[tibzejoker]

What

Guest registers are 32-bit words. A float result is bitcast to u32 when it is written to a register, and back to f32 when it is read, so emitted modules carry pairs of OpBitcast that return to the type of the first one's source. This change makes a bitcast back to that type return the source itself. The source dominates every use of the first bitcast, so the module stays valid.

To support this, SpirvModule records the result type and the bitcast source of each function instruction. Both Unary and EmitNative bitcasts look them up.

Tested

Windows 11, MinGW GCC 15.2, RTX 4090 (616.92):

• New test agc_bitcast_round_trip. It recompiles v_cvt_f32_u32 results added with v_add_f32 and requires that no bitcast goes back to the type of a bitcast's source. It fails on main (one of six bitcasts) and passes with this change.
• ctest -R ^agc_: 727/728. The failure is agc_driver_bda_device, which fails on main too.
• GTA V (PPSA04264): 72 compute requests dumped with APS5_DUMP_SHADERS were dispatched on synthetic guest memory with main and with this change. Every byte they wrote was identical. SPIR-V words over these 72 went from 1,930,912 to 1,898,324 (-1.7 %); for example compute 0x143978f00 went from 8,545 to 8,457 words, and 0x25b390800 from 647,759 to 642,443.

This is a small size win: every round trip removed is one instruction the driver no longer compiles.

Performance capture

GTA V (PPSA04264), open world from the same save, RTX 4090, one run each, same session. Base = main 47a266c + #6488 + #6848 + a local bench aid (keyboard pad input while an IME session is open), because GTA V does not run on plain main; c0206c244d15 is this PR on top of it. Traces: fork release perf-7673-7677-20261019.

These frame times do not measure this change. The trace buffer of queue worker 0 fills after about 3.7 s (trace_dropped), and in the open world frame time follows where the save respawns. This PR shortens shader compiles: the SPIR-V word counts and the cold compile times under Tested are the measurement.

{"commit":"6537538e26d6","dirty":false,"env":{"APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"330:60"},"gpu":"NVIDIA GeForce RTX 4090","driver":"NVIDIA 616.92","os":"Windows 10.0.26200","resolution":"3840x2160","window":{"spec":"seconds 330:60","first_present":4184,"ms":60074.435},"presents":416,"frame_ms":{"p50":129.613,"p90":188.599,"p99":328.411,"max":548.842},"gpu_busy_ms_per_present":47.351,"worker_busy_pct":{"queue worker 0x0":3120.864,"queue worker 0x20":33.818,"queue worker 0x28":19.713,"queue worker 0x38":29.948,"queue worker 0x30":12.619},"top_phases_ms_per_frame":{"Driver.EndSubmission":4370.075,"Driver.Packet":271.688,"Driver.Packet.memory_wait":136.191,"Driver.Packet.draw":114.943,"Driver.Draw":113.961,"Driver.Draw.shader_compile_and_link":62.672,"Shader.DrawStage":48.422,"Driver.Draw.draw_enqueued":39.814},"trace_dropped":true,"trace_dropped_threads":{"queue worker 0x0":{"events_full_after_s":3.742,"events_lost":3289625,"totals_full":false}},"trace":"aps5-trace-6537538e26d6-44248.json.gz"}

{"commit":"c0206c244d15","dirty":false,"env":{"APS5_TRACE_DIR":"<path>","APS5_TRACE_SECONDS":"330:60"},"gpu":"NVIDIA GeForce RTX 4090","driver":"NVIDIA 616.92","os":"Windows 10.0.26200","resolution":"3840x2160","window":{"spec":"seconds 330:60","first_present":3955,"ms":60187.407},"presents":451,"frame_ms":{"p50":126.146,"p90":156.364,"p99":281.649,"max":769.832},"gpu_busy_ms_per_present":49.259,"worker_busy_pct":{"queue worker 0x0":3440.583,"queue worker 0x30":15.325,"queue worker 0x38":20.011,"queue worker 0x20":38.649,"queue worker 0x28":22.082},"top_phases_ms_per_frame":{"Driver.EndSubmission":4465.061,"Driver.Packet":250.534,"Driver.Packet.memory_wait":125.851,"Driver.Packet.draw":102.510,"Driver.Draw":101.584,"Driver.Draw.shader_compile_and_link":63.901,"Shader.DrawStage":49.320,"Driver.Draw.draw_enqueued":26.262},"trace_dropped":true,"trace_dropped_threads":{"queue worker 0x0":{"events_full_after_s":3.442,"events_lost":3713441,"totals_full":false}},"trace":"aps5-trace-c0206c244d15-28644.json.gz"}

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none
• Unblocks: none
• GPUs tested: NVIDIA GeForce RTX 4090 (616.92, Windows 11); AMD Radeon 860M (32.0.13058.8001, Windows 11, by @LeandroLP23)
• AI-assisted: yes (Claude Code)

🤖 Generated with Claude Code

https://claude.ai/code/session_01XBXM4pgkMHYEtqN1EQni2P

[github-actions]

This pull request is labeled perf, so it needs the summary line of a capture on main and one with the change, from the same title, scene and machine, as the Performance report form asks. Please paste both aps5-summary.jsonl lines, attach the two traces and add the output of tools/trace_digest.py.

[github-actions]

Other open pull requests touch the same code. Merging one will break the other, or both implement the same thing. If this one builds on them, list them in Depends on.

Pull request	Conflicting files (lines on main)	Same exports	Same new files
PR 9508	core/libs/prx/libSceAgcDriver/CMakeLists.txt L2536
PR 10973	core/libs/prx/libSceAgcDriver/CMakeLists.txt L2536

Some comparisons could not be completed because a pull request conflicts with the base branch. Review these shared files manually before concluding that there is no overlap.

Pull request	Shared files not compared
PR 1368	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 1395	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 1896	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 2434	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 5021	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 5371	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 5925	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6187	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6188	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6192	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6209	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6235	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 6280	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6300	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6412	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6481	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6642	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6743	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6808	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 6959	core/libs/prx/libSceAgcDriver/CMakeLists.txt core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 7048	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 7201	core/libs/prx/libSceAgcDriver/CMakeLists.txt core/shader/recompiler/SpirvBackend/include/SpirvBackend/SpirvEmitterHelpers.hpp core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 7275	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 7457	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 7514	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 7604	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 7642	core/libs/prx/libSceAgcDriver/CMakeLists.txt core/shader/recompiler/SpirvBackend/include/SpirvBackend/SpirvEmitterHelpers.hpp core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 8014	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 8161	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 8246	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 8484	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 8542	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 8592	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 8707	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 8874	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9020	core/libs/prx/libSceAgcDriver/CMakeLists.txt core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 9176	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9248	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9287	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9289	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9335	core/libs/prx/libSceAgcDriver/CMakeLists.txt core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 9356	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 9387	core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 9700	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9748	core/libs/prx/libSceAgcDriver/CMakeLists.txt core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 9852	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9892	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9897	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9898	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 9989	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 10193	core/libs/prx/libSceAgcDriver/CMakeLists.txt core/shader/recompiler/SpirvBackend/include/SpirvBackend/SpirvEmitterHelpers.hpp core/shader/recompiler/SpirvBackend/src/SpirvModuleEmitter.cpp
PR 10233	core/libs/prx/libSceAgcDriver/CMakeLists.txt
PR 10442	core/libs/prx/libSceAgcDriver/CMakeLists.txt

[LeandroLP23]

AMD run, since the description covers an RTX 4090. Radeon 860M (RDNA 3.5, driver 32.0.13058.8001), Windows 11 26300, WinLibs GCC 15.2.0 r7, Release, BUILD_TESTING=ON. Build: main 8a456334f with #7677, #7673, #7671, #7666, #7676, #7675, #7672, #7668, #7664 and #7658 merged together (each merge clean, built once).

agc_bitcast_round_trip and agc_driver_bda_device pass 3 runs of 3 on AMD's proprietary driver.

AI-assisted

[tibzejoker]

About the performance-report request: I captured the requested warm-cache traces (GTA V PPSA04264, open world, RTX 4090). Two things make them a poor fit for this change:

• With the trace on, the queue worker's event buffer fills within about 3.7 s in this scene (trace_dropped: true, about 3.3 M events lost), so the per-phase totals are not usable. Only the frame percentiles survive.
• In the open world, the frame time depends mostly on where the save respawns: the same base build ran at 9.3 FPS in its warm-up and 5.5 FPS in its traced run. Shader compilation inside the traced minute also varies between runs (62 to 97 ms per frame of Driver.Draw.shader_compile_and_link), depending on which new shaders that run meets.

This change only makes modules smaller. Its effect is on pipeline creation, so I measured that directly: cold NVIDIA pipeline_create, median of 3 agc_compute_pipeline_replay runs, each with a distinct OpSourceExtension so the driver cache misses. The numbers are in the description. If you'd rather have a specific trace or scene, tell me which and I'll capture it. The three traces I took are here, for completeness: https://github.com/tibzejoker/AnyPS5/releases/tag/perf-7673-7677-20261019

AI-assisted (Claude Code).

[DavidAmsellem]

First Linux run of this PR, on AMD with RADV. Head c07b6bdc0 merged onto main 6bef96df7 (clean merge), built in full and compared with the same main, on the same machine.

AMD Radeon RX 7900 XT (RADV NAVI31, gfx1100), Mesa 25.2.8, Linux Mint 22.3 (Ubuntu 24.04 base), kernel 7.0.0-38, GCC 13.3, Release.

Full suite (ANYPS5_REQUIRE_VULKAN=1 ctest -j8) at 56990699d: 1713/1714 with this PR, against 1712/1713 on main. The one failure is the same on both, so this PR adds none: agc_driver_guest_memory_lifetime (#7784).

Comparing the JUnit output with main: the PR's new entry, agc_bitcast_round_trip, runs and passes. No other entry is added, and none is skipped or removed.

This checks correctness only; I did not measure performance.

GPUs tested: AMD Radeon RX 7900 XT (RADV NAVI31, gfx1100), Mesa 25.2.8, Linux Mint 22.3 / kernel 7.0.0-38

AI-assisted: yes

[tibzejoker]

Thanks @DavidAmsellem for the RADV run on gfx1100: same single failure as main (#7784), and agc_bitcast_round_trip runs and passes.

AI-assisted: yes (Claude Code)