gh-blackbearreloaded/ps5-opengl
Watch
1
0
Fork
You've already forked ps5-opengl
 0
Code Issues Pull requests Releases 10 Activity
10 releases 14 tags
  • sdk-0.1.0-perf20260909-g47-hfr-sdl2-focused 50edf9396c

    PS5 OpenGL + SDL2 SDK - native 1440p120 / 4K120 preview Pre-release

    archiver-bot released this 2026-09-09 05:45:54 -05:00 | 655 commits to main since this release

    Experimental OpenGL 3.3 Core + SDL2 SDKs for native 1440p120 and 4K120 output. These are the exact preverified G47 archives, not fresh CI rebuilds. Older releases and assets remain unchanged.

    Choose a profile

    Archive profile 30-second ImGui result Captured console HDMI
    1440p120 (2560×1440) 3,597 frames, 119.880050 FPS 2560×1440 at 119.88 Hz; restored to 59.94 Hz
    2160p120 (3840×2160 / 4K) 3,597 frames, 119.879451 FPS 3840×2160 at 119.88 Hz; restored to 59.94 Hz

    Each profile separately passed an SDL2 native run with 180 frames and two exact pixel probes. Each ImGui run also passed two pixel probes. All four cycles closed cleanly and passed service-health checks. SDL FPS was not measured.

    Each .tar.gz contains compiled GL/EGL and SDL2 libraries, headers, Make/pkg-config/CMake metadata, source examples, complete project/dependency sources, licenses, manifests and exact-binary provenance. After extraction, all 242 file checksums, both library manifests and five GL/SDL consumer compile/link checks passed per archive. Host regressions passed.

    Verification and use

    Download the archive for the intended fixed profile and its matching .sha256 file. Verify before extraction into a new directory:

    sha256sum --check ps5-opengl-sdk-<version>.tar.gz.sha256
tar -xzf ps5-opengl-sdk-<version>.tar.gz
cd ps5-opengl-sdk-<version>
sha256sum --check SHA256SUMS

    Use the complete matching sdk/ and sdl2/ prefixes; do not mix profile libraries. The public payload SDK and native-app boilerplate are separate prerequisites. No ready-to-launch app, toolchain binary, firmware module, console-enablement code or raw console log is included.

    See the current HFR bundle guide and limitations. The frozen archives retain their original documentation/provenance, including their then-local publication status. These release notes and the repository overview record the later publication; no archive was rewritten just to update wording.

    Exact identities

    Profile Archive SHA-256 Bytes Packaging source snapshot
    1440p120 8c0a91b6951eb96e075464e751ca727c764826e857301518d916244f3ca25538 174,370,464 2309e5ca9868fae7b111ad5596bee103e60a4c8e
    2160p120 38ee215ba2429b0f7ff6e96d3375c674ee2380e629111a15c1fa01cc59d6069e 174,370,158 e41dea9d6799c8ad0c4ebe053fcb83174c5ecc55

    Both retain runtime build source 23a594c3a5fda4135599d0dea8d2bf4dcef47a39. The release tag identifies the later publication/documentation source, not a new binary build. The perf20260908 archive names are intentionally preserved. G47 removes actual private build paths through a recorded runtime rebuild and debug stripping of dependency copies; its reviewed LLVM address-significance metadata exception is documented in the bundled provenance. No general optimization-equivalence claim is made.

    Scope, not certification

    • One recorded firmware-6.02 console and Hisense 55U78N on HDMI4. Native 1440p required the owner's matching output selection. Console HDMI logs are not independent per-run TV measurements or proof of every displayed frame.
    • No new sampled/full CTS acceptance for these bytes. The historical 39,544 accounted results (37,404 Pass + 2,140 reviewed NotSupported) belong to a different runtime. This is not Khronos certification or a guarantee of arbitrary game performance.
    • SDL2 supports one fixed-size window and one Core 3.3 context. A separate G42 physical controller/reconnect case passed at 1440p; the four release-qualification cycles above do not establish physical-input acceptance for every SDK/device combination.
    • Broader formats, mips and layers can retain CPU staging. ImGui's 120-class throughput is not a general 3D/game result. Separate high-resolution memory/lifecycle checks are documented with their own identities; first-window soak cadence remains a partial pass. Multi-hour sessions, cross-firmware behavior, suspend/resume, device loss and exhaustive OOM remain unqualified.
    • The older G25 + SDL2 1080p60 preview retains its own 204-execution sample and bounded stability evidence. The new HFR downloads do not replace those results.
    Downloads