[Celegans12]

What

A program with more than 32 distinct buffer descriptors (V#s) throws buffer resource limit exceeded. The 32 is a recompiler constant, ShaderInfo::MaxBuffers, not a Vulkan limit, so this PR drops it.

• Each buffer is a constant-indexed element of one storage buffer array (no descriptor indexing); ShaderResources already throws above the storage-buffer limits: 1,048,576 per stage on an RTX 3080 Laptop, 1,015,808 on lavapipe, 200 on Intel UHD.
• buildPrepare counted only storage buffers against maxPerStageResources (also 200 on Intel UHD) and now adds the stage's sampled and storage images, not samplers. A pixel shader's colour attachments still aren't counted (TechnicalDebt).
• The SPIR-V emitter's memoryByteOffsets is sized from the binding layout, not 32 entries. When the user data and offsets no longer fit in the push constants, the whole block moves to the shader data buffer, as before.
• PPSA26344 v1.008: the setup compute shader 0x80003c3100 reads two constant buffers, writes counts into 52 more and throws that message on main.
• perf(agc): baseline shader recompilation on registration #639 adds its own AddBuffer limit and per-stage count; whichever lands second drops the duplicate. feat(libSceAgcDriver): back the global data share with driver-owned guest memory #285, feat(agc): single-sample sample masks as a per-quad-pixel discard #448, feat(agc): implement image and sampler tables indexed at runtime offsets #757-feat(agc): implement deferred faults for reads through null SRT pointers #759, feat(agc): implement min and max sampler reduction modes #815, refactor(recompiler): share the SPIR-V emission failure helper #832, feat(agc): implement 8 and 8_8 srgb image loads through unorm views #866, feat(recompiler): implement 64-bit image atomics #896, feat(agc): implement image_load through image tables #920, feat(agc): implement sint storage images through uint views #944, fix(shader): throw for constant lane reads outside the host subgroup #997, perf(shader): share the BDA byte read fallback between read sites #1043 and perf(agc): size the storage and sampled texture caches to the device #1059 edit other lines of these files and merge cleanly.

Tested

• Windows 11, MinGW-w64 GCC 15.2.0, based on main db3e659: configure -DBUILD_TESTING=ON, full build, --target libs, ctest:
• 383/383 tests pass (ctest --output-on-failure)
• Same build on lavapipe (Mesa 26.2.4 llvmpipe, 8-wide subgroups, VK_DRIVER_FILES): 382/382 pass; agc_driver_flat_store is left out there because it takes several minutes on Windows llvmpipe
• New agc_driver_many_buffers reads and writes 52 buffers (push constants) and 124 (shader data), three in four not 16-byte aligned.
• agc_driver_graphics: two buffers and a sampled or storage image throw at maxPerStageResources 2, not 3; samplers don't count. Both fail on main.
• PPSA26344 v1.008: VALIDATION_PENDING
• Not tested: Linux (CI), console behaviour.

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 in TechnicalDebt
• 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
• AI-assisted: yes

[adriadam10]

Checked on RADV (merged onto current main, 395/395): no other per-buffer 32-entry table or mask is left, both new tests fail without the change, and agc_driver_many_buffers runs in 1.4 s cold on CI's lavapipe. Merging.