MayaFlux 1.0.0
Digital-First Multimedia Processing Framework
Loading...
Searching...
No Matches
Layouts.hpp File Reference

The push constant layouts of the tensor kernels, their flag bits and limits. More...

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes

struct  MayaFlux::Yantra::TensorKernels::AddPC
 Push constants of nn_add. More...
struct  MayaFlux::Yantra::TensorKernels::AttentionMixPC
 Push constants of nn_attention_mix. More...
struct  MayaFlux::Yantra::TensorKernels::AttentionScoresPC
 Push constants of nn_attention_scores. More...
struct  MayaFlux::Yantra::TensorKernels::Conv1dPC
 Push constants of nn_conv1d. More...
struct  MayaFlux::Yantra::TensorKernels::FrameAudioPC
 Push constants of nn_frame_audio. More...
struct  MayaFlux::Yantra::TensorKernels::GroupedConv1dPC
 Push constants of nn_conv1d_grouped. More...
struct  MayaFlux::Yantra::TensorKernels::LstmPC
 Push constants of nn_lstm_sequence. More...
struct  MayaFlux::Yantra::TensorKernels::MagnitudePC
 Push constants of nn_magnitude. More...
struct  MayaFlux::Yantra::TensorKernels::NormalizePC
 Push constants of nn_normalize. More...
struct  MayaFlux::Yantra::TensorKernels::SoftmaxPC
 Push constants of nn_softmax. More...

Namespaces

namespace  MayaFlux
 Main namespace for the Maya Flux audio engine.
namespace  MayaFlux::Yantra
namespace  MayaFlux::Yantra::TensorKernels

Variables

constexpr const char * MayaFlux::Yantra::TensorKernels::add_shader = "nn_add.comp.spv"
constexpr const char * MayaFlux::Yantra::TensorKernels::attention_mix_shader = "nn_attention_mix.comp.spv"
constexpr const char * MayaFlux::Yantra::TensorKernels::attention_scores_shader = "nn_attention_scores.comp.spv"
constexpr uint32_t MayaFlux::Yantra::TensorKernels::conv1d_activation_shift = 2
 nn_conv1d and nn_conv1d_grouped flag bits 2 and 3: the output activation, 0 none, 1 ReLU, 2 sigmoid, 3 GELU.
constexpr uint32_t MayaFlux::Yantra::TensorKernels::conv1d_flag_bias = 1
 nn_conv1d flag bit: add the bias.
constexpr uint32_t MayaFlux::Yantra::TensorKernels::conv1d_flag_relu_input = 2
 nn_conv1d flag bit: apply ReLU to the input.
constexpr const char * MayaFlux::Yantra::TensorKernels::conv1d_shader = "nn_conv1d.comp.spv"
constexpr uint32_t MayaFlux::Yantra::TensorKernels::element_workgroup = 64
 Invocations per workgroup of the element-wise kernels.
constexpr const char * MayaFlux::Yantra::TensorKernels::frame_audio_shader = "nn_frame_audio.comp.spv"
constexpr const char * MayaFlux::Yantra::TensorKernels::grouped_conv1d_shader = "nn_conv1d_grouped.comp.spv"
constexpr uint32_t MayaFlux::Yantra::TensorKernels::lstm_max_hidden = 256
 Widest hidden state nn_lstm_sequence supports.
constexpr const char * MayaFlux::Yantra::TensorKernels::lstm_sequence_shader = "nn_lstm_sequence.comp.spv"
constexpr uint32_t MayaFlux::Yantra::TensorKernels::lstm_workgroup = 512
 Invocations of the one workgroup nn_lstm_sequence runs in.
constexpr const char * MayaFlux::Yantra::TensorKernels::magnitude_shader = "nn_magnitude.comp.spv"
constexpr uint32_t MayaFlux::Yantra::TensorKernels::max_groups_per_axis = 32768
 Most workgroups the stages put on one axis; a larger dispatch continues along y.
constexpr uint32_t MayaFlux::Yantra::TensorKernels::normalize_activation_shift = 3
 nn_normalize flag bits 3 and 4: the output activation, 0 none, 3 GELU.
constexpr uint32_t MayaFlux::Yantra::TensorKernels::normalize_flag_affine = 2
 nn_normalize flag bit: multiply by weight and add bias.
constexpr uint32_t MayaFlux::Yantra::TensorKernels::normalize_flag_affine_outer = 4
 nn_normalize flag bit: weight and bias are indexed by the outer index instead of the position along extent.
constexpr uint32_t MayaFlux::Yantra::TensorKernels::normalize_flag_apply = 1
 nn_normalize flag bit: write the normalised values instead of measuring the rows.
constexpr const char * MayaFlux::Yantra::TensorKernels::normalize_shader = "nn_normalize.comp.spv"
constexpr const char * MayaFlux::Yantra::TensorKernels::softmax_shader = "nn_softmax.comp.spv"

Detailed Description

The push constant layouts of the tensor kernels, their flag bits and limits.

This is the interface of the kernels: a layout does not change once released. A kernel that needs different constants is a new kernel with its own struct, and fields are added only by taking a reserved word.

Definition in file Layouts.hpp.