Value | Description |
val abs : int32 -> int32 |
Returns the absolute value of the argument
|
val add : int32 -> int32 -> int32 |
Returns the sum of a and b
|
val bits_of_float : float -> int32 |
Converts a 64-bit float to a raw 32-bit representation
|
val bits_of_float32 : float32 -> int32 |
Converts a 32-bit float to a raw 32-bit representation
|
val compare : int32 -> int32 -> int |
Compares a and b and returns 1 if a > b, -1 if b < a and 0 if a = b
|
val div : int32 -> int32 -> int32 |
Returns a divided by b
|
val float32_of_bits : int32 -> float32 |
Converts a raw 32-bit representation to a 32-bit float
|
val float_of_bits : int32 -> float |
Converts a raw 32-bit representation to a 64-bit float
|
val logand : int32 -> int32 -> int32 |
Combines the binary representation of a and b by bitwise and
|
val lognot : int32 -> int32 |
Returns the bitwise logical negation of a
|
val logor : int32 -> int32 -> int32 |
Combines the binary representation of a and b by bitwise or
|
val logxor : int32 -> int32 -> int32 |
Combines the binary representation of a and b by bitwise xor
|
val max_int : int32 |
Returns the largest 32-bit signed integer
|
val min_int : int32 |
Returns the smallest 32-bit signed integer
|
val minus_one : int32 |
The value minus one as a System.Int32
|
val mul : int32 -> int32 -> int32 |
Returns a multiplied by b
|
val neg : int32 -> int32 |
Returns -a
|
val of_float : float -> int32 |
Converts a 64-bit float to a 32-bit integer
|
val of_float32 : float32 -> int32 |
Converts a 32-bit float to a 32-bit integer
|
val of_int : int -> int32 |
Converts a 32-bit integer to a 32-bit integer (included for ML compatability)
|
val of_int64 : int64 -> int32 |
Converts a 64-bit unsigned integer to a 32-bit integer
|
val of_nativeint : nativeint -> int32 |
Converts a 32-bit unsigned integer to a 32-bit integer
|
val of_string : string -> int32 |
Converts a string to a 32-bit integer
|
val of_uint32 : uint32 -> int32 |
Converts a 32-bit unsigned integer to a 32-bit integer
|
val one : int32 |
The value one as a System.Int32
|
val pred : int32 -> int32 |
Returns the predeccessor of the argument
|
val rem : int32 -> int32 -> int32 |
Returns the remainder of a divided by b
|
val shift_left : int32 -> int -> int32 |
Shifts the binary representation a by n bits to the left
|
val shift_right : int32 -> int -> int32 |
Shifts the binary representation a by n bits to the right; high-order empty bits are set to the sign bit
|
val shift_right_logical : int32 -> int -> int32 |
Shifts the binary representation a by n bits to the right; high-order bits are zero-filled
|
val sub : int32 -> int32 -> int32 |
Returns a minus b
|
val succ : int32 -> int32 |
Returns the successor of the argument
|
val to_float : int32 -> float |
Converts a 32-bit integer to a 64-bit float
|
val to_float32 : int32 -> float32 |
Converts a 32-bit integer to a 32-bit float
|
val to_int : int32 -> int |
Converts a 32-bit integer to a 32-bit integer (included for ML compatability)
|
val to_int64 : int32 -> int64 |
Converts a 32-bit unsigned integer to a 64-bit integer
|
val to_nativeint : int32 -> nativeint |
Converts a 32-bit unsigned integer to a 32-bit integer
|
val to_string : int32 -> string |
Converts a 32-bit integer to a string
|
val to_uint32 : int32 -> uint32 |
Converts a 32-bit integer to a 32-bit unsigned integer
|
val zero : int32 |
The value zero as a System.Int32
|