Prospero.Eden Encore

Encore is not an official upstream release or a promise that every newer upstream commit has been included.

• Dynarmic remains the production JIT. Experimental cross-core shared JIT and anticipatory compilation remain disabled in stable profiles due to unqualified concurrency/lifetime behavior.
• Vulkan/RADV and OpenGL are independent graphics routes. The native graphics build uses pinned source adaptations, tests, source fingerprints and archive checks rather than treating source presence as proof of compiled integration.
• Shader cache allocation and file size do not prove hit rate. Cold pipeline creation, long CPU waits and shader-cache reuse must be measured separately.
• Memory package declarations, sparse page tables, committed guest backing and JIT code-cache policies are not interchangeable. The single native PS5 release now selects demand-backed sparse RX/RW JIT aliases, with physical backing reclaimed only after a quiesced full-cache invalidation. Host checks exist; runtime memory savings and stability on firmware 13.60 remain unqualified, and physical memory is not unlimited.
• Diagnostics can distinguish a running compositor from actual guest progression; detailed capture has costs and is not automatically enabled in quiet gameplay.

See Architecture, Performance roadmap and Integration audit.

Encore began on 4 October 2026, first in the historical Prospero.Eden-Encore fork, then in this standalone repository from 30 October 2026. It is an extensive continuation of the original PS5 port and Eden engine work, not just a logo/name change. Original upstream authors retain full attribution.

The following is a detailed technical change inventory: delivered source changes, test infrastructure, hardware observations and experimental investigations are explicitly distinguished. Some foundations already existed in ProsperoEden and were retained or extended, not invented from scratch. For individually tracked source files, revisions, commits and decisions—including no-change audits and reverted experiments—see project history, active-source catalog, engineering audit, F07 functions and the complete commit history. Auditing a function does not imply optimizing it; passing CI does not imply faster games.

1. Native PS5 platform, release process and project structure

• Concentrated real-console testing on PS5 firmware 13.60; deliberately did not certify unsupported firmware.
• Preserved and adapted the native PS5 application, Eden emulator core, Dynarmic, native OpenGL and PS5 Vulkan/Mesa/RADV integrations.
• Standardized the launcher/application identity PPSA99008 and native installation, package layout, shared storage and user-data paths.
• Consolidated around one release profile and one final installable package rather than multiple DEV/experimental variants.
• Added guarded startup, failure cleanup, bounded log/crash paths, recovery and checks of the actually selected native code paths.
• Refined packaging and reproducibility, exact-SHA executable and symbols matching, optional FFPFSC, and the prohibition of nested ZIPs.
• Selected/pinned upstream engine fixes, compiler/runtime adaptations, SDK/Mesa/Vulkan/OpenGL dependencies and reproducible source manifests.
• Audited transitive source integration and proved which upstream/backport overrides are selected in the compiled archive; rejected incomplete or unverified integrations.
• Migrated the large vendor tree to an immutable, pinned Git submodule; preserved matching dependency sources and licenses.
• Compacted Git history while retaining recovery archives and provenance; migrated to an independent public repository with restored branches/tags/releases where applicable.
• Established fast/full host gates, native build/package gates, static security analysis (including CodeQL), checksum/source checks and firmware-specific console-test checklists.
• Kept code review, build results and on-device qualification separate, including a distinct record for source-only and experimental patches.
• Expanded source-level regression harnesses, sanitizers, O0/O2 fixtures, GPU/HLE emulation models, exact compiled-source checks and historical audit registers.
• References: native build, release policy, standalone migration, vendor migration, repository history.

2. Game formats, executable loading, library and storage

• Extended newer NSO handling with ZBIC/zstd, alongside existing LZ4 segment decompression, with validation of compressed executable boundaries.
• Retained and extended discovery/selection of NSP, XCI, NRO, NSO, NCA, NAX and KIP. Being selectable is not a promise of successful execution without complete supporting content.
• Added experimental NSZ/XCZ (NCZ) game/update/DLC reading, block-compressed random access, bounded metadata reads and controlled solid-file extraction/cache behavior.
• Added extracted ExeFS directory discovery when genuine main and main.npdm are present; rejected unsafe symlinks.
• Improved invalid-file rejection, compression and decompression error recovery, size limits, path validation, cancellation and temporary storage accounting.
• Consolidated the internal/external data-root layout for keys, firmware, ROMs, updates, mods, user settings, game caches and save import/export.
• Added safer migration/reset paths that do not silently erase stored preferences, saves or content.
• Reviewed title metadata, game updates, runtime file access, firmware-dependent behavior and assorted Eden HLE/loader backports.
• References: content-format matrix, installation, upstream gaps.

3. CPU, Dynarmic, JIT, memory and fastmem

• Investigated AArch64/AArch32 Dynarmic, guest thread progression, exclusive/atomic operations, JIT compile/invalidation waits, scheduling and PS5 logical processor allocation.
• Introduced the one selected sparse JIT RX/RW allocator with paired virtual aliases and physical memory committed in 2 MiB units on demand.
• Added alias/probe preflight at title launch and fail-closed behavior rather than silently falling back to the old dense native JIT.
• Implemented physical backing reclamation after complete quiescent JIT invalidation, preserving RX/RW permissions and guest code lifetime.
• Optimized sparse table capacity, scan bounds, reservation epochs, positive/negative hit memoization, cross-arena reuse, alternative lookup slots and RX→RW alias lookup.
• Improved page-map ownership, guest memory access bounds, empty-span handling, range/alignment checks, physical-backed-page accounting and invalidation granularity.
• Reduced repeated fastmem Map/Unmap/Protect updates for unchanged pages and permissions, one-page and cross-granule cases; preserved error/failure retries and alias coherency.
• Audited JIT pressure, cache sizing, guest RAM vs PS5 physical allocations, temporary memory, GPU memory and high-resolution/OOM interactions.
• Evaluated—but did not enable in the shipping profile—shared cross-core JIT, speculative/background translation, unsafe DMA reordering and other concurrency-sensitive experiments.
• Reduced some proved thread/fiber stack and short-span allocation costs; these optimizations do not imply extra console RAM or guaranteed FPS.
• References: PS5 architecture, JIT lifecycle, CPU/JIT migration, F07 tracker.

4. Maxwell, GPU queues, synchronization and native memory

• Reviewed/fixed portions of Maxwell DMA, host1x, GPU scheduler, command FIFOs, transfer extents, fence publication, queue-slot lifetimes and safe stop/restart behavior.
• Applied bounded batching of selected DMA work, buffer invalidations and GPU copy operations without bypassing dependencies or replaying guest commands.
• Improved dirty-span tracking, guest-write coherency, texture/buffer registration, fast small-page ownership probes and safe range checks.
• Audited/guarded indirect draw bindings, null descriptors, vertex/index buffers, compute graphics-resource slot limits and shader-local bounds.
• Reduced GPU-queue wakeups, unnecessary synchronization and certain allocations where source-extracted tests establish equivalent execution.
• Added checks separating actual guest progress from a compositor/GPU worker that continues presenting stale frames.
• References: DMA work, GPU source audit, engine regression register.

5. Vulkan/RADV shaders, pipelines, rasterization and textures

• Integrated and source-pinned native Mesa/RADV and Vulkan adaptations, shader/compiler compatibility work and independent OpenGL fallback.
• Optimized verified shader/pipeline lookup hot paths, vertex-shader identities, small multi-key compute-pipeline memo tables and invalidation-aware hit/miss caches.
• Improved release/acquire publication of ready pipelines, safe failure of optional pipeline optimizations and descriptor preflight/reuse while preserving ordered draws.
• Reduced repeated work and some lock contention in shader caches, pipeline creation, descriptor rings, rasterizer and sampler state.
• Addressed TIC/TSC sampler/texture mappings, invalid/empty texture views, image aliases, framebuffer-view selection and cache invalidation.
• Refined image upload/download staging, asynchronous decode and unswizzle, BCn/3D texture batches, failure rollback, GPU-dirty state and texture eviction safeguards.
• Fixed or safeguarded DrawTexture state invalidation, blit/copy/resolve boundaries, render-target changes, raw framebuffer stride/pitch, missing guest framebuffer and presentation freshness.
• Improved Vulkan swapchain acquire/present recreation, surface-loss bounded retries, synchronization of in-flight layers and resource retirement by real GPU timeline ticks.
• Coalesced Vulkan presenter/worker wakeups and drain notifications, reduced unnecessary lock hold during GPU recycler allocation and avoided redundant AA rebuilds.
• Optimized frequent one-page guest-buffer/texture registration and ownership checks, repeated vertex classification and framebuffer pixel-size lookup.
• Reviewed 38 specific F07 functions across eight source files covering cold shaders, pipeline waits, rasterizer/buffer contention and HUD/texture freshness.
• Still open: first-use guest IR→SPIR-V translation can block; a Vulkan worker may wait for a required pipeline; the black/delayed HUD and full-match frame pacing need console evidence.
• References: F07 function-by-function log, remaining optimization register, F07 cross-system audit, HUD priorities, Mesa/Vulkan review.

6. HLE, IPC, display, audio, input and networking

• Reviewed guest HLE kernel/service scheduling, IPC waits, event handoffs, slab bounds, thread stop/restart and service-owner lifetime.
• Investigated NVNflinger / BufferQueue / SurfaceFlinger / NvDisp / display layers, distinguishing a true game freeze from an image/present problem.
• Audited audio initialization, PCM overflow, ring buffers, mixer/upmix paths and native output threading; reduced selected redundant audio copies/allocations.
• Improved native DualSense polling and mappings: batched immutable inverse mapping, precomputed stick curves/trigger LUTs, deadzones, button-edge semantics and vibration lock boundaries.
• Removed repeated controller scratch clearing, unnecessary sample copies and quiet-poll diagnostic work; retained guest action mapping isolation.
• Reviewed guest networking and DNS/socket behavior (including NanoDNS), access policies, bootstrap and runtime errors without promising Nintendo online support.
• Improved quiet/crash-only logging so diagnostics are available without continuously doing high-cost per-frame work.
• References: input/glyph pipeline, guest IPC findings, network audit, audio/engine inventory.

7. Launcher, controller-first UX, media and translations

• Expanded the TV-first Home, Library and Recents presentation, game metadata, artworks, banners, screenshots, quick settings and offline fallback.
• Added/extended animated 3D game-case launch artwork with authentic front/back scans where present, and native reconstruction from available art for missing sides.
• Introduced/refined Minimum / Recommended / High / Ultra / Custom per-title video presets, override rules, manual graphics options and recovery-oriented Safe Launch for a single session.
• Added launch-time settings for internal resolution, Vulkan/OpenGL, filtering, sharpening and AA, with conservative memory constraints rather than unqualified 4× promises.
• Expanded launcher catalogs to 29 languages and improved complex/bidirectional text layout, fallback fonts, wrapping and glyph drawing.
• Tuned font fitting, font scratch reuse, audio ring copy behavior, controller navigation, scene transitions and costly UI/media work.
• Added/reviewed PS5-profile-avatar read-only, async retrieval paths; availability on every real console remains unverified.
• Investigated PlayStation in-game glyph assets, per-game mods and physical button mapping; universal PlayStation glyph replacement is not implemented.
• Improved accessibility guidance, high-contrast/reduced-motion support, safe settings reset, and user-data migration confirmation.
• Updated screenshots and preview disclaimers: the Home screenshot uses real renderer output but fictional sample games.
• References: launcher audit, game-case media, accessibility, glyph research.

8. Stability work, profiling and tested games

• Built frame-time/window diagnostics for FPS drops, long presentation stalls, GPU queue pressure, shader cold misses, JIT waits, mutex contention and guest IPC progression.
• Investigated FC27 match play, half-time/pause/end-of-match crashes or freezes, intermittent black/late HUD overlays, cold shader spikes and user control profiles.
• Used The Legend of Zelda: Breath of the Wild as a separate gameplay/memory comparator; this is not universal game compatibility proof.
• Recorded quiet-vs-detailed-logging and old-vs-cleared-shader-cache runs; showed that detailed logging itself can materially affect pacing.
• Investigated anti-aliasing/upscaling memory pressure and OOM at aggressive resolution settings; retained conservative device safeguards.
• Defined P0 blockers: (1) first-use shader translation, (2) ordered Vulkan pipeline wait, (3) buffer/texture cache contention, (4) HUD image/texture freshness.
• Tracked JIT, HLE, presentation, textures, glyphs and input as related—but distinct—owners rather than claiming a single FPS knob solves all.
• Retained regression checks and incident reports for frame pacing, guest-progress stalls and GPU presentation despite no game progress.
• References: FC27 master tracker, real-console analysis, freeze recovery, closure criteria.

9. Planned or experimental work — not part of a verified PS5 release

• Frame generation: native frame pacing/stability first; future exploration of temporal upscaling, frame interpolation and optional frame generation. Do not treat these as available now.
• FSR/AA/120 Hz: deeper spatial/temporal upscaling, image quality, x2/x4 AA and higher refresh are design/research targets, subject to title and memory limits.
• Switch 2 / NSW2: a separate research branch explores loaders, newer CPU/shader formats, homebrew NSO/IR/Dynarmic paths and GPU adaptations. It is not merged into the PS5 release, and no stable Switch 2 game compatibility is claimed.
• Cross-core JIT, speculative compilation, deferred GPU paths: preserved as research/disabled experiments where concurrency or correctness proof is missing.
• In-game PS button glyph replacement, avatar reliability, per-game input persistence and game-specific 60 FPS unlocks: active investigation, not universal shipping promises.
• References: performance roadmap, branch policy, glyph research.

A complete change history is not a single FPS claim: source reviews, reverted prototypes, CI-only fixes and documentation changes are recorded in MEMORY.md and commits. The list above summarizes the entire scope of Encore's work since its inception; the linked technical registers provide the individual-file detail. Real device qualification is deliberately narrower than the implementation inventory.

• Stability, frame pacing, input latency and image freshness come first. Shipping frame interpolation on top of freezes would make the counter look better, not fix the game.
• Encore intends to offer a better-managed frame-generation solution than the earlier implementation, but it is not currently a verified shipping feature. Research includes temporal/spatial upscaling and optional interpolation once the underlying rendering path is reliable.
• This is a short-term priority following core optimizations, not a fixed release date. See performance roadmap.

• Not in the current PS5 release. The NSW2 research branch and its feasibility report cover loading, shader/CPU compatibility and homebrew execution paths.
• A successful source test or homebrew harness is not proof that a commercial Switch 2 title runs. No stable compatibility, 30 FPS Mario Kart World performance or supported Switch 2 game-file matrix is claimed.
• The existing Switch 1 path remains the main development focus on PS5. Research: branch policy.

• Firmware 13.60 is the primary version with device testing, not universal compatibility.
• FC27 has reached match gameplay but still exhibits shader/frametime spikes, HUD errors and pause/half-time/end-match freezes in documented tests.
• Zelda: Breath of the Wild has shown comparatively good sessions, but individual games, settings and builds vary.
• There is no general guarantee of a steady 30 FPS, 60 FPS, 120 Hz or flawless controls. See console findings.

• Both engine and frontend have meaningful code changes: JIT memory and fastmem, Vulkan caching/synchronization, textures, game loading, HLE, controller input and launcher.
• Host regression passes and fewer operations do not prove PS5 FPS gains. The exact console gain is still pending comparable real-device tests.
• The current priority is stable frame times and correct game progression, not an inflated FPS counter.

• Existing graphics settings include resolution, filters/upscaling and AA choices, but a title's native rate, the emulator's frame pacing and the PS5 memory budget impose separate limits.
• Better FSR, temporal techniques, 60/120 Hz handling and higher AA quality are planned/researched. They are not promises of universally working or unlocked frame rates.
• Aggressive internal scaling can exhaust graphics memory. See roadmap.

• Recommended/tested for Encore: 13.60. This identifies our evidence base, not a recommendation to update your console.
• Earlier firmware may work, depending on the PS5 exploit/homebrew setup, filesystem access, graphics driver and native APIs. Minimum firmware: unknown; no older Encore version has been hardware-qualified.
• Original ProsperoEden reports 6.02 and 12.70 tests of its Lapy filesystem helper (upstream notes), but Encore uses a different v1.000.040-derived filesystem-access mechanism (fork notes). Those component tests do not prove Encore compatibility.
• Newer firmware remains unverified. Check the release notes and exact tested SHA.

• Library containers: NSP, XCI.
• Compressed containers: experimental NSZ, XCZ, with format and hardware-qualification limitations.
• Direct selections: NRO, NSO, NCA, NAX, KIP, plus correctly structured extracted ExeFS directories.
• LZ4 and ZBIC/zstd are executable-segment compression methods, not games or installable file extensions. Some selected files cannot run without companion content. Full matrix.

• The launcher supports controller-first navigation, DualSense mapping, deadzones and vibration options.
• The game's own saved action mapping and in-game Switch button artwork are separate. PlayStation glyph replacement is not universal; some in-match custom mapping still needs validation.
• See glyph and input research.

• GitHub Actions verifies source/host tests, security and native packaging as separate gates. An artifact is not necessarily a tested stable release.
• Prefer a published Release with explicit device-test notes and the matching SHA/checksum.
• A console session is required to confirm real gameplay, frame pacing, HUD, controls and safe return to the launcher.

• Encore is an unofficial community continuation, with independent changes based on the credited original ProsperoEden and Eden work.
• No commercial games, firmware, proprietary keys or unauthorized downloads are distributed. Use content and system files you have the legal right to use.
• See credits, licenses and installation.

• No automatic upstream merges. The original ProsperoEden v1.000.040 is Encore’s documented baseline; subsequent native-port and Eden updates are reviewed and integrated selectively, preserving PS5 Vulkan/JIT compatibility and testing. A newer upstream release is not necessarily part of Encore.
• No built-in updater is available in the current shipping release. Feature milestone R21 describes manual checking, an optional startup check, verified Encore-release downloads, user confirmation and safe rollback; it is a specification, not a released feature. Full roadmap.
• For now, use Encore Releases and the upgrade guide. Do not overwrite Encore with an original ProsperoEden ZIP merely because it has a higher version: application ID PPSA99008 and storage paths can overlap, although code and migration behavior differ.
• References: original native ProsperoEden, upstream gap audit, native-version review.

• Many commits are deliberately small source changes, regression fixtures or CI corrections. They are not one new feature per commit.
• Remaining high-priority work concerns first-use shaders, Vulkan pipeline waits, texture/buffer cache contention, HUD freshness, guest freezes and real-console confirmation.
• Work is progressing quickly, but there is no verified release date. Updates appear in commits, Actions, and Releases.