struct dnnl::ukernel::brgemm#

Overview#

BRGeMM ukernel. More…

#include <dnnl_ukernel.hpp>

struct brgemm: public dnnl::handle
{
    // construction

    brgemm();

    brgemm(
        memory::dim M,
        memory::dim N,
        memory::dim K,
        memory::dim batch_size,
        memory::dim lda,
        memory::dim ldb,
        memory::dim ldc,
        memory::data_type a_dt,
        memory::data_type b_dt,
        memory::data_type c_dt,
        bool allow_empty = false
        );

    // methods

    void set_add_C(bool add_C);

    void set_post_ops(
        memory::dim ldd,
        memory::data_type d_dt,
        const post_ops& po = default_post_ops()
        );

    void set_A_scales(int a_scale_mask);
    void set_B_scales(int b_scale_mask);
    void set_D_scales(int d_scale_mask);
    void finalize();
    size_t get_scratchpad_size() const;
    bool is_execute_postops_valid() const;
    void set_hw_context() const;
    void generate();

    void execute(
        const void* A,
        const void* B,
        const std::vector<std::pair<memory::dim, memory::dim>>& A_B_offsets,
        void* C,
        void* scratchpad
        ) const;

    void execute(
        const void* A,
        const void* B,
        const std::vector<std::pair<memory::dim, memory::dim>>& A_B_offsets,
        const void* C,
        void* D,
        void* scratchpad,
        const attr_params& params = default_attr_params()
        ) const;

    static pack_type get_B_pack_type(memory::data_type a_dt, memory::data_type b_dt);
    static void release_hw_context();
    static const post_ops& default_post_ops();
    static const attr_params& default_attr_params();
};

Inherited Members#

public:
    // methods

    handle<T, traits>& operator = (const handle<T, traits>&);
    handle<T, traits>& operator = (handle<T, traits>&&);
    void reset(T t, bool weak = false);
    T get(bool allow_empty = false) const;
    operator T () const;
    operator bool () const;
    bool operator == (const handle<T, traits>& other) const;
    bool operator != (const handle& other) const;

Detailed Documentation#

BRGeMM ukernel.

Construction#

brgemm()

Default constructor. Produces an empty object.

brgemm(
    memory::dim M,
    memory::dim N,
    memory::dim K,
    memory::dim batch_size,
    memory::dim lda,
    memory::dim ldb,
    memory::dim ldc,
    memory::data_type a_dt,
    memory::data_type b_dt,
    memory::data_type c_dt,
    bool allow_empty = false
    )

Constructs a BRGeMM ukernel object.

Operates by the following formula: C = [A x B].

Parameters:

M	Dimension M of tensor A.
N	Dimension N of tensor B.
K	Dimension K of tensors A and B.
batch_size	Number of batches to process.
lda	Leading dimension of tensor A.
ldb	Leading dimension of tensor B.
ldc	Leading dimension of tensor C.
a_dt	Data type of tensor A.
b_dt	Data type of tensor B.
c_dt	Data type of tensor C.
allow_empty	A flag signifying whether construction is allowed to fail without throwing an exception. In this case an empty object will be produced. This flag is optional and defaults to false.

Methods#

void set_add_C(bool add_C)

Sets adding an intermediate result to the output tensor C instead of writing: C += [A x B].

Parameters:

add_C

Value to indicate addition. false to skip addition, and true to apply addition.

void set_post_ops(
    memory::dim ldd,
    memory::data_type d_dt,
    const post_ops& po = default_post_ops()
    )

Sets post-operations to a BRGeMM ukernel object: D = post-operations(C).

Post-operations applies if one of the following holds:

Non-empty post-operations are specified.
Output data type d_dt is different from accumulation data type c_dt.

Parameters:

ldd	Leading dimension of tensor D.
d_dt	Data type of tensor D.
po	Primitive post-operation attributes to extend the kernel operations.

void set_A_scales(int a_scale_mask)

Sets tensor A scales mask to a BRGeMM ukernel object.

For quantization flavor tensor A scales apply to accumulation buffer once C is ready.

Parameters:

a_scale_mask

Tensor A scale mask. Can be 0 only.

void set_B_scales(int b_scale_mask)

Sets tensor B scales mask to a BRGeMM ukernel object.

For quantization flavor tensor B scales apply to accumulation buffer once C is ready.

Parameters:

b_scale_mask

Tensor B scale mask. Can be 0 and 2 only.

void set_D_scales(int d_scale_mask)

Sets tensor D scales mask to a BRGeMM ukernel object.

For quantization flavor tensor D scales apply after all post-ops are applied.

Parameters:

d_scale_mask

Tensor D scale mask. Can be 0 only.

void finalize()

Finalizes initialization of a BRGeMM ukernel object.

This step must be performed prior to querying information from the object.

size_t get_scratchpad_size() const

Returns the size of a scratchpad memory needed for the BRGeMM ukernel object.

bool is_execute_postops_valid() const

Returns the flag indicating when the call to execute with post operations is valid.

True is for a valid call, false, otherwise.

void set_hw_context() const

Initializes the hardware-specific context.

Affects the global state for all BRGeMM ukernel objects. If no initialization required, returns.

void generate()

Generates an executable part of BRGeMM ukernel object.

void execute(
    const void* A,
    const void* B,
    const std::vector<std::pair<memory::dim, memory::dim>>& A_B_offsets,
    void* C,
    void* scratchpad
    ) const

Executes a BRGeMM ukernel object.

Parameters:

A	Base pointer to a tensor A.
B	Base pointer to a tensor B.
A_B_offsets	Vector of pairs of tensors A and B offsets for each batch. The number of batches must coincide with the `batch_size` value passed at object construction stage.
C	Pointer to a tensor C (accumulation buffer).
scratchpad	Pointer to a scratchpad buffer.

void execute(
    const void* A,
    const void* B,
    const std::vector<std::pair<memory::dim, memory::dim>>& A_B_offsets,
    const void* C,
    void* D,
    void* scratchpad,
    const attr_params& params = default_attr_params()
    ) const

Executes a BRGeMM ukernel object with post operations.

Parameters:

A	Base pointer to a tensor A.
B	Base pointer to a tensor B.
A_B_offsets	Vector of pairs of tensors A and B offsets for each batch. The number of batches must coincide with the `batch_size` value passed at object construction stage.
C	Pointer to a tensor C (accumulation buffer).
D	Pointer to a tensor D (output buffer).
scratchpad	Pointer to a scratchpad buffer.
params	Post-op memory arguments. Must be passed If binary post-op or scales were set.

static pack_type get_B_pack_type(memory::data_type a_dt, memory::data_type b_dt)

Returns the packing type expected by a tensor B of a BRGeMM ukernel object.

Parameters:

a_dt	Data type of tensor A.
b_dt	Data type of tensor B.

static void release_hw_context()

Releases the hardware-specific context.

Affects the global state for all BRGeMM ukernel objects. Must be used after all the execution calls to BRGeMM ukernel objects.

static const post_ops& default_post_ops()

Returns a constant reference to a static instance of default constructed primitive post-operations attribute.

static const attr_params& default_attr_params()

Returns a constant reference to a static instance of default constructed ukernel attributes parameters.

struct dnnl::ukernel::brgemm

Contents

struct dnnl::ukernel::brgemm#

Overview#

Inherited Members#

Detailed Documentation#

Construction#

Methods#