A float inline constant (240-248) read by a 64-bit operand is the f64 bit
pattern for integer B64 opcodes too: AMD's compilers emit
v_lshrrev_b64 v[4:5], v0, 1.0 for 0x3ff0000000000000 >> lane. ReadU32Pair
returned {f32 bits, 0}, so only F64-typed reads had the right value.
Based on TheCruZ's 284deab28 (closed#1219).
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
V_PK_MIN_F16 and V_PK_MAX_F16 return the quieted NaN for a signaling NaN
operand and the other operand for a quiet one (per half, as V_MIN_F16 and
V_MAX_F16 do). The signaling test masked the half with 0x7c00 and checked
for a nonzero payload, so a quiet NaN with payload bits such as 0x7e01 was
taken as signaling and returned. Mask with 0x7e00 so the quiet bit must be
clear.
Two more vectors in PackedMinMaxF16NanAndSignedZeroEdges cover it.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The 16-bit encodings of the float inline constants are fixed: the exact
halves of 0.5, 1.0, 2.0 and 4.0 and their negations, and 0x3118 for
1/(2*pi). Read16LaneBits converted the 32-bit value at run time instead,
through PackHalf2x16, which the AMD driver rounds toward zero, so
1/(2*pi) became 0x3117. Use the encodings directly.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* shader: support DS_CMPST_RTN_B32
Assassin's Creed Valhalla (PPSA01490, #281) stops on DS opcode 0x30 in a
compute shader. DS compares DATA and stores DATA2, the opposite of
BUFFER_ATOMIC_CMPSWAP, so the 32-bit atomic path swaps the two for this
opcode, as the 64-bit form does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* shader: reject unsupported GDS compare-store variant
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: nmzik <Nmzik@mail.ru>
A patched TLS load calls TlsRegStub and SafeCall, which moved RSP past the
red zone with sub/add rsp,0x80. ADD and SUB write the arithmetic flags, so
the sub ran before SafeCall's pushfq saved them and the add ran after its
popfq restored them: guest code that compares, loads fs:[0] and then
branches on the flags could branch the wrong way. lea moves RSP without
touching the flags. The entry lea plus a two-byte nop keeps every patched
offset in both stubs; the exit lea is one byte longer and moves only ret.
TlsStubsPreserveFlags calls the stubs with two flag patterns and checks
the status flags, the output register and RAX; it fails with the old
bytes.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* tests: cover the 16-bit integer VOPC compares
V_CMP_{EQ,LE,GT,NE}_I16 and V_CMP_{EQ,LE,GT,NE,GE}_U16 decode and translate but
no test exercised them. VectorCompareI16U16Ops compares low halves that
differ in sign interpretation (0x8000 vs 0x7fff) with both the signed and the
unsigned opcodes, and uses operands whose high halves differ to check that only
the low 16 bits are compared. It runs in the default list and with
--int16-compare-only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* tests: cover the 16-bit integer VOP3 ALU ops
V_SUB_NC_U16, V_ADD_NC_I16, V_ASHRREV_I16, V_MAX_U16 and V_MIN_U16 decode and
translate but no test exercised them. Vop3Int16Ops checks wraparound without
clamp, the shift amount taken from the low four bits of src0, unsigned
ordering for max/min on 0x8000 vs 0x7fff, that the high halves of the sources
are ignored and that the high half of the destination is preserved. It runs in
the default list and with --int16-alu-only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* tests: cover the packed 16-bit integer VOP3P ops
V_PK_MAD_I16, V_PK_MAD_U16, V_PK_MUL_LO_U16, V_PK_ADD_I16, V_PK_SUB_U16,
V_PK_LSHLREV_B16, V_PK_LSHRREV_B16, V_PK_ASHRREV_I16, V_PK_MIN_I16, V_PK_MAX_U16
and V_PK_MIN_U16 decode and translate but no test exercised them. Vop3pInt16Ops
runs each on both halves with the default half selection: wraparound without
clamp, shift amounts from the low four bits of each src0 half, and min/max on
0x8000 vs 0x0001 so signed and unsigned ordering give different results. It
runs in the default list and with --int16-alu-only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
TestConcurrentBackingReads (7e011fa5) builds its reader pairs with
std::pair {TryReadBacking, TryReadSparseBacking}. Class template argument
deduction then also considers pair's const T1& / const T2& constructors, which
deduce function types, and libstdc++ and libc++ reject a pair member of
function type ("data member instantiated with function type"), so
virtual_memory_allocation_tests no longer compiles on Linux and macOS. MSVC's
library accepts it, which is why Windows still builds.
Take the functions' addresses so the pairs hold function pointers.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* kernel: apply clang-format to pthread.cpp
Output of the repository's pre-commit clang-format hook (v22.1.3) on
this file; no code change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* kernel: fix the macOS build of host thread naming
199d0a9d (#1106) names guest threads on the host with the two-argument
glibc pthread_setname_np(thread, name). KYTY_PLATFORM is
KYTY_PLATFORM_LINUX on macOS too, so the macOS job now fails with "no
matching function for call to 'pthread_setname_np'": Apple's version
takes only the name and renames the calling thread. This code already
runs on the new thread itself (it passes pthread_self()), so use the
one-argument form on Apple with Apple's 63-character limit and keep the
Linux call unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Remove formatting
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: nmzik <Nmzik@mail.ru>
CMake 4 warns "Compatibility with CMake < 3.10 will be removed" for every
fetched project that declares an older minimum. zydis and its bundled
zycore both declare cmake_minimum_required(VERSION 3.9), so each configure
prints two deprecation warnings. The existing floor of 3.5 does not lift
them past the threshold; 3.10 does.
Existing build directories keep the cached 3.5 until reconfigured with
-U CMAKE_POLICY_VERSION_MINIMUM or a fresh cache.
Tested on Windows 11, clang-cl 19.1.5, CMake 4.4.3: configure prints 0
deprecation warnings (2 with the old floor); virtual_memory_allocation_tests
builds and links zydis/xbyak; ctest guest_red_zone_patcher passes.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Nawid3333andnmzik
