Videodec2 opens FFmpeg with apply_cropping = 0, so the frame it reports is
the coded picture and its crop fields describe the cropping rectangle. The
coded size was derived from the frame size and the crop, which counted the
crop a second time when it added up to a multiple of 16: a picture cropped
four units at the top and four at the bottom was reported 16 rows taller
than the coded picture, and every picture was then refused for a frame
buffer that was sized for the coded picture. The frame size is the coded
size, so the crop only fills the crop offsets.
The fixture's sequence parameter set is also decoded with a crop of four
units at the top and four at the bottom now, which covers the case in the
guest test.
The recompiler asked for DenormPreserve only for 16-bit and 64-bit floats,
so a device that flushes 32-bit denormals by default (RADV, MoltenVK) flushed
them whatever the guest's float mode said. target.float32DenormPreserve is set
from shaderDenormPreserveFloat32 when denormBehaviorIndependence allows a 32-bit
mode, is part of the cache key and of the request serialization (version 41),
and makes the module declare DenormPreserve 32 unless the 32-bit flush is
already taken in hardware. APS5_NO_HW_F32_PRESERVE=1 turns it off.
sceKernelGetProcessTimeCounter counted nanoseconds and
sceKernelGetProcessTimeCounterFrequency returned 1 GHz, while
sceKernelGetTscFrequency reports the host TSC rate. On the console the
process time counter runs at the TSC frequency, so a title may scale
the counter by sceKernelGetTscFrequency. PPSA34501 does, and its clock
ran 3.4 times slow on a 3.4 GHz host: it paced its 60 Hz frames to
17.6 fps and its game time and audio drifted apart.
The counter now ticks at sceKernelGetTscFrequency, over the same
monotonic clock and APS5_TIME_SCALE, and its frequency is that rate.
With APS5_DUMP_FRAMES, FrameDumpWriter::Enqueue threw "frame dump queue exceeded 64 MiB" when the
writer thread fell more than 64 MiB behind, which failed the presentation and ended the title.
#4320 made that rare at the default APS5_DUMP_SCALE=4; at APS5_DUMP_SCALE=1 a 1920x1080 title
reaches it within nine frames and a 3840x2160 one on the third.
Enqueue now waits on the writer's condition variable until the frame fits, and the writer
notifies after each file; a single frame larger than the queue still throws. The VideoOut frame
dump queue test gets a wait mode at scale 1 whose third frame has to wait for the held first one.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
The change removed a technical-debt entry from origin/main that documents four
libSceVoiceChat exports it does not touch, left the new include out of the
quoted include group, put a forward declaration away from the other type
declarations, and cited a prototype source without a line anchor.
gethostbyname_r and sysinfo are names of host C library functions. Defining
them in libScePosixForWebKit made the library's own call to
::gethostbyname_r in Resolver.cpp bind to the new declaration, and the
guest_resolver test failed with "gethostbyname_r not implemented". Export
the two NIDs through a helper with the library's own prefix, the way
libSceAmpr re-exports the libc functions.
AnyPS5 resolves a title's imports by NID, so a title that imports a system
function that no AnyPS5 library provides fails to load with an unresolved
import. This declares 222 such exports across 11 PS5 system libraries that
AnyPS5 implements but does not provide.
Each declaration calls NotImplemented_nid_no_patch and returns 0; none of
them implements the function. A declaration with a credible prototype uses
it and cites the source with a pinned commit; the rest are declared without
parameters, which resolves the import without making an ABI claim. The
technical-debt files record the limitation per library.
The list is the intersection of the export table of the console's library
(function symbols only), the pinned aerolib names and the functions not yet
defined anywhere in the repo. Every name recomputes to its NID with
tools/nid_names.py.
libSceAudio3d and libSceVoice are left out because #8331 declares them.
On Windows, freopen reopened a stream through _wfreopen, which opens the file without FILE_SHARE_DELETE, so remove or rename of the file failed while the stream was open, unlike fopen streams and POSIX unlink/rename of an open file. freopen now closes the previous stream and opens the new file through the same CreateFileW path fopen uses, so the reopened stream shares delete access. Host standard streams and an empty filename keep the _wfreopen path. Reference: FreeBSD 11 unlink(2) and rename(2) succeed on a file that is open.
A needed library with no file in --libs or --modules is reported, but when the
file sits elsewhere in the converted title's folder the report gave no way to
find it. import_audit now looks for it, case-insensitively, up to four levels
below each registry's folder and names the directory to pass with --modules.
Co-Authored-By: Claude Sonnet 5.5 <[email protected]>
Dean