device_backend.hpp File Reference

Abstract interface isolating vendor-specific memory/compute operations from Tensor/ComputationGraph. More...

#include <cstddef>
#include <cstdint>

Include dependency graph for device_backend.hpp:

Go to the source code of this file.

Classes
struct	pulsatrix::RecurrentCellArgs
	Operand pointers for DeviceBackend::recurrent_cell (passed to kernels by value). More...
struct	pulsatrix::SsmPassArgs
	Operand pointers and dims for DeviceBackend::ssm_pass (passed to kernels by value). More...
struct	pulsatrix::ConvGeometry
	Window geometry for DeviceBackend::im2col / col2im_add: kernel size, stride and zero padding per axis. Defaults are stride 1 and no padding. Passed to kernels by value. More...
struct	pulsatrix::RlRowArgs
	Operand pointers and dims for DeviceBackend::rl_rows (passed to kernels by value). More...
class	pulsatrix::DeviceBackend
	Vendor-agnostic compute/memory backend. CPUBackend, CUDABackend (Phase 1.5), and HIPBackend (Phase 1.6) all implement this contract; Tensor and ComputationGraph depend only on this interface, never on a concrete backend's types. More...

Namespaces
namespace	pulsatrix

Enumerations
enum class	pulsatrix::DeviceType { pulsatrix::Cpu , pulsatrix::Cuda , pulsatrix::Hip }
	Which physical device a Tensor's buffer resides on. More...
enum class	pulsatrix::CopyDirection { pulsatrix::HostToDevice , pulsatrix::DeviceToHost , pulsatrix::DeviceToDevice , pulsatrix::HostToHost }
	Direction of a DeviceBackend::copy() call. More...
enum class	pulsatrix::ElementwiseOp {   pulsatrix::Relu , pulsatrix::Neg , pulsatrix::Tanh , pulsatrix::Sigmoid ,   pulsatrix::Silu , pulsatrix::Exp }
	Unary elementwise operations supported by DeviceBackend::elementwise(). More...
enum class	pulsatrix::LrpGate { pulsatrix::None , pulsatrix::Positive , pulsatrix::Negative }
	Elementwise boolean gate for DeviceBackend::lrp_stabilized_divide(). More...
enum class	pulsatrix::LogicOp {   pulsatrix::ConjunctionForward , pulsatrix::ConjunctionBackward , pulsatrix::ConjunctionLrp , pulsatrix::DisjunctionForward ,   pulsatrix::DisjunctionBackward , pulsatrix::DisjunctionLrp }
	Elementwise passes of the fuzzy-logic modules, for DeviceBackend::logic_pointwise. More...
enum class	pulsatrix::RecurrentCellOp {   pulsatrix::RnnBackward , pulsatrix::LstmForward , pulsatrix::LstmBackward , pulsatrix::LstmLrp ,   pulsatrix::GruBackward , pulsatrix::GruLrp }
	Fused per-element recurrent-cell passes, for DeviceBackend::recurrent_cell. Slots (in[] / out[]), all (rows x hidden) per timestep unless noted: More...
enum class	pulsatrix::SsmPassOp {   pulsatrix::MambaForward , pulsatrix::MambaBackward , pulsatrix::MambaGradBC , pulsatrix::MambaLrp ,   pulsatrix::RwkvTokenShift , pulsatrix::RwkvForward , pulsatrix::RwkvBackward , pulsatrix::RwkvShiftBackward ,   pulsatrix::RwkvLrp , pulsatrix::RwkvShiftLrp , pulsatrix::StabilizedDiv , pulsatrix::RetnetForward ,   pulsatrix::RetnetStateGrad , pulsatrix::RetnetGradQK , pulsatrix::RetnetGradV , pulsatrix::RetnetScores ,   pulsatrix::RetnetReadout , pulsatrix::RetnetLrpInput , pulsatrix::ReverseTimeSum }
	Fused passes of the state-space / linear-recurrence modules (MambaModule, RWKVModule, RetNetModule), for DeviceBackend::ssm_pass. Dims come from SsmPassArgs: n batch, l sequence length, d d_model (or C for ReverseTimeSum), s Mamba's state_size / RetNet's key_dim. Sequences are (n, l, X) row-major; "states" buffers are (n, l + 1, ...) with the zero initial state at index 0. Lanes and slots (in[] -> out[]): More...
enum class	pulsatrix::RlRowOp {   pulsatrix::DqnLoss , pulsatrix::DqnGrad , pulsatrix::PgLoss , pulsatrix::PgGrad ,   pulsatrix::PpoLoss , pulsatrix::PpoGrad , pulsatrix::DqnTarget , pulsatrix::PolyakBlend }
	Fused per-row reinforcement-learning passes, for DeviceBackend::rl_rows. One lane per batch row (per element for PolyakBlend); rows x cols from RlRowArgs. Index slots hold validated whole-number action indices as floats. Slots (in[] -> out[]): More...

Detailed Description

Abstract interface isolating vendor-specific memory/compute operations from Tensor/ComputationGraph.