Scalar Arithmetic Operations

This document describes scalar arithmetic operations from the MLIR arith dialect.

Important: PTO AS only supports scalar operations from the arith dialect. Vector and tensor operations are not supported.

Total Operations: 47

Note: The operations arith.scaling_extf and arith.scaling_truncf are not included as they are not supported in PTO AS.


Integer Arithmetic Operations

arith.addi - Integer Addition

Description: Integer addition with optional overflow flags.

Syntax:

%result = arith.addi %lhs, %rhs : i32
%result = arith.addi %lhs, %rhs overflow<nsw, nuw> : i32

Example:

// Scalar addition
%a = arith.addi %b, %c : i64

arith.subi - Integer Subtraction

Syntax:

%result = arith.subi %lhs, %rhs : i32

Example:

// Scalar subtraction
%a = arith.subi %b, %c : i32

arith.muli - Integer Multiplication

Syntax:

%result = arith.muli %lhs, %rhs : i32

Example:

// Scalar multiplication
%a = arith.muli %b, %c : i64

arith.divsi - Signed Integer Division

Description: Signed integer division, rounds towards zero.

Syntax:

%result = arith.divsi %lhs, %rhs : i32

Example:

// Scalar signed division
%a = arith.divsi %b, %c : i32

arith.divui - Unsigned Integer Division

Syntax:

%result = arith.divui %lhs, %rhs : i32

Example:

// Scalar unsigned division
%a = arith.divui %b, %c : i32

arith.remsi - Signed Integer Remainder

Syntax:

%result = arith.remsi %lhs, %rhs : i32

Example:

// Scalar signed remainder
%a = arith.remsi %b, %c : i32

arith.remui - Unsigned Integer Remainder

Syntax:

%result = arith.remui %lhs, %rhs : i32

Example:

// Scalar unsigned remainder
%a = arith.remui %b, %c : i32

arith.ceildivsi - Ceiling Division (Signed)

Description: Signed integer division rounding towards positive infinity.

Syntax:

%result = arith.ceildivsi %lhs, %rhs : i32

Example:

// Scalar ceiling division
%a = arith.ceildivsi %b, %c : i32

arith.ceildivui - Ceiling Division (Unsigned)

Syntax:

%result = arith.ceildivui %lhs, %rhs : i32

Example:

// Scalar ceiling division (unsigned)
%a = arith.ceildivui %b, %c : i32

arith.floordivsi - Floor Division (Signed)

Description: Signed integer division rounding towards negative infinity.

Syntax:

%result = arith.floordivsi %lhs, %rhs : i32

Example:

// Scalar floor division
%a = arith.floordivsi %b, %c : i32

Floating-Point Arithmetic Operations

arith.addf - Floating-Point Addition

Syntax:

%result = arith.addf %lhs, %rhs : f32
%result = arith.addf %lhs, %rhs fastmath<fast> : f32

Example:

// Scalar addition
%a = arith.addf %b, %c : f64

arith.subf - Floating-Point Subtraction

Syntax:

%result = arith.subf %lhs, %rhs : f32

Example:

// Scalar subtraction
%a = arith.subf %b, %c : f32

arith.mulf - Floating-Point Multiplication

Syntax:

%result = arith.mulf %lhs, %rhs : f32

Example:

// Scalar multiplication
%a = arith.mulf %b, %c : f64

arith.divf - Floating-Point Division

Syntax:

%result = arith.divf %lhs, %rhs : f32

Example:

// Scalar division
%a = arith.divf %b, %c : f32

arith.remf - Floating-Point Remainder

Syntax:

%result = arith.remf %lhs, %rhs : f32

Example:

// Scalar remainder
%a = arith.remf %b, %c : f64

arith.negf - Floating-Point Negation

Syntax:

%result = arith.negf %operand : f32

Example:

// Scalar negation
%a = arith.negf %b : f32

Bitwise Operations

arith.andi - Bitwise AND

Syntax:

%result = arith.andi %lhs, %rhs : i32

Example:

// Scalar bitwise AND
%a = arith.andi %b, %c : i32

arith.ori - Bitwise OR

Syntax:

%result = arith.ori %lhs, %rhs : i32

Example:

// Scalar bitwise OR
%a = arith.ori %b, %c : i64

arith.xori - Bitwise XOR

Syntax:

%result = arith.xori %lhs, %rhs : i32

Example:

// Scalar bitwise XOR
%a = arith.xori %b, %c : i32

Shift Operations

arith.shli - Shift Left

Syntax:

%result = arith.shli %lhs, %rhs : i32

Example:

// Scalar shift left
%a = arith.shli %b, %c : i32

arith.shrsi - Arithmetic Shift Right (Signed)

Description: Arithmetic right shift (sign-extended).

Syntax:

%result = arith.shrsi %lhs, %rhs : i32

Example:

// Scalar arithmetic shift right
%a = arith.shrsi %b, %c : i32

arith.shrui - Logical Shift Right (Unsigned)

Description: Logical right shift (zero-extended).

Syntax:

%result = arith.shrui %lhs, %rhs : i32

Example:

// Scalar logical shift right
%a = arith.shrui %b, %c : i32

Comparison Operations

arith.cmpi - Integer Comparison

Description: Compare two integers with specified predicate.

Syntax:

%result = arith.cmpi <predicate>, %lhs, %rhs : i32

Predicates: - Signed: slt, sle, sgt, sge - Unsigned: ult, ule, ugt, uge - Equality: eq, ne

Example:

// Scalar comparison
%cmp = arith.cmpi slt, %a, %b : i32
%eq = arith.cmpi eq, %x, %y : i64

arith.cmpf - Floating-Point Comparison

Description: Compare two floats with specified predicate.

Syntax:

%result = arith.cmpf <predicate>, %lhs, %rhs : f32

Predicates: - Ordered: oeq, one, olt, ole, ogt, oge, ord - Unordered: ueq, une, ult, ule, ugt, uge, uno - Always: true, false

Example:

// Scalar comparison
%cmp = arith.cmpf olt, %a, %b : f32
%eq = arith.cmpf oeq, %x, %y : f64

Min/Max Operations

arith.minsi - Minimum (Signed Integer)

Syntax:

%result = arith.minsi %lhs, %rhs : i32

Example:

// Scalar minimum
%min = arith.minsi %a, %b : i32

arith.minui - Minimum (Unsigned Integer)

Syntax:

%result = arith.minui %lhs, %rhs : i32

Example:

// Scalar minimum (unsigned)
%min = arith.minui %a, %b : i32

arith.maxsi - Maximum (Signed Integer)

Syntax:

%result = arith.maxsi %lhs, %rhs : i32

Example:

// Scalar maximum
%max = arith.maxsi %a, %b : i32

arith.maxui - Maximum (Unsigned Integer)

Syntax:

%result = arith.maxui %lhs, %rhs : i32

Example:

// Scalar maximum (unsigned)
%max = arith.maxui %a, %b : i32

arith.minimumf - Minimum (Float, propagates NaN)

Syntax:

%result = arith.minimumf %lhs, %rhs : f32

Example:

// Scalar minimum (propagates NaN)
%min = arith.minimumf %a, %b : f32

arith.maximumf - Maximum (Float, propagates NaN)

Syntax:

%result = arith.maximumf %lhs, %rhs : f32

Example:

// Scalar maximum (propagates NaN)
%max = arith.maximumf %a, %b : f32

arith.minnumf - Minimum (Float, ignores NaN)

Syntax:

%result = arith.minnumf %lhs, %rhs : f32

Example:

// Scalar minimum (ignores NaN)
%min = arith.minnumf %a, %b : f64

arith.maxnumf - Maximum (Float, ignores NaN)

Syntax:

%result = arith.maxnumf %lhs, %rhs : f32

Example:

// Scalar maximum (ignores NaN)
%max = arith.maxnumf %a, %b : f64

Type Conversion Operations

arith.extsi - Sign Extension

Description: Sign-extend integer to wider type.

Syntax:

%result = arith.extsi %in : i32 to i64

Example:

// Scalar sign extension
%wide = arith.extsi %narrow : i32 to i64

arith.extui - Zero Extension

Description: Zero-extend integer to wider type.

Syntax:

%result = arith.extui %in : i32 to i64

Example:

// Scalar zero extension
%wide = arith.extui %narrow : i32 to i64

arith.trunci - Truncate Integer

Description: Truncate integer to narrower type.

Syntax:

%result = arith.trunci %in : i64 to i32

Example:

// Scalar truncation
%narrow = arith.trunci %wide : i64 to i32

arith.extf - Extend Float

Description: Extend floating-point to wider type.

Syntax:

%result = arith.extf %in : f32 to f64

Example:

// Scalar float extension
%double = arith.extf %single : f32 to f64

arith.truncf - Truncate Float

Description: Truncate floating-point to narrower type.

Syntax:

%result = arith.truncf %in : f64 to f32

Example:

// Scalar float truncation
%single = arith.truncf %double : f64 to f32

arith.sitofp - Signed Integer to Float

Syntax:

%result = arith.sitofp %in : i32 to f32

Example:

// Scalar int to float
%fp = arith.sitofp %int : i32 to f32

arith.uitofp - Unsigned Integer to Float

Syntax:

%result = arith.uitofp %in : i32 to f32

Example:

// Scalar uint to float
%fp = arith.uitofp %uint : i32 to f32

arith.fptosi - Float to Signed Integer

Description: Convert float to signed integer (rounds towards zero).

Syntax:

%result = arith.fptosi %in : f32 to i32

Example:

// Scalar float to int
%int = arith.fptosi %fp : f32 to i32

arith.fptoui - Float to Unsigned Integer

Syntax:

%result = arith.fptoui %in : f32 to i32

Example:

// Scalar float to uint
%uint = arith.fptoui %fp : f32 to i32

arith.bitcast - Bitcast

Description: Reinterpret bits as different type (same bit width).

Syntax:

%result = arith.bitcast %in : f32 to i32

Example:

// Scalar bitcast
%bits = arith.bitcast %fp : f32 to i32

arith.index_cast - Index Cast (Signed)

Description: Cast between index type and integer types (sign-extended).

Syntax:

%result = arith.index_cast %in : i32 to index
%result = arith.index_cast %in : index to i64

Example:

// Scalar index cast
%idx = arith.index_cast %int : i32 to index
%int = arith.index_cast %idx : index to i64

arith.index_castui - Index Cast (Unsigned)

Description: Cast between index type and integer types (zero-extended).

Syntax:

%result = arith.index_castui %in : i32 to index

Example:

// Scalar index cast (unsigned)
%idx = arith.index_castui %uint : i32 to index

Special Operations

arith.select - Conditional Select

Syntax:

%result = arith.select %condition, %true_value, %false_value : i32

Example:

// Scalar select
%result = arith.select %cond, %a, %b : i32
%fp_result = arith.select %cond, %x, %y : f32

arith.constant - Constant Value

Syntax:

%result = arith.constant <value> : <type>

Example:

// Scalar constants
%c0 = arith.constant 0 : i32
%c1 = arith.constant 1 : i64
%pi = arith.constant 3.14159 : f32
%true = arith.constant true

Extended Arithmetic Operations

arith.addui_extended - Extended Unsigned Addition

Description: Unsigned addition with overflow flag.

Syntax:

%sum, %overflow = arith.addui_extended %lhs, %rhs : i32, i1

Example:

// Scalar extended addition
%sum, %overflow = arith.addui_extended %a, %b : i32, i1

arith.mulsi_extended - Extended Signed Multiplication

Description: Signed multiplication returning low and high bits.

Syntax:

%low, %high = arith.mulsi_extended %lhs, %rhs : i32

Example:

// Scalar extended multiplication
%low, %high = arith.mulsi_extended %a, %b : i32

arith.mului_extended - Extended Unsigned Multiplication

Syntax:

%low, %high = arith.mului_extended %lhs, %rhs : i32

Example:

// Scalar extended multiplication (unsigned)
%low, %high = arith.mului_extended %a, %b : i32