strtof and wcstof now set ERANGE for a decimal input whose nonzero
subnormal result is inexact, never for a hexadecimal nonzero subnormal,
and keep it for underflow to zero and overflow. This is what the libc.prx
bundled with PPSA08427 (J03547016) and PPSA03671 (J03848509) do when run
side by side with this libc. The earlier commit followed
libSceLibcInternal's _Stof, which leaves errno alone for hexadecimal under-
and overflow, but titles import these NIDs from their own libc.prx.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01RbLgECLQ2bpDM7oe2urGZK
PS5VR (PPSA99177) reads sysctl {CTL_KERN, KERN_PROC, KERN_PROC_ENV, pid}
for its own pid at start-up and stopped on "unsupported MIB 1.14.35.<pid>".
The call now returns the empty start-up environment that environ and
libc's getenv also begin with, and follows FreeBSD 11's
sysctl_kern_proc_env for the errors: EINVAL for a MIB without exactly one
pid, ESRCH for another pid or -1, EPERM for a write.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Both modules route every decode export through one worker. Its checks
run in this order: state, packet, size, PCM pointer, capacity, then a
priority outside 20..100 (+0xA), then the state check (+1). Decode passes
priority 40.
The float exports decode 16-bit PCM into capacity / 4 * 2 bytes of the
caller's buffer and widen each sample in place, dividing positive
samples by 32767 and the rest by 32768. They return the float byte count.
The OpusDec *AndTimeout exports take a timeout for the AJM batch wait and
a pointer that receives the batch ID only when that wait fails. The host
decode is synchronous, so neither is used; the debt entry records this.
Read from libSceOpusDec.sprx and libSceOpusCeltDec.sprx, firmware 12.00,
not measured on a console.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01SNqSvnFwRBz3D76n476yw6
Name the commands that provide the compiler, CMake and Ninja on macOS
next to the existing mention of AppleClang in the relinker-only section.
Co-Authored-By: Claude Fable 5.1 <[email protected]>
Mode 1 threw for every input since fe4016f61, which stopped Grand Theft Auto V (PPSA04264) right after sceVoiceInit. It now returns 0 for priorities 256-767 with non-zero affinity masks within 0x1fff, the range sceVideoOutOpen accepts for its service thread; no voice worker threads run, so the values reach no host scheduling, and other values still throw.
The trap checks unmask exceptions with feenableexcept and its relatives
and expect MXCSR to hold the new masks, and the enable and disable
results to report the traps enabled on both units. Under Rosetta 2
MXCSR keeps its six exception masks set, so no SSE trap is ever enabled
and every result reports none. The test now asks the host once whether
it keeps an unmasked MXCSR exception and, where it does not, expects
all six SSE masks set and no trap on both units; on other hosts the
expectations are the same as before. A debt entry records that the
trap functions change only the x87 masks under Rosetta.
Under Rosetta 2 the six MXCSR exception masks always read back set: a
write that unmasks an SSE exception is not kept, while the x87 masks and
the other MXCSR bits are. guest_floating_point unmasked the overflow
exception on both units and then expected MXCSR to equal the value it
wrote, so it failed on macOS although fesetexceptflag and feclearexcept
left the masks alone. The test now compares with the environment read
back after the write, which on other hosts is the value written. The
debt entry for the floating-point functions notes the Rosetta behaviour.
main now starts libSceNetCtl initialized, so GuestNetCtl's checks before
sceNetCtlInit expect argument errors; the NpToolkit register, unregister
and check join them with the errors and result their IPv4 counterparts give.
BunnyFl ran firmware 13.42's libSceHmd exports under an emulator. The
distortion and wide-near correction queries look the handle up before
checking initialization: a handle other than the one sceHmdOpen stored,
which stays 0 without a headset, returns HANDLE_INVALID even before
sceHmdInitialize. sceHmdGet2dVrCommand checks its arguments first:
PARAMETER_NULL for a null descriptor or command buffer or a null
pointer at descriptor offset 0, 8 or 0x20, and PARAMETER_INVALID for a
non-zero field at 0x34 or 0x38.
The queries and their sceHmdDistortion* aliases now answer in that
order. What handle 0 gets after initialization isn't measured, so it
throws NotImplemented_nid_no_patch instead of returning HANDLE_INVALID.
Each GCN instruction is re-encoded to its RDNA form and decoded by the
existing RDNA decoder, so the PS4 path shares the recompiler. Covers the
legacy VS stage (program at the VS slot, user SGPRs from s0, vertex and
instance index in v0-v3), fetch shaders linked as user-data calls, MIMG
(image_sample and friends), PS4 buffer descriptors (GCN DATA_FORMAT /
NUM_FORMAT, 44-bit base) and T# image descriptors (GCN tiling in word 7).
The GCN switch is keyed into the shader disk cache. PS5 behaviour is
unchanged: every PS4-only rule is gated on GcnGuestShaders().
A one-sample depth target under a multisampled rasterizer without MSAA_ENABLE is drawn as one sample at the pixel center, but DB_EQAA INTERPOLATE_COMP_Z threw on it. Gran Turismo 7 (PPSA01317) draws its 3840x2160 main pass this way after its EQAA pass (PA_SC_AA_CONFIG 0x0020c002, PA_SC_MODE_CNTL_0 0x2, DB_EQAA 0x00050000, a Z_32_FLOAT depth target with one sample) and stops on it.
On this path every sample is taken to be at the pixel center, so Z interpolated to any sample position is the center's Z, which the one-sample draw already tests and writes. INTERPOLATE_COMP_Z is now accepted there. INTERPOLATE_SRC_Z and overrasterization still throw. Neither PAL nor radeonsi sets the bit and its meaning is not documented, so the debt entry records this as the same inference.
A null or uninitialized attribute, and a kind outside 1 to 4, threw
std::invalid_argument. They now go through NotImplemented_nid_no_patch,
as pthread_rwlockattr_settype_np's unmeasured types do. FreeBSD 11
libthr stores an unknown kind unchecked and answers a null attribute
with -1 and errno EINVAL, so scePthreadMutexattrSettype's
SCE_KERNEL_ERROR_EINVAL would match neither answer.
hasStructuredConstructs tested every block of the graph against each merge header, and searched every block for a block's innermost loop, with dominator binary searches each time. Under APS5_FORCE_STRUCTURIZER_FALLBACK, Ghostrunner's compute shader 0x200c520000 becomes a 32,490-block reach-flag graph with 6,647 merge headers, and the check took 44 to 50 s of the 49 to 55 s.
The innermost loop is now searched among the loop headers only. Every other block failed the loop-header test first, so the candidates and their order are unchanged.
When constructsFollowDominance holds for the structural graph, the member checks visit only the blocks dominatedBlocks finds from the header, stopping at its merge, and for a loop also from its continue block, in block order. Only a block in the header's construct or in the loop's continue construct can fail a member check. A construct block is dominated by the header and not by the merge, so some path from the entry reaches it without the merge, and from the last header on that path every block is dominated by the header, so the walk reaches it. A continue-construct block is dominated by the continue block and is reached from it the same way. The walk finds every such block, and the blocks it skips passed every check, so the result is unchanged. Otherwise every block is checked as before, and APS5_NO_PERF_DOMINATOR_TREE keeps the old loop.
On an M4 Pro the check takes 0.27 s instead of 44 to 50 s on that graph, and the forced fallback takes 5.1 to 5.4 s instead of 49 to 55 s. The default structurization of the shader makes one check, which takes 0.03 s instead of 0.36 s.
recomputeAnalyses runs after every merge split, and its dominator and post-dominator fixpoints intersected a bitset of every block for every block until nothing changed. They now build the tree with the Cooper-Harvey-Kennedy iteration over a reverse post-order from a virtual root above the entry block, or above every exit for post-dominators, and write each block's sorted set as its parent's set plus the block itself.
The sets are the same. When every block is reachable from the virtual root, the greatest fixpoint of the old equations is the dominance relation of that rooted graph, which the tree describes. When a block is unreachable from the entry, or cannot reach an exit, the old fixpoint runs as before.
The post-dominator tree also keeps each block's immediate post-dominator, and FindNearestCommonPostDominator climbs it when both blocks have one: the sets are the tree's paths, so the deepest common block is the one the set intersection found. Blocks added or cloned after the analysis have no parent and take the set path, as before.
APS5_NO_PERF_DOMINATOR_TREE turns both off, and the structurizer and control-flow tests also run with it set. On an M4 Pro the structurization of Ghostrunner's compute shader 0x200c520000 takes 18 s instead of 37 s.
Federico Marras