The vertex entry set only the vertex and instance index VGPRs, leaving
the NGG merged wave info SGPR s3 at 0. NGG shaders build EXEC from it
(s_bfe_u64 exec, -1, <count in bits 16..22>), so Mesa's NGG vertex
shaders ran with an empty EXEC and rasterized nothing.
Each host invocation on the vertex path is one vertex of a full wave:
when the user data starts at s8 or later, set s3 to the wave size as
both vertex and primitive count, with wave index 0.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Store: negative, -0, NaN (quiet and low-payload), +-Inf, overflow to the
largest finite value, truncation and small-float denormals. Load: Inf,
denormal and largest finite codes. Values measured on gfx1035.
Co-Authored-By: Claude <[email protected]>
Format 43 packs R as a 10-bit and G and B as 11-bit unsigned floats, which
Vulkan has no format for. Keep it in an R32_UINT image like the other
packed formats without a host equivalent, and convert texels in the shader
on image loads and stores; sampling it still throws.
When one guest sampler is paired both with an image that needs point
filtering (sint, depth-bits or converted format) and with an ordinary
image, ResourceMaterializer::Apply appends a point-filtered copy of the
sampler after the original samplers. The resource snapshot only holds
the original samplers, so DescriptorBindingBuilder::Populate threw
"guest sampler index is out of range" for the copy, both in the sampler
binding and in the unnormalized-coordinate check.
SamplerResource::copyOf records the sampler a copy was made from, and
the guest sampler descriptor of a copy is read from that sampler; the
point-filter rewrite of word 2 still applies to the copy. The field is
part of the disk cache encoding: cache format version 8.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The integer emulation of the round-to-nearest-even float-to-half
conversion is exact but costly: a compute shader with about 3600
conversions never finished compiling on NVIDIA. When the device enables
shaderFloat16 and guarantees shaderRoundingModeRTEFloat16,
shaderDenormPreserveFloat16 and shaderSignedZeroInfNanPreserveFloat16,
it advertises Float16, RoundingModeRTE and DenormPreserve, and the
recompiler emits OpFConvert under the 16-bit RoundingModeRTE,
DenormPreserve and SignedZeroInfNanPreserve execution modes instead.
NaNs keep the emulated quiet NaN with the high payload bits. Devices
without these guarantees keep the emulation.
Graphics validation accepts the RoundingModeRTE, DenormPreserve and
SignedZeroInfNanPreserve execution modes for 16-, 32- and 64-bit floats,
so vertex and pixel shaders with a native conversion still draw.
Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
On GFX10.3 to GFX11.7, ACO lowers a subgroup-scope OpReadClockKHR to
s_getreg_b32 of SHADER_CYCLES, 20 bits zero-extended to 64, so s_memtime
wrapped every 2^20 cycles and went backwards. Detect those families from
the RADV device name and read the device clock there instead.
Fixes#985
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Resolve ASCII case differences in guest paths and mount aliases while preferring exact names and rejecting ambiguous fallback matches. Preserve new filenames and host roots. Five filesystem tests pass; Skyrim now loads its INI and shader archives, initializes Vulkan and uploads intro textures instead of dereferencing null shader metadata. Written with AI assistance.
Titles pass handle 0 (RTLD_DEFAULT) for a process-wide lookup, and the loader rejected it: the handle is
not in the module map, so every such lookup failed even when a loaded module exports the symbol. Handle 0
and -2 now take the default scope, searching the global modules by plain name, NID and `#guest` form, and
then the process-wide scope.
The process-wide scope of the real loader also contains glibc and libstdc++, so an unfiltered fallback
would answer `malloc`, `dlopen` and `pthread_create` where PS5 reports the symbol as missing: handle 0 is
the main program, whose lookup sees the program's exports, not the host C++ runtime. A result from the
process scope is therefore accepted only when `dladdr` attributes it to a converted or replacement module
(a `.prx` library or the executable), which keeps the fallback for guest symbols and removes the host
leak. `guest_dynamic_loader` asserts the three host names are not found; disabling the filter makes that
test fail.
The scripting heap that an earlier revision of this branch added has been dropped: the title reaches the
same boot milestone with the lookup failing honestly, so the export was an inference rather than a
requirement, and prosper's bring-up notes report `ESRCH` for that name on hardware.
DT_JMPREL and DT_PLTRELSZ were never range-checked against the input
file. With DT_OS_JMPREL the offset is used raw, so an offset outside the
image reached BuildDynamicSection and CompactPlt. When DT_PLTRELSZ is
zero the per-entry loops never run, so nothing rejected the table and
the relink produced an output executable.
Validate the declared range with the subtraction-based check already
used for DT_STRTAB and DT_STRSZ, next to it and before the table is
read.
A static branch into a skipped function tail is accepted when an
instruction decodes there within its executable segment, so a range that
ends inside an instruction no longer stops code analysis.
The Windows TLS rewriter skips a non-FS segment-prefixed instruction that
starts inside an earlier collected instruction: it is an alternative
decoding from a range that begins mid-instruction, not a TLS access.
A draw whose PA_SC_CONSERVATIVE_RASTERIZATION_CNTL (context 0x313) was
not the default 0x6000 threw in DecodeState and the draw precheck.
0x6001 sets OVER_RAST_ENABLE and now decodes to Vulkan's overestimation
mode, chained to the pipeline's rasterization state and its key.
The device enables VK_EXT_conservative_rasterization only when its
primitiveOverestimationSize is at most 1/256 pixel and it rasterizes
degenerate triangles; elsewhere the pipeline throws. Overestimation is
accepted only when triangles reach the rasterizer: VGT_GS_OUT_PRIM_TYPE
behind a geometry shader, VGT_TF_PARAM's output topology behind
tessellation, VGT_PRIMITIVE_TYPE otherwise. Other 0x313 values and
centroid interpolation still throw.
Adds the agc_driver_conservative_rasterization GPU test and extends
agc_driver_graphics.
Adrià