* graphics: give InlinePageOwnerList's const_iterator iterator traits
The Linux build failed in textureCache.cpp: std::any_of over the image
owners of a page needs std::iterator_traits, which libstdc++ only fills
from the iterator's own member types. MSVC's STL and libc++ don't
consult them there, so Windows and macOS built.
Declare the five member types. The category is input, not forward,
because operator* returns a copy.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* tests: assert InlinePageOwnerList's const_iterator has an input category
Only libstdc++ needs the iterator traits for std::any_of, so removing
them would compile on Windows and macOS and break only the Linux build.
The assert fails the build on every platform instead.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* tests: serve SRT walker test loads on macOS too
FakeGuestMemory only installed its fault handler on Linux and Windows,
so on macOS the walker's deliberately faulting loads crashed
scalar_provenance, resource_materialization and resource_tracking.
Share the Linux sigaction path with macOS, read the registers from the
macOS thread state, and catch SIGBUS as well as SIGSEGV there, like the
emulator's own handler.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
DecodePackedColorClear had no case for R8_UNORM/UINT, R8G8_UNORM/UINT and
R16_UINT, so a CMASK fast clear of such a render target was dropped and the
target kept its old contents. UFC 6 (PPSA23566) fast-clears an R8_UINT
region map this way every frame and then writes only part of it; a capture
showed the map never cleared. DWORD fills of these formats are decoded only
when every pixel in the word holds the same value.
Co-authored-by: 4wgycvn84p-afk <346802885+4wgycvn84p-afk@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* image: request extended usage only for storage images
* tests: restore block-texel-view support through the flag check loop order
---------
Co-authored-by: nmzik <Nmzik@mail.ru>
* shader: emit float min/max as host instructions on AMD GPUs
V_MIN_F32, V_MAX_F32, V_MIN3_F32 and V_MAX3_F32 were emulated with a
classify-and-select sequence of about a dozen instructions each, to give
the guest's NaN and signed-zero results on any host. AMD GPUs execute
GLSL.std.450 NMin/NMax with those same rules, so emit them directly when
the device is AMD and supports VK_KHR_shader_float_controls2. Each
instruction is decorated FPFastMathMode None, since
SignedZeroInfNanPreserve does not cover extended instructions. MED3 keeps
its exact emulation. KYTY_EMULATED_FLOAT_MIN_MAX=1 keeps the emulation.
In SAROS the two heaviest compute shaders get 37% and 59% faster on an
RX 9070 XT (RADV), about 4 ms of GPU time per frame.
Adds VectorMinMaxF32RuntimeEdges (runtime operands, so nothing is folded)
and a --float-min-max-only mode that runs it and
VectorMinMaxF32NanAndSignedZeroEdges with the emulation and, where
supported, with host instructions.
Developed with AI assistance (Claude Code).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* shader: gate host float min/max on RADV and check the emitted instructions
* shader: drop float controls 2 from host float min/max
* shader: drop the float min/max env override and atomic flag
* shader: pass host capabilities to the SPIR-V emitter as a profile built once per device
* shader: enable host float min/max on AMD GPUs
---------
Co-authored-by: arda <ardaeska@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: nmzik <Nmzik@mail.ru>
nmzikandstrxno
Mr. Big BritchesandClaude Opus 5.5
Arda Sahinandnmzik
