Intel® oneAPI Deep Neural Network Developer Guide and Reference
A newer version of this document is available. Customers should click here to go to the newest version.
Visible to Intel only — GUID: GUID-0B0CD9D0-2D24-4F21-BC74-1F3E4C16BFD8
Visible to Intel only — GUID: GUID-0B0CD9D0-2D24-4F21-BC74-1F3E4C16BFD8
enum dnnl::query
Overview
Primitive descriptor query specification. More…
#include <dnnl.hpp> enum query { undef = dnnl_query_undef, engine = dnnl_query_engine, primitive_kind = dnnl_query_primitive_kind, num_of_inputs_s32 = dnnl_query_num_of_inputs_s32, num_of_outputs_s32 = dnnl_query_num_of_outputs_s32, time_estimate_f64 = dnnl_query_time_estimate_f64, memory_consumption_s64 = dnnl_query_memory_consumption_s64, scratchpad_engine = dnnl_query_scratchpad_engine, reorder_src_engine = dnnl_query_reorder_src_engine, reorder_dst_engine = dnnl_query_reorder_dst_engine, impl_info_str = dnnl_query_impl_info_str, prop_kind = dnnl_query_prop_kind, cache_blob_id_size_s64 = dnnl_query_cache_blob_id_size_s64, cache_blob_id = dnnl_query_cache_blob_id, strides = dnnl_query_strides, dilations = dnnl_query_dilations, padding_l = dnnl_query_padding_l, padding_r = dnnl_query_padding_r, epsilon_f32 = dnnl_query_epsilon_f32, flags = dnnl_query_flags, alg_kind = dnnl_query_alg_kind, alpha_f32 = dnnl_query_alpha_f32, beta_f32 = dnnl_query_beta_f32, axis_s32 = dnnl_query_axis_s32, local_size_s64 = dnnl_query_local_size_s64, k_f32 = dnnl_query_k_f32, p_f32 = dnnl_query_p_f32, factors = dnnl_query_factors, cell_kind = dnnl_query_cell_kind, direction = dnnl_query_direction, activation_kind = dnnl_query_activation_kind, kernel = dnnl_query_kernel, group_size_s64 = dnnl_query_group_size_s64, src_md = dnnl_query_src_md, diff_src_md = dnnl_query_diff_src_md, weights_md = dnnl_query_weights_md, diff_weights_md = dnnl_query_diff_weights_md, dst_md = dnnl_query_dst_md, diff_dst_md = dnnl_query_diff_dst_md, workspace_md = dnnl_query_workspace_md, scratchpad_md = dnnl_query_scratchpad_md, exec_arg_md = dnnl_query_exec_arg_md, ndims_s32 = dnnl_query_ndims_s32, dims = dnnl_query_dims, data_type = dnnl_query_data_type, submemory_offset_s64 = dnnl_query_submemory_offset_s64, padded_dims = dnnl_query_padded_dims, padded_offsets = dnnl_query_padded_offsets, format_kind = dnnl_query_format_kind, inner_nblks_s32 = dnnl_query_inner_nblks_s32, inner_blks = dnnl_query_inner_blks, inner_idxs = dnnl_query_inner_idxs, sparse_encoding = dnnl_query_sparse_encoding, nnz_s64 = dnnl_query_nnz_s64, num_handles_s32 = dnnl_query_num_handles_s32, };
Detailed Documentation
Primitive descriptor query specification.
In general, queries are not used with the C++ API because most queries are implemented as class members.
See dnnl_query_t for more information.
Enum Values
undef
no query
engine
execution engine
primitive_kind
primitive kind
num_of_inputs_s32
number of inputs expected
num_of_outputs_s32
number of outputs expected
time_estimate_f64
runtime estimation (seconds), unimplemented
memory_consumption_s64
memory required for scratchpad (bytes)
See also:
Primitive Attributes: Scratchpad
scratchpad_engine
scratchpad engine
engine to be used for creating scratchpad memory
reorder_src_engine
reorder source engine
reorder_dst_engine
reorder destination engine
impl_info_str
implementation name
prop_kind
propagation kind
cache_blob_id_size_s64
size of cache blob ID in bytes
cache_blob_id
cache blob ID (pointer to array)
strides
strides
dilations
dilations
padding_l
left padding
padding_r
right padding
epsilon_f32
epsilon
flags
flags
alg_kind
algorithm kind
alpha_f32
alpha
beta_f32
beta
axis_s32
axis
local_size_s64
LRN parameter local size.
k_f32
LRN parameter K.
p_f32
Reduction parameter P.
factors
Resampling parameter factors.
cell_kind
RNN parameter cell kind.
direction
RNN parameter direction.
activation_kind
RNN parameter activation kind.
kernel
Pooling parameter kernel.
group_size_s64
Shuffle parameter group size.
src_md
source memory desc
diff_src_md
source gradient (diff) memory desc
weights_md
weights memory descriptor desc
diff_weights_md
weights gradient (diff) memory desc
dst_md
destination memory desc
diff_dst_md
destination gradient (diff) memory desc
workspace_md
workspace memory desc
scratchpad_md
scratchpad memory desc
exec_arg_md
memory desc of an execute argument
ndims_s32
number of dimensions
dims
vector of dimensions
data_type
data type
submemory_offset_s64
submemory offset
padded_dims
vector of padded dimensions
padded_offsets
vector of padded offsets
format_kind
format kind
inner_nblks_s32
number of innermost blocks
inner_blks
vector of sizes of the innermost blocks
inner_idxs
vector of logical indices of the blocks
sparse_encoding
Sparse encoding.
nnz_s64
Number of non-zero entries.
num_handles_s32
Number of buffers required for a memory descriptor.