This reverts commit 780b33aea3, restoring the 13824 MiB direct memory pool.
With the pool at 12288 MiB, Astro Bot (PPSA21564) can no longer load its
level: DirectMemoryAllocator asserts "Can't allocate direct memory" for a
64 MB VideoPlayerTexture with 12228 of 12288 MB in use, then the process
faults about 26 s in (2 of 2 runs, RTX 4090, Linux, main cccdaa2734).
The same tree with only this merge reverted runs the full 150 s with no
direct memory assert (2 of 2 runs).
ReportAllThreads captures at most 128 other threads and prints how many
further threads were left out; the report did not list every thread of a
larger process.
GuestModuleBuilder keeps each module's directory relative to the module
parent under app0, as the Linux and Windows outputs do, so modules from
sce_modules, prx or a --module-dir do not land in app0/sce_module.
MacOsMachOPatcher::Patch discarded its lazyBinding argument, so a
module-free `--macos --lazy-binding` conversion exited 0 and wrote the
same eagerly bound Mach-O as a run without the flag. The Mach-O writer
puts every import in the LC_DYLD_INFO_ONLY bind stream and leaves the
lazy_bind stream empty, with no __la_symbol_ptr or stub helper, so the
macOS target cannot honour the flag. The patcher now throws "macOS
target does not support --lazy-binding", in the form of its
--windows-diagnostics check, so such a conversion exits 2 and writes no
output. Conversions with bundled modules were already refused by the
guest module builder.
USAGE.md's performance overlay section now gives F10's cycle (off, page
1, page 2, off) and ANYPS5_OVERLAY=2, a table with one row per line of
page 2 that names each source and the system it comes from, and what the
page costs: the trace totals it turns on while it shows, and the checks
and counters that run while the overlay is hidden or on page 1.
ENVIRONMENT.md's ANYPS5_OVERLAY row lists page 2.
The relinker binary carried no build identity: a release build and a
checkout of main were indistinguishable, so a report could not name the
build it came from. --version prints the revision git describe reports,
falling back to "unknown" when the tree has no git metadata, and needs
no input or output paths. The usage line and USAGE.md list the option.
Walk each macOS function import before the guest entry and pass it through TraceImport_nid_no_patch, as Linux and Windows already do. An x86-64 libc build records the call and prints it on SIGABRT.
Titles report a progressive trophy's progress with _UpdateTrophyProgress
events. Their udsNN.ucp rules feed $._trophy_progress into an int32
statistic with the max aggregation, filtered on $._trophy_id, and the
trophy's unlockCondition names that statistic. Those statistics and
rules were skipped and the events were dropped, so such trophies never
reported progress or unlocked (21 of Borderlands 4's 64).
- Read max statistics (largest value posted, from initialValue) and
int32 statistics (0 to 2147483647).
- Read a rule condition with one == property equal to an unsigned
integer, and apply the rule only to events whose property matches.
- Post _UpdateTrophyProgress through the statistic rules.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Only the main executable output was checked for being a directory. A
directory at a guest module, extra executable, registry, needed-library,
patch manifest, icon, desktop entry or diagnostics path failed only when
it was opened, in most cases after earlier outputs had been written.
checkOutputInputs now rejects an existing directory for every output
kind it checks, as 286cf28ec7 did for the main executable. Such runs
still exit 2, now before any output is written.
The command-line relinker skips an ELF module whose file name extends
another module's in the same directory when both have the same SONAME, a
dumper's backup such as libX.prx.esbak (#7380). That check lives in the
CLI's ReadModules, so the online converter, which hands every ELF in
sce_module/ straight to Conversion, still stops on such a layout with
"Ambiguous guest dependency name: libX.prx".
The check moves into the relinker core as RemoveBackupCopies, which
Conversion::Convert now runs before patching or converting modules and
reports through the progress messages, so the browser prints the same
"Skipping backup copy" lines. ReadModules uses the same function, so the
command line and --verify-patches behave as before.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
--frame-red-zone replaces a leaf's prologue and each exit with a 5-byte jump
to a stub, so it skipped leaves whose pushes or pop-and-ret exit were shorter
than that, and leaves with neither pushes nor an allocation. A short prologue
now takes in the straight-line instructions after it and a short exit those
before it, re-encoded in the stub with the moved displacement. Widening stops
at branches, rets, RIP-relative instructions, pushes and pops, and the leaf is
skipped when a branch targets the instructions taken in or the first
instruction of a widened prologue, or when a leaf without pushes pushes or
pops.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A traced run that ends without a crash or an abort never reports which
imports it called. With APS5_TRACE_CALLS_FILE set, the first traced import
starts a thread that rewrites that file every 10 seconds with every import
called so far, most called first.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The graphics pipeline store kept a fixed 1024 entries and the pipeline journal replayed at most
1024 records, and a replayed pipeline that found the store full was dropped. A title whose
working set is larger rebuilt the rest at the draw on every run, and evicted pipelines it kept
using were rebuilt again and again.
The bound now starts at 1024 and doubles up to 4096: a journal replay that does not fit grows it
instead of dropping the pipeline, and so does eviction churn (a ghost list of evicted keys, checked
every bound/32 evictions: more than half of them came back). It never shrinks during a run;
APS5_PIPELINE_STORE_ENTRIES still fixes it. The journal loads its last 4096 records by default
(APS5_PIPELINE_JOURNAL_ENTRIES). Resident pipelines are bounded by count, not bytes.
agc_driver_recorder_tests --benchmark-pipeline-store-memory [count] (opt-in, not registered)
builds count graphics pipelines with one fragment module each through CachedPipeline and prints
process private bytes and VK_EXT_memory_budget heap usage before and after: with the triangle
shaders about 59 KiB host and no device memory per pipeline on an RTX 4090, 100 KiB host and
8 KiB device-local on a UHD 770.
Tests: agc_pipeline_store_growth (fails on main), agc_pipeline_store_fixed_bound,
agc_pipeline_store_replay_growth (replays on one thread: the mock Vulkan is not thread-safe).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_017SNUYRd7sRKZrMYpquAsnU
Adrià