EmitPass reuses the decoder's persistent workers through RunOnDecodeWorkers
instead of creating and joining threads per batch. A failed trace write stops
the writer thread rather than leaving it ticking every 250 ms.
Release the GameMode request when the daemon narrows the CPU mask; the pin
concentrates Roblox's worker pool and is re-applied every few seconds, so
ending the request is the only thing that lifts it. `gamemode: on` keeps it.
Emit texture uploads on worker threads instead of serially inside
vkQueueSubmit, in two passes so mip 0 lands before the levels scaled from it.
Decode ETC2 through a persistent pool and lower the parallel floor to one
128x128 image, so small batches stop running single-threaded.
Document all three in docs/PERFORMANCE.md.
Compressed bytes are copied to heap scratch, decoded heap to heap across
up to 8 threads in block-row bands, and streamed into the staging buffer.
Decoding straight from uncached Vulkan memory caused 200-460 ms hitches.
The FPS trace now reports the longest frame gap per window.
Hosts without native ETC2 now report ETC2/EAC support. Uploads into emulated images are decoded to RGBA8/R16 staging buffers at submit. MOCKTAIL_DISABLE_ETC2_EMULATION=1 turns it off.
glook9001