Index

Symbols

! model example

7 segment display recognizer

A

ABI add8 (class), [1] AddModule (class) can_delegate (class variable) constructor example_input (class variable) forward (method)

ahead-of-time see AoT AI AoT see also ExecuTorch AoT aot_corev_compiler.py (file) API ATen

B

Backend (class) BackendInterface (class), [1] bash binutils BSD

BSP BSS Buck build-executorch-aot.sh (file) build-executorch-runtime.sh (file)

C

CMake CMakeLists.txt (file), [1] CORE-V CORE_V_MODELS (variable) corev_backend.py (file), [1], [2] corev_executor_runner.cpp (file), [1], [2], [3], [4] COREV_MODELS (variable) corev_pal.cpp (file), [1] corev_partioner.py (file) corev_partitioner.py (file), [1] corev_quantizer.py (file) CoreVBackend (class), [1], [2], [3], [4], [5] destroy (method), [1] doAdd (method), [1] execute (method), [1], [2] init (method), [1], [2] is_available (method), [1] preprocess (method), [1] registration with ExecuTorch runtime CoreVBackend.cpp (file), [1], [2], [3] CoreVBackend.h (file), [1], [2], [3]

CoreVPartitioner (class), [1] CoreVPayload (class), [1] constructor Data (member variable), [1] operatorId (member variable), [1] CoreVQuantizer (class), [1] CoreVSupportedOperators (class) CPU crt0.o CRT0_PATH (environment variable) CSR CV32E40Pv2 binary format (hexdump) board support package BSP Clang/LLVM GCC on FPGA tool chain Verilator model Verilator model; build script Verilator model; VCD trace

D

delegation ExecuTorch, [1] dequantization DLL

DLX DMA do_executorch (method) document licensing

E

ean16-1.0-aot (git tag) ean16-1.0-benchmarking (git tag) ean16-1.0-optimization (git tag) ean16-1.0-runtime (git tag) early clobber variables Edge Embecosm registered trademark error.h (file) ET_CHECK_MSG (macro) ET_LOG (macro) et_pal_abort (function) et_pal_allocate (function) et_pal_current_ticks (function) et_pal_emit_log_message (function) et_pal_free (function) et_pal_init (function) et_pal_ticks_to_ns_multiplier (function) EValue, [1], [2] evalue.h (file) example ! model Example neural network ExecuTorch, [1] 8-bit addition model, [1], [2] architecture ATen operators benchmarking see: ExecuTorch benchmarking customization delegation, [1], [2], [3] delegation see: ExecuTorch delegation Edge operators external documentation install_executorch.sh (file) install_requirements.sh (file) optimization see: ExecuTorch optimization partitioner; see: ExecuTorch; delegation prerequisites quantization, [1], [2] repository structure runtime runtime see: ExecuTorch runtime setting up submodules third party packages ExecuTorch AoT argument processing. arguments; see: ExecuTorch AoT arguments build system building, [1], [2] convenience script for building CORE-V backend CORE-V partitioner CORE-V quantizer debugging delegation delegation implementation dump_delegation_info (function), [1] example output files changed quantization running stages Executorch AoT backend driver ExecuTorch AoT arguments --debug --delegate, [1] --model_input --model_name --output --quantize --so_library -d -m -o -q -s

ExecuTorch benchmarking dynamic cost fixed costs float32 add model gprof impact of delegation nm command runtime driver setting up inputs techniques uint8 add model varying model size, [1] varying number of executions, [1] verifying results zeroing BSS, [1] ExecuTorch delegation stages ExecuTorch optimization, [1] customized runtime double buffer tiling list of techniques memory hierarchy multi-core support multi-core with tiling operator fusion quantization quantization ahead-of-time quantization at runtime ExecuTorch runtime --whole-archive backend driver build script building, [1], [2] CMake configuration, [1] CORE-V backend CORE-V backend declaration, [1] CORE-V backend definition, [1] driver initial method identification driver input data setup driver loader creation driver main function driver memory allocator, [1] driver memory buffers driver memory manager driver method load driver method metadata driver model execution driver model loading driver printf support linking without side-effects logging model conversion model inclusion performance, [1] physical abstraction layer physical abstraction layer (PAL) running, [1] stages executorch::backends::corev (namespace) executorch::runtime (namespace) external documentation ExecuTorch PyTorch

F

FLAGS_num_executions (variable), [1], [2] FlatBuffer, [1]

FPU FST

G

GCC GEMM

generate_corev_compile_spec (function), [1], [2] GNU GPT

H

hardware description language

hexdump HierarchicalAllocator (class)

I

interface.h (file)

IPO ISA

J

JSON

L

libgloss licensing document software

Linux LLM LLVM LTO

M

make Markdown Matplotlib MCU MemoryManager (class)

method_allocator_pool (variable) MIPS ML MMU MODEL_PATH (environment variable) models.py (file), [1]

N

neural network, [1] evaluation inference neural networks 7 segment display recognizer introduction key features linear layer ReLU running the 7-segment display recognizer running the example

Newlib Ninja nn.Module (class) constructor forward (method)

O

objcopy

OpenRISC

P

PAL partitioner see: delegation PIC pip platform.h (file), [1] prepare_input_tensors (method), [1]

PULP Python virtual environment PyTorch, [1], [2] data dumping with Matplotlib external documentation introduction textual data dumping

Q

Quantization

quantization ExecuTorch, [1]

R

register_backend (function) ReLU, [1] repositories Result (class)

RISC RISC-V riscv32-corev-elf.cmake (file) RUNNER_MODEL_PATH (environment variable)

S

save_pte_program (function) shape shared object library SIMD, [1], [2] sloccount

software licensing STL struct (package) SupportedOperatorsList (variable) SystemVerilog

T

target audience temp_allocator_pool (variable) tensor tiling

torch (package) torch.nn.Module (class) torch.uint8 (type) training courses

V

VCD Verilator, [1]

Verilog Vulkan

W

wheel

X

XNNPACK

Z

zip