GEMMInfo Class Reference

GEMM information class. More...

#include <GEMMInfo.h>

Public Member Functions
	GEMMInfo () noexcept
	Default constructor. More...
	GEMMInfo (bool is_a_reshaped, bool is_b_reshaped, bool reshape_b_only_on_first_run, int depth_output_gemm3d=0, bool reinterpret_input_as_3d=false, bool retain_internal_weights=false, GEMMLowpOutputStageInfo gemmlowp_output_stage=GEMMLowpOutputStageInfo(), bool fp_mixed_precision=false, bool fast_math=false, bool broadcast_bias=false, const ActivationLayerInfo &activation_info=ActivationLayerInfo(), const experimental::PostOpList< ITensorInfo * > &post_ops=experimental::PostOpList< ITensorInfo * >(), bool fixed_format=false, arm_compute::WeightFormat weight_format=arm_compute::WeightFormat::UNSPECIFIED) noexcept
	Constructor. More...
bool	is_a_reshaped () const
	Flag which specifies if the matrix A has been reshaped. More...
bool	is_b_reshaped () const
	Flag which specifies if the matrix B has been reshaped. More...
bool	reshape_b_only_on_first_run () const
	Flag which specifies if the reshape of matrix B should executed only for the first. More...
int	depth_output_gemm3d () const
	Depth of the output when GEMM output is reinterpreted as 3D tensor. More...
bool	reinterpret_input_as_3d () const
	Flag which specifies if the input tensor has to be reinterpreted as 3D. More...
bool	retain_internal_weights () const
	Flag which specifies if the weights tensor has to be retained from previous run. More...
GEMMLowpOutputStageInfo	gemmlowp_output_stage () const
	GEMMLowp output stage. More...
void	set_gemmlowp_output_stage (GEMMLowpOutputStageInfo &output_stage)
	Sets GEMMLowp output stage. More...
bool	fp_mixed_precision () const
	Flag which specifies if a wider accumulator should be used. More...
bool	fast_math () const
	Flag which specifies if a shorter accumulator to be used. More...
void	set_fast_math (bool fast_math)
	Set fast math flag. More...
bool	broadcast_bias () const
	Flag which specifies whether to broadcast the shape of the bias tensor. More...
bool	pretranspose_A () const
	Flag which specifies whether A should be pre-transposed if supported. More...
void	set_pretranspose_A (bool flag)
	Set pre-transpose A flag. More...
bool	pretranspose_B () const
	Flag which specifies whether b should be pre-transposed if supported. More...
void	set_pretranspose_B (bool flag)
	Set pre-transpose b flag. More...
ActivationLayerInfo	activation_info () const
	Activation layer to apply after the matrix multiplication. More...
void	set_activation_info (const ActivationLayerInfo &activation_info)
	Set activation layer info. More...
const experimental::PostOpList< ITensorInfo * > &	post_ops () const
	Post operations to apply after the matrix multiplication. More...
void	set_post_ops (const experimental::PostOpList< ITensorInfo * > &post_ops)
	Set post ops. More...
bool	fixed_format () const
	Flag which specifies if the GEMM operation is running fixed-format kernels. More...
void	set_fixed_format (bool fixed_format)
	Set fixed-format flag. More...
arm_compute::WeightFormat	weight_format () const
void	set_weight_format (arm_compute::WeightFormat weight_format)
	Set weight format to be used. More...

Detailed Description

GEMM information class.

This class stores the necessary information to compute GEMM functions

This object also contains the information about how matrix A and matrix B have been reshaped

Definition at line 64 of file GEMMInfo.h.

Constructor & Destructor Documentation

GEMMInfo() [1/2]

GEMMInfo ( )

inlinenoexcept

Default constructor.

Definition at line 68 of file GEMMInfo.h.

        : _is_a_reshaped(false),
          _is_b_reshaped(false),
          _reshape_b_only_on_first_run(true),
          _depth_output_gemm3d(0),
          _reinterpret_input_as_3d(false),
          _retain_internal_weights(false),
          _gemmlowp_output_stage(),
          _fast_math(false),
          _fp_mixed_precision(false),
          _broadcast_bias(false),
          _pretranspose_A(false),
          _pretranspose_B(false),
          _activation_info(),
          _post_ops(),
          _fixed_format(false),
          _weight_format(arm_compute::WeightFormat::UNSPECIFIED)
    {
    }

GEMMInfo() [2/2]

GEMMInfo ( bool  is_a_reshaped, bool  is_b_reshaped, bool  reshape_b_only_on_first_run, int  depth_output_gemm3d = 0, bool  reinterpret_input_as_3d = false, bool  retain_internal_weights = false, GEMMLowpOutputStageInfo  gemmlowp_output_stage = GEMMLowpOutputStageInfo(), bool  fp_mixed_precision = false, bool  fast_math = false, bool  broadcast_bias = false, const ActivationLayerInfo &  activation_info = ActivationLayerInfo(), const experimental::PostOpList< ITensorInfo * > &  post_ops = experimental::PostOpList<ITensorInfo *>(), bool  fixed_format = false, arm_compute::WeightFormat  weight_format = arm_compute::WeightFormat::UNSPECIFIED  )

inlinenoexcept

Constructor.

Parameters

[in]	is_a_reshaped	True if the matrix A has been reshaped
[in]	is_b_reshaped	True if the matrix B has been reshaped
[in]	reshape_b_only_on_first_run	Reshape matrix B only for the first run
[in]	depth_output_gemm3d	(Optional) Depth (third dimension) of the output tensor to be used with the GEMM3D kernel If 0 the output will not be reinterpreted as 3D. Default 0
[in]	reinterpret_input_as_3d	(Optional) Reinterpret the input as 3D tensor. (i.e. this flag should be set to true when GEMM is used to perform 1x1 convolutions with the NHWC data layout)
[in]	retain_internal_weights	(Optional) Retain the weights tensor from previous run
[in]	gemmlowp_output_stage	(Optional) GEMMLowp Output stage info
[in]	fp_mixed_precision	(Optional) Use wider accumulators (32 bit instead of 16 for FP16) to improve accuracy.
[in]	fast_math	(Optional) Use a data type of shorter width to improve performance
[in]	broadcast_bias	(Optional) Broadcast the shape of the bias tensor from a vector to a matrix.
[in]	activation_info	(Optional) Activation to apply after the matrix multiplication
[in]	post_ops	(Optional) A sequence of post operations that are performed after the main operation.
[in]	fixed_format	(Optional) Specify the selection of fixed format kernels for variable weights support in GEMM. These kernels expect the weights tensor to be in amemory format that is fixed by the kernel itself. For more information, see arm_compute::WeightFormat.
[in]	weight_format	(Optional) arm_gemm:WeightFormat enumeration requested by the user. Default is arm_compute::WeightFormat::UNSPECIFIED.

Definition at line 106 of file GEMMInfo.h.

        : _is_a_reshaped(is_a_reshaped),
          _is_b_reshaped(is_b_reshaped),
          _reshape_b_only_on_first_run(reshape_b_only_on_first_run),
          _depth_output_gemm3d(depth_output_gemm3d),
          _reinterpret_input_as_3d(reinterpret_input_as_3d),
          _retain_internal_weights(retain_internal_weights),
          _gemmlowp_output_stage(gemmlowp_output_stage),
          _fast_math(fast_math),
          _fp_mixed_precision(fp_mixed_precision),
          _broadcast_bias(broadcast_bias),
          _pretranspose_A(false),
          _pretranspose_B(false),
          _activation_info(activation_info),
          _post_ops(post_ops),
          _fixed_format(fixed_format),
          _weight_format(weight_format)
    {
    }

Member Function Documentation

activation_info()

ActivationLayerInfo activation_info ( ) const

inline

Activation layer to apply after the matrix multiplication.

Returns

ActivationLayerInfo object

Definition at line 262 of file GEMMInfo.h.

    {
        return _activation_info;
    }

Referenced by GEMMInfo::set_activation_info().

broadcast_bias()

bool broadcast_bias ( ) const

inline

Flag which specifies whether to broadcast the shape of the bias tensor.

Returns

True if the shape of the bias tensor is to be broadcasted.

Definition at line 222 of file GEMMInfo.h.

    {
        return _broadcast_bias;
    };

depth_output_gemm3d()

int depth_output_gemm3d ( ) const

inline

Depth of the output when GEMM output is reinterpreted as 3D tensor.

Returns

the depth of the output tensor

Definition at line 158 of file GEMMInfo.h.

    {
        return _depth_output_gemm3d;
    };

fast_math()

bool fast_math ( ) const

inline

Flag which specifies if a shorter accumulator to be used.

Returns

True if a shorter accumulator has to be used

Definition at line 206 of file GEMMInfo.h.

    {
        return _fast_math;
    };

Referenced by GEMMInfo::set_fast_math().

fixed_format()

bool fixed_format ( ) const

inline

Flag which specifies if the GEMM operation is running fixed-format kernels.

Returns

True if the GEMM operation is running fixed-format kernel else false.

Definition at line 294 of file GEMMInfo.h.

    {
        return _fixed_format;
    }

Referenced by GEMMInfo::set_fixed_format().

fp_mixed_precision()

bool fp_mixed_precision ( ) const

inline

Flag which specifies if a wider accumulator should be used.

Returns

True if a wider accumulator has to be used

Definition at line 198 of file GEMMInfo.h.

    {
        return _fp_mixed_precision;
    };

gemmlowp_output_stage()

GEMMLowpOutputStageInfo gemmlowp_output_stage ( ) const

inline

GEMMLowp output stage.

Returns

the GEMMLowp output stage info

Definition at line 182 of file GEMMInfo.h.

    {
        return _gemmlowp_output_stage;
    };

Referenced by ClGemmLowpMatrixMultiplyCore::prepare(), and CpuGemmLowpMatrixMultiplyCore::run().

is_a_reshaped()

bool is_a_reshaped ( ) const

inline

Flag which specifies if the matrix A has been reshaped.

Returns

True if the matrix A has been reshaped

Definition at line 132 of file GEMMInfo.h.

    {
        return _is_a_reshaped;
    };

is_b_reshaped()

bool is_b_reshaped ( ) const

inline

Flag which specifies if the matrix B has been reshaped.

Returns

True if the matrix B has been reshaped

Definition at line 140 of file GEMMInfo.h.

    {
        return _is_b_reshaped;
    };

post_ops()

const experimental::PostOpList<ITensorInfo *>& post_ops ( ) const

inline

Post operations to apply after the matrix multiplication.

Returns

experimental::PostOpList object

Definition at line 278 of file GEMMInfo.h.

    {
        return _post_ops;
    }

Referenced by GEMMInfo::set_post_ops().

pretranspose_A()

bool pretranspose_A ( ) const

inline

Flag which specifies whether A should be pre-transposed if supported.

Returns

True if A should be pre-transposed else false.

Definition at line 230 of file GEMMInfo.h.

    {
        return _pretranspose_A;
    };

pretranspose_B()

bool pretranspose_B ( ) const

inline

Flag which specifies whether b should be pre-transposed if supported.

Returns

True if b should be pre-transposed else false.

Definition at line 246 of file GEMMInfo.h.

    {
        return _pretranspose_B;
    };

reinterpret_input_as_3d()

bool reinterpret_input_as_3d ( ) const

inline

Flag which specifies if the input tensor has to be reinterpreted as 3D.

Returns

True if the input tensor has to be reinterpreted as 3D tensor

Definition at line 166 of file GEMMInfo.h.

    {
        return _reinterpret_input_as_3d;
    };

reshape_b_only_on_first_run()

bool reshape_b_only_on_first_run ( ) const

inline

Flag which specifies if the reshape of matrix B should executed only for the first.

Note

This flag could be set to TRUE when GEMM is used to accelerate convolution layer

Returns

True if the reshaped of matrix B happens only for the first run

Definition at line 150 of file GEMMInfo.h.

    {
        return _reshape_b_only_on_first_run;
    };

retain_internal_weights()

bool retain_internal_weights ( ) const

inline

Flag which specifies if the weights tensor has to be retained from previous run.

Returns

True if the weights tensor has to be retained

Definition at line 174 of file GEMMInfo.h.

    {
        return _retain_internal_weights;
    };

set_activation_info()

void set_activation_info ( const ActivationLayerInfo &  activation_info )

inline

Set activation layer info.

Parameters

[in]

activation_info

ActivationLayerInfo object to set

Definition at line 270 of file GEMMInfo.h.

    {
        _activation_info = activation_info;
    }

References GEMMInfo::activation_info().

set_fast_math()

void set_fast_math ( bool  fast_math )

inline

Set fast math flag.

Parameters

[in]

fast_math

Flag to set

Definition at line 214 of file GEMMInfo.h.

    {
        _fast_math = fast_math;
    }

References GEMMInfo::fast_math().

set_fixed_format()

void set_fixed_format ( bool  fixed_format )

inline

Set fixed-format flag.

Parameters

[in]

fixed_format

sets whether or not to use fixed-format kernels

Definition at line 303 of file GEMMInfo.h.

    {
        _fixed_format = fixed_format;
    }

References GEMMInfo::fixed_format().

set_gemmlowp_output_stage()

void set_gemmlowp_output_stage ( GEMMLowpOutputStageInfo &  output_stage )

inline

Sets GEMMLowp output stage.

Parameters

[in]

output_stage

Output stage to set

Definition at line 190 of file GEMMInfo.h.

    {
        _gemmlowp_output_stage = output_stage;
    };

References output_stage.

set_post_ops()

void set_post_ops ( const experimental::PostOpList< ITensorInfo * > &  post_ops )

inline

Set post ops.

Parameters

[in]

post_ops

experimental::PostOpList object to set

Definition at line 286 of file GEMMInfo.h.

    {
        _post_ops = post_ops;
    }

References GEMMInfo::post_ops().

set_pretranspose_A()

void set_pretranspose_A ( bool  flag )

inline

Set pre-transpose A flag.

Parameters

[in]

flag

Flag to set

Definition at line 238 of file GEMMInfo.h.

    {
        _pretranspose_A = flag;
    }

set_pretranspose_B()

void set_pretranspose_B ( bool  flag )

inline

Set pre-transpose b flag.

Parameters

[in]

flag

Flag to set

Definition at line 254 of file GEMMInfo.h.

    {
        _pretranspose_B = flag;
    }

set_weight_format()

void set_weight_format ( arm_compute::WeightFormat  weight_format )

inline

Set weight format to be used.

Parameters

[in]

weight_format

arm_compute::WeightFormat enumeration

Definition at line 317 of file GEMMInfo.h.

    {
        _weight_format = weight_format;
    }

References GEMMInfo::weight_format().

weight_format()

arm_compute::WeightFormat weight_format ( ) const

inline

Definition at line 308 of file GEMMInfo.h.

    {
        return _weight_format;
    }

Referenced by GEMMInfo::set_weight_format().

---

The documentation for this class was generated from the following file:

• arm_compute/function_info/GEMMInfo.h