#5097 added ShaderInfo::layerExported to the entry encoder, and this
series moves that line into encodeShaderInfo, which plan entries use
too. The entry round trip left the flag false, so an encoder and a
decoder that both skipped it still passed. Set it like the other
ShaderInfo flags: with layerExported dropped from both, the round trip
now fails with "entry round trip: shader info differs".
Co-Authored-By: Claude Opus 5.5 <[email protected]>
windres passes include directories to the preprocessor unquoted, so
relinker.rc failed to compile when the checkout path contains a space.
The resource is now an object library without the relinker's include
directories.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
SPIRV-Tools v2025.4 and the SPIRV-Headers revision before it predate
SPV_KHR_fma, so spirv-val rejects the FMAKHR capability (6030) and
spv::OpFmaKHR does not exist. v2025.5 is the first release with
SPV_KHR_fma validation (8f9381775). SPIRV-Headers moves to b824a462d,
the revision v2025.5's DEPS pins, which includes the SPV_KHR_fma
headers update (9268f3057).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A 10-bit resident display buffer (and an 8-bit one in the other channel
order or an sRGB type) went through vkCmdCopyImageToBuffer, the color
transfer's convert dispatch, vkCmdCopyBufferToImage into the scaler source
and a blit, all at the display's resolution, plus a full clear of the
swapchain image.
PresentationPass now draws such an image straight into the swapchain image
with one fragment pass: it reads the raw texels through an R32_UINT view,
decodes them as the guest-memory path does (the display's channel order, the
10-bit variant truncated to 8 bits), scales with a bilinear filter of the
decoded values and writes black outside the aspect-fit rectangle, so it
needs neither intermediate buffer nor a clear. The swapchain stays
B8G8R8A8_UNORM and now also allows color attachment use. A 16-bit float
display buffer keeps the convert chain: its encode goes through the
scanout threshold tables, and the pass refuses it.
The blit paths clear only the letterbox bands (blits from a 1x1 black
image) instead of the whole swapchain image, and the scaler source image is
only used by the guest-memory, uniform and 16-bit float paths. GPU frame
dumps of a drawn image are drawn by the same pass at the dump size,
sampling every n-th texel, so they hold the same bytes as before.
Test: agc_driver_presentation_pass compares the pass with the convert path
for both 8-bit and both 10-bit channel orders (identical unscaled and
letterboxed, within two levels of the blit filter scaled), checks that a
10-bit destination keeps every source bit, that dumps
equal the subsampled guest-path frame, that the band clears equal a full
clear, and that a 16-bit float display buffer is refused.
PS5CEMU-HAR (PPSA99360, ps5-payload-dev build, relinks with
--trap-sdk-syscall-stubs) imports pthread_setaffinity_np (5KWrg7-ZqvE)
from libkernel.prx. babyhuey's run with #3787 merged onto main e7c70a2d
(PR #3787, 2026-10-12T10:54:44Z) stops at it:
Windows: FAIL: unresolved ELF import 5KWrg7-ZqvE (pthread_setaffinity_np) from libkernel.prx
Linux: undefined symbol: 5KWrg7-ZqvE (pthread_setaffinity_np)
It follows FreeBSD 11.4 libthr thr_affinity.c (8c639945) over
kern_cpuset_setaffinity: another thread is looked up first (EINVAL for
null, ESRCH when unknown or finished, as _thr_find_thread without dead
threads), a size other than the 32-byte cpuset_t returns ERANGE, the
mask is clipped to sceKernelGetAvailableCpumask and an empty result
returns EDEADLK, as cpuset_setaffinity does. The affinity is stored on
the guest thread through scePthreadSetaffinity, for the caller and for
any live guest thread.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The plan and ShaderInfo encoders write these structures field by field,
and the sizeof asserts are their only reminder to update the encoders. A
field that fits in existing padding keeps the size: main's layerExported
landed in ShaderInfo with sizeof still 232, and the three bools #5089
adds to DescriptorSource::IndirectImage keep it at 28. A warm cache
would then hand back entries without the new field and nothing would
fail.
Count the fields of every aggregate the plan codec reaches (IrResourcePlan,
GuestRegister, MemoryInfo, DescriptorSource and IndirectImage,
ResourceBlock, SrtRead, UniformFill, UniformFillPlan, and ShaderInfo with
its resource and stage I/O records) by brace-initializing each with more
and more convertible-to-anything values until the initializer no longer
compiles, and assert the counts. Unlike the size asserts this does not
depend on the ABI, so it is checked on every platform. IrValue is a
class with private state and keeps only its size assert.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The round trip requires some fixture to plan a phi, and only one did:
the loop counter of "bindless sample in a loop latch", which reaches the
plan through the latch condition. The counter is not invariant, so the
CPU walk can never evaluate that condition. A plan that keeps only the
conditions the walk can evaluate, as #5181 aims to, would hold no phi
and fail the check with "no fixture plans a phi".
"descriptor invariant phi" loads the image descriptor from the same SRT
offset on both sides of a branch on an SRT dword. Its eight dwords are
phis of equivalent reads: the tracker keeps them, since it splits only
descriptor phis that are not invariant, and the walk evaluates them
through the invariant read. The plan stores them as descriptor dwords,
whichever conditions it keeps, so the phi encoding stays covered. The
fixture goes after the other compute fixtures, so they keep their seeds
and their damaged-entry checks.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
APS5_NO_PERF_DISK_PLANS=1 is listed as a neutral switch, so it stays out
of the variant key, and it exists to compare warm starts with and
without stored plans. Nothing tested either property: the cross-process
test ran with the switch unset only.
A fourth child process now sets the switch against the cache the first
process filled. It must run the front end for all 14 fixtures, neither
load nor store a plan, still hit all 14 stored variants, and produce the
same SPIR-V, bindings, push constants, snapshots, read traces and plans
as the first process.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The plan tests compared a decoded plan with the original only through
the codec itself (re-encoding must give the same bytes) and through
materialization. Neither catches a field that EncodePlan and DecodePlan
both leave out: the bytes still agree, and the walk does not read every
field. Leaving userDataBase, the plan's sRGB formats, the value ids or
several MemoryInfo fields (secondaryOffset, dataDwords, dataFormat,
planningOnly, idxen) out of both sides passed every test. The hand-built
plan also never materializes (its uniform values cannot be evaluated),
so its 1000 runtimes all failed with the same error, and no plan held an
immediate wider than 32 bits.
samePlan compares every field of two plans without the codec: the value
graph by storage index (opcode, type, id, flags, register, immediate,
arguments, phi blocks), the block ids, the roots, and the plain data
through the structures' own equality. The round trip now requires it
for the 14 fixtures and the hand-built plan, which sets every MemoryInfo
field, all indirect table fields, a second buffer, the whole uniform
fill and a 64-bit immediate to values other than their defaults. The
second process also builds each fixture's plan with its own front end
and requires it to equal the plan it used, so a stored plan must match
what the front end builds in another process, not only its own
re-encoding.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
DecodePlan checked every descriptor source index a plan holds except the
two inside an indirect image table: the heap source and the material
source. A damaged entry could therefore decode with a table that points
past the plan's descriptor sources, and only the materializer's own
bounds checks stood between it and an out-of-range read. The front end
interns both sources before the image source that names them, so a plan
it builds always passes the new check.
agc_shader_disk_cache now requires a plan whose table heap or material
source lies past the sources not to decode.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
getSource built ResourceProgram (PrepareResourceProgram and ExtractPlan)
for the first request of every shader a process saw, before
findOrCompileVariant looked in the disk cache. With a warm cache every
launch still decoded, built the CFG, structurized, translated and ran
SSA, the fold passes, SrtWalker and ResourceTracker for every shader;
only SPIR-V emission was skipped.
getSource now builds the plan key and tries LoadPlan first. On a hit it
sets the plan and leaves the program null: findOrCompileVariant already
rebuilds the program when no stored variant fits the layout. The loaded
plan takes the request's code address as its shader hash, as a plan
built in this process would. On a miss it builds the plan as before and
stores it. A plan the front end refuses is not stored. The disk is not
used when the cache is off, under the debug probe, with APS5_DUMP_IR
(which prints the front-end passes), or with APS5_NO_PERF_DISK_PLANS=1,
a cache-neutral switch for comparisons.
FrontEndBuilds() counts PrepareResourceProgram calls. The new
cross-process test prepares, captures and materializes 14 shader
fixtures in one process and repeats it in a second process sharing the
cache directory: the second process must run the front end 0 times, hit
14 plans and 14 variants, and produce byte-identical SPIR-V, bindings,
push constants, snapshots, read traces and plans. A damaged plan entry
is rejected and rebuilt alone with the same output and stored again; a
changed code word and another switch value miss.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The disk cache keeps compiled variants (SPIR-V, shader info, binding
layout) but not the IrResourcePlan the front end extracts for the
resource walk, so a process can only get a plan by running the whole
front end. Since #471 the plan owns its values (ownPlanValues clones the
graph behind every root into valueStorage and blockStorage), so it can
be written out and rebuilt.
- EncodePlan writes the value graph in storage order, each value as
LEB128 integers: opcode, type, id, flags, register, operand count,
the immediate if it has one, each argument as the zigzag distance to
the value's own index (0 for null) and, for phis, the index of each
predecessor block. The roots (descriptor source dwords, SRT reads,
control-flow conditions, uniform-fill values) are value indices; the
plain data (memory info, indirect image tables, control-flow edges,
materialization sources, guarded, clean and pure flat slots, the
flags, the ShaderInfo through the variant's encoder) follows. A plan
whose roots or arguments point outside its storage is not encoded.
Control-flow conditions can pull most of a program into the plan:
a synthetic 2,038-dword compute shader with a branch every 20 dwords
has 42,775 plan values, which take 518 KB this way and 2.5 MB with
fixed-width fields; its compiled variant entry is 935 KB.
- DecodePlan rebuilds the graph the way ownPlanValues does and rejects
an unknown opcode, an overlong integer, an index past the values or
blocks, a descriptor source, image, buffer or sampler past the
sources, a branch without two successors, an edge past the blocks,
and a uniform fill of more than four words.
- BuildPlanKey is the in-memory source key (RecompileCacheKey::Build and
the host subgroup size), every code word and the switch key, after a
plan entry tag. It leaves out the binding layout, so all layouts of a
shader share one plan.
- EncodePlanEntry/DecodePlanEntry use the variant's header and hashes;
LoadPlan and StorePlan use the same directory and background writer
and count plan hits, misses, writes and load failures on their own.
- Size asserts cover the plan and the structures it is written from.
- FormatVersion 30.
agc_shader_disk_cache now round-trips the plans of 14 in-tree shader
fixtures (SRT reads, guarded and chained pointers, control-flow
conditions, an indirect image table, a loop phi, a fragment shader) and
a hand-built plan with a uniform fill. For each it materializes the
original and the decoded plan over 1000 random user-data and memory
runtimes and requires the same snapshot, poison, null-root count and
read trace, or the same error. Truncated, bit-flipped and
other-format entries are rejected, malformed plans do not decode, and
the key follows the code, the target and the host subgroup size but not
the layout, addresses or user data.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A guest handler that called exit threw out of the libc exit path because the exiting thread waited on itself. Like FreeBSD, the nested call now runs the handlers that are still registered and ends the process with the status it passed.
encodeInfo and decodeInfo wrote the ShaderInfo fields inline between the
compiled-shader header and the binding layout, and EncodeEntry and
DecodeEntry built and checked the file header (magic, format, source
version, key and payload hashes) inline. Move both into helpers,
encodeShaderInfo/decodeShaderInfo and wrapEntry/unwrapEntry, so that a
second kind of entry can reuse them.
The bytes of a compiled variant entry do not change.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The cache key of relinked images has no time stamp or check sum, so a
different module of the same size loaded at the same base was served the
old tables. The loader now forgets the image when it frees the module, the
cache has a separate valid flag, and the sysmodule name cache is left out
of this change.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
A commit failure inside the guest arena threw and left the reserved range in use. The heap now resets and releases the range, and returns null with ENOMEM when the host is out of memory. reallocalign returns null with EINVAL or ENOMEM like the other allocator entry points.
A size the default heap rejects or the guest arena cannot place, a calloc product that overflows, and an invalid alignment threw through the guest call, which aborted the process. FreeBSD 11.4 returns null with ENOMEM, or EINVAL for an alignment that is not a power of two. Guest arena exhaustion now throws bad_alloc.
longjmp moves the Windows stack bounds onto the stack of the target jump buffer so that exceptions and stack probes keep working after a jump between guest stacks. siglongjmp restored the same registers without that step, so a jump from a context stack back to the thread stack left the thread bounds describing the stack it had left.
Add missing libc.prx symbols required by Unity IL2CPP runtime:
- Implement _Dtest double classification in Math.cpp
- Implement _Inf double constant in Math.cpp
- Implement sceLibcBacktraceGetBufferSize and sceLibcBacktraceSelf in Backtrace.cpp
- Implement _Cnd_signal in Threads.cpp
- Implement _ZNSt8ios_base7_AddstdEPS_ in RuntimeSupport.cpp
- Implement __pointer_type_info and __function_type_info vtables in TypeInfo.cpp
- Remove duplicate stubs from libSceAmpr/Export.cpp
- Add unit tests in GuestMath.cpp and GuestLibcThreads.cpp
- Document unverified behaviors in docs/dev/debt/libc.md
The cache from #5107 was keyed by Mesa version only, so a runner restored
llvmpipe code compiled on another runner's CPU and about 85 agc tests died
with SIGILL (Illegal instruction) on main.
Co-Authored-By: Claude <[email protected]>
A PS5 eboot can keep initial-exec TLS as R_X86_64_TPOFF64 GOT relocations
(SuperTuxKart, STKT00001, has 13, all without a symbol), which the
relocation allowlist rejected. Relocations against the executable's own
TLS block (no symbol, or a defined STT_TLS symbol folded into the addend)
now go to the Linux ELF as symbolless dynamic relocations that the
dynamic loader resolves against where it placed the block; one against an
import stays named and is typed STT_TLS. The Windows and macOS images
resolve TPOFF64 and DTPOFF64 statically against the block
AlignUp(p_memsz, p_align) below the thread pointer, the offset
WindowsTlsBuilder and MacOsImage use, and reject DTPMOD64 and imported
TLS. TPOFF32 stays unsupported.
New test executable_tls_relocations covers the Linux, Windows and macOS
outputs and the rejected cases.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
FreeBSD's sigwait hands the error number back as its return value and leaves errno alone; the guest export returned -1 and set errno.
(cherry picked from commit b627f6c9443dd94a6392959f33e52d837ee4cae4)
MathiasDevelopes