tag
23 posts
Measure slow Haiku installs by repository refresh, download, solver, disk transaction, packagefs activation, and scripts before intervening.
A prompt that's slow only inside repositories is usually running expensive Git status work on every redraw — profile it, isolate repository scale, cache safely.
OpenSSH can reuse one authenticated transport for later sessions — configure a private control socket, bounded persistence, and explicit cleanup.
Why the WSL ext4 disk, NTFS DrvFs, and ReFS Dev Drive serve genuinely different build and Windows-tool workflows, not interchangeable options.
How to measure FreeBSD boot delays with boottrace, inspect rcorder dependencies, isolate blocking services, and repair ordering safely.
A production-safe FreeBSD DTrace workflow for loading providers, narrowing probes, aggregating syscalls and stacks, and controlling tracing overhead.
Load average climbs while top shows idle CPU — this usually means processes stuck waiting on I/O, not CPU; here's how to find which one.
Understand io_uring queues, submissions, completions, workers, polling, resource registration, cancellation and security before benchmarking.
Why you can't find most of macOS's own system libraries as individual files on disk anymore, and how the shared cache changed launch performance.
Distinguishing normal, temporary Spotlight indexing load from a genuinely stuck reindex loop, and resolving the latter instead of tolerating it.
Reading the memory pressure gauge correctly, telling normal swap usage from a real problem, and identifying which process is actually responsible.
A technical comparison of CPU interpreters and dynamic recompilers, including dispatch cost, translation blocks, timing, exceptions, and cache invalidation.
Separate emulated-GPU shader generation from RetroArch post-processing, then test caches, official drivers, backend, and documented precompile modes.
Diagnose incorrect emulator speed by separating throttle and region settings, host overload, display synchronization, and intentional console slowdown.
Opening a new terminal tab shouldn't take a visibly annoying second or two — how to find which specific config line is responsible, rather than guessing.
Understanding what the Memory Compression process actually is before assuming it's a problem, versus identifying a genuine memory leak correctly.
Diagnosing whether high vmmem usage is a genuine leak or reclaimable page cache, and setting a deliberate .wslconfig memory ceiling that actually fits your workload.
Confirming that /mnt/c's cross-filesystem overhead is the actual cause of slow WSL2 file I/O, then moving the project to fix it structurally, not cosmetically.
WSL2's VM claims memory dynamically as Linux needs it and has historically been slow to release it back, plus what .wslconfig actually lets you control.
A practical toolkit for watching CPU, memory, disk, and network usage live — going beyond top to actually find what's causing a resource problem.
Trace emulator audio crackling through performance, buffer underruns, refresh synchronization, drivers, resampling, and wireless output latency.
Task Manager shows disk usage pinned at 100% with no obvious cause. Windows Search's indexer is a frequent culprit — here's how to confirm it and fix it properly.
Boot times creeping up usually trace to a specific cause — too many startup programs, a failing drive, or a delaying driver — not general 'Windows rot.'