Insight7

Add device information query functions through HAL → ins:: API → language bindings

Add --device all/--timer/--info CLI flags to demos across all languages for competition scoring

Add composite signal operators using existing primitives (no dedicated backend kernels needed)

How to port a cusignal Python/CUDA module to ins::signal C++ using Insight7 primitives and HAL

Profile and optimize binding language (Lua/Julia/Python) demo performance to match or exceed C++ baseline

busted loads _insight.so from ~/.luarocks/lib first, not build/ — must copy .so after rebuild

Julia C API bindings must check for nullptr return — silent failure otherwise

Lua sol2 lambdas and raw lua_pushcfunction must catch C++ exceptions — sol2 does NOT always auto-catch

Common issues in Python/Lua/Julia demos — shape()/numel/item/tostring/init/ins.abs() complex bug patterns

Adding __setitem__ to Array class — view-based assignment with fill_/copy_from_ via HAL dispatch, operator= semantics, non-contiguous output handling

Adding all Python dunder operators to pybind11 Array binding — arithmetic, comparison, bitwise, unary, conversion

Patterns for porting Python radar detection demo to C++, Lua, Julia — CLI args, framework APIs, 1-based indexing, missing bindings

Preallocated 2D array row assignment (pc[p] = array) silently fails for large arrays — use ins.stack instead

ins.fft(arr, n, axis=0) on 2D complex128 GPU arrays produces wrong results — use numpy FFT or axis=1 workaround

Exclude .md/.txt files from triggering build CI workflows — bindings/**/*.md should not trigger language_bindings or demo_tests

contiguous_copy kernel writes to wrong positions when output is a non-contiguous view — fast path and non-contiguous path both ignore output strides

Replace element-by-element cudaMemcpy loop with cudaMemcpy2D for batched strided copy (2390x speedup on typical sizes)

Fix CUDA kernel runtime bugs discovered through test alignment (missing dtypes, wrong dims, incorrect parameter reading).

CUDA kernels that iterate over input array elements must add src->offset when computing source byte address, or the kernel reads from byte 0 instead of the view's start.

full kernel ignores out->offset, writing to parent array start instead of view position — affects fill_() on views

Make Insight7 installable via pip/luarocks/Julia Pkg with backend .so auto-discovery

Replace single-plan FFT cache with multi-plan cache to avoid ~5ms cuFFT plan recreation when alternating between FFT sizes

Template for pure Insight7 demos with MF FFT caching, device-aware init, --plot matplotlib rendering, and no numpy in core path

Pure Insight7 target extraction from CFAR detection map: nonzero + energy sort + clustering + per-range dedup, avoiding numpy dependency.

Backend discovery for pip/luarocks installs — add_backend_search_path + LD_LIBRARY_PATH + ctypes pre-load

Create luarocks rockspec with cmake build, dynamic backend .so copying, cross-platform

Implement NumPy-style partial indexing in C++ core — a[1] on multi-dim returns slice view, all languages benefit

Implement PaddlePaddle-style lazy GPU default device in C++ core, all languages inherit

Common Lua binding API issues — comparison operators, item/get, __tostring, scalar inputs

Common Julia binding API issues — from_data column-major, getindex, Base.abs shadowing, load_backend, complex dtypes, axis reversal

Fix CI test failures from parallel test suites competing for shared temp directories

Python native __getitem__ with int, tuple (mixed int/slice), and slice support for NumPy-style indexing

ins::init() with optional auto-discover CPU+GPU backends, dynamic-only loading

Wrapping an external C++ library (matplotplusplus) with Array conversion, conditional compilation, and separate CI

CMake mechanism for idempotent, reproducible patches to third-party dependencies (inspired by PaddlePaddle)

Systematic approach to debug and fix multi-job CI failures from user-provided log files

Handle Julia column-major ↔ Insight row-major layout: from_data dim reversal, axis conversion, to_data reshape

Fix CI workflow test paths after tests/bindings/ migration to per-module structure

Fix CI workflows not triggering on PRs when new directories are added (backends/**, tests/**, etc.)

Common fixes for CUDA kernel build errors (atomicAdd, void*, missing headers, half-precision).