{"id":137528,"date":"2026-10-04T11:45:40","date_gmt":"2026-10-04T09:45:40","guid":{"rendered":"https:\/\/www.emu-france.com\/?p=137528"},"modified":"2026-10-04T12:54:40","modified_gmt":"2026-10-04T10:54:40","slug":"consoles-de-salon-prosper-v0-2-3","status":"publish","type":"post","link":"https:\/\/www.emu-france.com\/?p=137528","title":{"rendered":"[Consoles de salon] Prosper v0.2.3"},"content":{"rendered":"

Prosper (ps5ys) est une couche de compatibilit\u00e9 open-source permettant de faire tourner des logiciels et des jeux PS5 sous Linux et Windows.<\/p>\n

 <\/p>\n

Ce n’est pas un \u00e9mulateur: la PS5 et les PC partageant la m\u00eame architecture processeur (x86-64), le code des jeux tourne nativement sur le processeur du PC. Prosper agit comme Wine ou Proton en traduisant les appels syst\u00e8me, les biblioth\u00e8ques internes et l’API graphique (AGC vers Vulkan) de l’OS de la PS5 pour les rendre lisibles par votre ordinateur.
\n <\/p>\n

\"\"<\/a>
\n <\/p>\n

Les changements sont les suivants:
\n <\/p>\n

– tests: pin the CFG-dispatcher re-arm in the entry-M0 token containment (#3203) by @mattias800 in #3230
\n– shader_conformance: replace scan.py’s ported MIMG length rules with the real decoder by @mattias800 in #3232
\n– agc: sceAgcAcbDispatchIndirect carries an ADDRESS, not an offset \u2014 the async-compute ring has no base register by @mattias800 in #3233
\n– worktree hygiene: refs\/stash is shared by every worktree — park work per-worktree instead (#3174) by @mattias800 in #3234
\n– recompiler: IMAGE_LOAD_MIP reads its NSA address encoding, and tail-packed mip chains materialize by @mattias800 in #3237
\n– wt_stash: fix the hook installer on Windows\/MSYS and make it prove git honours the hook by @mattias800 in #3238
\n– recompiler: a CFG region whose entry block defines VCC before reading it has a dead entry value (#3231) by @mattias800 in #3236
\n– sync: scePthreadCondTimedwait spends its relative timeout on the condvar’s OWN clock, and the Windows tick attribution in #3056 is wrong by @mattias800 in #3235
\n– render backend: the persistent-resource caches are one domain with one owner (#2953) by @mattias800 in #3241
\n– kernel: select()\/pselect() stop sleeping on the Windows tick, and their guest timeout stops passing through a 32-bit long by @mattias800 in #3244
\n– windows: the import trampoline places floats from the handler’s own signature (#2955) by @mattias800 in #3242
\n– zerowatch: a one-window verdict was published as a claim about the range — profile it, and follow the content (#3142) by @mattias800 in #3243
\n– render backend: a readback needs an availability operation, not just a fence (#2944) by @mattias800 in #3250
\n– ci: the macOS\/Rosetta sem_timedwait red was a 12 ms ceiling, and a green run must carry its margin by @mattias800 in #3247
\n– zerowatch: the tail of #3243’s review that the merge missed — including a silent 239-byte overread by @mattias800 in #3252
\n– test_game_compute hosts write-watch promotion, and the promotion decision is now counted where it is taken (#3156, #3155) by @mattias800 in #3251
\n– live compute: a dispatch result the host maps needs an availability op, not just a fence (#3249) by @mattias800 in #3254
\n– render backend: order the mip-assembly clear against its copies, and stop two diagnostics answering wrongly (#3248) by @mattias800 in #3256
\n– docs: record the falsified Stray RTT-authority hypothesis in Ruled out (#3140) by @mattias800 in #3258
\n– tools\/ci: a merge gate that reads check buckets and verifies the head (#3259) by @mattias800 in #3260
\n– env: the write-watch knobs refuse a typo instead of quietly picking another setting (#3253) by @mattias800 in #3268
\n– APR read builders: refuse what cannot be derived, and stop reporting an undelivered segment as delivered by @mattias800 in #3262
\n– windows: a variadic HLE handler gets the guest’s own frame, not a signature (#3246) by @mattias800 in #3266
\n– #2945 re-measured: 3,875 replays give one hash per arm \u2014 and a campaign tool that answers UNDECIDED instead of \u00ab\u00a0fixed\u00a0\u00bb by @mattias800 in #3270
\n– APR measure-size: retire the compatibility read per container once a submit re-delivers it by @mattias800 in #3263
\n– docs: update GTA5_STATUS.md with gameplay framerate optimization findings by @mattias800 in #3274
\n– perf: optimize GTA V gameplay framerate to 21+ FPS by @mattias800 in #3275
\n– fix(gta5): address #3276 follow-ups (env parsing & readback buffer sizing) by @mattias800 in #3277
\n– DO NOT MERGE \u2014 HTILE invalidation: breaks GTA V characters\/lighting (#3264) by @mattias800 in #3278
\n– fix(render): preserve in-flight frame depth under byte-preserving HTILE writebacks by @mattias800 in #3281
\n– perf(render): evict older targets in copy_persistent_color_target when over capacity by @mattias800 in #3280
\n– perf(compute, render): expand direct compute-to-graphics image import aliasing for Float16 (#3282) by @mattias800 in #3286
\n– perf(render): memoize mid-submit re-decoded textures to prevent repeated CPU de-tiling (#3283) by @mattias800 in #3287
\n– perf(texture): vectorize tile_surface (sw64kb_copy) to accelerate compute writeback (#3284) by @mattias800 in #3288
\n– perf(compute): skip redundant seed and writeback for untouched storage images by @mattias800 in #3289
\n– perf(compute): fast-decline unmaterializable compute image binding shapes (#3290) by @mattias800 in #3292
\n– fix(scripts): fast-route and pure performance graphics mode for GTA V benchmark (#3301) by @mattias800 in #3302
\n– hle: close the CALL side of the guest-ABI boundary (#3272, #3273) by @mattias800 in #3293
\n– env: refuse a malformed value on the 23 PROSPER_* knobs where 0 is a setting, not \u00ab\u00a0off\u00a0\u00bb (#3267) by @mattias800 in #3294
\n– perf(compute): cache sub-1MB and multi-level sampled textures in persistent compute image cache (#3291) by @mattias800 in #3305
\n– hle: answer sceHttpUriBuild and sceHttpTerm honestly, and validate the HTTP context id (#2930) by @mattias800 in #3295
\n– tests: test_trap_linux asserted The Messenger’s import count, not its own subject (#2997) by @mattias800 in #3296
\n– tests: stop deliberate-death children from writing core dumps \u2014 ulimit -c 0 does not (#3269) by @mattias800 in #3306
\n– gpu: resolve an image op’s descriptor by class, not by first match at the key (#3126) by @mattias800 in #3313
\n– perf(renderer): the frontend texture materializer’s time is the allocator, not the decode by @mattias800 in #3309
\n– recompiler: the CFG dispatcher must not stamp the entry-M0 token at the entry block by @mattias800 in #3312
\n– gpu: say whether the SGPR route was CONSULTED, and give the mapping arms teeth (#3126) by @mattias800 in #3315
\n– docs: instrument trap 257 \u2014 a running-total census read before the run reached the state by @mattias800 in #3323
\n– perf(compute): partition why the compute->graphics image borrow declines (#3307) by @mattias800 in #3320
\n– diag: aim RenderDoc headlessly by frame ordinal or elapsed time (#3321) by @mattias800 in #3322
\n– diag: let a RenderDoc capture span several guest presents (#3321) by @mattias800 in #3326
\n– perf(compute): pool the per-dispatch full-surface intermediates, and name what layout_ms\/watch_ms really are by @mattias800 in #3317
\n– docs(stray): record five falsified explanations for the black title screen by @mattias800 in #3330
\n– diag: census guest VA->physical across all colour slots; prosper renders INTO Stray’s scanout buffers (#2932) by @mattias800 in #3332
\n– screenshot: the picture is not the provenance \u2014 fix the stale comment that misled two readers, and record trap 258 by @mattias800 in #3334
\n– diag: let the compute census speak for small populations; the skipped-dispatch hypothesis is falsified on Stray’s title screen (#2932) by @mattias800 in #3335
\n– chore: point every operational reference at main; master is deleted and archived (#3333) by @mattias800 in #3336
\n– perf(gpu,compute): multi-layer writeback, near-full coverage bypass, and bulk maps caching by @mattias800 in #3328
\n– perf(renderer): route sampled Float32 narrowing through the vectorised packer (#1811) by @mattias800 in #3339
\n– fault reports: label a guest rip by its own module, not as an unbounded eboot offset by @mattias800 in #3341
\n– COMPATIBILITY: #2908 is not shared with Khazan \u2014 Sifu does not import libScePsml by @mattias800 in #3351
\n– docs: instrument trap 260 \u2014 a filter that silently changes the population before the check by @mattias800 in #3352
\n– libScePsml: report NOT IMPLEMENTED instead of SCE_OK \u2014 Khazan stops aborting 8.6 s into every boot by @mattias800 in #3347
\n– charter: a green CI GPU-execution job cannot clear a RADV failure \u2014 CI has no GPU by @mattias800 in #3356
\n– traps: a pgrep -f self-match can disable a wait loop’s own abort guard (trap 135) by @mattias800 in #3357
\n– orchestration: refresh the checkpoint and record the 2026-09-05 objective by @mattias800 in #3359
\n– docs(syberia): the per-dispatch re-upload loop, and a stale Ruled-out row corrected by @mattias800 in #3361
\n– traps: grep -c’s zero-match exit corrupts || guards too, and only when the answer is zero (row 40) by @mattias800 in #3362
\n– sceAvPlayerJumpToTime: publish the seek-completion notification the guest waits for by @mattias800 in #3348
\n– Little Nightmares III: name the flooding draws, fix two instruments that were reporting nothing by @mattias800 in #3363
\n– gpu_replay: let –inspect-only print the whole index list, not just 16 by @mattias800 in #3367
\n– render: PROSPER_BUFVERIFY re-reads uploaded buffers from device memory by @mattias800 in #3370
\n– Record the four-title Unity shear as a cross-implementation split, and add PROSPER_INDEX_ECHO by @mattias800 in #3375
\n– Settle #3374’s fork: RADV drops indexed draws, and module UB cannot be the mechanism by @mattias800 in #3377
\n– Withdraw #3374: the Unity shear does not reproduce from a clean build by @mattias800 in #3380
\n– Add restricted autonomous scheduler profiling access by @mattias800 in #3383
\n– gpu: hash finalized shader compile keys once by @mattias800 in #3385
\n– gpu: retain export watch registrations across full writebacks by @mattias800 in #3386
\n– gpu: avoid redundant snapshots for exact write-only image aliases by @mattias800 in #3389
\n– gpu: make storage-image seed proofs alias-group aware by @mattias800 in #3392
\n– memory: reconcile shared write-watch alias protections by @mattias800 in #3394
\n– gpu: stream per-draw shader property scans by @mattias800 in #3395
\n– gpu: remove per-pass texture-cache statistics traversal by @mattias800 in #3396
\n– Add verified debugging and profiling workflows by @mattias800 in #3382
\n– pixel_history: answer \u00ab\u00a0why is this pixel this colour\u00a0\u00bb in one call by @mattias800 in #3402
\n– render: keep texture staging buffers mapped between frames (Sonic Frontiers +6.3%) by @mattias800 in #3406
\n– compute: parallelise the sampled texel-conversion arms (Sonic Frontiers +18.7%) by @mattias800 in #3408
\n– docs: record the ten falsified Sonic Frontiers performance hypotheses by @mattias800 in #3409
\n– docs: instrument trap 270 (nid_census –registered), and the BLOG entry #3408 owed by @mattias800 in #3412
\n– docs: trap 271 \u2014 the trap-41 deletion check false-positives on a moved base by @mattias800 in #3413
\n– Restore Sonic audio, retain render targets on GPU, and adopt Vulkan 1.4 by @mattias800 in #3418
\n– Reuse graphics pipelines and keep RGBA8 integer copies on the GPU by @mattias800 in #3420
\n– Remove unused color-readback waits and detect stale session instructions by @mattias800 in #3421
\n– Persist graphics pipeline caches across app launches by @mattias800 in #3424
\n– Keep guest clocks advancing during graphics stalls by @mattias800 in #3427
\n– Quantize layered FP16 texture channels directly to UNORM8 by @mattias800 in #3428
\n– Honor HTILE broadcast clears during Blue Prince FMV composition by @mattias800 in #3430
\n– Detile sampled FP16 cubes within graphics submissions by @mattias800 in #3431
\n– Measure SDL consumption demand separately from input queue gaps by @mattias800 in #3433
\n– Reuse mapped GPU detile input buffers after submission completion by @mattias800 in #3434
\n– Convert packed sampled render targets on the GPU by @mattias800 in #3437
\n– Retain mapped detile inputs under staging budget pressure by @mattias800 in #3438
\n– Retile exact compute image results on the GPU by @mattias800 in #3439
\n– Preserve primary compute buffers under result-cache pressure by @mattias800 in #3440
\n– Derive the session-start worktree root relatively instead of from git path output by @mattias800 in #3444
\n– Record Grand Theft Auto V at rung 3 \u2014 the world has rendered since #2996 by @mattias800 in #3445
\n– perf(render,compute): descriptor set layout cache, storage pack parallelization, and parallel buffer upload by @mattias800 in #3442
\n– Fix selected-mip array cache snapshots after storage writeback by @mattias800 in #3457
\n– perf(renderer): avoid per-upload CPU topology probes by @mattias800 in #3459
\n– fix(capture): move owned bundle writing off the guest flip path by @mattias800 in #3460
\n– Print every compute timer the capture carries, with the remainder named by @mattias800 in #3451
\n– Timestamp the GpuFlip event so flip pacing can read a real run by @mattias800 in #3453
\n– fix(gpu): preserve guest authority across retained storage writes by @mattias800 in #3463
\n– perf(gpu): retile standard 3D storage images on the device by @mattias800 in #3465
\n– perf(memory): reduce write-watch publication overhead by @mattias800 in #3469
\n– tools(shader_inspect): add –wave-reasons, an offline fragment wave-width census (#3464) by @mattias800 in #3466
\n– fix(gpu): isolate storage seed proofs by executable and launch by @mattias800 in #3472
\n– tools(shader_inspect): skip_survey.py \u2014 31 refused shaders across 5 of 9 titles, all wave-any (#3464 W7) by @mattias800 in #3473
\n– Preserve dynamic storage inputs and GPU-seed renderer targets by @mattias800 in #3475
\n– tools(shader_inspect): baseline guard for refused fragment shaders (#3464 W6) by @mattias800 in #3474
\n– perf(gpu): retain completed renderer-seeded storage results by @mattias800 in #3477
\n– perf(gpu): omit proven read-only storage image writeback by @mattias800 in #3478
\n– fix(gpu): preserve authority across overlapping output writebacks by @mattias800 in #3481
\n– perf(capture): stream bundle files with checked atomic replacement by @mattias800 in #3483
\n– fix(gpu): validate DCC metadata before borrowing retained images by @mattias800 in #3484
\n– feat(audio): trace output retirement and final consumption demand by @mattias800 in #3485
\n– perf(rtt): remove per-pixel division from exact nearest-neighbor scaling by @mattias800 in #3487
\n– perf(gpu): retile native 16-bit storage arrays in the compute submission by @mattias800 in #3488
\n– fix(capture): break the exit deadlock between the owned writer and the CRT onexit table by @mattias800 in #3492
\n– fix(capture): retry a bundle install past Windows’ transient replace window by @mattias800 in #3495
\n– fix(audio): –volume is an amplifier after the guest’s mixer, not a rival writer of it by @mattias800 in #3490
\n– fix(gpu): admit fragment programs at native wave32 when the module is provably width-independent (#3464 W5) by @mattias800 in #3480
\n– libkernel: register the memory-pool and configured-flexible-memory calls \u2014 fixes NINJA GAIDEN 4’s null deref (#3502) by @mattias800 in #3505
\n– loader: link a title’s own middleware from the dump root, and skip only duplicate BUILDS \u2014 Darksiders II reaches its title screen (#3497) by @mattias800 in #3508
\n– diag: a guest-state dump that cannot run must say so, not print success (#3509) by @mattias800 in #3510
\n– gpu: derive a RectList’s fourth corner from the geometry, not from a fixed index (#3507) by @mattias800 in #3511
\n– docs: reframe the README around the experiment; move build steps to BUILDING.md by @mattias800 in #3513
\n– fs+libkernel: map guest \/download0 and register sceKernelCancelEventFlag \u2014 BlazBlue Entropy Effect X reaches its title screen by @mattias800 in #3512
\n– tools\/refactor: survey which refactoring tool can act on each large file by @mattias800 in #3514
\n– tools\/refactor: promote_internal wrote headers that cannot compile, three ways by @mattias800 in #3519
\n– tools\/refactor: a split needs a visible declaration, which linkage does not provide by @mattias800 in #3524
\n– tools\/refactor: a name in a comment is not a public symbol by @mattias800 in #3525
\n– tools\/refactor: merge into an existing header instead of replacing it by @mattias800 in #3526
\n– gpu: split gpu_capture.cpp and rdna2_to_spirv.cpp into eleven files, none over 1,624 lines by @mattias800 in #3522
\n– docs(gta5): the world runs at 7-8 FPS, not 21 \u2014 the 21 was a phase with no world in it by @mattias800 in #3528
\n– hle: sceKernelMprotect must round rather than refuse, plus the two missing APR\/memory-pool NIDs (#3498, #3506) by @mattias800 in #3530
\n– worktree_reclaim: removal status must not land on the –json stdout (#3397) by @mattias800 in #3534
\n– agc: a mistyped render\/present cadence must not become a divide by zero by @mattias800 in #3537
\n– perf: sample the Windows working set and commit charge into the F8 capture by @mattias800 in #3535
\n– agc: register sceAgcDcbDrawIndirect and build its real packet (#2929) by @mattias800 in #3542
\n– ci: prove a PR body was published rather than trusting gh exit codes (#2918) by @mattias800 in #3536
\n– env: PROSPER_WAITCAP must not remove its own cap, and a typo must not pick the lever’s OFF arm by @mattias800 in #3538
\n– fix(recompiler): OpAtomicLoad is not a memory write in the wave-width analysis by @mattias800 in #3547
\n– Enable image robustness on the compute backend’s own Vulkan device (#3531) by @mattias800 in #3546
\n– Honour the V# DST_SEL channel routing on MUBUF format fetches by @mattias800 in #3552
\n– libSceHttp2: stop reporting a request nobody sent as SCE_OK (#2894) by @mattias800 in #3544
\n– libSceHttp: implement the request path instead of answering SCE_OK (#2930) by @mattias800 in #3551
\n– hle(savedata): persist the param block instead of answering SCE_OK and storing nothing by @mattias800 in #3553
\n– gpu_replay: say which shader dump writes SPIR-V and which writes guest RDNA2 by @mattias800 in #3539
\n– loader: unresolved DATA imports must not bind to a code stub, and AddAmprEvent must keep the guest’s udata (#3529, #3500) by @mattias800 in #3541
\n– flip_pacing_report: state the chance baseline the tick-aligned share is read against by @mattias800 in #3543
\n– Record which Windows session-start failure is which (#3426 verified fixed by #3444) by @mattias800 in #3555
\n– Start the session-start hook with an interpreter that exists on this platform by @mattias800 in #3556
\n– docs: Evergate renders correctly on Windows\/NVIDIA, and two candidates for the shear are dead by @mattias800 in #3558
\n– Save the compute pipeline cache on the frontend’s _Exit path by @mattias800 in #3557
\n– Pace guest flips at the rate the title asked for, not at the host refresh (#3379) by @mattias800 in #3559
\n– pixel_history: never read RenderDoc’s no-information sentinel as a black pixel by @mattias800 in #3566
\n– Report the guest flip rate over the wall clock, not over a filtered mean by @mattias800 in #3562
\n– Tell a backwards emission from a run boundary by its magnitude by @mattias800 in #3563
\n– Anchor a pad-script press’s start in flips and its hold in wall time (#3449) by @mattias800 in #3568
\n– pixel_history: a copy, blit or resolve is not a shader that wrote black by @mattias800 in #3570
\n– Count and report the device memory prosper holds on each heap by @mattias800 in #3565
\n– recompiler: declare SignedZeroInfNanPreserve so guest Inf\/NaN semantics survive the host driver by @mattias800 in #3561
\n– pixel_history: give the RenderDoc citations their provenance, and narrow one claim to what was measured by @mattias800 in #3584
\n– sweep(3573-3580): the eight correctness-sweep issues, one branch by @mattias800 in #3582
\n– tests: restore stderr, fix an inverted portability guard, and stop two comments over-claiming (#3598 #3600 #3602) by @mattias800 in #3604
\n– recompiler: a lane-local scalar must not satisfy the wave-uniformity proof (#3596) by @mattias800 in #3603
\n– perf(render): add opt-in validated buffer retention and cost diagnostics by @mattias800 in #3493
\n– perf(gpu): retile native packed 2D textures on the GPU by @mattias800 in #3613
\n– gpu: transfer bounded texture validation snapshots without a second copy by @mattias800 in #3614
\n– Reduce CPU work for shared mip-tail texture layouts by @mattias800 in #3617
\n– gpu: refuse a reserved DST_SEL encoding on an image descriptor by @mattias800 in #3605
\n– Batch: five correctness-sweep follow-ups (one commit per issue) by @mattias800 in #3618
\n– Preserve ordered writeback across private hosted outputs by @mattias800 in #3621
\n– Reclaim idle compute buffers for budgeted exact comparisons by @mattias800 in #3622
\n– perf(gpu): reduce key preparation and retain GPU result comparisons by @mattias800 in #3625
\n– hle(sync): resolve guest mutex ownership on the calling host thread TCB, and a trace that shows it by @mattias800 in #3624
\n– gpu: move FP16 2D texture and storage preparation to the GPU by @mattias800 in #3626
\n– hle(fiber): implement sceFiberRename instead of leaving it to the return-0 default by @mattias800 in #3628
\n– gpu: share overlapping graphics buffer uploads within a pass by @mattias800 in #3629
\n– gpu: replace a false comment about null descriptors, and justify the image DST_SEL screen by @mattias800 in #3627
\n– gpu: skip scalar-spill invalidation for registers with no spills by @mattias800 in #3631
\n– gpu: reuse validated complete fills and preserve clear semantics by @mattias800 in #3632
\n– hle: keep a fiber’s guest thread pointer with the host thread that resumes it by @mattias800 in #3637
\n– docs: record the Uncharted fiber\/TLS investigation and instrument trap 281 by @mattias800 in #3641
\n– gpu: fuse exact integer image readback and retile by @mattias800 in #3640
\n– tools: detect import stubs overwritten in place, and record Uncharted’s patched dump by @mattias800 in #3652
\n– gpu: reuse shader control analysis during resource folding by @mattias800 in #3663
\n– Pin flip_pacing_report’s three unpinned #3562 behaviour claims by @mattias800 in #3664
\n– savedata: reject GetMountInfo for an invalid or inactive mount by @mattias800 in #3665
\n– savedata: name the bound that actually makes the mount-point read safe by @mattias800 in #3667
\n– gpu: reuse resource-fold instruction facts with exact-input replay by @mattias800 in #3670
\n– savedata: make both umount entry points read their mount-point argument by @mattias800 in #3672
\n– Uncharted: fold the guest’s draws, stop zeroing its shaders, and recompile SRT-addressed resources by @mattias800 in #3673
\n– Avoid full Linux memory-map scans and completion-write starvation by @mattias800 in #3399
\n– Reuse render-target footprints across queued guest writes by @mattias800 in #3677
\n– Reduce buffer lookup allocation traffic and repeated shader analysis by @mattias800 in #3679
\n– Avoid repeated exact validation of stable small textures by @mattias800 in #3680
\n– perf(memory): answer write-watch queries from indexes instead of full scans by @mattias800 in #3682
\n– diag(compute): name why a retile declines, and split the memcpy out of layout_ms by @mattias800 in #3684
\n– perf(gpu): derive the result-compare ceiling from the memory topology by @mattias800 in #3685
\n– docs: rule out moving compute work off the GPU, which loses what moving it on gained by @mattias800 in #3688
\n– test(gpu): cover the result-compare shader, which decides whether a writeback is skipped by @mattias800 in #3689
\n– refactor(gpu): delete the dead compute-dispatch realizer, and count the writebacks it skips by @mattias800 in #3690
\n– docs(compute): measure the writeback skip fraction (82% on GTA V), and rule out the obvious control for the CPU-compare skip by @mattias800 in #3692
\n– docs(compute): correct two off-by-one citations in the census Ruled-out entry by @mattias800 in #3693
\n– docs(compute): the host result snapshot is never taken \u2014 measured, and #3687 is inverted by @mattias800 in #3694
\n– docs(charter): batch related work into one PR \u2014 a PR per commit costs everyone by @mattias800 in #3700
\n– perf(compute): upload only the bytes that differ \u2014 GTA V compute \u221220% by @mattias800 in #3697
\n– Fuse storage-image retile and exact result comparison by @mattias800 in #3698
\n– test(memory): negative control proving the write-watch index reduces work (#3681) by @mattias800 in #3702
\n– fix(perf): keep the F8 rate window and the detail population as distinct observations by @mattias800 in #3701
\n– perf(render): avoid superseded texture preparation by @mattias800 in #3703
\n– perf(render): avoid redundant encoded texture snapshots by @mattias800 in #3704
\n– perf(render): index the AGC shader registry and cache hot diagnostic gates by @mattias800 in #3705
\n– recompiler: a sampler’s FORCE_UNNORMALIZED must not gate IMAGE_LOAD_MIP (#2818) by @mattias800 in #3706
\n– gpu(compute): materialize a mip chain for a one-layer 2D_ARRAY binding (#2818) by @mattias800 in #3709
\n– docs(frontiers): five falsified hypotheses from the 2026-09-17 rendering pass by @mattias800 in #3707
\n– docs(frontiers): falsify the scene-width-compute explanation of the black world by @mattias800 in #3710
\n– docs(frontiers): the Cyber Space world IS rendered \u2014 it never reaches the screen by @mattias800 in #3711
\n– Sample per-draw diagnostic switches once per submit (\u221248% getenv), and falsify the fold-snapshot cost by @mattias800 in #3712
\n– Emit the DS st64 pair and report scanout registrations from one place by @mattias800 in #3715
\n– Resolve the trip-bound selectors once per shader-cache operation (fixes a key\/module coherence hole, \u221221.4% getenv) by @mattias800 in #3716
\n– tools: memcpy_probe \u2014 attribute block copies by call site and cycles, because perf cannot by @mattias800 in #3718
\n– Prove scalar branches workgroup-uniform so their regions may hold barriers by @mattias800 in #3717
\n– Reuse the per-pass command pool and its buffer instead of recreating both by @mattias800 in #3720
\n– Reject extent-incompatible cached RTTs when sampling reused addresses (#3436) by @mattias800 in #3722
\n– gpu: seed renderer color targets from compute storage image writeback by @mattias800 in #3723
\n– gpu: lower v_cvt SDWA sub-dwords, CUBE resinfo, and s_bitreplicate_b64 (#2790) by @mattias800 in #3724
\n– gpu: resolve entry-prefix indirect SMEM descriptors and SCC branches for workgroup barriers by @mattias800 in #3726
\n– Acquire the draw pass’s command pool first, so a failed acquisition can be reported by @mattias800 in #3725
\n– Publish non-mirrored compute pixels instead of re-validating a stale device image by @mattias800 in #3731
\n– A measured refactoring plan, and the two tool gaps found measuring it by @mattias800 in #3732
\n– hle: lift the shared service-call helpers into an internal header (step 1 of #3735) by @mattias800 in #3736
\n– refactor: split hle_service.cpp into per-library files by @mattias800 in #3737
\n– refactor: split the video decode and IME libraries out of hle_service.cpp by @mattias800 in #3738
\n– refactor: split libSceSaveData out of hle_service.cpp by @mattias800 in #3741
\n– refactor: split the NP \/ online family out of hle_service.cpp by @mattias800 in #3743
\n– compute: admit cube and query-only layered image bindings (and the frame does not change) by @mattias800 in #3740
\n– refactor(gpu): lift the shader-cache types out of gpu_executor.cpp by @mattias800 in #3744
\n– gpu: keep eligible compute publication and R11 seeds on the GPU by @mattias800 in #3746
\n– docs: Sonic’s blow-out is banded and enters at one submit; bloom falsified (trap 285) by @mattias800 in #3748
\n– docs: Sonic’s world renders \u2014 correct the charter, COMPATIBILITY and the replay README by @mattias800 in #3749
\n– gpu: publish same-image RGBA8 compute results after source seeding by @mattias800 in #3750
\n– refactor(gpu): give the SPIR-V emitter a folder that names what it emits by @mattias800 in #3747
\n– docs: three gpu_replay instruments that fail silently (trap 286) by @mattias800 in #3752
\n– perf(compute): publish packed R11 partial RTT results into their renderer image by @mattias800 in #3753
\n– perf(render): convert retained depth cubes from mapped readbacks by @mattias800 in #3754
\n– perf(present): account completed renderer producer delivery by @mattias800 in #3755
\n– perf(render): retain unchanged direct guest buffer ranges by @mattias800 in #3756
\n– renderer: bound retained range admission by @mattias800 in #3757
\n– perf(render): cache tracked buffer membership by @mattias800 in #3758
\n– perf: batch persistent resolve copies by @mattias800 in #3759
\n– perf(renderer): compact live buffer resource carriers by @mattias800 in #3761
\n– gpu_replay: refuse capture-only flags on a bundle instead of ignoring them by @mattias800 in #3762
\n– tools(evidence): colour-aware image likeness \u2014 every default metric here is achromatic by @mattias800 in #3760
\n– docs(sonic): withdraw the \u00ab\u00a0history buffer\u00a0\u00bb inference — it came from a replay flag, not the guest by @mattias800 in #3764
\n– render: batch shared buffer range uploads by @mattias800 in #3763
\n– gpu_replay: report small live targets in their native type by @mattias800 in #3766
\n– gpu: mirror completed RGBA16 compute targets by @mattias800 in #3768
\n– recompiler: name which representation a wave-mask operand is missing by @mattias800 in #3767
\n– gpu: share exact fragment analysis within each draw by @mattias800 in #3769
\n– gpu: borrow indexed draw words into backend by @mattias800 in #3770
\n– gpu: preserve lighting texture array layers and sampled mip chains by @mattias800 in #3771
\n– Admit Float32 array gathers with bounded retained depth snapshots by @mattias800 in #3790
\n– perf\/refactor tools: stop reporting what the tool could not see as a fact about the run by @mattias800 in #3772
\n– gpu_replay: RGBA8 conversion for every capture format, and –native-texel for chosen texels by @mattias800 in #3775
\n– hle\/net: define the libSceNet error encoding once (#3545) by @mattias800 in #3776
\n– live_compute: attribute an early-break dispatch’s time to the phase it broke in by @mattias800 in #3778
\n– prosper-app: –volume attenuates the launcher music too (#3499) by @mattias800 in #3777
\n– host\/tls: make Linux guest_tls_activate_thread idempotent (#3635) by @mattias800 in #3774
\n– host\/memory: one fault-safe guest<->host copy policy; service, audio and gpu helpers forward to it (#3734) by @mattias800 in #3789
\n– Register sceAvPlayerStartEx so a movie started with it actually plays (#3781) by @mattias800 in #3788
\n– gpu: give CachedShader a move pair so shader-cache insertions move (#3745) by @mattias800 in #3773
\n– test_game_compute: per-mode expectations under PROSPER_NO_SKIP_SEED, and register the mode by @mattias800 in #3785
\n– Enumerate save directories on Windows so sceSaveDataDirNameSearch sees them (#2760) by @mattias800 in #3780
\n– savedata: retain the requested allocation and report used\/free capacity (#3654) by @mattias800 in #3783
\n– gpu: run Wave64 fragment programs as partially populated waves on narrow hosts (GTA V world on NVIDIA, #3464) by @mattias800 in #3794
\n– perf tools: review follow-ups on #3772, and #3461’s banner wording by @mattias800 in #3793
\n– hle: carry the flipArg in bits 16+ of the VideoOut flip event (#3668) by @mattias800 in #3798
\n– tools: lanekill test compares worktree identity, not path spelling (#3730) by @mattias800 in #3799
\n– diagnostics: end-of-run reports that survive the _Exit\/_exit paths runs actually take by @mattias800 in #3782
\n– recompiler: compile barrier-phased complex compute kernels phase by phase (Metaphor 0x2281c74100) by @mattias800 in #3800
\n– docs(perf): hand off post-3770 roadmap evidence by @mattias800 in #3805
\n– Add opt-in Sonic depth-clear diagnostic controls by @mattias800 in #3803
\n– Record Outer Wilds live title-start checkpoint by @mattias800 in #3806
\n– Add opt-in publication and shader-fold cost probes by @mattias800 in #3808
\n– Identify depth surfaces and viewport ranges in prefix replay by @mattias800 in #3809
\n– Record all programmed MRT targets in provenance diagnostics by @mattias800 in #3810
\n– session_start test: pay the fresh runner’s first hook launch outside the timed calls (#3583) by @mattias800 in #3792
\n– Windows: decode sceVideodec2 access units (VP9, H.264) through Media Foundation by @mattias800 in #3786
\n– Metaphor: ReFantazio reaches its title screen \u2014 libSceSigninDialog lifecycle and the glyph advance by @mattias800 in #3802
\n– Trace Kena image bindings and bound retained-target diagnostics by @mattias800 in #3813
\n– Recompiler: take s_bfm_b64’s fragment lane from guest_lane_id(), not linear_localid (#3816) by @mattias800 in #3817
\n– Windows: let a fixed MAP_DIRECT replace pages prosper lazily committed (#3812) by @mattias800 in #3814
\n– docs: Kena: Bridge of Spirits status, route and compatibility row (rung 2) by @mattias800 in #3818
\n– Run the diff-span exactness test on every platform: lift the byte helpers out of live_compute.cpp (#3699) by @mattias800 in #3796
\n– perf: attribute allocation sites and bound per-draw resource growth by @mattias800 in #3821
\n– gpu: run the partial-wave fragment tier on every native-vote title, not only GTA V (#3797) by @mattias800 in #3822
\n– gpu: transfer eager draws in ordered submissions by @mattias800 in #3823
\n– gpu: avoid double-offsetting folded zero-stride vertex fetches by @mattias800 in #3825
\n– prosper-app: capture only frames that reached the window by @mattias800 in #3826
\n– Bind F9 BMP to the captured closing flip and producer submit by @mattias800 in #3827
\n– Keep F9 bundle open through the presented producer submit by @mattias800 in #3830
\n– Expose backend texture preparation and binding in F8 captures by @mattias800 in #3833
\n– diag: trace Kena’s ordered title scene and compositor by @mattias800 in #3834
\n– capture: retain instance counts for failed graphics draws by @mattias800 in #3836
\n– Preserve 3D color target views through Kena draw captures by @mattias800 in #3837
\n– Prepare merged-NGG mesh workgroup rendering by @mattias800 in #3838
\n– Preserve failed-draw volume activity and vertex retry inputs by @mattias800 in #3839
\n– Probe Kena merged-NGG workgroup with captured inputs by @mattias800 in #3840
\n– Probe Kena merged-NGG launch inputs without admitting live draw by @mattias800 in #3841
\n– Retain bounded 3D color targets for direct volume sampling by @mattias800 in #3842
\n– Add guarded native Wave64 diagnostic for merged NGG captures by @mattias800 in #3843
\n– Resolve physical aliases across split guest mappings by @mattias800 in #3844
\n– Trace native NGG probe phases and fix bounded DPP reduction by @mattias800 in #3845
\n– Test 32-slice retained volume handoff through bounded views by @mattias800 in #3846
\n– Bound renderer guest reads by committed mapping prefix by @mattias800 in #3847
\n– Trace Kena NGG launch state and indexed export layers by @mattias800 in #3848
\n– Guard layered mesh PARAM interpolation into retained volumes by @mattias800 in #3850
\n– Expose failed draw raster state and input snapshots in gpu_replay by @mattias800 in #3851
\n– Capture live ABI for failed fragment retries by @mattias800 in #3853
\n– Seed fragment ancillary layer from rasterized primitive by @mattias800 in #3854
\n– Let pixel history inspect a capture-local image by @mattias800 in #3855
\n– Measure and relieve backend texture cache pressure by @mattias800 in #3856
\n– Use exact one-layer array storage for 2D compute outputs by @mattias800 in #3858
\n– Speed up retained depth-cube conversion on AVX2 CPUs by @mattias800 in #3859
\n– gpu\/recompiler: preserve packed RTZ conversion classes by @mattias800 in #3860
\n– perf: record F8 renderer submit wall spans by @mattias800 in #3861
\n– hle\/agc: preserve older interpolant identity mapping by @mattias800 in #3862
\n– Reuse released CPU RTT snapshots across resolution changes by @mattias800 in #3863
\n– Batch index arena uploads before draw command recording by @mattias800 in #3864
\n– gpu: bind use-proven raw two-word scalar data loads by @mattias800 in #3865
\n– Instrument why the render path drops and copies work, and fix the RTT snapshot pool by @mattias800 in #3869
\n– Outer Wilds frame-gap attribution: the AVX2 index-expansion kernel loses, and its switch goes with it by @mattias800 in #3866
\n– Offscreen prosper-app runs adopt GPU present (enable VK_EXT_headless_surface) by @mattias800 in #3874
\n– –fps: read the GPU-present content sample from cached host memory by @mattias800 in #3875
\n– Evict idle resident textures while an ordered batch is pending (Outer Wilds ~2x) by @mattias800 in #3876
\n– renderer: index volume RTT candidates instead of walking the RTT cache per texture reference (+18% GTA V fps) by @mattias800 in #3877
\n– docs: rule out per-pass object recycling and the index-scan early exit as frame-time levers by @mattias800 in #3878
\n– VideoOut: pace flips outside the handle lock by @mattias800 in #3879
\n– renderer: gather retained depth arrays on the GPU instead of reading them back (Sonic Frontiers +21-37% flips\/s) by @mattias800 in #3882
\n– renderer: upload guest BCn textures natively (4-8x smaller texture footprint; ~1.5x Outer Wilds at an 8 GB-card budget) by @mattias800 in #3883
\n– gpu: name every decline that sends a compute result through the CPU by @mattias800 in #3872
\n– gpu: memoize per-program compute dispatch facts (Sonic Frontiers +6% flips\/s) by @mattias800 in #3886
\n– renderer: size the texture cache from VK_EXT_memory_budget headroom (8 GB card: 7.3 -> 10.5 fps on Outer Wilds RGBA8) by @mattias800 in #3887
\n– Fix GTA V black menus and missing HUD: a volume claims guest authority only with a valid renderer image (#3842 regression) by @mattias800 in #3889
\n– Performance alarms: an always-on ledger and six [perf-alarm] rules that announce prosper’s own performance problems (#3891 phases 1-2) by @mattias800 in #3894
\n– renderer: GPU-only images and buffers prefer a DEVICE_LOCAL memory type (#3888) by @mattias800 in #3896
\n– Performance alarms phases 3-4: per-site drop reasons, skipped-dispatches \/ host-copy-pressure \/ shader-compile rules, –fps marker (#3891) by @mattias800 in #3898
\n– Keep GTA V’s UI under capture and diagnostic modes: a denied volume producer releases its claim (#3890) by @mattias800 in #3899
\n– renderer: retain guest decodes of an identity-only dim-5 RTT shell (Sonic Frontiers menus +8-23% flips\/s) by @mattias800 in #3900
\n– renderer: retry GPU-only allocations on another memory type when VRAM runs out; gpu-memory-off-device alarm (#3897) by @mattias800 in #3904
\n– renderer: read never-rendered depth-array layers from guest memory instead of dropping the draw (#3893) by @mattias800 in #3903
\n– Keep live GPU targets under PROSPER_GPU_CAPTURE; read back on demand for the captured submit (#3895) by @mattias800 in #3909
\n– Keep fully retained depth cubes on the GPU by @mattias800 in #3910
\n– renderer: release the transient pool before spilling; move spilled retained resources back to VRAM (#3902, #3905) by @mattias800 in #3911
\n– renderer: widest slice span per pass, padded depth identities, MSAA colour footprints for guest depth layers (#3906) by @mattias800 in #3912
\n– renderer: never bind null memory to a depth target or texture when every memory type fails (#3901) by @mattias800 in #3913
\n– Keep exact compute RTT results in renderer images by @mattias800 in #3914
\n– perf alarms: six rules from the #3891 queue by @mattias800 in #3916
\n– capture: a deferred capture seeds the targets its submit writes from their pre-submit pixels (#3908) by @mattias800 in #3917
\n– Count effective colour readback requests by @mattias800 in #3918
\n– live renderer: carve the submit callback into named pieces (#3892, part 1) by @mattias800 in #3920
\n– Keep exact compute RTT results when a read-only alias samples the source by @mattias800 in #3921
\n– live renderer: the pass loop, backend draws and backend timing leave live_renderer.cpp (#3892, part 2) by @mattias800 in #3923
\n– present: GPU-present compute-written display buffers, and name every GPU-present decline (#3915) by @mattias800 in #3924
\n– Capture deferred depth before execution and avoid false scanout oracles (#3908) by @mattias800 in #3925
\n– live renderer: resolve, prelude work and diagnostics leave the submit callback (#3892, part 3) by @mattias800 in #3926
\n– compute: retain RTT authority with folded sampled aliases (#3871) by @mattias800 in #3927
\n– perf-alarm: host-copy-per-flip, present-slot-trouble, and a guest-scanout transfer site (#3891) by @mattias800 in #3928
\n– compute: retain DCC export watch across forced image seed by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3931
\n– compute: avoid forced-DCC source copy with writeback watch by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3932
\n– live renderer: final-span present selection, the single-framebuffer path and the MRT census leave the callback (#3892, part 4) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3934
\n– compute: detile exact tiled storage seeds into upload staging by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3935
\n– Avoid redundant GPU-write alias scans for single-VA watches by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3936
\n– Bound physical write-watch invalidation to affected pages by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3937
\n– perf: retain first compute-produced render targets on GPU by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3938
\n– video: add libSceVdecsw unit test and fix decoder lifecycle leaks (#3669) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3930
\n– gpu: resolve exact split image descriptors across scalar loads by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3939
\n– tools: promote_internal merge includes (#3922); check_diag_gates follows context members (#3919) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3940
\n– compute: mirror one-layer 2D_ARRAY storage results into the renderer image (#3929) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3941
\n– perf: reduce recurring compute guest writeback through bounded direct-memory alias by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3942
\n– perf-alarm: rtt-destination-refused, color-target-count-ceiling, exact-result census (#3891) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3943
\n– docs: GTA V ruled out — declined LUT volume pass is not #3907’s black world by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3945
\n– recompiler: reject numeric reads of raw wide SMEM placeholders by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3944
\n– refactor: move the pass loop’s small inline diagnostics into pass_diagnostics (#3892) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3947
\n– gpu: observe indirect CPU argument reads in compute authority census by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3946
\n– perf: let a compute full-overwrite destination evict at the colour-target bound (#3873) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3949
\n– tools: check_diag_gates follows a same-name context pass-through (#3892, #3919) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3954
\n– perf-alarm: executor GPU fence waits vs device time, gpu-sync-wait (#3948 stage 0) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3953
\n– gpu: back proven raw scalar data in Uncharted producer shaders by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3956
\n– refactor: draw_resources’ buffer branch becomes materialize_buffer_resource (#3892) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3955
\n– gpu: honor Wave64 mask width in x16 descriptor proof by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3958
\n– gpu: restore descriptor-only register-offset wide loads by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3960
\n– gpu: scope raw-wide SMEM refusal to compute; count shader-recompile draw drops by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3959
\n– perf-alarm: gpu-present-stalled rule; ceiling reports its refusal arm; split destination-allocation-failed by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3962
\n– gpu: speed up 8-byte tiled detiling with paired AVX2 loads by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3952
\n– gpu: reject unbacked nested wide scalar data by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3961
\n– Keep high-dispatch F8 compute captures complete by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3963
\n– perf-alarm: gpu-present-stalled stays quiet while the window cannot be presented (#3891) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3965
\n– gpu: defer the graphics batch wait before a compute dispatch (#3948 stage 2) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3964
\n– perf: rank measured compute storage detiles by elapsed time by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3966
\n– gpu: guard one-hop nested scalar data with dispatch snapshots by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3969
\n– gpu: a pass that writes no colour must not replace its target’s CPU copy (#3907) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3967
\n– test: pin when the live renderer callback’s prelude statics initialise (#3892) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3968
\n– test: distinguish skipped-callback decode-budget reads by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3971
\n– diagnostics: flag incomplete GPU device-time coverage by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3970
\n– docs: record GTA deferred-wait collapse controls by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3975
\n– Back fold-proven register-offset raw wide scalar loads by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3972
\n– tools: retain config provenance through cached reference getters by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3976
\n– diagnostics: detect colourless CPU pass publications by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3977
\n– refactor: separate live-renderer process prelude state by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3974
\n– Back memory-fed scalar offsets for raw wide data loads by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3979
\n– diagnostics: distinguish readback attempt volume from duration by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3980
\n– diagnostics: qualify draw exit snapshots by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3981
\n– refactor: keep renderer prelude in calling-thread state by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3982
\n– diagnostics: retain graphics queue-call origins by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3983
\n– tools: isolate declared test member facts in diagnostic scans by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3985
\n– refactor(live): return image resource status before extraction by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3986
\n– Clarify validation-prefix and target-cache alarm evidence by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3988
\n– Preserve declared reference-output diagnostic associations by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3989
\n– diagnostics: announce unsupported Wave64 shaders on every platform by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3993
\n– refactor(live): extract image resource materialization by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3991
\n– gpu: portable multi-wave Wave64 scalar mask reductions by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3997
\n– gpu: portable Wave64 read-first-lane broadcast and probe barrier safety by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3999
\n– Recheck submit visibility before each completion drain pop by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4004
\n– gpu: count numeric MBCNT source bits independently of fragment helpers by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4005
\n– diagnostics: observe direct WaitRegMem predicates by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4001
\n– gpu: implement numeric WQM nibble widening and exact SCC by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4006
\n– Observe direct texture comparison extents and outcomes by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4009
\n– gpu: preserve vote-derived addresses in fragment width proof by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4008
\n– gpu: lower proved fragment votes without game allowlists by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4010
\n– gpu: prove initialized logical-wave fragment vote traces by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4012
\n– gpu: prove width-independent masked fragment selections by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4014
\n– fix(recompiler): mask inactive vector compares without poison propagation by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4020
\n– Record first launches for Wolverine, 007, Onimusha, Hades II and Silent Hill 2 by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4026
\n– fix(replay): preserve requested fragment wave width by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4028
\n– hle: bound sceKernelSync to guest storage filesystems by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4029
\n– fix(frontend): build F9 capture matcher without Vulkan headers by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4037
\n– fix(kernel): return the available guest CPU mask by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4034
\n– fix(recompiler): lower normal-magnitude predicates with integer bits by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4052
\n– fix(playgo): enumerate validated manifest chunk identifiers by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4035
\n– memory: keep automatic mappings clear of fixed startup arenas by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4046
\n– fix(capture): retain realized guest fragment wave width by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4051
\n– loader: return errors for failed module sizing and seek by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4045
\n– kernel: return EBUSY when a semaphore poll cannot acquire by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4047
\n– docs: record title-wave handoff and Silent Hill setup captures by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4050
\n– loader: reject images without a PT_LOAD segment by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4055
\n– kernel: reject null thread joins before calling the host by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4057
\n– gpu: retain producing fragment FLOAT_MODE and exact zero comparisons by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4061
\n– loader: reject unresolved SELF load-segment backing by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4067
\n– diagnostics: announce unverified fragment F32 ADD\/MUL mode semantics by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4070
\n– hle(graphics): derive OutputStatus resolution from the advertised mode by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4058
\n– Place ordinary automatic maps above fixed startup arenas by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4069
\n– gpu: identify the first failed fragment Wave64 vote in diagnostics by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4080
\n– hle: implement missing C11 try-lock and timed waits by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4072
\n– fix(abi): preserve variadic AL across bare import tail jumps by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4085
\n– build(windows): opt-in native-Clang TLS profile and lifetime regression by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4081
\n– feat(gpu): record faithful fragment compiler cases for offline replay by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4073
\n– gpu: implement portable compute DS_BPERMUTE gather by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4091
\n– Back per-draw raw x4\/x8 scalar loads with owned snapshots by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/3987
\n– loader: reject IL2CPP symbols beyond the loaded image by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4088
\n– run-windows: use ccache when present, fail fast without a Vulkan SDK by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4074
\n– ci: add ccache to the build jobs by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4077
\n– loader: recover runtime stub append failures (#3554 item 2) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4097
\n– gpu: resolve eligible indirect compute dispatches on the device (#3656) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4083
\n– gpu: retain explicit F32 transport profile through compiler and replay by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4092
\n– gpu\/recompiler: execute owned Wave64 fragment packets and quad WQM by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4107
\n– tools\/env: use forward-slash repo paths on Windows by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4112
\n– ci: cancel superseded runs, main-only ccache saves, parallel ctest, timeouts, docs-only gate by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4116
\n– gpu\/capture: retain producing fragment launch context by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4113
\n– hle: preserve SDK inline APR read output slots by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4102
\n– docs: Assassin’s Creed Black Flag Resynced status and compatibility row by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4132
\n– docs: refresh the five-game title investigation checkpoint by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4135
\n– gpu: execute MBCNT from owned logical64 fragment packets by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4114
\n– tools: report actual nid census input scope by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4105
\n– tools: document task skills and check their metadata paths by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4111
\n– macOS\/Rosetta: fix 6 of the core-job failures (#4043) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4124
\n– tests: fix schedstat_probe_logic on Windows and run it serially by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4143
\n– gpu: collect bounded owned raster quads from real DrawItems by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4128
\n– app: diagnose missing Vulkan driver instead of terminating by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4144
\n– docs: retain the raster-quad scope evidence boundary by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4154
\n– fs: use 64-bit offsets for lseek, fseek and ftell on Windows by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4137
\n– sync ring: record the guest call site of every event (PROSPER_SYNC_RING_CALLERS) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4149
\n– ampr: write the APR submit result slots as 32-bit values by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4146
\n– ci: uv\/ruff\/pytest python-quality ratchet on changed files by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4127
\n– hle\/kernel: timed ERRORCHECK self-relock waits out the deadline, then ETIMEDOUT (libthr) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4150
\n– loader: reject runtime DATA imports without a Program base by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4152
\n– libkernel: implement GetOperationMode, the coredump handler pair and PRT apertures by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4153
\n– docs: add diagnostics roadmap by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4151
\n– docs: explain the Black Flag init-function fault and the stack-check failure by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4160
\n– diagnostics: distinguish early persistent-color readback refusals by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4108
\n– present: report frame-time tail (p90\/p95\/p99) and 1% low by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4156
\n– tools: apply safe ruff autofixes across prosper Python (stacked on #4127) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4130
\n– gpu: back one-hop nested x4\/x8 loads in direct graphics stages by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4104
\n– AGC: submit multi-DCB arrays as one ordered batch by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4158
\n– tools: add module docstrings to Python files lacking one (stacked on #4130) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4134
\n– gpu: report consumed SMEM words in targeted fold traces by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4118
\n– tools: fix ruff bugbear findings B904\/B023\/B007 (stacked on #4134) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4136
\n– test(hle): cover AMPR userdata registration and completion by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4168
\n– kernel: give Windows guests a readable, validated ScePthread handle by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4129
\n– hle: register scePthreadAttrGetstack by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4166
\n– fix(vkval): resolve native Windows validation probes by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4163
\n– recompiler: lower numeric S_FF1_I32_B32 by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4063
\n– prosper-app: exit non-zero when the guest faulted; report worker-thread and host faults by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4145
\n– tools: record run conditions and refuse to compare mismatched runs by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4165
\n– gpu: retain keyless buffer consumer PCs beside matching vertex buffers by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4169
\n– gpu: prove scalar scratch initialization in owned fragment packets by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4164
\n– tests: measured 300 s TIMEOUT for the fragment-packet oracle tests by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4177
\n– tests: run c11_sync_bookkeeping on Windows too by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4178
\n– tests: construct the lock test’s contention instead of relying on the scheduler by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4185
\n– tools: drop the fragment-compiler reader-policy hash gate by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4186
\n– ci: collect seven more unittest-style tests via pythonpath (stacked on #4170) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4183
\n– ci: add gitleaks secret-scan workflow by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4182
\n– docs: record that the Black Flag guest now runs and why every draw is dropped by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4196
\n– gpu: PROSPER_GPU_BREADCRUMBS names where the GPU stopped on a device loss by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4173
\n– tests: resolve the vkval scan test’s temp root (macOS \/var symlink) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4205
\n– docs: refresh title-wave integration and validation checkpoint by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4171
\n– test: use relative SDK pointers in buffer-publication fixture by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4181
\n– gpu: retain producing fragment entry facts in shipping packet preparation by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4190
\n– prosper-app: keep the grab drain and expiry poll at the top of the app loop by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4206
\n– tests: adopt GoogleTest, migrate 213 test files, enforce it for new tests by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4155
\n– fix: reject unusable capture comparisons and preserve graphics identity by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4172
\n– tests: GoogleTest migration follow-ups (Linux build fix, vendored gtest, live guards) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4207
\n– tools\/ci: architecture ratchet gate (#4198) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4199
\n– Agent rules: architecture\/performance ratchets and steady-state perf invariants (#4193) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4204
\n– hle: skip zero-length segments in sceAgcDriverSubmitMultiDcbs by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4194
\n– perf-alarm: print where the frame budget goes at exit (frame-breakdown) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4188
\n– tools\/ci: structural ratchet rules, baseline path containment, platform seam and target tree docs (#4198) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4215
\n– gpu: execute owned logical64 fragment resource chains transactionally by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4212
\n– ci: clang-format + clang-tidy, gated on changed lines only (#4210) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4217
\n– gpu: prepare the producing PS USER_SGPR entry prefix by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4213
\n– ci: fix Windows App ctest syntax error in skip regex by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4226
\n– ci: fix main build breaks (gtest registration, Vulkan-gated diagnostics tests) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4222
\n– hle: key APR completion delivery on a per-binding dialect (stacked on #4174) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4200
\n– gpu\/recompiler: integer packet RCP SQRT and direct RSQ by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4232
\n– tests: expect EVFILT_AMPR (-25) in AMPR completion tests by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4237
\n– ci: register test_indirect_dispatch_device_route as a gtest on Linux by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4233
\n– host: Windows data write-watchpoint on every guest thread (PROSPER_HWWATCH_ABS) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4227
\n– test: update raw_register_wide_data for capture v69 tail by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4238
\n– ci: fix arch_ratchet_gate and ratchet_hook on Windows MinGW (^{commit} mangled by MSYS git) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4239
\n– gpu: separate packet VGPR scratch from per-lane entry values by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4236
\n– docs: organize prosper\/docs into topic folders by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4229
\n– docs: archive resolved docs, rewrite ROADMAP as current direction by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4231
\n– gpu: cache original fragment kernels with owned per-wave data by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4241
\n– gpu: admit a closed early-out block past the first s_endpgm in the split-T# proof (stacked on #4219) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4234
\n– tests: migrate 62 more tests to GoogleTest (Linux\/macOS\/Vulkan-gated, CI-verified) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4216
\n– Fix APR userdata lifetime across equeue handle reuse by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4175
\n– tests: exhaustive RDNA2 decode sweep; fix gather4 and D16 register-footprint under-counts by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4228
\n– gpu: preserve descriptor routing across read-only FLAT loads by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4252
\n– docs: allow concurrent compatibility and correctness checks by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4255
\n– fix(gpu): produce distinct perspective pull-model planes by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4245
\n– Fix APR result records and separate completed submit IDs by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4258
\n– fix(loader): validate initial Linux stubs before mapping by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4259
\n– Make fence build journal samples fault safe by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4261
\n– Materialize actual saved Wave64 mask words in owned packets by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4244
\n– loader: run unimported auto-linked PRX init at sceKernelLoadStartModule (#4139) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4221
\n– fix(symbols): expose tied IL2CPP method candidates by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4264
\n– Fix owned nested x2 numeric VCC inputs for compute by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4263
\n– docs: add 007 First Light and Silksong progress entries by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4266
\n– gpu: execute owned per-wave raw x4\/x8 graphics loads by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4270
\n– gpu: demand only reaching initial packet EXEC VCC SCC by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4249
\n– fix(host): recover Windows guest faults with explicit native state by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4267
\n– fix(host): single-gate PROSPER_HWWATCH_MAX warning (unbreak Docs diag-gate job) by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4282
\n– Silksong: restore all menu text (storage-image DST_SEL), implement its 9 missing calls, keep refused shaders by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4290
\n– fix(host): define module initializer callback and descriptor ABI by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4286
\n– gpu: explicit architectural EXP observation and completion-safe owned packet events by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4256
\n– test(savedata): release icon readback handles before replacement by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4294
\n– hle(memory): query Darwin direct readable memory window via mach_vm_region by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4288
\n– docs: support varied tracker headings, unadorned rungs, and unrecorded oracle state by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4289
\n– Fix #4276 stale contract tests; hand compute BGRA render targets in guest order (#4291) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4297
\n– fix(gpu): bound original refused-shader observations by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4298
\n– Lower unsigned compute DPP row maxima with guarded CFG sources by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4271
\n– gpu: migrate test_rdna2_decode to GTest; pin SOP1\/VOP3 predicate tables to LLVM by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4303
\n– gpu(recompiler): fix RDNA2 register-footprint counts for byte\/short stores and FLAT D16 loads by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4312
\n– test: use SDK-relative registers in compute producer fixtures by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4329
\n– gpu: connect original fragment Wave64 to device draw transactions by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4281
\n– test(host): guard guest FS swap stub frame layout by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4327
\n– test(apr): call the APR stub with the guest ABI and install it once per process by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4285
\n– gpu: finalize ordered rendering after actual snapshot splits by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4284
\n– test: add dump-independent NID golden vectors by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4333
\n– test(hle): pthread key teardown conformance through the scePthread HLE by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4307
\n– test(gpu): pin fail-visible refusals for six unlowered opcodes by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4316
\n– hle\/libc: call guest qsort\/bsearch comparators in the guest ABI; strlcpy\/qsort\/bsearch HLE tests by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4295
\n– test(hle): sscanf contract tests through the registered HLE handler by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4299
\n– gpu: retain original scalar-buffer read demands and descriptor origins by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4342
\n– test(hle): printf\/scanf\/strtod contract tests through the registered HLE handlers by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4310
\n– hle\/fs: parse guest fopen modes before the host CRT; fopen mode tests by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4313
\n– test: migrate messenger guard + compare_exec_masking to GTest by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4335
\n– test: cover direct-memory map failure rollback by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4336
\n– test(ci): sanitize SELF and ELF loader regressions by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4340
\n– test(hle): cover BatchMap partial failure prefix by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4343
\n– fix(loader): reject PT_LOAD filesz larger than memsz by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4347
\n– test(hle): cover MapDirectMemory2 failure state by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4346
\n– memory: prove fault safety of committed Windows direct-section ranges by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4350
\n– Add guarded floating DPP row rotation and exact native CFG routing by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4315
\n– docs: record bounded Sonic HUD fetch findings by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4353
\n– gpu: issue ordered read points for checked graphics RAW sources by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4332
\n– gpu: reassemble split quad scopes so llvmpipe can run owned fragment waves (#4277) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4354
\n– docs(gpu): cite both guards behind the split-scope key’s primitive identity by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4361
\n– diagnostics: expose bounded Windows native HLE wait provenance by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4337
\n– test(ci): sanitize synthetic loader and relocation tests by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4360
\n– docs: add the stub-platform-arm debt, open questions and an APR placement recommendation to the target tree by @yuriolive in https:\/\/github.com\/mattias800\/prosper\/pull\/4363
\n– ci: fix main’s remaining Linux and Windows CI failures by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4368
\n– fix(gpu): recognise Silksong’s DCC-decompress helper; its 1.0 export painted the boot screen white by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4359
\n– fix(macos): gpu_replay did not compile on macOS; build it in the macOS App job by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4364
\n– fix(hle): macOS GC stop-the-world burst overflowed the pending-exception table (Silksong scene-load hang) by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4351
\n– test(fixtures): retry scratch creation when a peer removes the shared root by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4371
\n– macOS: AvPlayer backend on AVFoundation + VideoToolbox; skip an FFmpeg that cannot link for the target arch by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4326
\n– ci(ratchet_hook): derive the checkout root from –show-cdup, not –show-toplevel by @mattias800 in https:\/\/github.com\/mattias800\/prosper\/pull\/4373\n<\/div>\n

 <\/p>\n

T\u00e9l\u00e9charger Prosper v0.2.3 (85.4 Mo)<\/a><\/p>\n

Site Officiel<\/a><\/p>\n

En savoir plus…<\/a><\/p>\n

<\/div><\/div>","protected":false},"excerpt":{"rendered":"

Prosper (ps5ys) est une couche de compatibilit\u00e9 open-source permettant de faire tourner des logiciels et des jeux PS5 sous Linux et Windows.   Ce n’est pas un \u00e9mulateur: la PS5 et les PC partageant la m\u00eame architecture processeur (x86-64), le …<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2},"jetpack_post_was_ever_published":false},"categories":[1],"tags":[],"class_list":["post-137528","post","type-post","status-publish","format-standard","hentry","category-par-defaut","uentry","postonpage-1","odd","post-author-Jets"],"jetpack_publicize_connections":[],"jetpack_shortlink":"https:\/\/wp.me\/pKi2R-zMc","jetpack_likes_enabled":false,"jetpack_sharing_enabled":true,"jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/www.emu-france.com\/index.php?rest_route=\/wp\/v2\/posts\/137528","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.emu-france.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.emu-france.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.emu-france.com\/index.php?rest_route=\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.emu-france.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=137528"}],"version-history":[{"count":12,"href":"https:\/\/www.emu-france.com\/index.php?rest_route=\/wp\/v2\/posts\/137528\/revisions"}],"predecessor-version":[{"id":137570,"href":"https:\/\/www.emu-france.com\/index.php?rest_route=\/wp\/v2\/posts\/137528\/revisions\/137570"}],"wp:attachment":[{"href":"https:\/\/www.emu-france.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=137528"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.emu-france.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=137528"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.emu-france.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=137528"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}