All arithmetic operations in MMIX are executed strictly according to the basic MMIX format

Every integer arithmetic instruction has two modifications - for signed numbers and unsigned ones (mnemonics for the last have additional letter U). The main difference between these kinds of operations is that "unsigned" instructions don't test the results for overflow. The inequality for multiplication and division is more essential, but we shall not consider it here. And as usual in MMIX instruction set there are two serial codes for each operation; it allows to do this operation with register $Z or unsigned byte constant Z. So MMIX has 4 modifications for each integer arithmetic operation.

Not all arithmetic operations are presented in the tables below - some additional instructions exist in the comprehensive MMIX set. For example, 16ADD operation which calculates 16*$Y+$Z. Such commands may be very useful for practical programming, but can be omitted during primary familiarization without any loss.

Main integer arithmetic operations

Code	Mnemonic	Name	Comments
18	MUL $X, $Y, $Z	multiply	the signed product of register Y by register Z is placed into register X; the result is tested for overflow
19	MUL $X, $Y, Z	multiply	the signed product of register Y by constant Z is placed into register X; the result is tested for overflow
1A	MULU $X, $Y, $Z	multiply unsigned	the lower 64 bits of unsigned product of $Y by register Z is placed into $X (the upper 64 bit - in special register rH)
1B	MULU $X, $Y, Z	multiply unsigned	the lower 64 bits of unsigned product of $Y by constant Z is placed into $X (the upper 64 bit - in special register rH)
1C	DIV $X, $Y, $Z	divide	the signed quotient of register Y by register Z is placed into register X; the result is tested for overflow
1D	DIV $X, $Y, Z	divide	the signed quotient of register Y by constant Z is placed into register X; the result is tested for overflow
1E	DIVU $X, $Y, $Z	divide unsigned	combined number from rD and $Y is divided by register Z; the result is stored into register X (remainder - in rR)
1F	DIVU $X, $Y, Z	divide unsigned	combined number from rD and $Y is divided by constant Z; the result is stored into register X (remainder - in rR)
20	ADD $X, $Y, $Z	add	the signed sum of register Y with register Z is placed into register X; the result is tested for overflow
21	ADD $X, $Y, Z	add	the signed sum of register Y with constant Z is placed into register X; the result is tested for overflow
22	ADDU $X, $Y, $Z	add unsigned	the unsigned sum of register Y with register Z is placed into register X; the result is not tested for overflow
23	ADDU $X, $Y, Z	add unsigned	the unsigned sum of register Y with constant Z is placed into register X; the result is not tested for overflow
24	SUB $X, $Y, $Z	subtract	the signed difference between register Y and register Z is placed into register X; the result is tested for overflow
25	SUB $X, $Y, Z	subtract	the signed difference between register Y and constant Z is placed into register X; the result is tested for overflow
26	SUBU $X, $Y, $Z	subtract unsigned	the unsigned difference between register Y and register Z is placed into register X; the result is not tested for overflow
27	SUBU $X, $Y, Z	subtract unsigned	the unsigned difference between register Y and constant Z is placed into register X; the result is not tested for overflow

Main float arithmetic operations

Code	Mnemonic	Name	Comments
04	FADD $X, $Y, $Z	float add	the floating point sum of registers Y and Z is placed into register X
06	FSUB $X, $Y, $Z	float subtract	the floating point difference of registers Y and Z is placed into register X
10	FMUL $X, $Y, $Z	float multiply	the floating point product of registers Y and Z is placed into register X
14	FDIV $X, $Y, $Z	float divide	the floating point quotient of registers Y and Z is placed into register X

Current rounding mode is used for most of float point instructions such as FIX or FSQR (square root). But this regime can be directly set in the Y field
Y=1: ROUND_OFF
Y=2: ROUND_UP
Y=3: ROUND_DOWN
Y=4: ROUND_NEAR

Converting numeric types

Code	Mnemonic	Name	Comments
05	FIX $X, $Z	convert floating to fixed	overflow is testing
07	FIXU $X, $Z	convert floating to fixed unsigned	overflow is not testing
08	FLOT $X, $Z	convert fixed to floating	convert signed integer value from register Z to float and store to $X
09	FLOT $X, Z	convert fixed to floating	convert signed integer value of constant Z to float and store to $X
0A	FLOTU $X, $Z	convert fixed to floating unsigned	convert unsigned integer value from register Z to float and store to $X
0B	FLOTU $X, Z	convert fixed to floating unsigned	convert unsigned integer value of constant Z to float and store to $X
02	FUN X$, $Y, Z$	float unordered	check $Y and $Z for NAN and reset $X to 0 if both are not NAN (1 otherwise)

Examples:

• instruction ADD $2, $2, 10 adds 10 to register 2
• instruction SUB $1, $2, $3 stores the difference $2 - $3 to register 1
• instruction FMUL $1, $2, $3 multiply float values from registers 2 and 3; the result is save to register 1
• instruction FIX $1, 00, $2 is equivalent to Pascal operator $1:=trunc($2) and FIX $1, 04, $2 - to $1:=round($2)

• signed and unsigned data
• floating point numbers

• compare instructions