‘mathExt’ Dialect

Extended Math dialect.

Operations

mathExt.ilogb (mathExt::IlogbOp)

An exponent components of a floating-point number

Syntax:

operation ::= `mathExt.ilogb` $operand (`fastmath` `` $fastmath^)?
              attr-dict `:` type($result)

ilogb(x) := floor(log2(abs(x)))

The ilogb operation takes one operand of floating point type (i.e., scalar, tensor or vector) and returns one result of the same type. It has no standard attributes.

Traits: AlwaysSpeculatableImplTrait, Elementwise, SameOperandsAndResultType, Scalarizable, Tensorizable, Vectorizable

Interfaces: ArithFastMathInterface, ConditionallySpeculatable, InferTypeOpInterface, NoMemoryEffect (MemoryEffectOpInterface), VectorUnrollOpInterface

Effects: MemoryEffects::Effect{}

Attributes

Attribute	MLIR Type	Description
fastmath	::mlir::arith::FastMathFlagsAttr	Floating point fast math flags {{% markdown %}}Enum cases: * none (`none`) * reassoc (`reassoc`) * nnan (`nnan`) * ninf (`ninf`) * nsz (`nsz`) * arcp (`arcp`) * contract (`contract`) * afn (`afn`) * fast (`fast`){{% /markdown %}}

Operands

Operand	Description
operand	floating-point-like

Results

Result	Description
result	floating-point-like

mathExt.ldep (mathExt::LdexpOp)

A fractional components of a floating-point number

Syntax:

operation ::= `mathExt.ldep` $lhs `,` $rhs (`fastmath` `` $fastmath^)?
              attr-dict `:` type($result)

ldep(x) = x * (ilogb(x) + 1)^(-1)

The ldexp operation takes two operands of floating point type (i.e., scalar, tensor or vector) and returns one result of the same type. Operands must have the same type.

Traits: AlwaysSpeculatableImplTrait, Elementwise, SameOperandsAndResultType, Scalarizable, Tensorizable, Vectorizable

Interfaces: ArithFastMathInterface, ConditionallySpeculatable, InferTypeOpInterface, NoMemoryEffect (MemoryEffectOpInterface), VectorUnrollOpInterface

Effects: MemoryEffects::Effect{}

Attributes

Attribute	MLIR Type	Description
fastmath	::mlir::arith::FastMathFlagsAttr	Floating point fast math flags {{% markdown %}}Enum cases: * none (`none`) * reassoc (`reassoc`) * nnan (`nnan`) * ninf (`ninf`) * nsz (`nsz`) * arcp (`arcp`) * contract (`contract`) * afn (`afn`) * fast (`fast`){{% /markdown %}}

Operands

Operand	Description
lhs	floating-point-like
rhs	floating-point-like

Results

Result	Description
result	floating-point-like

Enums

CmpFPredicate

allowed 64-bit signless integer cases: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15

Cases

Symbol	Value	String
AlwaysFalse	0	false
OEQ	1	oeq
OGT	2	ogt
OGE	3	oge
OLT	4	olt
OLE	5	ole
ONE	6	one
ORD	7	ord
UEQ	8	ueq
UGT	9	ugt
UGE	10	uge
ULT	11	ult
ULE	12	ule
UNE	13	une
UNO	14	uno
AlwaysTrue	15	true

CmpIPredicate

allowed 64-bit signless integer cases: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9

Cases

Symbol	Value	String
eq	0	eq
ne	1	ne
slt	2	slt
sle	3	sle
sgt	4	sgt
sge	5	sge
ult	6	ult
ule	7	ule
ugt	8	ugt
uge	9	uge

IntegerOverflowFlags

Integer overflow arith flags

Cases

Symbol	Value	String
none	0	none
nsw	1	nsw
nuw	2	nuw

RoundingMode

Floating point rounding mode

Cases

Symbol	Value	String
to_nearest_even	0	to_nearest_even
downward	1	downward
upward	2	upward
toward_zero	3	toward_zero
to_nearest_away	4	to_nearest_away

AtomicRMWKind

allowed 64-bit signless integer cases: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14

Cases

Symbol	Value	String
addf	0	addf
addi	1	addi
assign	2	assign
maximumf	3	maximumf
maxs	4	maxs
maxu	5	maxu
minimumf	6	minimumf
mins	7	mins
minu	8	minu
mulf	9	mulf
muli	10	muli
ori	11	ori
andi	12	andi
maxnumf	13	maxnumf
minnumf	14	minnumf

FastMathFlags

Floating point fast math flags

Cases

Symbol	Value	String
none	0	none
reassoc	1	reassoc
nnan	2	nnan
ninf	4	ninf
nsz	8	nsz
arcp	16	arcp
contract	32	contract
afn	64	afn
fast	127	fast