exex: performance optimizations

[rabarbra]

No description provided.

[Copilot]

Pull request overview

This PR focuses on reducing CPU, allocation, and UI-latency costs across both the CLI dump paths and the TUI by introducing lighter-weight “layout-only” loads, more streaming/early-exit behavior, and better async job cancellation/ignore-stale-result handling.

Changes:

• Add “layout-only” binary loading and use it for CLI output views that don’t need full symbol/reloc/DWARF parsing (plus perfreport measurement refinements).
• Improve TUI responsiveness via cooperative cancellation, caching, and stale-result guards for async scans (search/xref/syscalls/cpufeatures/demangle).
• Optimize dump/scanning hot paths (strings streaming, syscall/CPU-feature scans, symbol lookup iteration) and add/extend targeted tests/benchmarks.

Reviewed changes

Copilot reviewed 47 out of 47 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
tools/perfreport/main.go	Refines per-stage perf measurement to use per-run prepared files and clarifies cold vs warm measurements.
main.go	Adds layout-only open option for CLI -o views that can use it.
internal/ui/xref.go	Adds xref caching, cancellation, and stale-file guards for async xref scans.
internal/ui/view_symbols.go	Defers/guards symbol list construction and avoids lowercasing work when filter is empty.
internal/ui/view_relocs.go	Fixes relocs view to use its own filter model (not libs filter) throughout.
internal/ui/view_disasm.go	Invalidates/aborts derived async disasm jobs when disasm image state changes.
internal/ui/update.go	Tightens async message handling (file identity checks), adjusts prewarm behavior, refreshes modal symbol names on demangle changes.
internal/ui/syscalls.go	Adds cancellation, caching, stale-file guards, and cancel behavior for syscall scans (incl. full scope).
internal/ui/search.go	Adds explicit disasm search cancellation channel teardown.
internal/ui/search_disasm.go	Adds cancellation + stale-file checks to disasm search, plus faster ASCII-folded matching path.
internal/ui/render_bench_test.go	Adds benchmark for Relocs view rendering.
internal/ui/newviews_test.go	Extends relocs/libs filter regression coverage.
internal/ui/new.go	Initializes a dedicated reloc filter and defers initial symbol recompute.
internal/ui/model.go	Adds symbols “ready” state, relocsState struct, and search cancel channel storage.
internal/ui/key_dispatch.go	Adds Esc handling to cancel full syscall scan; uses reloc-specific filter capture.
internal/ui/goto.go	Avoids sorting/allocation by using bounded ranked buckets for symbol matches.
internal/ui/goto_test.go	Adds tests for goto ranking/cap behavior.
internal/ui/disasm_decode.go	Tags async disasm decode results with file identity for stale-result ignores.
internal/ui/cpufeatures.go	Adds cancellation + caching for CPU feature scan modal and stale-file guards.
internal/ui/chrome.go	Ensures symbols are prepared when switching into Symbols mode.
internal/ui/async_test.go	Adds tests ensuring stale async messages are ignored and cancellation behaves correctly.
internal/explorer/navigation.go	Refines default exec address selection using section properties (exec + file bytes).
internal/explorer/disasm_service.go	Switches to RangeFunc-style decoding to reduce work and early-stop cleanly.
internal/dump/syscalls.go	Adds cancelable full syscall collection and refines chunk scanning boundaries/cancellation.
internal/dump/syscalls_test.go	Adds coverage for x86 localized syscall scanning and chunk-boundary behavior.
internal/dump/perf_bench_test.go	Improves bench timing and adds CPU-features benchmark; uses layout-only where appropriate.
internal/dump/dump.go	Adds ViewNeedsLayoutOnly, improves Disasm symbol labeling via range iterator, and rewrites Strings output to stream/avoid copies.
internal/dump/cpufeatures.go	Adds cancelable CPU feature scanning and faster ARM64/x86 decode paths using x/arch decoders.
internal/dump/cpufeatures_test.go	Adds tests verifying raw-decode classification matches existing text-based classifier behavior.
internal/disasm/decoder.go	Avoids full Unicode lowercasing in common mnemonic paths; reduces allocations in immediate formatting when unchanged.
internal/cpufeat/cpufeat.go	Uses ASCII-only lowering fast-path for mnemonic extraction.
internal/binfile/symsort.go	Simplifies symbol sort to true in-place sort to reduce memory pressure.
internal/binfile/symbols_test.go	Adds tests for SymbolAt alias preference and range iterator parity.
internal/binfile/strings.go	Adds streaming string scan API with bounded memory and parallel chunk scan with ordered emit.
internal/binfile/strings_test.go	Adds ScanStrings tests (ordering, no-cache, closed-pipe behavior).
internal/binfile/reloc.go	Makes HasRelocs cheaper when availability is known; clears builder after first build; adds format-level “has relocs” helpers.
internal/binfile/reloc_test.go	Adds tests for reloc availability short-circuiting and Mach-O reloc symbol-name retention.
internal/binfile/pe.go	Adds layout-only early return; defers DWARF/reloc builds with availability checks.
internal/binfile/parse_bench_test.go	Closes files in open benchmark and adds layout-only open benchmark.
internal/binfile/model.go	Adds layoutOnly flag, refactors symByAddr to index-based structure + prefix max ends, adds source-files-only DWARF path, and range iterator.
internal/binfile/macho.go	Adds layout-only Mach-O parser, fat-header-lite parsing, reloc availability + lazy build, and relocatable synthetic address handling.
internal/binfile/macho_test.go	Adds tests for layout-only equivalence, synthetic section addrs for objects, and fat-slice bounds checks.
internal/binfile/loader.go	Introduces WithLayoutOnly option and skips heavy finalization/overview when enabled.
internal/binfile/image.go	Makes HasExecCode cheap via section scan instead of building exec image.
internal/binfile/elf.go	Moves segment/LMA parse earlier and adds DWARF/reloc availability + lazy build support; supports layout-only early return.
docs/ROADMAP.md	Marks roadmap items complete.
docs/hyper.sh	Parameterizes exex binary path for benchmarking.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.