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A newer version of this document is available.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.
enum dnnl_alg_kind_t
Overview
Kinds of algorithms. More…
#include <dnnl_types.h>
enum dnnl_alg_kind_t
{
dnnl_alg_kind_undef,
dnnl_convolution_direct = 0x1,
dnnl_convolution_winograd = 0x2,
dnnl_convolution_auto = 0x3,
dnnl_deconvolution_direct = 0xa,
dnnl_deconvolution_winograd = 0xb,
dnnl_eltwise_relu = 0x20,
dnnl_eltwise_tanh,
dnnl_eltwise_elu,
dnnl_eltwise_square,
dnnl_eltwise_abs,
dnnl_eltwise_sqrt,
dnnl_eltwise_linear,
dnnl_eltwise_soft_relu,
dnnl_eltwise_hardsigmoid,
dnnl_eltwise_logistic,
dnnl_eltwise_exp,
dnnl_eltwise_gelu_tanh,
dnnl_eltwise_swish,
dnnl_eltwise_log,
dnnl_eltwise_clip,
dnnl_eltwise_clip_v2,
dnnl_eltwise_pow,
dnnl_eltwise_gelu_erf,
dnnl_eltwise_round,
dnnl_eltwise_mish,
dnnl_eltwise_hardswish,
dnnl_eltwise_relu_use_dst_for_bwd = 0x100,
dnnl_eltwise_tanh_use_dst_for_bwd,
dnnl_eltwise_elu_use_dst_for_bwd,
dnnl_eltwise_sqrt_use_dst_for_bwd,
dnnl_eltwise_logistic_use_dst_for_bwd,
dnnl_eltwise_exp_use_dst_for_bwd,
dnnl_eltwise_clip_v2_use_dst_for_bwd,
dnnl_pooling_max = 0x1ff,
dnnl_pooling_avg_include_padding = 0x2ff,
dnnl_pooling_avg_exclude_padding = 0x3ff,
dnnl_lrn_across_channels = 0xaff,
dnnl_lrn_within_channel = 0xbff,
dnnl_vanilla_rnn = 0x1fff,
dnnl_vanilla_lstm = 0x2fff,
dnnl_vanilla_gru = 0x3fff,
dnnl_lbr_gru = 0x4fff,
dnnl_vanilla_augru = 0x5fff,
dnnl_lbr_augru = 0x6fff,
dnnl_binary_add = 0x1fff0,
dnnl_binary_mul = 0x1fff1,
dnnl_binary_max = 0x1fff2,
dnnl_binary_min = 0x1fff3,
dnnl_binary_div = 0x1fff4,
dnnl_binary_sub = 0x1fff5,
dnnl_binary_ge = 0x1fff6,
dnnl_binary_gt = 0x1fff7,
dnnl_binary_le = 0x1fff8,
dnnl_binary_lt = 0x1fff9,
dnnl_binary_eq = 0x1fffa,
dnnl_binary_ne = 0x1fffb,
dnnl_resampling_nearest = 0x2fff0,
dnnl_resampling_linear = 0x2fff1,
dnnl_reduction_max,
dnnl_reduction_min,
dnnl_reduction_sum,
dnnl_reduction_mul,
dnnl_reduction_mean,
dnnl_reduction_norm_lp_max,
dnnl_reduction_norm_lp_sum,
dnnl_reduction_norm_lp_power_p_max,
dnnl_reduction_norm_lp_power_p_sum,
dnnl_softmax_accurate = 0x30000,
dnnl_softmax_log,
};Detailed Documentation
Kinds of algorithms.
Enum Values
dnnl_convolution_directDirect convolution.
dnnl_convolution_winogradWinograd convolution.
dnnl_convolution_autoConvolution algorithm(either direct or Winograd) is chosen just in time.
dnnl_deconvolution_directDirect deconvolution.
dnnl_deconvolution_winogradWinograd deconvolution.
dnnl_eltwise_reluEltwise: ReLU.
dnnl_eltwise_tanhEltwise: hyperbolic tangent non-linearity (tanh)
dnnl_eltwise_eluEltwise: exponential linear unit (elu)
dnnl_eltwise_squareEltwise: square.
dnnl_eltwise_absEltwise: abs.
dnnl_eltwise_sqrtEltwise: square root.
dnnl_eltwise_linearEltwise: linear.
dnnl_eltwise_soft_reluEltwise: soft_relu.
dnnl_eltwise_hardsigmoidEltwise: hardsigmoid.
dnnl_eltwise_logisticEltwise: logistic.
dnnl_eltwise_expEltwise: exponent.
dnnl_eltwise_gelu_tanhEltwise: gelu.
dnnl_eltwise_swishEltwise: swish.
dnnl_eltwise_logEltwise: natural logarithm.
dnnl_eltwise_clipEltwise: clip.
dnnl_eltwise_clip_v2Eltwise: clip version 2.
dnnl_eltwise_powEltwise: pow.
dnnl_eltwise_gelu_erfEltwise: erf-based gelu.
dnnl_eltwise_roundEltwise: round.
dnnl_eltwise_mishEltwise: mish.
dnnl_eltwise_hardswishEltwise: hardswish.
dnnl_eltwise_relu_use_dst_for_bwdEltwise: ReLU (dst for backward)
dnnl_eltwise_tanh_use_dst_for_bwdEltwise: hyperbolic tangent non-linearity (tanh) (dst for backward)
dnnl_eltwise_elu_use_dst_for_bwdEltwise: exponential linear unit (elu) (dst for backward)
dnnl_eltwise_sqrt_use_dst_for_bwdEltwise: square root (dst for backward)
dnnl_eltwise_logistic_use_dst_for_bwdEltwise: logistic (dst for backward)
dnnl_eltwise_exp_use_dst_for_bwdEltwise: exp (dst for backward)
dnnl_eltwise_clip_v2_use_dst_for_bwdEltwise: clip version 2 (dst for backward)
dnnl_pooling_maxMax pooling.
dnnl_pooling_avg_include_paddingAverage pooling include padding.
dnnl_pooling_avg_exclude_paddingAverage pooling exclude padding.
dnnl_lrn_across_channelsLocal response normalization (LRN) across multiple channels.
dnnl_lrn_within_channelLRN within a single channel.
dnnl_vanilla_rnnRNN cell.
dnnl_vanilla_lstmLSTM cell.
dnnl_vanilla_gruGRU cell.
dnnl_lbr_gruGRU cell with linear before reset.
Modification of original GRU cell. Differs from dnnl_vanilla_gru in how the new memory gate is calculated:
Primitive expects 4 biases on input: 
dnnl_vanilla_augruAUGRU cell.
dnnl_lbr_augruAUGRU cell with linear before reset.
dnnl_binary_addBinary add.
dnnl_binary_mulBinary mul.
dnnl_binary_maxBinary max.
dnnl_binary_minBinary min.
dnnl_binary_divBinary div.
dnnl_binary_subBinary sub.
dnnl_binary_geBinary greater or equal.
dnnl_binary_gtBinary greater than.
dnnl_binary_leBinary less or equal.
dnnl_binary_ltBinary less than.
dnnl_binary_eqBinary equal.
dnnl_binary_neBinary not equal.
dnnl_resampling_nearestNearest Neighbor Resampling Method.
dnnl_resampling_linearLinear Resampling Method.
dnnl_reduction_maxReduction using max.
dnnl_reduction_minReduction using min.
dnnl_reduction_sumReduction using sum.
dnnl_reduction_mulReduction using mul.
dnnl_reduction_meanReduction using mean.
dnnl_reduction_norm_lp_maxReduction using lp norm.
dnnl_reduction_norm_lp_sumReduction using lp norm.
dnnl_reduction_norm_lp_power_p_maxReduction using lp norm without final pth-root.
dnnl_reduction_norm_lp_power_p_sumReduction using lp norm without final pth-root.
dnnl_softmax_accurateSoftmax.
dnnl_softmax_logLogsoftmax.