sceAgcGetIsTrinityMode takes a pointer to a one-byte flag; the only
known caller ignores the return value. The export took no argument and
never wrote the flag. That caller's flag is in zero-initialised .bss,
so this corrects the signature; it does not fix a visible failure.
The function now stores false, as sceKernelIsTrinityMode reports a base
PS5, returns 0 and throws on a null pointer, whose error code is
unknown. It moves from Unimplemented.cpp to Misc/src/Platform.cpp, and
the new agc_platform_tests ctest target covers it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
EmitDppTargetLane threw "DPP control 336 is not supported" for the
DPP16 row_share controls 0x150-0x15f. row_share:sel now reads lane
sel of the reading lane's own row, (lane & ~15) | sel, as the RDNA2
ISA defines it. The source never leaves the row, so the target is
always valid and an inactive source follows the existing fi and
bound_ctrl rule, as for row_xmask.
The new agc_driver_dpp16_row_share_xmask execution test covers
v_mov_b32, v_add_nc_u32, v_sub_nc_u32 and v_max_f32 with row_share
and row_xmask under a partial exec mask, with fi, bound_ctrl and
row/bank masks, as wave32 and wave64; row_xmask had no test before.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
DecodeState and the draw precheck rejected PA_CL_CLIP_CNTL bit 24
(DX_LINEAR_ATTR_CLIP_ENA), so every draw that sets it was skipped.
Vulkan requires clipped NoPerspective outputs to vary linearly in
framebuffer space, and RADV and radeonsi set this bit for every draw
they emit. With the bit set the console therefore clips the way the
host driver already does, and the pipeline needs no extra state.
Clip space, depth clamping and the viewport depth range decode as
with the bit clear. Every other unsupported bit still rejects the
draw, with or without bit 24.
TechnicalDebt records that a clear bit, which was already accepted,
gets the same clipping.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
ForgetDirectMemory erased the physical pages but left g_directMappings entries pointing at the dead backing, so sceKernelVirtualQuery kept reporting is_direct with a stale physical offset. Erase (and split, on the phys axis) the affected mapping tracking on release; mapped views are untouched, so no new faults are introduced.
Any other live use of the packed register keeps throwing, as on main, instead of
reading SampleId << 8 | Layer << 16 with the other bits zero. The field range
check no longer underflows for an extract that starts past a field's end.
Adds the pixel_ancillary recompiler test.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
Bit-field extracts of the packed ancillary VGPR become gl_SampleID and gl_Layer reads. A pixel
shader reading gl_Layer needs the Geometry capability instead of ShaderLayer and SPIR-V 1.5, and
the Geometry capability is available whenever the geometryShader feature is. 32-bit
SPI_SHADER_Z_FORMATs and 32_ABGR are accepted with the depth and sample-mask export enables.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
sceAjmBatchStart records the batch id it hands out and sceAjmBatchWait
removes it. Waiting on a batch id that was never issued or was already
waited on returns INVALID_BATCH instead of 0, as in shadPS4.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
GLOBAL opcode 0x34 subtracts the data register from a global dword and
clamps the result to zero, returning the previous value. It only exists
in the global segment, so the flat form throws in the decoder, and the
form without glc throws like buffer_atomic_csub does. The clamped
subtract is now a shared helper used by both opcodes.
Single rounding of the whole sum, f32 accumulator denormals flushed, NaN
picked on the bits, clamp ignored and the v_dot2c accumulator read from
the lane itself under DPP, all as measured on hardware.
Co-Authored-By: Claude <[email protected]>
AMD hardware compares a sample_c reference against a color texture's red
channel; Vulkan defines comparison sampling only for depth formats, so the
driver refused any compare-sampled texture that is not R32 float or R16
unorm. PPSA21567 compares the same LESSEQUAL, clamp-to-border (opaque
white) sampler against a shadow map and against a 1x1 RGBA8 texture.
For a compare-sampled image whose format is not natively comparable, the
specialization records the paired sampler's compare function, filter,
addressing and border, and the view's reference clamp kind; the image then
binds as an ordinary color texture. The backend fetches the red channel,
applies wrap, clamp-to-edge or clamp-to-border addressing with the
sampler's texel-centre convention (border taps take the border's red),
compares, and blends four comparisons for bilinear filtering. Real depth
formats keep the native path.
Only sample_c / sample_c_lz on 2D and 2D array views, float/unorm/snorm
formats, wrap, clamp-to-edge or clamp-to-border, point or bilinear at one
level are implemented; half-border, the border color table and everything
else on this path throw. The reference clamp for unorm/snorm color formats
is not documented by AMD and is recorded in the technical debt.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The output modifier was a separate multiply by 2, 4 or 0.5. On NVIDIA
that multiply behaves as if folded into the instruction's own v_mul_f32
despite NoContraction, so a product that overflows comes back finite
after div:2 (FLT_MAX * 2.0 div:2 gave FLT_MAX instead of +inf).
Scale a normal result by adding to or subtracting from its exponent
field instead: mul:2/mul:4 saturate to signed infinity when the exponent
overflows, div:2 keeps the existing signed-zero flush below 2^-125, and
infinities and NaNs pass through unchanged. Over all 2^32 inputs the
results are bit-identical to the previous IEEE multiply and selects.
Titles that list libSceUsbd.prx as a needed module stop in the
startup stub with "Failed to load module: libs\libSceUsbd.prx"
(GetLastError 126), even when they import no function from it.
Add libSceUsbd with an Export.cpp that contains only APS5_DUMMY_FUN,
like libkernel_sys, libkernel_web, libSceAgcVsh, libSceAjm and
libSceNgs2. The module now loads, exports no sceUsbd function and
guesses no signature, so a title that does import one still stops
with "FAIL: unresolved ELF import" for that NID.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
IntrusivePtr<T> is three pointers: object, deleter, LibContext. The default
constructors zero them, the copy constructors copy them and take a reference
on the object, operator-> and operator* return the object, and the destructors
drop a reference and hand the object to the deleter when it was the last one.
The last-reference release of a pointer without a deleter needs the object's
destructor and the LibContext memory pool and still throws.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The constructor was a silent stub that left the guest object
uninitialised and c_str threw. The constructor now writes the 0x20-byte
object, c_str returns the buffer or an empty string, and the destructor
throws for a non-null buffer, whose release needs the LibContext
allocator.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
