RADV clears a multisampled DCC target to a color other than 0000, 0001,
1110 or 1111 with radv_clear_dcc_comp_to_single: a compute dispatch stores
the color at the first pixel of every DCC block through a uint storage view,
and the DCC keys become 0x10. The driver refused that view of a resident
target, made a separate target in the view's format otherwise, and couldn't
import 0x10 keys.
The dispatch is now recognized by its code and not run. Its clear color is
recorded for the surface its image descriptor names, and 0x10 keys on a
multisampled target of that surface, size and sample count import as a clear
to that color, for texels up to 32 bits.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
When a title has configured 119.88 Hz output (mode 15) and the window moves
to a display below 120 Hz, the presenter fell back to std::terminate via a
require() call. Replace the abort with a graceful fallback: reset outputMode
to mode 1 (59.94 Hz), fire the output-mode event so the title is notified,
and continue presenting.
This matches the HDMI output mode renegotiation a real display change
requires: the system falls back to a compatible mode and notifies the
title, rather than terminating the process.
After the fallback:
- sceVideoOutGetOutputStatus reports refreshRate 3 (59.94 Hz)
- sceVideoOutGetResolutionStatus reports refreshRate 3
- vblanks run at 59.94 Hz (every-other-tick skip resumes)
- sceVideoOutIsOutputSupported(mode=15) returns the unsupported error
- sceVideoOutConfigureOutput(mode=15) succeeds again once the window is
on a 120 Hz display
The console's exact behaviour when a 120 Hz display is lost has not been
measured; this fallback is recorded as unverified in the debt file.
Fixes#7877
libSceVideodec's five exports were stubs that threw. Decoder creation,
decoding, flushing, resetting and deleting are implemented with the
structure layouts of shadPS4's PS4 library, and the decode runs through
libSceVideodec2, which decodes H.264 with FFmpeg. The picture is copied
into the caller's frame buffer with the pitch aligned to 64 and the coded
height, the chroma after the luma rows, and the picture information holds
the timestamps and the crop offsets.
A guest test decodes the H.264 fixture through the library and checks the
pictures against the reference hashes.
AnyPS5 resolves a title's imports by NID, so a title that imports a system
function no library provides fails to load with an unresolved import. The
PS5 3.20 export inventory lists 193 such exports across 37 PS5 system
libraries AnyPS5 implements but does not provide.
Declare them as throwing stubs so those titles load and reach their next
blocker. Each declaration calls NotImplemented_nid_no_patch and returns 0;
none of them implements the function. Declarations with a credible
prototype use it and cite the source with a pinned commit, and the rest
are declared without parameters, which resolves the import without making
an ABI claim. The technical-debt files record the limitation per library.
libSceMbus and libSceSystemService are left out because pull request #7821
declares them. libSceAgc is left out because its exports are versioned, so
the 3.20 NID names a different symbol from the one AnyPS5 ships.
libSceAvPlayer.native is left out because it compiles libSceAvPlayer's
Export.cpp.
Names and NIDs come from the PS5 3.20 export inventory at
https://github.com/DNNDHH/PS5-3.20_Libs/tree/b2a77c09cd71c01f34c790cfbe98c751a13262ad
The agc_driver_indirect_thread_dimensions_cpu registration sat at the end of
the address-passes test block, where other open changes add their tests too.
It now has its own BUILD_TESTING block after the execution test name check,
with the same properties.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
The guest socket layer relies on a few Linux calls:
- The equeue wait uses poll with the timeout rounded up to a millisecond;
macOS has no ppoll.
- A readable stream socket whose peer closed is reported with POLLHUP by
peeking one byte; macOS has no POLLRDHUP.
- FIONREAD on a datagram socket, and the byte count that fstat and kevent
report for it, is SO_NREAD, the size of the next datagram; macOS adds the
source address to FIONREAD.
- socketpair throws for SOCK_SEQPACKET, which has no AF_UNIX pair on macOS.
- gethostbyname replaces gethostbyname_r, which macOS lacks.
The UDP test sends on its connected socket without an address, as #7475
does, because macOS, like FreeBSD, fails sendto with an address on a
connected socket with EISCONN. The log test uses a datagram pair with
enlarged buffers instead of SOCK_SEQPACKET.
sscanf, vsscanf, sscanf_s and vsscanf_s threw on Linux for %ls, %lc, %l[, %S and %C, because glibc stores a 32-bit wchar_t per character into the guest's 16-bit buffer. ScanGuest, the per-directive scanner the _s forms already use, now scans those fields itself: the input is decoded as UTF-8, mirroring the narrow printf family's UTF-8 output, and each character is stored as one 16-bit unit, the width and the _s size counting characters; %ls skips leading white space and stops at the next, %l[ matches an ASCII scanset with ranges and negation, and a %lc field that meets the end of the input early keeps what it read, as FreeBSD's convert_wchar does. sscanf and vsscanf route a format with a wide conversion through ScanGuest and keep the host path otherwise. The UTF-8 decoding and scanset matching live in ScanWideField.cpp, which throws for input that is not UTF-8, a character outside the BMP, a scanset member outside ASCII and positional arguments next to a wide conversion; wide conversions in fscanf still throw.
A PS4 package's sce_sys/playgo-chunk.dat starts with the magic "plgo"
instead of the PS5 "plgx", so scePlayGo threw on every PS4 title that
ships one. Up to the chunk count the two headers have the same layout: a
32-bit magic, then 16-bit version major, version minor, image count and
chunk count, so the count is the little-endian 16-bit word at 0x0A in
both. The table is now read for either magic; a zero count is still
rejected.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A draw with no pixel program was rejected as writing color whenever
CB_TARGET_MASK and CB_SHADER_MASK overlap, even with CB_COLOR_CONTROL
mode 0 (CB_DISABLE). DecodeState already drops every color export in
that mode, so such a draw writes no color: a depth or stencil
decompress pass draws this way, leaving the masks of the previous draw.
An unwritten CB_COLOR_CONTROL still leaves the decision to the masks.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The first hourly pass over every open pull request after #5976 was
cancelled at the 10-minute limit (run 39160516271).
Co-Authored-By: Claude Opus 5.5 <[email protected]>
title_ids took every title ID from the Unblocks field, so a field that
starts with "none" and then names a title for context ("none by
itself; foundation for PPSA27836 in #6570") got the unblocks title
label. A field that starts with "none" now names no title.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Queue 0's draw committer records the draws the packet worker hands
over. When the worker's APS5_SAMPLE_WORKER stacks wait in
DrawPipeline::Enqueue, the committer is the thread that bounds the
frame, and only APS5_SAMPLE_THREADS sees it: 40 addresses, 12 frames
deep, without whole stacks.
APS5_SAMPLE_COMMITTER=<file> attaches the worker's sampler to the
committer when its thread starts and writes the same two files. Both
use one start function, and StopWorkerSampler stops both.
