|
MayaFlux 1.0.0
Digital-First Multimedia Processing Framework
|
The push constant layouts of the tensor kernels, their flag bits and limits. More...
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" |
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.