

Reading and writing data

The following instructions make it possible to get and to set a point coordinate, to measure the distance between two points, and to determine the current settings for pixels per em and point size.

Get Coordinate projected onto the projection_vector

GC[a]
Code Range	0x46 - 0x47
a	0: use current position of point p
	1: use the position of point p in the original outline
Pops	p: point number (ULONG)
Pushes	value: coordinate location (F26Dot6)
Uses	zp2, projection_vector

Measures the coordinate value of point p on the current projection_vector and pushes the value onto the stack.

Example

The following example shows that the value returned by GC is dependent upon the current position of the projection_vector. Note that point p is at the position (300,420) in the coordinate grid.

GC[1] 9

Sets Coordinate From the Stack using projection_vector and freedom_vector

SCFS[ ]
Code Range	0x48
Pops	value: distance from origin to move point (F26Dot6)
	p: point number (ULONG)
Pushes	-
Uses	zp2, freedom_vector, projection_vector

Moves point p from its current position along the freedom_vector so that its component along the projection_vector becomes the value popped off the stack.

Measure Distance

MD[a]
Code Range	0x49 - 0x4A
a	0: measure distance in grid-fitted outline
	1: measure distance in original outline
Pops	p1: point number (ULONG)
	p2: point number (ULONG)
Pushes	distance (F26Dot6)
Uses	zp1 with point p1, zp0 with point p2, projection_vector

Measures the distance between outline point p1 and outline point p2. The value returned is in pixels (F26Dot6) If distance is negative, it was measured against the projection vector. Reversing the order in which the points are listed will change the sign of the result.

Example:

In the illustration below MD[1] between points 25 and 31 will return a smaller value than MD[0] at 10 pixels per em on a 72 dpi device. The difference is due to the effects of grid-fitting which, at this size, stretches out the counter.

Measure Pixels Per EM

MPPEM[ ]
Code Range	0x4B
Pops	-
Pushes	ppem: pixels per em (ULONG)

This instruction pushes the number of pixels per em onto the stack. Pixels per em is a function of the resolution of the rendering device and the current point size and the current transformation matrix. This instruction looks at the projection_vector and returns the number of pixels per em in that direction.

Measure Point Size

MPS[ ]
Code Range	0x4C
Pops	-
Pushes	pointSize: the size in points of the current glyph (F26Dot6)

Pushes the current point size onto the stack.

Measure point size can be used to obtain a value which serves as the basis for choosing whether to branch to an alternative path through the instruction stream. It makes it possible to treat point sizes below or above a certain threshold differently

Managing outlines

The following set of instructions make it possible to move the points that make up a glyph outline. They are the instructions that accomplish the actual work of grid-fitting. They include instructions to move points, shift points or groups of points, flip points from off to on the curve or vice versa, and to interpolate points

FLIP PoinT

FLIPPT[ ]
Code Range	0x80
Pops	p: point number (ULONG)
Pushes	-
Uses	loop, p is referenced in zp0

Flips points that are off the curve so that they are on the curve and points that are on the curve so that they are off the curve. The point is not marked as touched. The result of a FLIPPT instruction is that the contour describing part of a glyph outline is redefined

FLIP RanGe ON

FLIPRGON[ ]
Code Range	0x81
Pops	highpoint: highest point number in range of points to be flipped (ULONG)
	lowpoint: lowest point number in range of points to be flipped (ULONG)
Pushes	-

Flips a range of points beginning with lowpoint and ending with highpoint so that any off the curve points become on the curve points. The points are not marked as touched

Example:

Will make all off curve points between point 0 and point 5 into on curve points as shown

FLIP RanGe OFF

FLIPRGOFF[ ]
Code Range	0x82
Pops	highpoint: highest point number in range of points to be flipped (ULONG)
	lowpoint: lowest point number in range of points to be flipped (ULONG)
Pushes	-

Flips a range of points beginning with lowpoint and ending with highpoint so that any off the curve points become on the curve points. The points are not marked as touched.

NOTE: This instruction changes the curve but the position of the points is unaffected. Accordingly, points affected by this instruction are not marked as touched.

Example:

SHift Point by the last point

SHP[a]
Code Range	0x32 - 0x33
a	0: uses rp2 in the zone pointed to by zp1
	1: uses rp1 in the zone pointed to by zp0
Pops	p: point to be shifted (ULONG)
Pushes	-
Uses	zp0 with rp1 or zp1 with rp2 depending on flag
	zp2 with point p
	loop, freedom_vector, projection_vector

Shift point p by the same amount that the reference point has been shifted. Point p is shifted along the freedom_vector so that the distance between the new position of point p and the current position of point p is the same as the distance between the current position of the reference point and the original position of the reference point.

NOTE: Point p is shifted from its current position, not its original position. The distance that the reference point has shifted is measured between its current position and the original position.

In the illustration below rp is the original position of the reference point, rp' is the current position of the reference point, p is the original position of point p, p' is the current position, p" the position after it is shifted by the SHP instruction. (White indicates original position, gray is current position, black is position to which this instruction moves a point

SHift Contour by the last point

SHC[a]
Code Range	0x34 - 0x35
a	0: uses rp2 in the zone pointed to by zp1
	1: uses rp1 in the zone pointed to by zp0
Pops	c: contour to be shifted (ULONG)
Pushes	-
Uses	zp0 with rp1 or zp1 with rp2 depending on flag
	zp2 with contour c
	freedom_vector, projection_vector

Shifts every point on contour c by the same amount that the reference point has been shifted. Each point is shifted along the freedom_vector so that the distance between the new position of the point and the old position of that point is the same as the distance between the current position of the reference point and the original position of the reference point. The distance is measured along the projection_vector. If the reference point is one of the points defining the contour, the reference point is not moved by this instruction.

This instruction is similar to SHP, but every point on the contour is shifted.

SHift Zone by the last pt

SHZ[a]
Code Range	0x36 - 0x37
a	0: the reference point rp2 is in the zone pointed to by zp1
	1: the reference point rp1 is in the zone pointed to by zp0
Pops	e: zone to be shifted (ULONG)
Pushes	-
Uses	zp0 with rp1 or zp1 with rp2 depending on flag
	freedom_vector, projection_vector

Shift the points in the specified zone (Z1 or Z0) by the same amount that the reference point has been shifted. The points in the zone are shifted along the freedom_vector so that the distance between the new position of the shifted points and their old position is the same as the distance between the current position of the reference point and the original position of the reference point.

SHZ[a] uses zp0 with rp1 or zp1 with rp2. This instruction is similar to SHC, but all points in the zone are shifted, not just the points on a single contour.

SHift point by a PIXel amount

SHPIX[ ]
Code Range	0x38
Pops	amount: magnitude of the shift (F26Dot6)
	p1, p2,...pn: points to be shifted (ULONG)
Pushes	-
Uses	zp2, loop, freedom_vector

Shifts the points specified by the amount stated. When the loop variable is used, the amount to be shifted is put onto the stack only once. That is, if loop = 3, then the contents of the top of the stack should be point p1, point p2, point p3, amount. The value amount is expressed in sixty-fourths of a pixel.

SHPIX is unique in relying solely on the direction of the freedom_vector. It makes no use of the projection_vector. Measurement is made in the direction of the freedom_vector.

Example

The instruction shifts points 27, 28, and 29 by 80/64 or 1.25 pixels in the direction of the freedom vector. The distance is measured in the direction of the freedom_vector; the projection vector is ignored.

SHPIX[]

Move Stack Indirect Relative Point

MSIRP[a]
Code Range	0x3A - 0x3B
a	0: Do not set rp0 to p
	1: Set rp0 to p
Pops	d: distance (F26Dot6)
	p: point number (ULONG)
Pushes	-
Uses	zp1 with point p and zp0 with rp0, freedom_vector, projection_vector.
Sets	After it has moved the point this instruction sets rp1 = rp0, rp2 = point p, and if a=1, rp0 is set to point p.

Makes the distance between a point p and rp0 equal to the value specified on the stack. The distance on the stack is in fractional pixels (F26Dot6). An MSIRP has the same effect as a MIRP instruction except that it takes its value from the stack rather than the Control Value Table. As a result, the cut_in does not affect the results of a MSIRP. Additionally, MSIRP is unaffected by the round_state.

Move Direct Absolute Point

MDAP[a]
Code Range	0x2E - 0x2F
a	0: do not round the value
	1: round the value
Pops	p: point number (ULONG)
Pushes	-
Sets	rp0 = rp1 = point p
Uses	zp0, round_state, projection_vector, freedom_vector.

Sets the reference points rp0 and rp1 equal to point p. If a=1, this instruction rounds point p to the grid point specified by the state variable round_state. If a=0, it simply marks the point as touched in the direction(s) specified by the current freedom_vector. This command is often used to set points in the twilight zone.

Example:

MDAP[0]

When a=0, the point is simply marked as touched and the values of rp0 and rp1 set to point p.

MDAP[1] assuming that the round_state is round to grid and the freedom_vector is a shown.

Move Indirect Absolute Point

MIAP[a]
Code Range	0x3E - 0x3F
a	0: don’t round the distance and don’t look at the control_value_cut_in
	1: round the distance and look at the control_value_cut_in
Pops	n: CVT entry number (ULONG)
	p: point number (ULONG)
Pushes	-
Sets	rp0 = rp1 = point p
Uses	zp0, round_state, control_value_cut_in, freedom_vector, projection_vector

Moves point p to the absolute coordinate position specified by the nth Control Value Table entry. The coordinate is measured along the current projection_vector. If a=1, the position will be rounded as specified by round_state. If a=1, and if the device space difference between the CVT value and the original position is greater than the control_value_cut_in, then the original position will be rounded (instead of the CVT value.)

Rounding is done as if the entire coordinate system has been rotated to be consistent with the projection_vector. That is, if round_state is set to 1, and the projection_vector and freedom_vector are at a 45° angle to the x-axis, then a MIAP[1] of a point to 2.9 pixels will round to 3.0 pixels along the projection_vector.

The a Boolean above controls both rounding and the use of the control_value_cut_in. If you would like the meaning of this Boolean to specify only whether or not the MIAP[ ] instruction should look at the control_value_cut_in value, use the ROFF[ ] instruction to turn off rounding.

This instruction can be used to create Twilight Zone points.

Example:

MIAP[1] 4 7

case 1:

rounding is OFF

The point is moved to the position specified in the CVT.

case 2:

The cut_in test succeeds and rounding is RTG.

The value in the CVT is subjected to the rounding rule and then the point is moved to the rounded position.

case 3:

The cut_in test fails and rounding is OFF.

Here the point is not moved.

case 4:

The cut_in test fails and rounding is RTG.

In this case the point is moved to the nearest grid position.

Move Direct Relative Point

MDRP[abcde]
Code Range	0xC0 - 0xDF
a	0: do not set rp0 to point p after move
	1: do set rp0 to point p after move
b	0: do not keep distance greater than or equal to minimum_distance
	1: keep distance greater than or equal to minimum_distance
c	0: do not round distance
	1: round the distance
de	distance type for engine characteristic compensation
Pops	p: point number (ULONG)
Pushes	-
Sets	after point p is moved, rp1 is set equal to rp0, rp2 is set equal to point p; if the a flag is set to TRUE, rp0 is set equal to point
Uses	zp0 with rp0 and zp1 with point p, round_state, single_width_value, single_width_cut_in, freedom_vector, projection_vector.

MDRP moves point p along the freedom_vector so that the distance from its new position to the current position of rp0 is the same as the distance between the two points in the original uninstructed outline, and then adjusts it to be consistent with the Boolean settings. Note that it is only the original positions of rp0 and point p and the current position of rp0 that determine the new position of point p along the freedom_vector.

MDRP is typically used to control the width or height of a glyph feature using a value which comes from the original outline. Since MDRP uses a direct measurement and does not reference the control_value_cut_in, it is used to control measurements that are unique to the glyph being instructed. Where there is a need to coordinate the control of a point with the treatment of points in other glyphs in the font, a MIRP instruction is needed.

Though MDRP does not refer to the CVT, its effect does depend upon the single-width cut-in value. If the device space distance between the measured value taken from the uninstructed outline and the single_width_value is less than the single_width_cut_in, the single_width_value will be used in preference to the outline distance. In other words, if the two distances are sufficiently close (differ by less than the single_width_cut_in), the single_width_value will be used.

The setting of the round_state Graphics State variable will determine whether and how the distance of point p from point q is rounded. If the round bit is not set, the value will be unrounded. If the round bit is set, the effect will depend upon the choice of rounding state. The value of the minimum distance variable is the smallest possible value the distance between two points can be rounded to.

Distances measured with the MDRP instruction must be either black, white or gray. Indicating this value in Booleans de allows the interpreter to compensate for engine characteristics as needed. The value de specifies the distance type as described in the chapter, “Instructing Glyphs.” Three values are possible: Gray=0, Black=1, White=2.

Example 1:

Graphics State Settings

rp0	7
rp1	?
rp2	?

Case 1:

rp0	7
rp1	7
rp2	8

Case 2:

rp0	8
rp1	7
rp2	8

Example 2:

Point p is moved so that its distance from rp1 is the same as it was in the original outline.

Move Indirect Relative Point

MIRP[abcde]
Code Range	0xE0 - 0xFF
a	0: Do not set rp0 to p
	1: Set rp0 to p
b	0: Do not keep distance greater than or equal to minimum_distance
	1: Keep distance greater than or equal to minimum_distance
c	0: Do not round the distance and do not look at the control_value_cut_in
	1: Round the distance and look at the control_value_cut_in value
de	distance type for engine characteristic compensation
Pops	n: CVT entry number (ULONG)
	p: point number (ULONG)
Pushes	-
Uses	zp0 with rp0 and zp1 with point p. round_state, control_value_cut_in, single_width_value, single_width_cut_in, freedom_vector, projection_vector
Sets	After it has moved the point this instruction sets rp1 = rp0, rp2 = point p, and if a = 1, rp0 is set to point p.

A MIRP instruction makes it possible to preserve the distance between two points subject to a number of qualifications. Depending upon the setting of Boolean flag b, the distance can be kept greater than or equal to the value established by the minimum_distance state variable. Similarly, the instruction can be set to round the distance according to the round_state graphics state variable. The value of the minimum distance variable is the smallest possible value the distance between two points can be rounded to. Additionally, if the c Boolean is set, the MIRP instruction acts subject to the control_value_cut_in. If the difference between the actual measurement and the value in the CVT is sufficiently small (less than the cut_in_value), the CVT value will be used and not the actual value. If the device space difference between this distance from the CVT and the single_width_value is smaller than the single_width_cut_in, then use the single_width_value rather than the outline or Control Value Table distance.

MIRP measures distance relative to point rp0. More formally, MIRP moves point p along the freedom_vector so that the distance from p to rp0 is equal to the distance stated in the reference CVT entry (assuming that the cut_in test succeeds)

The c Boolean above controls both rounding and the use of Control Value Table entries. If you would like the meaning of this Boolean to specify only whether or not the MIRP[ ] instruction should look at the control_value_cut_in, use the ROFF[ ] instruction to turn off rounding. In this manner, it is possible to specify rounding off and no cut_in.

The value de specifies the distance type as described in th chapter, “Instructing Glyphs.” Three values are possible: Gray=0, Black=1, White=2.

Example 1:

MIRP[00110] 3 17

case 1:

The cut_in test succeeds and rounding is off.

The point is moved so that the distance from RP0 is equal to that given in CVT entry 17.

case 2:

The cut_in test succeeds and rounding is set to RTG.

The distance in the CVT is rounded and the point is moved by the rounded distance.

case 3:

The cut_in test fails and the round_state is OFF.

The point is not moved.

case 4:

The cut_in test fails and the round_state is RTG.

The current position of the point is rounded to the grid.

ALIGN Relative Point

ALIGNRP[ ]
Code Range	0x3C
Pops	p: point number (ULONG
Pushes	-
Uses	zp1 with point p, zp0 with rp0, loop, freedom_vector, projection_vector.

Reduces the distance between rp0 and point p to zero. Since distance is measured along the projection_vector and movement is along the freedom_vector, the effect of the instruction is to align points.

case 1:

case 2:

moves point p to the InterSECTion of two lines

ISECT[ ]
Code Range	0x0F
Pops	b1: end point of line 2 (ULONG)
	b0: start point of line 2 (ULONG)
	a1: end point of line 1 (ULONG)
	a0: start point of line 1 (ULONG)
	p: point to move (ULONG)
Pushes	-
Uses	zp2 with point p, zp1 with line A, zp0 with line B

Puts point p at the intersection of the lines A and B. The points a0 and a1 define line A. Similarly, b0 and b1 define line B. ISECT ignores the freedom_vector in moving point p.

Example:

ISECT[ ] 21 9 5 4 7

Example:

ISECT[ ] 21 9 5 4 7

ALIGN Points

ALIGNPTS[ ]
Code Range	0x27
Pops	p1: point number (ULONG)
	p2: point number (ULONG)
Pushes	-
Uses	zp1 with point p1, zp0 with point p2, freedom_vector, projection_vector.

Makes the distance between point 1 and point 2 zero by moving both along the freedom_vector to the average of both their projections along the projection_vector.

Example:

ALIGNPTS[] 3 7

Interpolate Point by the last relative stretch

IP[ ]
Code Range	0x39
Pops	p: point number (ULONG)
Pushes	-
Uses	zp0 with rp1, zp1 with rp2, zp2 with point p, loop, freedom_vector, projection_vector

Moves point p so that its relationship to rp1 and rp2 is the same as it was in the original uninstructed outline. Measurements are made along the projection_vector, and movement to satisfy the interpolation relationship is constrained to be along the freedom_vector. This instruction is illegal if rp1 and rp2 have the same position on the projection_vector.

In the example shown, assume that the points referenced by rp1 and rp2 are moved as shown. An IP instruction is then used to preserve their relative relationship with point p. After the IP the following should be true

• D(p, rp1)/D(p',rp1') = D(p,rp2)/D(p', rp2')

In other words, the relative distance is preserved.

UnTouch Point

UTP[ ]
Code Range	0x29
Pops	p: point number (ULONG)
Pushes	-
Uses	zp0 with point p, freedom_vector

Marks point p as untouched. A point may be touched in the x direction, the y direction, both, or neither. This instruction uses the current freedom_vector to determine whether to untouch the point in the x-direction, the y direction, or both. Points that are marked as untouched will be moved by an IUP (interpolate untouched points) instruction. Using UTP you can ensure that a point will be affected by IUP even if it was previously touched.

Interpolate Untouched Points through the outline

IUP[a]
Code Range	0x30 - 0x31
a	0: interpolate in the y-direction
	1: interpolate in the x-direction
Pops	-
Pushes	-
Uses	zp2, freedom_vector, projection_vector

Considers a glyph contour by contour, moving any untouched points in each contour that are between a pair of touched points. If the coordinates of an untouched point were originally between those of the touched pair, it is linearly interpolated between the new coordinates, otherwise the untouched point is shifted by the amount the nearest touched point is shifted.

This instruction operates on points in the glyph zone pointed to by zp2. This zone should almost always be zone 1. Applying IUP to zone 0 is an error.

Consider three consecutive points all on the same contour. Two of the three points, p1 and p3 have been touched. Point p2 is untouched. The effect of an IUP in the x-direction is to move point p2 so that is in the same relative position to points p1 and p3 before they were moved.

The IUP instruction does not touch the points it moves. Thus the untouched points affected by an IUP instruction will be affected by subsequent IUP instructions unless they are touched by an intervening instruction. In this case, the first interpolation is ignored and the point is moved based on its original position.

Managing exceptions

DELTA instructions can be used to alter the outline of a glyph at a particular size. They are generally used to turn on or off specific pixels. Delta instructions work by moving points (DELTAP’s) or by changing a value in the Control Value Table (DELTAC’s).

More formally, the DELTA instructions take a variable number of arguments from the stack and allow the use of an exception of the form: at size x apply the movement d to point p (or at size x add or subtract an amount less than or equal to the Control Value Table entry c). DELTAs take a list of exceptions of the form: relative ppem value, the magnitude of the exception and the point number to which the exception is to be applied.

Each DELTA instruction works on a range of sizes as specified below. As a result, sizes are specified in relative pixels per em (ppem), that is relative to the delta_base. The default value for delta_base is 9 ppem. To set delta_base to another value, use the SDB instruction.

The DELTAP1 and DELTAC1 instructions allow values to be changed for glyphs at 9 through 24 ppem, assuming the default value for delta_base. Lowering the value for delta_base allows you to invoke exceptions at a smaller number of ppem.

DELTAP2 and DELTAC2 are triggered at 16 ppem higher than the value set for DELTAP1 and DELTAC1, and consequently the formula for the relative ppem is

ppem - 16 - delta_base.

DELTAP3 and DELTAC3 are triggered at 16 ppem higher than the value set for DELTAP2 and DELTAC2, or 32 ppem higher than the value set for DELTAP1 and DELTAC1, and consequently the formula for the relative ppem is:

ppem - 32 - delta_base.

DELTA*1	delta_base through delta_base + 15 ppem
DELTA*2	delta_base + 16 ppem through delta_base + 31 ppem
DELTA*3	delta_base + 32 ppem through delta_base + 47 ppem

In specifying a DELTA instruction, the high 4 bits of arg₁ describe the relative ppem value that will activate the exception.

The low 4 bits of arg₁ describe the magnitude of the exception. The amount the point moves is a function of the exception stated and the Graphics State variable delta_shift. To set delta_shift, use the SDS instruction

In the descriptions of the instructions that follow, p_i is a point number, c_i is a Control Value Table entry number and arg_i is a byte composed of two parts: the relative ppem (ppem - delta_base) and the magnitude of the exception.

Increasing the delta_shift will allow for more fine control over pixel movement at the sacrifice of total range of movement. A step is the minimal amount that a delta instruction can move a point. Points can be moved in integral multiples of steps.

The size of a step is 1 divided by 2 to the power delta_shift. The range of movement produced by a given delta_shift can be calculated by taking the number of steps allowed (16) and dividing it by 2 to the power delta_shift. For example, a delta_shift equal to 2 allows the smallest movement to be ± 1/4 pixel (because 2² equals 4) and the largest movement to be ± 2 pixels (16/4 = 4 pixels of movement). A delta_shift of 5 allows the smallest movement to be ± 1/32 pixel (because 2⁵ equals 32), but the largest movement is limited to ± 1/4 pixel. (16/32 = 1/2 a pixel of movement).

Internally, the value obtained for the exception is stored as a 4 bit binary number. As a result, the desired output range must be converted to a number between 0 and 15 before being converted to binary. Here is the internal remapping table for the DELTA instructions.

NOTE: that zero is lacking in the output range.

Number of Steps	→ Exception
-8	0
-7	1
-6	2
-5	3
-4	4
-3	5
-2	6
-1	7
1	8
2	9
3	10
4	11
5	12
6	13
7	14
8	15

DELTA exception P1

DELTAP1[ ]
Code Range	0x5D
Pops	n: number of pairs of exception specifications and points (ULONG)
	p1, arg1, p2, arg2, ..., pnn argn: n pairs of exception specifications and points (pairs of ULONGs)
Pushes	-
Uses	zp0, delta_base, delta_shift

DELTAP1 moves the specified points at the size and by the amount specified in the paired argument. An arbitrary number of points and arguments can be specified.

The grouping [p_i, arg_i] can be executed n times. The value of argi may vary between iterations.

DELTA exception P2

DELTAP2[ ]
Code Range	0x71
Pops	n: number of pairs of exception specifications and points (ULONG)
	p1, arg1, p2, arg2, ..., pnn argn: n pairs of exception specifications and points (pairs of ULONGs)
Pushes	-
Uses	zp0, delta_shift, delta_base

DELTAP2 moves the specified points at the size and by the amount specified in the paired argument. An arbitrary number of points and arguments can be specified.

The grouping [p_i, arg_i] can be executed n times. The value of arg_i may vary between iterations.

DELTA exception P3

DELTAP3[ ]
Code Range	0x72
Pops	n: number of pairs of exception specifications and points (ULONG)
	p1, arg1, p2, arg2, ..., pnn argn: n pairs of exception specifications and points (pairs of ULONGs)
Pushes	-
Uses	zp0, delta_base, delta_shift

DELTAP3 moves the specified points at the size and by the amount specified in the paired argument. An arbitrary number of point and arguments can be specified.

The grouping [p_i, arg_i] can be executed n times. The value of arg_i may vary between iterations.

DELTA exception C1

DELTAC1[ ]
Code Range	0x73
Pops	n: number of pairs of exception specifications and CVT entry numbers (ULONG)
	c1, arg1, c2, arg2,..., cn, argn: (pairs of ULONGs)
Pushes	-

DELTAC1 changes the value in each CVT entry specified at the size and by the amount specified in its paired argument.

The grouping [c_i, arg_i] can be executed n times. The value of arg_i may vary between iterations.

DELTA exception C2

DELTAC2[ ]
Code Range	0x74
Pops	n: number of pairs of exception specifications and CVT entry numbers (ULONG)
	c1, arg1, c2, arg2,..., cn, argn: (pairs of ULONGs)
Pushes	-

DELTAC2 changes the value in each CVT entry specified at the size and by the amount specified in its paired argument.

The grouping [c_i, arg_i] can be executed n times. The value of arg_i may vary between iterations.

DELTA exception C3

DELTAC3[ ]
Code Range	0x75
Pops	n: number of pairs of exception specifications and CVT entry numbers (ULONG)
	c1, arg1, c2, arg2,..., cn, argn: (pairs of ULONGs)
Pushes	-

DELTAC3 changes the value in each CVT entry specified at the size and by the amount specified in its paired argument.

The grouping [c_i, arg_i] can be executed n times. The value of arg_i may vary between iterations.

Example of DELTA exceptions

Assume that you want to move point 15 of your glyph 1/8 of a pixel along the freedom_vector at 12 pixels per em. Assume that delta_base has the default value 9 and delta_shift the default value 3.

To specify that the exception should be made at 12 ppem, you subtract the delta_base, which is 9, from 12 and store the result, which is 3, in the high nibble of arg_i.

To specify that the point is to be moved 1/8 of a pixel, multiply 1/8 by 2 raised to the power delta_shift. In other words, you multiply 1/8 by 2 raised to the third power (or 8)yielding 1. This value must be mapped to an internal value which using the table shown is 8.

Putting these two results together yields a 3 in the high nibble and an 8 in the low nibble or 56 (00111000, in binary).

To obtain this single exception, the top of the stack is: 56, 15, 1.

    (iteration)
    (point number)
    (arg1: ppem and magnitude)

Now if the interpreter executes

DELTAP1[ ]

then this instruction will move point 15 of the glyph (at 12 ppem) 1/8 of a pixel along the freedom_vector.

Managing the stack

The following set of instructions make it possible to manage elements on the stack. They make it possible to duplicate the element at the top of the stack, remove the top element from the stack, clear the stack, swap the top two stack elements, determine the number of elements currently on the stack, copy a specified element to the top of the stack, move a specified element to the top of the stack, and rearrange the order of the top three elements on the stack.

Duplicate top stack element

DUP[ ]
Code Range	0x20
Pops	e: stack element (ULONG)
Pushes	e, e (two ULONGs)

Duplicates the element at the top of the stack.

POP top stack element

POP[ ]
Code Range	0x21
Pops	e: stack element (ULONG)
Pushes	-

Pops the top element of the stack.

Clear the entire stack

CLEAR[ ]
Code Range	0x22
Pops	all the items on the stack (ULONGs)
Pushes	-

Clears all elements from the stack.

SWAP the top two elements on the stack

SWAP[ ]
Code Range	0x23
Pops	e2: stack element (ULONG)
	e1: stack element (ULONG)
Pushes	e1, e2 (pair of ULONGs)

Swaps the top two elements of the stack making the old top element the second from the top and the old second element the top element.

Returns the DEPTH of the stack

DEPTH[ ]
Code Range	0x24
Pops	-
Pushes	n: number of elements (ULONG)

Pushes n, the number of elements currently in the stack onto the stack.

Example:

Copy the INDEXed element to the top of the stack

CINDEX[ ]
Code Range	0x25
Pops	k : stack element number
Pushes	ek: indexed element (ULONG)

Puts a copy of the kth stack element on the top of the stack.

Example:

Move the INDEXed element to the top of the stack

MINDEX[ ]
Code Range	0x26
Pops	k : stack element number
Pushes	ek: indexed element

Moves the indexed element to the top of the stack.

MINDEX[ ]

ROLL the top three stack elements

ROLL[ ]
Code Range	0x8a
Pops	a, b, c (top three stack elements)
Pushes	b, a, c (elements reordered)

Performs a circular shift of the top three objects on the stack with the effect being to move the third element to the top of the stack and to move the first two elements down one position. ROLL is equivalent to MINDEX[ ] 3.

Managing the flow of control

This section describes those instructions that make it possible to alter the sequence in which items in the instruction stream are executed. The IF and JMP instructions and their variants work by testing the value of an element on the stack and changing the value of the instruction pointer accordingly.

IF test

IF[ ]
Code Range	0x58
Pops	e: stack element (ULONG)
Pushes	-

Tests the element popped off the stack: if it is zero (FALSE), the instruction pointer is jumped to the next ELSE or EIF instruction in the instruction stream. If the element at the top of the stack is nonzero (TRUE), the next instruction in the instruction stream is executed. Execution continues until an ELSE instruction is encountered or an EIF instruction ends the IF. If an else statement is found before the EIF, the instruction pointer is moved to the EIF statement.

case 1:

Element at top of stack is TRUE; instruction pointer is unaffected. IF terminates with EIF.

case 2:

Element at top of stack is TRUE. The instruction stream is sequentially executed until ELSE is encountered whereupon the instruction pointer jumps to the EIF statement that terminates the IF.

case 3:

Element at the top of the stack is FALSE; instruction pointer is moved to the ELSE statement; instructions are then executed sequentially; EIF ends the IF statement.

ELSE

ELSE[ ]
Code Range	0x1B
Pops	-
Pushes	-

Marks the start of the sequence of instructions that are to be executed if an IF instruction encounters a FALSE value on the stack. This sequence of instructions is terminated with an EIF instruction.

End IF

EIF[ ]
Code Range	0x59
Pops	-
Pushes	-

Marks the end of an IF[ ] instruction.

Jump Relative On True

JROT[ ]
Code Range	0x78
Pops	b: Boolean (ULONG)
	offset: number of bytes to move instruction pointer (LONG)
Pushes	-

Obtains an offset and tests a Boolean value. If the Boolean is TRUE, the signed offset will be added to the instruction pointer and execution will be resumed at the address obtained. Otherwise, the jump is not taken. The jump is relative to the position of the instruction itself. That is, the instruction pointer is still pointing at the JROT[ ] instruction when offset is added to obtain the new address.

Example:

case 1:

Boolean is FALSE.

case 2:

Boolean is TRUE.

JuMP

JMPR[ ]
Code Range	0x1C
Pops	offset: number of bytes to move instruction pointer (LONG)
Pushes	-

The signed offset is added to the instruction pointer and execution is resumed at the new location in the instruction steam. The jump is relative to the position of the instruction itself. That is, the instruction pointer is still pointing at the JROT[ ] instruction when offset is added to obtain the new address.

Jump Relative On False

JROF[ ]
Code Range	0x79
Pops	e: stack element (ULONG)
	offset: number of bytes to move instruction pointer (LONG)
Pushes	-

In the case where the Boolean is FALSE, the signed offset will be added to the instruction pointer and execution will be resumed there; otherwise, the jump is not taken. The jump is relative to the position of the instruction itself. That is, the instruction pointer is still pointing at the JROT[ ] instruction when offset is added to obtain the new address.

case 1:

element is FALSE.

case 2:

element is TRUE.

Logical functions

The TrueType instruction set includes a set of logical functions that can be used to test the value of a stack element or to compare the values of two stack elements. The logical functions compare 32 bit values (ULONG) and return a Boolean value to the top of the stack.

To easily remember the order in which stack values are handled during logical operations, imagine writing the stack values from left to right, starting with the bottom value. Then insert the operator between the two furthest right elements. For example:

GT a,b would be interpreted as (b>a):

c b > a

Less Than

LT[ ]
Code Range	0x50
Pops	e2: stack element (ULONG)
	e1: stack element (ULONG)
Pushes	Boolean value (ULONG in the range [0,1])

First pops e2, then pops e1 off the stack and compares them: if e1 is less than e2, 1, signifying TRUE, is pushed onto the stack. If e1 is not less than e2, 0, signifying FALSE, is placed onto the stack.

Example:

LT[ ]

Less Than or Equal

LTEQ[ ]
Code Range	0x51
Pops	e2: stack element (ULONG)
	e1: stack element (ULONG)
Pushes	Boolean value (ULONG in the range [0,1])

Pops e2 and e1 off the stack and compares them. If e1 is less than or equal to e2, 1, signifying TRUE, is pushed onto the stack. If e1 is not less than or equal to e2, 0, signifying FALSE, is placed onto the stack.

Example:

LTEQ[ ]

Greater Than

GT[ ]
Code Range	0x52
Pops	e2: stack element (ULONG)
	e1: stack element (ULONG)
Pushes	Boolean value (ULONG in the range [0,1])

First pops e2 then pops e1 off the stack and compares them. If e1 is greater than e2, 1, signifying TRUE, is pushed onto the stack. If e1 is not greater than e2, 0, signifying FALSE, is placed onto the stack.

Example:

GT[ ]

Greater Than or Equal

GTEQ[ ]
Code Range	0x53
Pops	e2: stack element (ULONG)
	e1: stack element (ULONG)
Pushes	Boolean value (ULONG in the range [0,1])

Pops e1 and e2 off the stack and compares them. If e1 is greater than or equal to e2, 1, signifying TRUE, is pushed onto the stack. If e1 is not greater than or equal to e2, 0, signifying FALSE, is placed onto the stack.

Example:

GTEQ[ ]

EQual

EQ[ ]
Code Range	0x54
Pops	e2: stack element (ULONG)
	e1: stack element (ULONG)
Pushes	Boolean value (ULONG in the range [0,1])

Pops e1 and e2 off the stack and compares them. If they are equal, 1, signifying TRUE is pushed onto the stack. If they are not equal, 0, signifying FALSE is placed onto the stack.

Example:

EQ[ ]

Not EQual

NEQ[ ]
Code Range	0x55
Pops	e2: stack element (ULONG)
	e1: stack element (ULONG)
Pushes	Boolean value (ULONG in the range [0,1])

Pops e1 and e2 from the stack and compares them. If they are not equal, 1, signifying TRUE, is pushed onto the stack. If they are equal, 0, signifying FALSE, is placed on the stack.

Example:

NEQ[ ]

ODD

ODD[ ]
Code Range	0x56
Pops	e1: stack element (F26Dot6)
	Boolean value
Pushes	Boolean value (ULONG in the range [0,1])
Uses	round_state

Tests whether the number at the top of the stack is odd. Pops e1 from the stack and rounds it as specified by the round_state before testing it. After the value is rounded, it is shifted from a fixed point value to an integer value (any fractional values are ignored). If the integer value is odd, one, signifying TRUE, is pushed onto the stack. If it is even, zero, signifying FALSE is placed onto the stack.

Example:

ODD[ ]

This example assumes that round_state is RTG.

EVEN

EVEN[ ]
Code Range	0x57
Pops	e1: stack element (F26Dot6)
Pushes	Boolean value (ULONG in the range [0,1])
Uses	round_state

Tests whether the number at the top of the stack is even. Pops e1 off the stack and rounds it as specified by the round_state before testing it. If the rounded number is even, one, signifying TRUE, is pushed onto the stack if it is odd, zero, signifying FALSE, is placed onto the stack.

Example:

EVEN[ ]

This example assumes that round_state is RTG.

logical AND

AND[ ]
Code Range	0x5A
Pops	e1: stack element (ULONG)
	e2: stack element (ULONG)
Pushes	( e1 and e2 ): logical and of e1 and e2 (ULONG)

Pops e1 and e2 off the stack and pushes onto the stack the result of a logical and of the two elements. Zero is returned if either or both of the elements are FALSE (have the value zero). One is returned if both elements are TRUE (have a non zero value).

Example:

case 1:

AND[ ]

case 2:

AND[ ]

logical OR

OR[ ]
Code Range	0x5B
Pops	e1: stack element (ULONG)
	e2: stack element (ULONG)
Pushes	( e1 or e2 ): logical and of e1 and e2 (ULONG)

Pops e1 and e2 off the stack and pushes onto the stack the result of a logical or operation between the two elements. Zero is returned if both of the elements are FALSE. One is returned if either both of the elements are TRUE.

Example:

case 1:

OR[ ]

case 2:

OR[ ]

logical NOT

NOT[ ]
Code Range	0x5C
Pops	e: stack element (ULONG)
Pushes	(not e): logical negation of e (ULONG)

Pops e off the stack and returns the result of a logical NOT operation performed on e. If originally zero, one is pushed onto the stack if originally nonzero, zero is pushed onto the stack.

Example:

case 1:

case 2:

Arithmetic and math instructions

These instructions perform arithmetic on stack values. Values are treated as signed (two’s complement) 26.6 fixed-point numbers (F26Dot6) and give results in the same form. There is no overflow or underflow protection for these instructions.

To easily remember the order in which stack values are handled during arithmetic operations, imagine writing the stack values from left to right, starting with the bottom value. Then insert the operator between the two furthest right elements. For example:

subtract a,b would be interpreted as (b-a): c b - a

ADD

ADD[ ]
Code Range	0x60
Pops	n1, n2 (F26Dot6)
Pushes	(n2 + n1)

Pops n1 and n2 off the stack and pushes the sum of the two elements onto the stack.

SUBtract

SUB[ ]
Code Range	0x61
Pops	n1, n2 (F26Dot6)
Pushes	(n2 - n1): difference

Pops n1 and n2 off the stack and pushes the difference between the two elements onto the stack.

DIVide

DIV[ ]
Code Range	0x62
Pops	n1: divisor (F26Dot6)
	n2: dividend (F26Dot6)
Pushes	n2/n1 (F26Dot6)

Pops n1 and n2 off the stack and pushes onto the stack the quotient obtained by dividing n2 by n1. Note that this truncates rather than rounds the value. The TrueType Rasterizer v.1.7 and later will catch any division-by-zero errors.

MULtiply

MUL[ ]
Code Range	0x63
Pops	n1, n2: multiplier and multiplicand (F26Dot6)
Pushes	n1 * n2 (F26Dot6)

Pops n1 and n2 off the stack and pushes onto the stack the product of the two elements.

ABSolute value

ABS[ ]
Code Range	0x64
Pops	n
Pushes	|n|: absolute value of n (F26Dot6)

Pops n off the stack and pushes onto the stack the absolute value of n.

Example:

case 1:

case 2:

NEGate

NEG[ ]
Code Range	0x65
Pops	n1
Pushes	-n1: negation of n1 (F26Dot6)

This instruction pops n1 off the stack and pushes onto the stack the negated value of n1.

NEG[ ]

FLOOR

FLOOR[ ]
Code Range	0x66
Pops	n1: number whose floor is desired (F26Dot6)
Pushes	n : floor of n1 (F26Dot6)

Pops n1 and returns n, the greatest integer value less than or equal to n1.

Example:

FLOOR[ ]

case 1:

case 2:

case 3:

CEILING

CEILING[ ]
Code Range	0x67
Pops	n1: number whose ceiling is desired (F26Dot6)
Pushes	n: ceiling of n1 (F26Dot6)

Pops n1 and returns n, the least integer value greater than or equal to n1. For instance, the ceiling of 15 is 15, but the ceiling of 15.3 is 16. The ceiling of -0.8 is 0. (n is the least integer value greater than or equal to n1)

Example:

CEILING[ ]

case 1:

case 2:

MAXimum of top two stack elements

MAX[ ]
Code Range	0x8B
Pops	e1: stack element (ULONG)
	e2: stack element (ULONG)
Pushes	maximum of e1 and e2

Pops two elements, e1 and e2, from the stack and pushes the larger of these two quantities onto the stack.

MINimum of top two stack elements

MIN[ ]
Code Range	0x8C
Pops	e1: stack element (ULONG)
	e2: stack element (ULONG)
Pushes	minimum of e1 and e2

Pops two elements, e1 and e2, from the stack and pushes the smaller of these two quantities onto the stack.

Compensating for the engine characteristics

The following two functions make it possible to compensate for the engine characteristic. Each takes value and make the compensation. In addition to the engine compensation, ROUND, rounds the value according to the round_state. NROUND only compensates for the engine.

ROUND value

ROUND[ab]
Flags	ab: distance type for engine characteristic compensation
Pops	n1
Pushes	n2
Code	0x68 - 0x6B

Rounds a value according to the state variable round_state while compensating for the engine. n1 is popped off the stack and, depending on the engine characteristics, is increased or decreased by a set amount. The number obtained is then rounded and pushed back onto the stack as n2.

The value ab specifies the distance type as described in the chapter, “Instructing Glyphs.” Three values are possible: Gray=0, Black=1, White=2.

No ROUNDing of value

NROUND[ab]
Flags	ab: distance type for engine characteristic compensation
Pops	n1
Pushes	n2
Code	0x6C - 0x6F

NROUND[ab] does the same operation as ROUND[ab] (above), except that it does not round the result obtained after compensating for the engine characteristics. n1 is popped off the stack and, depending on the engine characteristics, increases or decreases by a set amount. This figure is then pushed back onto the stack as n2.

The value ab specifies the distance type as described in the chapter, “Instructing Glyphs.” Three values are possible: Gray=0, Black=1, White=2.

Defining and using functions and instructions

The following instructions make it possible to define and use both functions and new instructions.

In addition to a simple call function, there is a loop and call function.

An IDEF or instruction definition call makes it possible to patch old scalers in order to add newly defined instructions.

Function DEFinition

FDEF[ ]
Code Range	0x2C
Pops	f: function identifier number (integer in the range 0 to n-1 where n is specified in the 'maxp' table )
Pushes	-

Marks the start of a function definition. The argument f is a number that uniquely identifies this function. A function definition can appear only in the Font Program or the CVT program. Functions may not exceed 64K in size

END Function definition

ENDF[ ]
Code Range	0x2D
Pops	-
Pushes	-

Marks the end of a function definition or an instruction definition.

CALL function

CALL[ ]
Code Range	0x2B
Pops	f: function identifier number (integer in the range 0 to n-1 where n is specified in the 'maxp' table )
Pushes	-

Calls the function identified by the number f.

LOOP and CALL function

LOOPCALL[ ]
Code Range	0x2A
Pops	f: function number integer in the range 0 to n-1 where n is specified in the 'maxp' table
	count: number of times to call the function (signed word)
Pushes	-

Calls the function f, count number of times.

Example:

Assume the Font Program contains this:

PUSHB(000), 17	push 17 onto the stack
FDEF[ ]	start defining function 17
...	contents of function
...
ENDF[ ]	end the definition
PUSHB(000),5	push count
PUSHB(000),17	push function number
LOOPCALL[ ]	call function17, 5 times

Instruction DEFinition

IDEF[ ]
Code Range	0x89
Pops	opcode (8 bit code padded with zeroes to ULONG)
Pushes	-

Begins the definition of an instruction. The instruction definition terminates when at ENDF, which is encountered in the instruction stream. Subsequent executions of the opcode popped will be directed to the contents of this instruction definition (IDEF). IDEFs should be defined in the Font Program or the CVT Program. An IDEF affects only undefined opcodes. If the opcode in question is already defined, the interpreter will ignore the IDEF. This is to be used as a patching mechanism for future instructions. Instructions may not exceed 64K in size.

Debugging

The TrueType instruction set provides the following instruction as an aid to debugging.

DEBUG call

DEBUG[ ]
Code Range	0x4F
Pops	number (ULONG)
Pushes	-

This instruction is only for debugging purposes and should not be a part of a finished font. Some implementations may not support this instruction.

Miscellaneous instructions

The following instruction obtains information about the current glyph and the scaler version.

GET INFOrmation

GETINFO[ ]
Code Range	0x88
Pops	selector (integer)
Pushes	result (integer)

GETINFO is used to obtain data about the font scaler version and the characteristics of the current glyph. The instruction pops a selector used to determine the type of information desired and pushes a result onto the stack.

A selector value of 1 indicates that the scaler version number is the desired information, a value of 2 is a request for glyph rotation status, a value of 4 asks whether the glyph has been stretched, a value of 32 asks whether the glyph is being rasterized for grayscale. (Looking at this another way, setting bit 0 asks for the scaler version number, setting bit 1 asks for the rotation information, setting bit 2 asks for stretching information, setting bit 5 asks for grayscale information. To request information on two or more of these set the appropriate bits.)

Selector bits 3 and 4 are reserved for Apple Computer.

The following selectors are for ClearType information. Selector bit 6 indicates that ClearType is enabled. Selector bit 7 indicates that “compatible-width” ClearType is enabled. Selector bit 8 indicates that symmetrical smoothing ClearType is enabled. And, selector bit 9 indicates that ClearType is processing the LCD stripes in BGR (Blue, Green, Red) order.

The result pushed onto the stack contains the requested information. More precisely, bits 0 through 7 comprise the Scaler version number. Version 1 is Macintosh System 6 INIT, version 2 is Macintosh System 7, version 3 is Windows 3.1, version 33 is MS rasterizer v.1.5, version 34 is MS rasterizer v.1.6, version 35 is MS rasterizer v.1.7, version 36 is MS rasterizer 1.6+, version, version 37 is MS rasterizer 1.8, version 38 is MS rasterizer 1.9). Version numbers 0 and 4 through 255 are reserved.

Note: Before hinting vertical phantom points or asking for grayscale information, use GETINFO to check that the rasterizer in use is MS rasterizer v.1.7 or later. Before checking if ClearType is enabled, make sure that the rasterizer is number is 36 or greater. For the other ClearType bits, make sure that the rasterizer is number 37 or greater.

Bit 8 is set to 1 if the current glyph has been rotated. It is 0 otherwise. Setting bit 9 to 1 indicates that the glyph has been stretched. It is 0 otherwise.

Bit 12 is set to 1 if the rasterization is done for grayscale (for MS rasterizer v.1.7 and later).

Bit 13 is set to 1 if ClearType is enabled, otherwise it is 0 (for MS rasterizer v.1.6+ or MS rasterizer v.1.8 and later).

Bit 14 is set to 1 if ClearType widths are compatible with bi-level and grayscale widths. It is 0 if the widths are the natural ClearType widths (for MS rasterizer 1.8 and later).

Bit 15 is set to 1 if ClearType symmetrical smoothing is being used. It is set to zero otherwise (for MS rasterizer 1.8 and later).

Bit 16 is set to 1 if ClearType is processing the LCD stripes in BGR (Blue, Green, Red) order. Otherwise if it is 0, in RGB (Red, Green, Blue) order (for MS rasterizer 1.8 and later).

When the selector is set to request more than one piece of information, that information is OR’d together and pushed onto the stack. For example, a selector value of 6 requests both information on rotation and stretching and will result in the setting of both bits 8 and 9.