'mhlo' Dialeto

Operações

mhlo.abs (mhlo::AbsOp)

Operação de abdômen

Sintaxe:

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

Executa operação abs elemento a elemento no tensor operand e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , Elementwise , SameOperandsAndResultShape

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

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.acos (mhlo::AcosOp)

Operação Acos

Sintaxe:

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

Executa a operação acos elemento a elemento no tensor operand e produz um tensor result .

Exemplo:

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

Características: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Operandos:

Resultados:

mhlo.acosh (mhlo::AcoshOp)

Operação Acosh

Sintaxe:

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

Executa a operação acosh elemento a elemento no tensor operand e produz um tensor result .

Exemplo:

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

Características: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Operandos:

Resultados:

mhlo.add (mhlo::AddOp)

Adicionar operação

Sintaxe:

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

Executa a adição elemento a elemento de dois tensores lhs e rhs e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , Commutative , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.add_dependency (mhlo::AddDependencyOp)

Operação AddDependency

Sintaxe:

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

Esta operação é privada do compilador XLA, portanto ainda não possui uma especificação.

Informalmente, esta operação utiliza dois operandos: um operando de dados e um token. A saída da operação é o operando de dados. Quando usada com AfterAll, esta operação permite ordenar operações sem efeitos colaterais (aquelas que não produzem valores de token).

Exemplo:

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

Traços: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.after_all (mhlo::AfterAllOp)

Operação AfterAll

Sintaxe:

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

Garante que as operações que produzem as inputs sejam executadas antes de quaisquer operações que dependam do result .

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

Exemplo:

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

Traços: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.all_gather (mhlo::AllGatherOp)

Operação AllGather

Dentro de cada grupo de processos na grade de processos, concatena os valores do tensor de operando de cada processo ao longo all_gather_dim e produz um tensor de resultado. O computation é aplicado separadamente para cada operando em operands , produzindo um resultado por operando.

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

Exemplo:

%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>

Características: SameOperandsAndResultElementType

Atributos:

Operandos:

Resultados:

mhlo.all_reduce (mhlo::AllReduceOp)

Operação AllReduce

Dentro de cada grupo de processos na grade de processos, aplica-se um computation de função de redução aos valores de um tensor de operando de cada processo e produz-se um tensor de resultado. O computation é aplicado separadamente para cada operando em operands , produzindo um resultado por operando.

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

Exemplo:

%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>

Características: InferTensorType , SingleBlockImplicitTerminator<ReturnOp> , SingleBlock

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Atributos:

Operandos:

Resultados:

mhlo.all_to_all (mhlo::AllToAllOp)

Operação AllToAll

Dentro de cada grupo de processos na grade de processos, divide os valores do tensor do operand ao longo split_dimension em partes, espalha as partes divididas entre os processos, concatena as partes dispersas ao longo concat_dimension e produz um tensor result .

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

Exemplo:

%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>

Características: AlwaysSpeculatableImplTrait , InferTensorType , SameOperandsElementType , SameOperandsShape , SameVariadicOperandSize

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.and (mhlo::AndOp)

E operação

Sintaxe:

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

Executa AND elemento a elemento de dois tensores lhs e rhs e produz um tensor result

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , Commutative , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.asin (mhlo::AsinOp)

Operação Asin

Sintaxe:

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

Executa uma operação asin elemento a elemento no tensor operand e produz um tensor result .

Exemplo:

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

Características: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Operandos:

Resultados:

mhlo.asinh (mhlo::AsinhOp)

Operação Asinh

Sintaxe:

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

Executa uma operação asinh elemento a elemento no tensor operand e produz um tensor result .

Exemplo:

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

Características: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Operandos:

Resultados:

mhlo.async_done (mhlo::AsyncDoneOp)

Operação AsyncDone

Esta operação é privada do compilador XLA, portanto ainda não possui uma especificação.

Informalmente, esta operação bloqueia até o final de uma computação assíncrona. Ela retorna o resultado final da computação assíncrona.

Consulte a documentação do AsyncStart para obter mais informações.

Interfaces: InferTypeOpInterface

Operandos:

Resultados:

mhlo.async_start (mhlo::AsyncStartOp)

Operação AsyncStart

Esta operação é privada do compilador XLA, portanto ainda não possui uma especificação.

Informalmente, essa operação inicia uma computação assíncrona.

Isso é usado quando há funções que contêm esperas assíncronas (como DMAs) e computação on-thread. Por exemplo, uma função pode consistir em uma computação, uma DMA, outra computação, uma segunda DMA e uma computação final. Isso seria representado como um async_start seguido por um async_update e um async_done. O async_start faria a primeira computação on-thread e então iniciaria a DMA. O async_update esperaria a DMA ser concluída, caso ainda não tivesse sido concluída, então executaria a segunda computação na função e iniciaria a segunda DMA. Finalmente, o async_done esperaria por esta última DMA e então executaria a última computação que precisa ser executada on-thread e retornaria o resultado dessa computação final.

operands são passados ​​diretamente para a computação. called_computation é a função que será executada de forma assíncrona. execution_thread é o nome da thread na qual ela será executada. A thread principal é chamada de "main". Todas as threads têm nomes.

Isso retorna todo o estado necessário entre operações assíncronas. Após a atribuição do buffer, os valores retornados representam o espaço necessário para armazenar a entrada, os resultados e quaisquer blocos de anotações necessários ou editados pela operação assíncrona.

Atributos:

Operandos:

Resultados:

mhlo.async_update (mhlo::AsyncUpdateOp)

Operação AsyncUpdate

Esta operação é privada do compilador XLA, portanto ainda não possui uma especificação.

Informalmente, esta operação bloqueia uma computação assíncrona até que haja uma barreira de sincronização. Isso retorna bundle após a operação.

Consulte a documentação do AsyncStart para obter mais informações.

Interfaces: InferTypeOpInterface

Operandos:

Resultados:

mhlo.atan2 (mhlo::Atan2Op)

Operação Atan2

Sintaxe:

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

Executa a operação atan2 elemento a elemento nos tensores lhs e rhs e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.atanh (mhlo::AtanhOp)

Operação Atanh

Sintaxe:

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

Executa a operação atanh elemento a elemento no tensor operand e produz um tensor result .

Exemplo:

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

Características: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Operandos:

Resultados:

mhlo.batch_norm_grad (mhlo::BatchNormGradOp)

Operação BatchNormGrad

Calcula gradientes de várias entradas de BatchNormTrainingOp retropropagando de grad_output e produz tensores grad_operand , grad_scale e grad_offset .

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

Exemplo:

%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>)

Características: AlwaysSpeculatableImplTrait , InferTensorType

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.batch_norm_inference (mhlo::BatchNormInferenceOp)

Operação BatchNormInference

Normaliza o tensor operand em todas as dimensões, exceto na dimensão feature_index , e produz um tensor result .

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

Exemplo:

%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>

Características: AlwaysSpeculatableImplTrait , InferTensorType

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.batch_norm_training (mhlo::BatchNormTrainingOp)

Operação BatchNormTraining

Calcula a média e a variância entre dimensões de lote e espaciais e normaliza o tensor operand para cada recurso na dimensão feature_index e produz tensores output , batch_mean e batch_var .

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

Exemplo:

%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>)

Características: AlwaysSpeculatableImplTrait , InferTensorType

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.bitcast (mhlo::BitcastOp)

Operação Bitcast

Sintaxe:

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

Esta operação é privada do compilador XLA, portanto ainda não possui uma especificação.

Informalmente, essa operação altera a forma da entrada de forma que o arranjo físico dos elementos permanece inalterado.

Esta operação precisa de informações de layout para dar sentido ao "arranjo físico dos elementos", e o suporte ao layout no MHLO está atualmente em andamento.

Exemplo:

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

Traços: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.bitcast_convert (mhlo::BitcastConvertOp)

Operação BitcastConvert

Sintaxe:

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

Executa uma operação de bitcast no tensor operand e produz um tensor result onde os bits de todo o tensor operand são reinterpretados usando o tipo do tensor result .

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

Exemplo:

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

Traços: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.broadcast (mhlo::BroadcastOp)

Operação de transmissão

Esta operação está saindo do StableHLO, portanto não está incluída na especificação: https://github.com/openxla/stablehlo/issues/3

Informalmente, esta operação faz a mesma coisa que a transmissão do XLA: https://www.tensorflow.org/xla/operation_semantics#broadcast

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , InferTensorType , SameOperandsAndResultElementType

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.broadcast_in_dim (mhlo::BroadcastInDimOp)

Operação BroadcastInDim

Expande as dimensões e/ou classificação de um tensor de entrada duplicando os dados no tensor operand e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , HLO_CompatibleOperandsAndResultElementType

Interfaces: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.case (mhlo::CaseOp)

Operação de caso

Produz a saída da execução de exatamente uma function de branches dependendo do valor do index .

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

Exemplo:

%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>)

Características: RecursiveMemoryEffects , SingleBlockImplicitTerminator<ReturnOp> , SingleBlock

Interfaces: InferTypeOpInterface

Operandos:

Resultados:

mhlo.cbrt (mhlo::CbrtOp)

Operação Cbrt

Sintaxe:

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

Executa operação de raiz cúbica elemento a elemento no tensor operand e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.ceil (mhlo::CeilOp)

Operação de teto

Sintaxe:

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

Executa o teto do tensor operand elemento a elemento e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.cholesky (mhlo::CholeskyOp)

Operação Cholesky

Calcula a decomposição de Cholesky de um lote de matrizes.

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , InferTensorType , SameOperandsAndResultElementType

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.clamp (mhlo::ClampOp)

Operação de fixação

Sintaxe:

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

Fixa cada elemento do tensor operand entre um valor mínimo e máximo e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , HLO_BroadcastingElementwise , InferTensorType , SameOperandsAndResultElementType

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

Efeitos: MemoryEffects::Effect{}

Operandos:

Resultados:

mhlo.collective_broadcast (mhlo::Opção_de_transmissão_coletiva)

Operação de transmissão coletiva

Dentro de cada grupo de processos na grade de processos, envie o valor do tensor do operand do processo de origem para os processos de destino e produza um tensor result .

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

Exemplo:

%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>

Características: CompatibleOperandsAndResultType

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Atributos:

Operandos:

Resultados:

mhlo.collective_permute (mhlo::OpçãoDePermutaColetiva)

Operação CollectivePermute

Dentro de cada grupo de processos na grade de processos, envia o valor do tensor do operand do processo de origem para o processo de destino e produz um tensor result .

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

Exemplo:

%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>

Características: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.compare (mhlo::CompareOp)

Comparar operação

Sintaxe:

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

Executa comparação elemento a elemento dos tensores lhs e rhs de acordo com comparison_direction e compare_type e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , Elementwise , InferTensorType , SameOperandsAndResultShape , SameOperandsElementType

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

Efeitos: MemoryEffects::Effect{}

Atributos:

Operandos:

Resultados:

mhlo.complex (mhlo::ComplexOp)

Operação complexa

Sintaxe:

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

Executa a conversão elemento a elemento para um valor complexo a partir de um par de valores reais e imaginários, lhs e rhs , e produz um tensor result .

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

Exemplo:

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

Características: AlwaysSpeculatableImplTrait , Elementwise , SameOperandsAndResultShape , SameOperandsElementType

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

Efeitos: MemoryEffects::Effect{}

Operands:

Resultados:

mhlo.composite (mhlo::CompositeOp)

Operação composta

Sintaxe:

operation ::= `mhlo.composite` $name $inputs 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

Exemplo:

%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

Atributos:

Operands:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

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

Exemplo:

%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{}

Atributos:

Resultados:

mhlo.convert (mhlo::ConvertOp)

Convert operation

Sintaxe:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , Elementwise , SameOperandsAndResultShape

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Operands:

Resultados:

mhlo.convolution (mhlo::ConvolutionOp)

Operação de convolução

Sintaxe:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

mhlo.copy (mhlo::CopyOp)

Copy operation

Sintaxe:

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.

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise

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

Effects: MemoryEffects::Effect{}

Atributos:

Operands:

Resultados:

mhlo.cosh (mhlo::CoshOp)

Cosh operation

Sintaxe:

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.

Exemplo:

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

Traits: CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

Interfaces: InferShapedTypeOpInterface , InferTypeOpInterface

Operands:

Resultados:

mhlo.cosine (mhlo::CosineOp)

Cosine operation

Sintaxe:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Effects: MemoryEffects::Effect{}

Atributos:

Operands:

Resultados:

mhlo.count_leading_zeros (mhlo::ClzOp)

Clz operation

Sintaxe:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Effects: MemoryEffects::Effect{}

Operands:

Resultados:

mhlo.create_token (mhlo::CreateTokenOp)

CreateToken operation

Sintaxe:

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

Exemplo:

%output = mhlo.create_token : !mhlo.token

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

mhlo.custom_call (mhlo::CustomCallOp)

CustomCall operation

Sintaxe:

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

Exemplo:

%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

Atributos:

Operands:

Resultados:

mhlo.divide (mhlo::DivOp)

Div operation

Sintaxe:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Effects: MemoryEffects::Effect{}

Operands:

Resultados:

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{}

Atributos:

Operands:

Resultados:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Atributos:

Operands:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait

Interfaces: ConditionallySpeculatable , InferShapedTypeOpInterface , NoMemoryEffect (MemoryEffectOpInterface)

Effects: MemoryEffects::Effect{}

Atributos:

Operands:

Resultados:

mhlo.dynamic_pad (mhlo::DynamicPadOp)

DynamicPad operation

Sintaxe:

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:

Resultados:

mhlo.dynamic_reshape (mhlo::DynamicReshapeOp)

DynamicReshape operation

Sintaxe:

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

Exemplo:

%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:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

mhlo.dynamic_update_slice (mhlo::DynamicUpdateSliceOp)

DynamicUpdateSlice operation

Sintaxe:

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

Exemplo:

%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:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

mhlo.erf (mhlo::ErfOp)

Erf operation

Sintaxe:

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.

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Effects: MemoryEffects::Effect{}

Operands:

Resultados:

mhlo.exponential (mhlo::ExpOp)

Exp operation

Sintaxe:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Effects: MemoryEffects::Effect{}

Atributos:

Operands:

Resultados:

mhlo.exponential_minus_one (mhlo::Expm1Op)

Expm1 operation

Sintaxe:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Effects: MemoryEffects::Effect{}

Atributos:

Operands:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

mhlo.floor (mhlo::FloorOp)

Floor operation

Sintaxe:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , CompatibleOperandsAndResultType , Elementwise , SameOperandsAndResultShape

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

Effects: MemoryEffects::Effect{}

Operands:

Resultados:

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.

Atributos:

Operands:

Resultados:

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

Exemplo:

%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{}

Atributos:

Operands:

Resultados:

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

Exemplo:

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

Traits: AlwaysSpeculatableImplTrait , InferTensorType

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

Effects: MemoryEffects::Effect{}

Atributos:

Operands:

Resultados:

mhlo.get_tuple_element (mhlo::GetTupleElementOp)

GetTupleElement operation

Sintaxe:

operation ::= `mhlo.get_tuple_element` $operand `[` $index `]` attr-dict `:` functional-type(operands, results)