An indirect dispatch with USE_THREAD_DIMENSIONS read its thread counts on the
CPU, which waits for the recorded work that stores them. The argument kernel now
turns them into group counts on the GPU and copies the counts into the
ShaderData thread limit, so the dispatch stays on the GPU path.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The text filter follows KytyPS5's contract; the external keyboard filter is
stored on the session and is not called for the host input path, which delivers
no key data.
Replace the 0x80B8xxxx common-dialog codes with the module's 0x8134xxxx
codes, validate the sceLoginDialogOpen param, and write the module's
result words from sceLoginDialogGetResult.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A second sceVoiceInit returns 0x804E0802, and sceVoiceStart,
sceVoiceCreatePort and sceVoiceGetPortInfo return 0x804E0801 before
checking any argument while the library is not initialized, as the
shipped module does.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Add edge case test rows in guest_integer_builtins for zero shifts, cross-word shift boundaries, zero bitcounts, comparison symmetry, arithmetic identities, division boundary and wrapping cases.
A short AMD-only instruction takes the following instructions into its stub until its site fits a jump. The loop did not check which instructions an earlier 256-bit store island had already claimed, so with --to-rosetta two trampolines covered the same bytes: for je; jmp; vmovups [rdi],ymm0; rcpps xmm1,xmm2 followed by adds, the RCPPS site at +8 overlapped the island at +11, the patched code crashed, and the real patchers failed with 'AMD-only site bytes changed before patching'. A claimed follower now ends the attempt: an optional instruction such as RCPPS or RDPID is kept, and any other instruction fails with an error naming the moved follower.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
A short AMD-only instruction lowered through a stub takes the following instructions with it. When one of them is an AMD-only instruction that is replaced in place (MOVNTSS, MOVNTSD, MONITORX, MWAITX, and RDSEED and CLWB for Rosetta), MatchSequence found no out-of-line lowering for it and the conversion failed with 'AMD-only instruction sequence has no out-of-line lowering', for example on extrq xmm0,xmm1 followed by movntss [rdi],xmm0. A following instruction with an in-place replacement now has that replacement moved into the stub, with its RIP-relative displacement relocated. Out-of-line emission for 256-bit store windows is unchanged.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
sceJpegDecParseHeader and sceJpegDecDecode threw for every down scale
other than 1. SharpProspero's JpegDec.cs documents the factors 1, 2, 4
and 8 and output dimensions that account for the down scale.
Decoder::Jpeg gains DecodeScaled, which decodes through libjpeg-turbo's
scaled inverse DCT, and ScaledDimension, its rounded-up output size.
sceJpegDecParseHeader reports the scaled output size; sceJpegDecDecode
checks the pitch and image buffer against it, writes the scaled image
and packs the output size into its result. A down scale of 1 still
decodes as before. Decode modes other than 0, other down scales and a
down-scaled CMYK or YCCK image still throw.
getifaddrs listed only the host's IPv4 and IPv6 addresses, so the
per-interface AF_LINK entries FreeBSD's getifaddrs puts first were
missing, and with them the interface index, hardware address, MTU and
counters and every interface without an address.
On Linux each interface now starts with an AF_LINK entry, followed by
its addresses, as FreeBSD 11.4's getifaddrs orders them. Its address is
a sockaddr_dl with the host index, type, name and hardware address
(none for a loopback), sized as FreeBSD's if_attach sizes it. Its
ifa_data is a 152-byte FreeBSD if_data filled as if_data_copy fills it:
type, address and header length, link state, MTU, speed and the host's
packet, byte, error, drop and multicast counters from sysfs, or the
host list's 32-bit counters when sysfs cannot be read. Fields with no
host equivalent are 0 and listed in the debt entry. Windows is
unchanged.
Write the unattenuated linear gain 1.0f into *gain before checking the port handle in sceAudioInGetGain, and return 0 for open ports or SCE_AUDIO_IN_ERROR_INVALID_HANDLE (0x80260101) for closed or out-of-range handles in both sceAudioInGetGain and sceAudioInGetRerouteCount.
guest_sce_net skips its IPv6 checks when sending to ::1 fails and
HostIpv6LoopbackUnavailable confirms the host cannot reach it. The host
probe accepts EADDRNOTAVAIL, ENETUNREACH and EHOSTUNREACH, but the guest
check accepted only ENETUNREACH (51). On a Linux host without ::1 the
send can fail with EADDRNOTAVAIL instead, which the guest sees as 49,
and the test aborted in CheckUnspecifiedIpv6 instead of skipping.
Accept the guest numbers of the same three errors (49, 51 and 65).
MinGW's std::filesystem::equivalent returns false for hard links whose
names give different permissions, such as .elf and .exe. The temporary
hard-link probe that worked around it failed on read-only files and on
directories that refuse the link. Compare the volume serial number and
file index from GetFileInformationByHandle instead.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
On RADV, importing host memory and reading it from the GPU marks the
pages as written in the write watch. The test forces
APS5_WRITE_WATCH_IMPORTS=watch, so the vouch refuses the table and
supplyWords is never called. Probe host-import write protection on
Linux, as the Windows path already does, and skip when a GPU read
reports writes or the probe fails.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Yuri Olive