'mhlo' Dialetto

Operazioni

mhlo.abs (mhlo::AbsOp)

Intervento addominale

Sintassi:

operation ::= `mhlo.abs` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Esegue l'operazione di valore assoluto elemento per elemento sul tensore operand e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#abs

Esempio:

%result = mhlo.abs %operand : tensor<3xi32>

Caratteristiche: AlwaysSpeculatableImplTrait , Elementwise , SameOperandsAndResultShape

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.acos (mhlo::AcosOp)

Operazione Acos

Sintassi:

operation ::= `mhlo.acos` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Esegue l'operazione acos elemento per elemento sul tensore operand e produce un tensore result .

Esempio:

%result = mhlo.acos %operand : tensor<2x2xf32>

Caratteristiche: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: InferShapedTypeOpInterface , InferTypeOpInterface

Operandi:

Risultati:

mhlo.acosh (mhlo::AcoshOp)

Operazione Acoh

Sintassi:

operation ::= `mhlo.acosh` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Esegue l'operazione acosh elemento per elemento sul tensore operand e produce un tensore result .

Esempio:

%result = mhlo.acosh %operand : tensor<2x2xf32>

Caratteristiche: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: InferShapedTypeOpInterface , InferTypeOpInterface

Operandi:

Risultati:

mhlo.add (mhlo::AddOp)

Aggiungere l'operazione

Sintassi:

operation ::= `mhlo.add` $lhs `,` $rhs attr-dict
              `:` custom<SameOperandsAndResultType>(type($lhs), type($rhs), type($result))

Esegue la somma elemento per elemento di due tensori, lhs e rhs , e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#add

Esempio:

%result = mhlo.add %lhs, %rhs : tensor<2x2xi32>

Caratteristiche: AlwaysSpeculatableImplTrait , Commutative , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.add_dependency (mhlo::AddDependencyOp)

Operazione AddDependency

Sintassi:

operation ::= `mhlo.add_dependency` operands attr-dict `:` functional-type(operands, results)

Questa operazione è privata del compilatore XLA, quindi non ha ancora una specifica.

In termini informali, questa operazione ha due operandi: un operando dati e un token. L'output dell'operazione è l'operando dati. Se utilizzata con AfterAll, questa operazione consente di ordinare le operazioni senza effetti collaterali (quelle che non producono valori token).

Esempio:

%1 = mhlo.add_dependency %arg0, %0 : (tensor<3x4xf32>, !mhlo.token) -> tensor<3x4xf32>

Caratteristiche: AlwaysSpeculatableImplTrait

Interfacce: ConditionallySpeculatable , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.after_all (mhlo::AfterAllOp)

Dopotutto l'operazione

Sintassi:

operation ::= `mhlo.after_all` $inputs attr-dict
              `:` custom<VariadicSameOperandsAndResultType>(ref($inputs), type($inputs), type($result))

Garantisce che le operazioni che producono gli inputs vengano eseguite prima di qualsiasi operazione che dipenda dal result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#after_all

Esempio:

%result = mhlo.after_all %input0, %input1 : !mhlo.token

Caratteristiche: AlwaysSpeculatableImplTrait

Interfacce: ConditionallySpeculatable , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.all_gather (mhlo::AllGatherOp)

Operazione AllGather

All'interno di ciascun gruppo di processi nella griglia di processi, vengono concatenati i valori del tensore operando di ciascun processo lungo all_gather_dim e viene prodotto un tensore di risultato. Il computation viene applicato separatamente per ciascun operando in operands , producendo un risultato per operando.

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#all_gather

Esempio:

%result = "mhlo.all_gather"(%operand) {
  all_gather_dim = 1 : i64,
  replica_groups = dense<[[0, 1]]> : tensor<1x2xi64>
  // channel_id = 0
  channel_handle = #mhlo.channel_handle<handle = 0, type = 0>,
  // use_global_device_ids = false
} : (tensor<2x2xf32>) -> tensor<2x4xf32>

Caratteristiche: SameOperandsAndResultElementType

Attributi:

Operandi:

Risultati:

mhlo.all_reduce (mhlo::AllReduceOp)

Operazione AllReduce

All'interno di ciascun gruppo di processi nella griglia dei processi, viene applicata una computation di riduzione ai valori di un tensore operando di ciascun processo e viene prodotto un tensore di risultato. Il computation viene applicato separatamente per ciascun operando in operands , producendo un risultato per operando.

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#all_reduce

Esempio:

%result = "mhlo.all_reduce"(%operand) ({
  ^bb0(%arg0: tensor<f32>, %arg1: tensor<f32>):
    %0 = mhlo.add %arg1, %arg2 : tensor<f32>
    mhlo.return %0 : tensor<f32>
}) {
  replica_groups = dense<[[0, 1]]> : tensor<1x2xi64>
  // channel_id = 0
  channel_handle = #mhlo.channel_handle<handle = 0, type = 0>
  // use_global_device_ids = false
} : (tensor<4xf32>) -> tensor<4xf32>

Caratteristiche: InferTensorType , SingleBlockImplicitTerminator<ReturnOp> , SingleBlock

Interfacce: InferShapedTypeOpInterface , InferTypeOpInterface

Attributi:

Operandi:

Risultati:

mhlo.all_to_all (mhlo::AllToAllOp)

Operazione AllToAll

All'interno di ciascun gruppo di processi nella griglia di processi, suddivide i valori del tensore operand lungo split_dimension in parti, distribuisce le parti suddivise tra i processi, concatena le parti distribuite lungo concat_dimension e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#all_to_all

Esempio:

%result = "mhlo.all_to_all"(%operand) {
  split_dimension = 1 : i64,
  concat_dimension = 0 : i64,
  split_count = 2 : i64,
  replica_groups = dense<[[0, 1]]> : tensor<1x2xi64>
} : (tensor<2x4xf32>) -> tensor<4x2xf32>

Caratteristiche: AlwaysSpeculatableImplTrait , InferTensorType , SameOperandsElementType , SameOperandsShape , SameVariadicOperandSize

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.and (mhlo::AndOp)

E operazione

Sintassi:

operation ::= `mhlo.and` $lhs `,` $rhs attr-dict
              `:` custom<SameOperandsAndResultType>(type($lhs), type($rhs), type($result))

Esegue l'operazione AND elemento per elemento di due tensori, lhs e rhs , e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#and

Esempio:

%result = mhlo.and %lhs, %rhs : tensor<2x2xi32>

Caratteristiche: AlwaysSpeculatableImplTrait , Commutative , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.asin (mhlo::AsinOp)

operazione Asin

Sintassi:

operation ::= `mhlo.asin` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Esegue l'operazione ASIN elemento per elemento sul tensore operand e produce un tensore result .

Esempio:

%result = mhlo.asin %operand : tensor<2x2xf32>

Caratteristiche: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: InferShapedTypeOpInterface , InferTypeOpInterface

Operandi:

Risultati:

mhlo.asinh (mhlo::AsinhOp)

operazione Asinh

Sintassi:

operation ::= `mhlo.asinh` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Esegue l'operazione asinh elemento per elemento sul tensore operand e produce un tensore result .

Esempio:

%result = mhlo.asinh %operand : tensor<2x2xf32>

Caratteristiche: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: InferShapedTypeOpInterface , InferTypeOpInterface

Operandi:

Risultati:

mhlo.async_done (mhlo::AsyncDoneOp)

Operazione AsyncDone completata

Questa operazione è privata del compilatore XLA, quindi non ha ancora una specifica.

In termini informali, questa operazione si blocca fino al termine di un calcolo asincrono. Restituisce il risultato finale del calcolo asincrono.

Per ulteriori informazioni, consultare la documentazione di AsyncStart.

Interfacce: InferTypeOpInterface

Operandi:

Risultati:

mhlo.async_start (mhlo::AsyncStartOp)

Operazione AsyncStart

Questa operazione è privata del compilatore XLA, quindi non ha ancora una specifica.

In termini informali, questa operazione avvia un calcolo asincrono.

Questo approccio viene utilizzato quando sono presenti funzioni che contengono sia attese asincrone (come le operazioni DMA) sia calcoli on-thread. Ad esempio, una funzione potrebbe essere composta da un calcolo, un'operazione DMA, un altro calcolo, un secondo DMA e un calcolo finale. Questo verrebbe rappresentato da un `async_start` seguito da un `async_update` e un `async_done`. L'`async_start` esegue il primo calcolo on-thread e poi avvia il DMA. L'`async_update` attende il completamento del DMA, se non è ancora terminato, quindi esegue il secondo calcolo nella funzione e avvia il secondo DMA. Infine, l'`async_done` attende il completamento di quest'ultimo DMA, quindi esegue l'ultimo calcolo che deve essere eseguito on-thread e restituisce il risultato di tale calcolo finale.

operands vengono passati direttamente al calcolo. called_computation è la funzione che verrà eseguita in modo asincrono. execution_thread è il nome del thread in cui verrà eseguita. Il thread principale si chiama "main". Tutti i thread hanno un nome.

Questa funzione restituisce tutto lo stato necessario tra le operazioni asincrone. Dopo l'assegnazione del buffer, i valori restituiti rappresentano lo spazio necessario per contenere l'input, i risultati e qualsiasi blocco di lavoro necessario o modificato dall'operazione asincrona.

Attributi:

Operandi:

Risultati:

mhlo.async_update (mhlo::AsyncUpdateOp)

Operazione AsyncUpdate

Questa operazione è privata del compilatore XLA, quindi non ha ancora una specifica.

In termini informali, questa operazione si blocca su un calcolo asincrono fino a una barriera di sincronizzazione. Restituisce il bundle dopo l'operazione.

Per ulteriori informazioni, consultare la documentazione di AsyncStart.

Interfacce: InferTypeOpInterface

Operandi:

Risultati:

mhlo.atan2 (mhlo::Atan2Op)

Operazione Atan2

Sintassi:

operation ::= `mhlo.atan2` $lhs `,` $rhs attr-dict
              `:` custom<SameOperandsAndResultType>(type($lhs), type($rhs), type($result))

Esegue l'operazione atan2 elemento per elemento sul lhs e rhs del tensore e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#atan2

Esempio:

%result = mhlo.atan2 %lhs, %rhs : tensor<3xf32>

Caratteristiche: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.atanh (mhlo::AtanhOp)

Operazione Atanh

Sintassi:

operation ::= `mhlo.atanh` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Esegue l'operazione atanh elemento per elemento sul tensore operand e produce un tensore result .

Esempio:

%result = mhlo.atanh %operand : tensor<2x2xf32>

Caratteristiche: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: InferShapedTypeOpInterface , InferTypeOpInterface

Operandi:

Risultati:

mhlo.batch_norm_grad (mhlo::BatchNormGradOp)

Operazione BatchNormGrad

Calcola i gradienti di diversi input di BatchNormTrainingOp retropropagando da grad_output e produce i tensori grad_operand , grad_scale e grad_offset .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#batch_norm_grad

Esempio:

%grad_operand, %grad_scale, %grad_offset =
"mhlo.batch_norm_grad"(%operand, %scale, %mean, %variance, %grad_output) {
  epsilon = 0.0 : f32,
  feature_index = 2 : i64
} : (tensor<2x2x2xf32>, tensor<2xf32>, tensor<2xf32>, tensor<2xf32>,
    tensor<2x2x2xf32>) -> (tensor<2x2x2xf32>, tensor<2xf32>, tensor<2xf32>)

Caratteristiche: AlwaysSpeculatableImplTrait , InferTensorType

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.batch_norm_inference (mhlo::BatchNormInferenceOp)

Operazione BatchNormInference

Normalizza il tensore operand rispetto a tutte le dimensioni tranne la dimensione feature_index e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#batch_norm_inference

Esempio:

%result = "mhlo.batch_norm_inference"(%operand, %scale, %offset, %mean, %variance) {
  epsilon = 0.0 : f32,
  feature_index = 2 : i64
} : (tensor<2x2x2xf32>, tensor<2xf32>, tensor<2xf32>, tensor<2xf32>, tensor<2xf32>) -> tensor<2x2x2xf32>

Caratteristiche: AlwaysSpeculatableImplTrait , InferTensorType

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.batch_norm_training (mhlo::BatchNormTrainingOp)

Operazione BatchNormTraining

Calcola la media e la varianza tra le dimensioni batch e spaziali e normalizza il tensore operand , per ogni caratteristica nella dimensione feature_index e produce i tensori output batch_mean e batch_var .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#batch_norm_training

Esempio:

%output, %batch_mean, %batch_var = "mhlo.batch_norm_training"(%operand, %scale, %offset) {
  epsilon = 0.0 : f32,
  feature_index = 2 : i64
} : (tensor<2x2x2xf32>, tensor<2xf32>, tensor<2xf32>) -> (tensor<2x2x2xf32>, tensor<2xf32>, tensor<2xf32>)

Caratteristiche: AlwaysSpeculatableImplTrait , InferTensorType

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.bitcast (mhlo::BitcastOp)

Operazione bitcast

Sintassi:

operation ::= `mhlo.bitcast` operands attr-dict `:` functional-type(operands, results)

Questa operazione è privata del compilatore XLA, quindi non ha ancora una specifica.

In termini informali, questa operazione modifica la forma dell'input in modo tale che la disposizione fisica degli elementi rimanga invariata.

Questa operazione richiede informazioni sul layout per comprendere la "disposizione fisica degli elementi", e il supporto per il layout in MHLO è attualmente in fase di sviluppo.

Esempio:

%0 = mhlo.bitcast %arg0 : (tensor<3x4xf32>) -> tensor<3x4x1xf32>

Caratteristiche: AlwaysSpeculatableImplTrait

Interfacce: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.bitcast_convert (mhlo::BitcastConvertOp)

Operazione BitcastConvert

Sintassi:

operation ::= `mhlo.bitcast_convert` operands attr-dict `:` functional-type(operands, results)

Esegue un'operazione di conversione di tipo (bitcast) sul tensore operand e produce un tensore result in cui i bit dell'intero tensore operand vengono reinterpretati utilizzando il tipo del tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#bitcast_convert

Esempio:

%result = mhlo.bitcast_convert %operand : (tensor<2xf32>) -> tensor<2x4xi8>

Caratteristiche: AlwaysSpeculatableImplTrait

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.broadcast (mhlo::BroadcastOp)

Trasmissione

Questa operazione sta per essere rimossa da StableHLO, quindi non è inclusa nella specifica: https://github.com/openxla/stablehlo/issues/3

In termini informali, questa operazione fa la stessa cosa di Broadcast di XLA: https://www.tensorflow.org/xla/operation_semantics#broadcast

Esempio:

%result = mhlo.broadcast %operand, sizes = [1, 2] : (tensor<3xi32>) -> tensor<1x2x3xi32>

Caratteristiche: AlwaysSpeculatableImplTrait , InferTensorType , SameOperandsAndResultElementType

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.broadcast_in_dim (mhlo::BroadcastInDimOp)

Operazione BroadcastInDim

Espande le dimensioni e/o il rango di un tensore di input duplicando i dati nel tensore operand e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#broadcast_in_dim

Esempio:

%result = mhlo.broadcast_in_dim %operand, dims = [2, 1] : (tensor<1x3xi32>) -> tensor<2x3x2xi32>

Caratteristiche: AlwaysSpeculatableImplTrait , HLO_CompatibleOperandsAndResultElementType

Interfacce: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.case (mhlo::CaseOp)

operazione di caso

Produce l'output derivante dall'esecuzione di una sola function dai branches a seconda del valore index .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#case

Esempio:

%result0, %result1 = "mhlo.case"(%index) ({
  mhlo.return %result_branch0, %result_branch0 : tensor<2xi64>, tensor<2xi64>
}, {
  mhlo.return %result_branch1, %result_branch1 : tensor<2xi64>, tensor<2xi64>
}) : (tensor<i32>) -> (tensor<2xi64>, tensor<2xi64>)

Caratteristiche: RecursiveMemoryEffects , SingleBlockImplicitTerminator<ReturnOp> , SingleBlock

Interfacce: InferTypeOpInterface

Operandi:

Risultati:

mhlo.cbrt (mhlo::CbrtOp)

operazione CBRT

Sintassi:

operation ::= `mhlo.cbrt` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Esegue l'operazione di radice cubica elemento per elemento sul tensore operand e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#cbrt

Esempio:

%result = mhlo.cbrt %operand : tensor<4xf32>

Caratteristiche: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.ceil (mhlo::CeilOp)

Operazione a soffitto

Sintassi:

operation ::= `mhlo.ceil` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Esegue l'operazione di troncamento elemento per elemento del tensore operand e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#ceil

Esempio:

%result = mhlo.ceil %operand : tensor<5xf32>

Caratteristiche: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.cholesky (mhlo::CholeskyOp)

operazione colesica

Calcola la decomposizione di Cholesky di un insieme di matrici.

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#cholesky

Esempio:

%result = mhlo.cholesky %a, lower = true : tensor<3x3xf32>

Caratteristiche: AlwaysSpeculatableImplTrait , InferTensorType , SameOperandsAndResultElementType

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.clamp (mhlo::ClampOp)

operazione di bloccaggio

Sintassi:

operation ::= `mhlo.clamp` $min `,` $operand `,` $max attr-dict
              `:` custom<SameOperandsAndResultType>(type($min), type($operand), type($max), type($result))

Limita ogni elemento del tensore operand tra un valore minimo e uno massimo e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#clamp

Esempio:

%result = mhlo.clamp %min, %operand, %max : tensor<3xi32>

Caratteristiche: AlwaysSpeculatableImplTrait , HLO_BroadcastingElementwise , InferTensorType , SameOperandsAndResultElementType

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Operandi:

Risultati:

mhlo.collective_broadcast (mhlo::CollectiveBroadcastOp)

Operazione di trasmissione collettiva

All'interno di ciascun gruppo di processi nella griglia di processi, inviare il valore del tensore operand dal processo sorgente ai processi di destinazione e produrre un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#collective_broadcast

Esempio:

%result = "mhlo.collective_broadcast"(%operand) {
  replica_groups = dense<[[0, 1]]> : tensor<1x2xi64>,
  channel_handle = #mhlo.channel_handle<handle = 0, type = 0>
} : (tensor<1x2xi64>) -> tensor<1x2xi64>

Caratteristiche: CompatibleOperandsAndResultType

Interfacce: InferShapedTypeOpInterface , InferTypeOpInterface

Attributi:

Operandi:

Risultati:

mhlo.collective_permute (mhlo::CollectivePermuteOp)

Operazione CollectivePermute

All'interno di ciascun gruppo di processi nella griglia dei processi, invia il valore del tensore operand dal processo sorgente al processo di destinazione e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#collective_permute

Esempio:

%result = "mhlo.collective_permute"(%operand) {
  source_target_pairs = dense<[[0, 1], [1, 2]]> : tensor<2x2xi64>,
  // channel_id = 0
  channel_handle = #mhlo.channel_handle<handle = 0, type = 0>
} : (tensor<4x2xf32>) -> tensor<4x2xf32>

Caratteristiche: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Risultati:

mhlo.compare (mhlo::CompareOp)

Confrontare l'operazione

Sintassi:

operation ::= `mhlo.compare` $comparison_direction `,` $lhs `,` $rhs (`,` $compare_type^)?
              attr-dict `:` functional-type(operands, results)

Esegue un confronto elemento per elemento dei tensori lhs e rhs in base a comparison_direction e compare_type e produce un tensore result .

Vedi: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#compare

Esempio:

%result = mhlo.compare LT, %lhs, %rhs, FLOAT : (tensor<2xf32>, tensor<2xf32>) -> tensor<2xi1>

Caratteristiche: AlwaysSpeculatableImplTrait , Elementwise , InferTensorType , SameOperandsAndResultShape , SameOperandsElementType

Interfacce: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effetti: MemoryEffects::Effect{}

Attributi:

Operandi:

Results:

mhlo.complex (mhlo::ComplexOp)

Complex operation

Sintassi:

operation ::= `mhlo.complex` operands attr-dict
              `:` custom<ComplexOpType>(type($lhs), type($rhs), type($result))

Performs element-wise conversion to a complex value from a pair of real and imaginary values, lhs and rhs , and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#complex

Esempio:

%result = mhlo.complex %lhs, %rhs : tensor<2xcomplex<f32>>

Traits: AlwaysSpeculatableImplTrait , Elementwise , SameOperandsAndResultShape , SameOperandsElementType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.composite (mhlo::CompositeOp)

Composite operation

Sintassi:

operation ::= `mhlo.composite` $name operands (` ` `(` $composite_regions^ `)`)? attr-dict
              `:` functional-type(operands, results)

Encapsulates an operation made up (composed) of other StableHLO operations, taking inputs and composite_attributes and producing results . The semantics of the op are implemented by the decomposition attribute. The composite op can be replaced with its decomposition without changing program semantics. In cases where inlining the decomposition does not provide the same op semantics, prefer using custom_call .

The version field (defaults to 0 ) is used to denote when a composite's semantics change.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#composite

Esempio:

%results = mhlo.composite "my.op" %arg0, %arg1 {
  decomposition = @my_op,
  composite_attributes = { my_attribute = "my_value" },
  version = 1 : i32
} : (tensor<f32>, tensor<f32>) -> tensor<f32>

Interfaces: SymbolUserOpInterface

Attributes:

Operands:

Results:

mhlo.concatenate (mhlo::ConcatenateOp)

Concatenate operation

Concatenates a variadic number of tensors in inputs along dimension dimension in the same order as the given arguments and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#concatenate

Esempio:

%result = mhlo.concatenate %input0, %input1, dim = 0 : (tensor<3x2xi64>, tensor<1x2xi64>) -> tensor<4x2xi64>

Traits: AlwaysSpeculatableImplTrait , SameOperandsAndResultElementType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.constant (mhlo::ConstantOp)

Constant operation

Produces an output tensor from a constant value .

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#constant

Esempio:

%output = mhlo.constant dense<[[0.0, 1.0], [2.0, 3.0]]> : tensor<2x2xf32>

Traits: AlwaysSpeculatableImplTrait , ConstantLike

Interfaces: ConditionallySpeculatable , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Results:

mhlo.convert (mhlo::ConvertOp)

Convert operation

Sintassi:

operation ::= `mhlo.convert` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Performs an element-wise conversion from one element type to another on operand tensor and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#convert

Esempio:

%result = mhlo.convert %operand : (tensor<3xi32>) -> tensor<3xcomplex<f32>>

Traits: AlwaysSpeculatableImplTrait , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.convolution (mhlo::ConvolutionOp)

Convolution operation

Sintassi:

operation ::= `mhlo.convolution` `(`operands`)`
              `dim_numbers` `=` custom<ConvolutionDimensions>($dimension_numbers) `,`
              `window` `=` `{` custom<WindowAttributes>($window_strides, $padding,
              $lhs_dilation, $rhs_dilation,
              $window_reversal) `}`
              attr-dict `:` functional-type(operands, results)

Computes dot products between windows of lhs and slices of rhs and produces result .

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#convolution

Esempio:

%result = "mhlo.convolution"(%lhs, %rhs) {
  window_strides = dense<4> : tensor<2xi64>,
  padding = dense<0> : tensor<2x2xi64>,
  lhs_dilation = dense<2> : tensor<2xi64>,
  rhs_dilation = dense<1> : tensor<2xi64>,
  window_reversal = dense<false> : tensor<2xi1>,
  dimension_numbers = #mhlo.conv<[b, 0, 1, f]x[0, 1, i, o]->[b, 0, 1, f]>,
  feature_group_count = 1 : i64,
  batch_group_count = 1 : i64,
  precision_config = [#stablehlo<precision DEFAULT>, #stablehlo<precision DEFAULT>]
} : (tensor<1x4x4x1xi32>, tensor<3x3x1x1xi32>) -> tensor<1x2x2x1xi32>

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.copy (mhlo::CopyOp)

Copy operation

Sintassi:

operation ::= `mhlo.copy` operands attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

This operation is private to the XLA compiler, so it is does not yet have a specification.

Informally, this operation a copy of operand . Depending on the metadata attached to the operation, it can behave quite differently from a no-op.

Esempio:

%0 = mhlo.copy %arg0 : tensor<f32>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.cosh (mhlo::CoshOp)

Cosh operation

Sintassi:

operation ::= `mhlo.cosh` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Performs element-wise cosh operation on operand tensor and produces a result tensor.

Esempio:

%result = mhlo.cosh %operand : tensor<2x2xf32>

Traits: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Operands:

Results:

mhlo.cosine (mhlo::CosineOp)

Cosine operation

Sintassi:

operation ::= `mhlo.cosine` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Performs element-wise cosine operation on operand tensor and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#cosine

Esempio:

%result = mhlo.cosine %operand : tensor<2xf32>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.count_leading_zeros (mhlo::ClzOp)

Clz operation

Sintassi:

operation ::= `mhlo.count_leading_zeros` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Performs element-wise count of the number of leading zero bits in the operand tensor and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#count_leading_zeros

Esempio:

%result = mhlo.count_leading_zeros %operand : tensor<2x2xi8>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.create_token (mhlo::CreateTokenOp)

CreateToken operation

Sintassi:

operation ::= `mhlo.create_token` attr-dict `:` type(results)

This operation is on its way out of StableHLO, so it is not included in the specification: https://github.com/openxla/stablehlo/issues/3

Informally, this operation does the same thing as AfterAllOp with 0 inputs: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#after_all

Esempio:

%output = mhlo.create_token : !mhlo.token

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Results:

mhlo.cross-replica-sum (mhlo::CrossReplicaSumOp)

CrossReplicaSum operation

This operation is on its way out of StableHLO, so it is not included in the specification: https://github.com/openxla/stablehlo/issues/3

Informally, this operation does the same thing as AllReduceOp with channel_id = 0 , use_global_device_ids = false and computation implementing addition: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#all_reduce

Esempio:

%result = "mhlo.cross-replica-sum"(%operand) {
  replica_groups = dense<[[0, 1]]> : tensor<1x2xi64>
} : (tensor<4xf32>) -> tensor<4xf32>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.custom_call (mhlo::CustomCallOp)

CustomCall operation

Sintassi:

operation ::= `mhlo.custom_call` custom<CustomCallTarget>($call_target_name) `(` $inputs `)`
              attr-dict `:` functional-type(operands, results)

Encapsulates an implementation-defined operation call_target_name that takes inputs and called_computations and produces results .

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#custom_call

Esempio:

%results = "mhlo.custom_call"(%input0) {
  call_target_name = "foo",
  has_side_effect = false,
  backend_config = "bar",
  api_version = 1 : i32,
  called_computations = [@foo]
} : (tensor<f32>) -> tensor<f32>

A custom call invokes code external to XLA. The `inputs` are passed to the
external code, and the external code is expected to produce a result of the
given type. The exact mechanism is backend-specific. For example, in the CPU
backend, a call instruction is emitted which targets a symbol with the name
`call_target_name`.

If XLA runtime is enabled for a backend, then custom calls use the runtime
custom call calling convention to call into the external functions. This
calling convention defines an ABI for encoding arguments, attributes and
results.

Depending on the API version there are two ways to pass extra bits of static
information to the external function:

1. For `API_VERSION_TYPED_FFI` custom calls `backend_config` must be a
   dictionary attribute, that will be encoded according to the custom call
   calling convention and passed to the external function as the attributes
   argument. External code is expected to use declarative bindings (see
   `xla/runtime/custom_call.h`) to decode them at run time. These custom
   calls are only supported if XLA uses XLA runtime.

2. For previous API versions it is the user responsibility to encode extra
   bits of static information as a string `backend_config` attribute, and
   decode it at run time.

Interfaces: MemoryEffectOpInterface

Attributes:

Operands:

Results:

mhlo.divide (mhlo::DivOp)

Div operation

Sintassi:

operation ::= `mhlo.divide` $lhs `,` $rhs attr-dict
              `:` custom<SameOperandsAndResultType>(type($lhs), type($rhs), type($result))

Performs element-wise division of dividend lhs and divisor rhs tensors and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#divide

Esempio:

%result = mhlo.divide %lhs, %rhs : tensor<4xf32>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.domain (mhlo::DomainOp)

Domain operation

This operation is private to the XLA compiler, so it is does not yet have a specification.

Informally, these operations are used to group instructions with the same DomainMetadata property. ShardingMetadata is the main use case today to group instructions on the same device. Domain instructions provide two major benefits:

• Prevent unintentionally optimizing instructions across domains.
• Automatically assign the metadata of the instructions created in the domain. Without domain instructions, each HLO optimization pass would have to check and propagate the metadata, which would be easy to miss and also adds complexity to the compiler. Since domain instructions connect two different domains, each domain instruction is associated with two DomainMetadata -- one on the operand side and one on the user side of the domain.

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.dot (mhlo::DotOp)

Dot operation

This operation is on its way out of StableHLO, so it is not included in the specification: https://github.com/openxla/stablehlo/issues/3

Informally, this operation does the same thing as XLA's Dot: https://www.tensorflow.org/xla/operation_semantics#dot

Esempio:

%0 = mhlo.dot %arg0, %arg1 : (tensor<1x2xi32>, tensor<2x1xi32>) -> tensor<1x1xi32>

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.dot_general (mhlo::DotGeneralOp)

DotGeneral operation

Computes dot products between slices of lhs and slices of rhs and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#dot_general

Esempio:

%result = "mhlo.dot_general"(%lhs, %rhs) {
  dot_dimension_numbers = #mhlo.dot<
    lhs_batching_dimensions = [0],
    rhs_batching_dimensions = [0],
    lhs_contracting_dimensions = [2],
    rhs_contracting_dimensions = [1]
  >,
  precision_config = [#stablehlo<precision DEFAULT>, #stablehlo<precision DEFAULT>]
} : (tensor<2x2x2xi32>, tensor<2x2x2xi32>) -> tensor<2x2x2xi32>

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.dynamic_broadcast_in_dim (mhlo::DynamicBroadcastInDimOp)

DynamicBroadcastInDim operation

This operation is functionally identical to broadcast_in_dim op, but the result shape is specified dynamically via output_dimensions .

It also accepts optional attributes to express static knowledge about the expanding behavior of dimensions. If not specified, all dimensions are assumed to be possibly expanding. The sets of dimensions that are known to be expanding and the set of dimensions that are known to be non-expanding must be disjoint and they must be a subset of the operand's dimensions.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#dynamic_broadcast_in_dim

Esempio:

%operand = mhlo.constant dense<[[1, 2, 3]]> : tensor<1x3xi64>
%output_dimensions = mhlo.constant dense<[2, 3, 2]> : tensor<3xi64>
%result = "mhlo.dynamic_broadcast_in_dim"(%operand, %output_dimensions) {
  broadcast_dimensions = array<i64: 2, 1>,
  known_expanding_dimensions = array<i64: 0>,
  known_nonexpanding_dimensions = array<i64: 1>
} : (tensor<1x3xi64>, tensor<3xi64>) -> tensor<2x3x2xi64>

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.dynamic_conv (mhlo::DynamicConvOp)

DynamicConv operation

This operation is a work in progress, so it is not yet included in the specification: https://github.com/openxla/stablehlo/issues/8

Informally, this operation does the same thing as ConvolutionOp except that padding is specified dynamically via d_padding : https://github.com/openxla/stablehlo/blob/main/docs/spec.md#convolution

Esempio:

%result = "mhlo.dynamic_conv"(%lhs, %rhs, %d_padding) {
  window_strides = dense<4> : tensor<2xi64>,
  lhs_dilation = dense<2> : tensor<2xi64>,
  rhs_dilation = dense<1> : tensor<2xi64>,
  window_reversal = dense<false> : tensor<2xi1>,
  dimension_numbers = #mhlo.conv<[b, 0, 1, f]x[0, 1, i, o]->[b, 0, 1, f]>,
  feature_group_count = 1 : i64,
  batch_group_count = 1 : i64,
  precision_config = [#stablehlo<precision DEFAULT>, #stablehlo<precision DEFAULT>]
} : (tensor<1x4x4x1xi32>, tensor<3x3x1x1xi32>, tensor<2x2xi64>) -> tensor<1x2x2x1xi32>

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.dynamic_gather (mhlo::DynamicGatherOp)

DynamicGather operation

This operation is a work in progress, so it is not yet included in the specification: https://github.com/openxla/stablehlo/issues/8

Informally, this operation does the same thing as GatherOp except that slice_sizes are specified dynamically: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#gather

Esempio:

%result = "mhlo.dynamic_gather"(%operand, %start_indices, %slice_sizes) {
  dimension_numbers = #mhlo.gather<
    offset_dims = [2, 3],
    collapsed_slice_dims = [0],
    start_index_map = [0, 2],
    index_vector_dim = 2>,
  indices_are_sorted = false
} : (tensor<3x4x2xi32>, tensor<2x3x2xi64>, tensor<3xi64>) -> tensor<2x3x2x2xi32>

Traits: AlwaysSpeculatableImplTrait , InferTensorType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.dynamic_iota (mhlo::DynamicIotaOp)

DynamicIota operation

This operation is functionally identical to iota op, but the result shape is specified dynamically via output_shape .

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#dynamic_iota

Esempio:

%0 = mhlo.dynamic_iota %arg0, dim = 0 : (tensor<1xindex>) -> tensor<4xi32>

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.dynamic_pad (mhlo::DynamicPadOp)

DynamicPad operation

Sintassi:

operation ::= `mhlo.dynamic_pad` operands attr-dict `:` functional-type(operands, results)

Dynamically Pads the operand , with amount of padding added at low-end/high-end/interior is passed through input tensors.

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.dynamic_reshape (mhlo::DynamicReshapeOp)

DynamicReshape operation

Sintassi:

operation ::= `mhlo.dynamic_reshape` operands attr-dict `:` functional-type(operands, results)

This operation is functionally identical to reshape op, but the result shape is specified dynamically via output_shape .

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#dynamic_reshape

Esempio:

%output_shape = mhlo.constant dense<[3, 2]> : tensor<2xi64>
%result = mhlo.dynamic_reshape %operand, %output_shape : (tensor<2x3xi64>, tensor<2xi64>) -> tensor<3x2xi64>

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.dynamic_slice (mhlo::DynamicSliceOp)

DynamicSlice operation

Extracts a slice from the operand using dynamically-computed starting indices and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#dynamic_slice

Esempio:

%result = mhlo.dynamic_slice %operand, %start_indices0, %start_indices1, sizes = [2, 2]
  : (tensor<4x4xi32>, tensor<i64>, tensor<i64>) -> tensor<2x2xi32>

Traits: AlwaysSpeculatableImplTrait , InferTensorType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.dynamic_update_slice (mhlo::DynamicUpdateSliceOp)

DynamicUpdateSlice operation

Sintassi:

operation ::= `mhlo.dynamic_update_slice` operands attr-dict `:` functional-type(operands, results)

Produces a result tensor which is equal to the operand tensor except that the slice starting at start_indices is updated with the values in update .

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#dynamic_update_slice

Esempio:

%result = mhlo.dynamic_update_slice %operand, %update, %start_indices0, %start_indices1
  : (tensor<4x4xi32>, tensor<2x2xi32>, tensor<i64>, tensor<i64>) -> tensor<4x4xi32>

Traits: AlwaysSpeculatableImplTrait , InferTensorType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.einsum (mhlo::EinsumOp)

Einsum operation

This operation is on its way out of StableHLO, so it is not included in the specification: https://github.com/openxla/stablehlo/issues/3

Informally, this operation does the same thing as TF's einsum: https://www.tensorflow.org/api_docs/python/tf/einsum

Esempio:

%result = "mhlo.einsum"(%lhs, %rhs) {
  einsum_config = "ab,bc->ac"
} : (tensor<4x16xf32>, tensor<16x4xf32>) -> tensor<4x4xf32>

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.erf (mhlo::ErfOp)

Erf operation

Sintassi:

operation ::= `mhlo.erf` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Performs element-wise erf operation on operand tensor and produces a result tensor.

Esempio:

%result = mhlo.erf %operand : tensor<2x2xf32>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.exponential (mhlo::ExpOp)

Exp operation

Sintassi:

operation ::= `mhlo.exponential` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Performs element-wise exponential operation on operand tensor and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#exponential

Esempio:

%result = mhlo.exponential %operand : tensor<2x2xf64>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.exponential_minus_one (mhlo::Expm1Op)

Expm1 operation

Sintassi:

operation ::= `mhlo.exponential_minus_one` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Performs element-wise exponential minus one operation on operand tensor and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#exponential_minus_one

Esempio:

%result = mhlo.exponential_minus_one %operand : tensor<2xf32>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.fft (mhlo::FftOp)

Fft operation

Performs the forward and inverse Fourier transforms for real and complex inputs/outputs.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#fft

Esempio:

%result = mhlo.fft %operand, type = FFT, length = [4] : (tensor<4xcomplex<f32>>) -> tensor<4xcomplex<f32>>

Traits: AlwaysSpeculatableImplTrait , InferTensorType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.floor (mhlo::FloorOp)

Floor operation

Sintassi:

operation ::= `mhlo.floor` $operand attr-dict
              `:` custom<SameOperandsAndResultType>(type($operand), type($result))

Performs element-wise floor of operand tensor and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#floor

Esempio:

%result = mhlo.floor %operand : tensor<2xf32>

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Results:

mhlo.fusion (mhlo::FusionOp)

Fusion operation

This operation is private to the XLA compiler, so it is does not yet have a specification.

Informally, this operation consists of a group of basic ops (represented as a region attached to it). It serves as a hint to the backend that it is beneficial to emit the contained ops into a single loop nest or kernel.

Attributes:

Operands:

Results:

mhlo.gather (mhlo::GatherOp)

Gather operation

Gathers slices from operand tensor from offsets specified in start_indices and produces a result tensor.

See: https://github.com/openxla/stablehlo/blob/main/docs/spec.md#gather

Esempio:

%result = "mhlo.gather"(%operand, %start_indices) {
  dimension_numbers = #stablehlo.gather<
    offset_dims = [3, 4],
    collapsed_slice_dims = [1],
    operand_batching_dims = [0],
    start_indices_batching_dims = [1],
    start_index_map = [2, 1],
    index_vector_dim = 3>,
  slice_sizes = dense<[0, 2, 2]> : tensor<3xi64>,
  indices_are_sorted = false
} : (tensor<2x3x4x2xi64>, tensor<2x2x3x2xi64>) -> tensor<2x2x3x2x2xi64>

Traits: AlwaysSpeculatableImplTrait , InferTensorType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results:

mhlo.get_dimension_size (mhlo::GetDimensionSizeOp)

GetDimensionSize operation

Produces the size of the given dimension of the operand .

See https://github.com/openxla/stablehlo/blob/main/docs/spec.md#get_dimension_size

Esempio:

%result = mhlo.get_dimension_size %operand, dim = 1 : (tensor<2x3xf32>) -> tensor<i32>

Traits: AlwaysSpeculatableImplTrait , InferTensorType

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Attributes:

Operands:

Results: