@cpp-linter/clang-tidy-win32-x64 19.1.0 → 23.1.3

package/README.md

@@ -1,7 +1,7 @@
 # @cpp-linter/clang-tidy-win32-x64
 
-The Windows x64 binary of clang-tidy from LLVM 19.1.0, for [`@cpp-linter/clang-tidy`](https://www.npmjs.com/package/@cpp-linter/clang-tidy).
+The Windows x64 binary of clang-tidy from LLVM 23.1.3, for [`@cpp-linter/clang-tidy`](https://www.npmjs.com/package/@cpp-linter/clang-tidy).
 
 Install `@cpp-linter/clang-tidy` instead of this package: npm picks the binary for your platform.
 
-The binary is a static build from the release [`llvm-19.1.0-3`](https://github.com/cpp-linter/clang-tools-static-binaries/releases/tag/llvm-19.1.0-3) of [cpp-linter/clang-tools-static-binaries](https://github.com/cpp-linter/clang-tools-static-binaries). LLVM is licensed under the Apache License v2.0 with LLVM Exceptions (see `LICENSE.TXT`).
+The binary is a static build from the release [`llvm-23.1.3-3`](https://github.com/cpp-linter/clang-tools-static-binaries/releases/tag/llvm-23.1.3-3) of [cpp-linter/clang-tools-static-binaries](https://github.com/cpp-linter/clang-tools-static-binaries). LLVM is licensed under the Apache License v2.0 with LLVM Exceptions (see `LICENSE.TXT`).

package/bin/clang-tidy.exe

@@ -1,4 +1,4 @@
 [diffend] Oversized file quarantined before diffing.
 name: package/bin/clang-tidy.exe
-size: 63233536 bytes
-sha256: a2124d24b978fce834697c057b2a1afbe42043f9f92a03dcaf41df0c25f267c0
+size: 80119296 bytes
+sha256: a4285331cc4d87e49ef715059e4da27f2ad1aa626e95f9af804061924923b416

package/lib/clang/23/include/__clang_cuda_complex_builtins.h

@@ -0,0 +1,236 @@
+/*===-- __clang_cuda_complex_builtins - CUDA impls of runtime complex fns ---===
+ *
+ * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+ * See https://llvm.org/LICENSE.txt for license information.
+ * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+ *
+ *===-----------------------------------------------------------------------===
+ */
+
+#ifndef __CLANG_CUDA_COMPLEX_BUILTINS
+#define __CLANG_CUDA_COMPLEX_BUILTINS
+
+// This header defines __muldc3, __mulsc3, __divdc3, and __divsc3.  These are
+// libgcc functions that clang assumes are available when compiling c99 complex
+// operations.  (These implementations come from libc++, and have been modified
+// to work with CUDA and OpenMP target offloading [in C and C++ mode].)
+
+#pragma push_macro("__DEVICE__")
+#if defined(__OPENMP_NVPTX__) || defined(__OPENMP_AMDGCN__) ||                 \
+    defined(__OPENMP_SPIRV__)
+#pragma omp declare target
+#define __DEVICE__ __attribute__((noinline, nothrow, cold, weak))
+#else
+#define __DEVICE__ __device__ inline
+#endif
+
+#if defined(__NVPTX__)
+// FIXME: NVPTX should use generic builtins.
+#define _SCALBNd __nv_scalbn
+#define _SCALBNf __nv_scalbnf
+#define _LOGBd __nv_logb
+#define _LOGBf __nv_logbf
+#elif defined(__OPENMP_SPIRV__)
+#define _SCALBNd __spirv_ocl_ldexp
+#define _SCALBNf __spirv_ocl_ldexp
+#define _LOGBd __spirv_ocl_logb
+#define _LOGBf __spirv_ocl_logb
+#else
+#define _SCALBNd __builtin_scalbn
+#define _SCALBNf __builtin_scalbnf
+#define _LOGBd __builtin_logb
+#define _LOGBf __builtin_logbf
+#endif
+
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
+__DEVICE__ double _Complex __muldc3(double __a, double __b, double __c,
+                                    double __d) {
+  double __ac = __a * __c;
+  double __bd = __b * __d;
+  double __ad = __a * __d;
+  double __bc = __b * __c;
+  double _Complex z;
+  __real__(z) = __ac - __bd;
+  __imag__(z) = __ad + __bc;
+  if (__builtin_isnan(__real__(z)) && __builtin_isnan(__imag__(z))) {
+    int __recalc = 0;
+    if (__builtin_isinf(__a) || __builtin_isinf(__b)) {
+      __a = __builtin_copysign(__builtin_isinf(__a) ? 1 : 0, __a);
+      __b = __builtin_copysign(__builtin_isinf(__b) ? 1 : 0, __b);
+      if (__builtin_isnan(__c))
+        __c = __builtin_copysign(0, __c);
+      if (__builtin_isnan(__d))
+        __d = __builtin_copysign(0, __d);
+      __recalc = 1;
+    }
+    if (__builtin_isinf(__c) || __builtin_isinf(__d)) {
+      __c = __builtin_copysign(__builtin_isinf(__c) ? 1 : 0, __c);
+      __d = __builtin_copysign(__builtin_isinf(__d) ? 1 : 0, __d);
+      if (__builtin_isnan(__a))
+        __a = __builtin_copysign(0, __a);
+      if (__builtin_isnan(__b))
+        __b = __builtin_copysign(0, __b);
+      __recalc = 1;
+    }
+    if (!__recalc && (__builtin_isinf(__ac) || __builtin_isinf(__bd) ||
+                      __builtin_isinf(__ad) || __builtin_isinf(__bc))) {
+      if (__builtin_isnan(__a))
+        __a = __builtin_copysign(0, __a);
+      if (__builtin_isnan(__b))
+        __b = __builtin_copysign(0, __b);
+      if (__builtin_isnan(__c))
+        __c = __builtin_copysign(0, __c);
+      if (__builtin_isnan(__d))
+        __d = __builtin_copysign(0, __d);
+      __recalc = 1;
+    }
+    if (__recalc) {
+      // Can't use std::numeric_limits<double>::infinity() -- that doesn't have
+      // a device overload (and isn't constexpr before C++11, naturally).
+      __real__(z) = __builtin_huge_val() * (__a * __c - __b * __d);
+      __imag__(z) = __builtin_huge_val() * (__a * __d + __b * __c);
+    }
+  }
+  return z;
+}
+
+__DEVICE__ float _Complex __mulsc3(float __a, float __b, float __c, float __d) {
+  float __ac = __a * __c;
+  float __bd = __b * __d;
+  float __ad = __a * __d;
+  float __bc = __b * __c;
+  float _Complex z;
+  __real__(z) = __ac - __bd;
+  __imag__(z) = __ad + __bc;
+  if (__builtin_isnan(__real__(z)) && __builtin_isnan(__imag__(z))) {
+    int __recalc = 0;
+    if (__builtin_isinf(__a) || __builtin_isinf(__b)) {
+      __a = __builtin_copysignf(__builtin_isinf(__a) ? 1 : 0, __a);
+      __b = __builtin_copysignf(__builtin_isinf(__b) ? 1 : 0, __b);
+      if (__builtin_isnan(__c))
+        __c = __builtin_copysignf(0, __c);
+      if (__builtin_isnan(__d))
+        __d = __builtin_copysignf(0, __d);
+      __recalc = 1;
+    }
+    if (__builtin_isinf(__c) || __builtin_isinf(__d)) {
+      __c = __builtin_copysignf(__builtin_isinf(__c) ? 1 : 0, __c);
+      __d = __builtin_copysignf(__builtin_isinf(__d) ? 1 : 0, __d);
+      if (__builtin_isnan(__a))
+        __a = __builtin_copysignf(0, __a);
+      if (__builtin_isnan(__b))
+        __b = __builtin_copysignf(0, __b);
+      __recalc = 1;
+    }
+    if (!__recalc && (__builtin_isinf(__ac) || __builtin_isinf(__bd) ||
+                      __builtin_isinf(__ad) || __builtin_isinf(__bc))) {
+      if (__builtin_isnan(__a))
+        __a = __builtin_copysignf(0, __a);
+      if (__builtin_isnan(__b))
+        __b = __builtin_copysignf(0, __b);
+      if (__builtin_isnan(__c))
+        __c = __builtin_copysignf(0, __c);
+      if (__builtin_isnan(__d))
+        __d = __builtin_copysignf(0, __d);
+      __recalc = 1;
+    }
+    if (__recalc) {
+      __real__(z) = __builtin_huge_valf() * (__a * __c - __b * __d);
+      __imag__(z) = __builtin_huge_valf() * (__a * __d + __b * __c);
+    }
+  }
+  return z;
+}
+
+__DEVICE__ double _Complex __divdc3(double __a, double __b, double __c,
+                                    double __d) {
+  int __ilogbw = 0;
+  // Can't use std::max, because that's defined in <algorithm>, and we don't
+  // want to pull that in for every compile.  The CUDA headers define
+  // ::max(float, float) and ::max(double, double), which is sufficient for us.
+  double __logbw =
+      _LOGBd(__builtin_fmax(__builtin_fabs(__c), __builtin_fabs(__d)));
+  if (__builtin_isfinite(__logbw)) {
+    __ilogbw = (int)__logbw;
+    __c = _SCALBNd(__c, -__ilogbw);
+    __d = _SCALBNd(__d, -__ilogbw);
+  }
+  double __denom = __c * __c + __d * __d;
+  double _Complex z;
+  __real__(z) = _SCALBNd((__a * __c + __b * __d) / __denom, -__ilogbw);
+  __imag__(z) = _SCALBNd((__b * __c - __a * __d) / __denom, -__ilogbw);
+  if (__builtin_isnan(__real__(z)) && __builtin_isnan(__imag__(z))) {
+    if ((__denom == 0.0) && (!__builtin_isnan(__a) || !__builtin_isnan(__b))) {
+      __real__(z) = __builtin_copysign(__builtin_huge_val(), __c) * __a;
+      __imag__(z) = __builtin_copysign(__builtin_huge_val(), __c) * __b;
+    } else if ((__builtin_isinf(__a) || __builtin_isinf(__b)) &&
+               __builtin_isfinite(__c) && __builtin_isfinite(__d)) {
+      __a = __builtin_copysign(__builtin_isinf(__a) ? 1.0 : 0.0, __a);
+      __b = __builtin_copysign(__builtin_isinf(__b) ? 1.0 : 0.0, __b);
+      __real__(z) = __builtin_huge_val() * (__a * __c + __b * __d);
+      __imag__(z) = __builtin_huge_val() * (__b * __c - __a * __d);
+    } else if (__builtin_isinf(__logbw) && __logbw > 0.0 &&
+               __builtin_isfinite(__a) && __builtin_isfinite(__b)) {
+      __c = __builtin_copysign(__builtin_isinf(__c) ? 1.0 : 0.0, __c);
+      __d = __builtin_copysign(__builtin_isinf(__d) ? 1.0 : 0.0, __d);
+      __real__(z) = 0.0 * (__a * __c + __b * __d);
+      __imag__(z) = 0.0 * (__b * __c - __a * __d);
+    }
+  }
+  return z;
+}
+
+__DEVICE__ float _Complex __divsc3(float __a, float __b, float __c, float __d) {
+  int __ilogbw = 0;
+  float __logbw =
+      _LOGBf(__builtin_fmaxf(__builtin_fabsf(__c), __builtin_fabsf(__d)));
+  if (__builtin_isfinite(__logbw)) {
+    __ilogbw = (int)__logbw;
+    __c = _SCALBNf(__c, -__ilogbw);
+    __d = _SCALBNf(__d, -__ilogbw);
+  }
+  float __denom = __c * __c + __d * __d;
+  float _Complex z;
+  __real__(z) = _SCALBNf((__a * __c + __b * __d) / __denom, -__ilogbw);
+  __imag__(z) = _SCALBNf((__b * __c - __a * __d) / __denom, -__ilogbw);
+  if (__builtin_isnan(__real__(z)) && __builtin_isnan(__imag__(z))) {
+    if ((__denom == 0) && (!__builtin_isnan(__a) || !__builtin_isnan(__b))) {
+      __real__(z) = __builtin_copysignf(__builtin_huge_valf(), __c) * __a;
+      __imag__(z) = __builtin_copysignf(__builtin_huge_valf(), __c) * __b;
+    } else if ((__builtin_isinf(__a) || __builtin_isinf(__b)) &&
+               __builtin_isfinite(__c) && __builtin_isfinite(__d)) {
+      __a = __builtin_copysignf(__builtin_isinf(__a) ? 1 : 0, __a);
+      __b = __builtin_copysignf(__builtin_isinf(__b) ? 1 : 0, __b);
+      __real__(z) = __builtin_huge_valf() * (__a * __c + __b * __d);
+      __imag__(z) = __builtin_huge_valf() * (__b * __c - __a * __d);
+    } else if (__builtin_isinf(__logbw) && __logbw > 0 &&
+               __builtin_isfinite(__a) && __builtin_isfinite(__b)) {
+      __c = __builtin_copysignf(__builtin_isinf(__c) ? 1 : 0, __c);
+      __d = __builtin_copysignf(__builtin_isinf(__d) ? 1 : 0, __d);
+      __real__(z) = 0 * (__a * __c + __b * __d);
+      __imag__(z) = 0 * (__b * __c - __a * __d);
+    }
+  }
+  return z;
+}
+
+#if defined(__cplusplus)
+} // extern "C"
+#endif
+
+#undef _SCALBNd
+#undef _SCALBNf
+#undef _LOGBd
+#undef _LOGBf
+
+#if defined(__OPENMP_NVPTX__) || defined(__OPENMP_AMDGCN__) ||                 \
+    defined(__OPENMP_SPIRV__)
+#pragma omp end declare target
+#endif
+
+#pragma pop_macro("__DEVICE__")
+
+#endif // __CLANG_CUDA_COMPLEX_BUILTINS

package/lib/clang/{19 → 23}/include/__clang_cuda_device_functions.h

@@ -528,7 +528,21 @@ __DEVICE__ float __tanf(float __a) { return __nv_fast_tanf(__a); }
 __DEVICE__ void __threadfence(void) { __nvvm_membar_gl(); }
 __DEVICE__ void __threadfence_block(void) { __nvvm_membar_cta(); };
 __DEVICE__ void __threadfence_system(void) { __nvvm_membar_sys(); };
-__DEVICE__ void __trap(void) { __asm__ __volatile__("trap;"); }
+__DEVICE__ __attribute__((noreturn)) void __trap(void) { __builtin_trap(); }
+__DEVICE__ unsigned short
+__usAtomicCAS(unsigned short *__p, unsigned short __cmp, unsigned short __v) {
+  return __nvvm_atom_cas_gen_us(__p, __cmp, __v);
+}
+__DEVICE__ unsigned short __usAtomicCAS_block(unsigned short *__p,
+                                              unsigned short __cmp,
+                                              unsigned short __v) {
+  return __nvvm_atom_cta_cas_gen_us(__p, __cmp, __v);
+}
+__DEVICE__ unsigned short __usAtomicCAS_system(unsigned short *__p,
+                                               unsigned short __cmp,
+                                               unsigned short __v) {
+  return __nvvm_atom_sys_cas_gen_us(__p, __cmp, __v);
+}
 __DEVICE__ unsigned int __uAtomicAdd(unsigned int *__p, unsigned int __v) {
   return __nvvm_atom_add_gen_i((int *)__p, __v);
 }

package/lib/clang/{19 → 23}/include/__clang_cuda_intrinsics.h

@@ -449,36 +449,304 @@ inline __device__ double2 __ldg(const double2 *ptr) {
 inline __device__ unsigned __funnelshift_l(unsigned low32, unsigned high32,
                                            unsigned shiftWidth) {
   unsigned result;
-  asm("shf.l.wrap.b32 %0, %1, %2, %3;"
-      : "=r"(result)
-      : "r"(low32), "r"(high32), "r"(shiftWidth));
+  __asm__("shf.l.wrap.b32 %0, %1, %2, %3;"
+          : "=r"(result)
+          : "r"(low32), "r"(high32), "r"(shiftWidth));
   return result;
 }
 inline __device__ unsigned __funnelshift_lc(unsigned low32, unsigned high32,
                                             unsigned shiftWidth) {
   unsigned result;
-  asm("shf.l.clamp.b32 %0, %1, %2, %3;"
-      : "=r"(result)
-      : "r"(low32), "r"(high32), "r"(shiftWidth));
+  __asm__("shf.l.clamp.b32 %0, %1, %2, %3;"
+          : "=r"(result)
+          : "r"(low32), "r"(high32), "r"(shiftWidth));
   return result;
 }
 inline __device__ unsigned __funnelshift_r(unsigned low32, unsigned high32,
                                            unsigned shiftWidth) {
   unsigned result;
-  asm("shf.r.wrap.b32 %0, %1, %2, %3;"
-      : "=r"(result)
-      : "r"(low32), "r"(high32), "r"(shiftWidth));
+  __asm__("shf.r.wrap.b32 %0, %1, %2, %3;"
+          : "=r"(result)
+          : "r"(low32), "r"(high32), "r"(shiftWidth));
   return result;
 }
 inline __device__ unsigned __funnelshift_rc(unsigned low32, unsigned high32,
                                             unsigned shiftWidth) {
   unsigned ret;
-  asm("shf.r.clamp.b32 %0, %1, %2, %3;"
-      : "=r"(ret)
-      : "r"(low32), "r"(high32), "r"(shiftWidth));
+  __asm__("shf.r.clamp.b32 %0, %1, %2, %3;"
+          : "=r"(ret)
+          : "r"(low32), "r"(high32), "r"(shiftWidth));
   return ret;
 }
 
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+
+#pragma push_macro("__INTRINSIC_LOAD")
+#define __INTRINSIC_LOAD(__FnName, __AsmOp, __DeclType, __TmpType, __AsmType,  \
+                         __Volatile, __Clobber)                                \
+  inline __device__ __DeclType __FnName(const __DeclType *__ptr) {             \
+    __TmpType __ret;                                                           \
+    __asm__ __Volatile(__AsmOp " %0, [%1];"                                    \
+                       : __AsmType(__ret)                                      \
+                       : "l"(__ptr)__Clobber);                                 \
+    return (__DeclType)__ret;                                                  \
+  }
+
+#pragma push_macro("__INTRINSIC_LOAD2")
+#define __INTRINSIC_LOAD2(__FnName, __AsmOp, __DeclType, __TmpType, __AsmType, \
+                          __Volatile, __Clobber)                               \
+  inline __device__ __DeclType __FnName(const __DeclType *__ptr) {             \
+    __DeclType __ret;                                                          \
+    __TmpType __tmp;                                                           \
+    __asm__ __Volatile(__AsmOp " {%0,%1}, [%2];"                               \
+                       : __AsmType(__tmp.x), __AsmType(__tmp.y)                \
+                       : "l"(__ptr)__Clobber);                                 \
+    using __ElementType = decltype(__ret.x);                                   \
+    __ret.x = (__ElementType)__tmp.x;                                          \
+    __ret.y = (__ElementType)__tmp.y;                                          \
+    return __ret;                                                              \
+  }
+
+#pragma push_macro("__INTRINSIC_LOAD4")
+#define __INTRINSIC_LOAD4(__FnName, __AsmOp, __DeclType, __TmpType, __AsmType, \
+                          __Volatile, __Clobber)                               \
+  inline __device__ __DeclType __FnName(const __DeclType *__ptr) {             \
+    __DeclType __ret;                                                          \
+    __TmpType __tmp;                                                           \
+    __asm__ __Volatile(__AsmOp " {%0,%1,%2,%3}, [%4];"                         \
+                       : __AsmType(__tmp.x), __AsmType(__tmp.y),               \
+                         __AsmType(__tmp.z), __AsmType(__tmp.w)                \
+                       : "l"(__ptr)__Clobber);                                 \
+    using __ElementType = decltype(__ret.x);                                   \
+    __ret.x = (__ElementType)__tmp.x;                                          \
+    __ret.y = (__ElementType)__tmp.y;                                          \
+    __ret.z = (__ElementType)__tmp.z;                                          \
+    __ret.w = (__ElementType)__tmp.w;                                          \
+    return __ret;                                                              \
+  }
+
+#pragma push_macro("__INTRINSIC_LOAD_LONG")
+#define __INTRINSIC_LOAD_LONG(__Mode)                                          \
+  inline __device__ long __ld##__Mode(const long *__ptr) {                     \
+    if (__SIZEOF_LONG__ == __SIZEOF_LONG_LONG__) {                             \
+      return (long)__ld##__Mode((const long long *)__ptr);                     \
+    } else {                                                                   \
+      return (long)__ld##__Mode((const int *)__ptr);                           \
+    }                                                                          \
+  }
+
+#pragma push_macro("__INTRINSIC_LOAD_ULONG")
+#define __INTRINSIC_LOAD_ULONG(__Mode)                                         \
+  inline __device__ unsigned long __ld##__Mode(const unsigned long *__ptr) {   \
+    if (__SIZEOF_LONG__ == __SIZEOF_LONG_LONG__) {                             \
+      return (unsigned long)__ld##__Mode((const unsigned long long *)__ptr);   \
+    } else {                                                                   \
+      return (unsigned long)__ld##__Mode((const unsigned int *)__ptr);         \
+    }                                                                          \
+  }
+
+#define __INTRINSIC_LOAD_FAMILY(__Mode, __Volatile, __Clobber)                 \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".s8", char,             \
+                   unsigned int, "=r", __Volatile, __Clobber)                  \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".s8", signed char,      \
+                   unsigned int, "=r", __Volatile, __Clobber)                  \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".s16", short,           \
+                   unsigned short, "=h", __Volatile, __Clobber)                \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".s32", int,             \
+                   unsigned int, "=r", __Volatile, __Clobber)                  \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".s64", long long,       \
+                   unsigned long long, "=l", __Volatile, __Clobber)            \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.s8", char2, int2,  \
+                    "=r", __Volatile, __Clobber)                               \
+  __INTRINSIC_LOAD4(__ld##__Mode, "ld.global." #__Mode ".v4.s8", char4, int4,  \
+                    "=r", __Volatile, __Clobber)                               \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.s16", short2,      \
+                    short2, "=h", __Volatile, __Clobber)                       \
+  __INTRINSIC_LOAD4(__ld##__Mode, "ld.global." #__Mode ".v4.s16", short4,      \
+                    short4, "=h", __Volatile, __Clobber)                       \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.s32", int2, int2,  \
+                    "=r", __Volatile, __Clobber)                               \
+  __INTRINSIC_LOAD4(__ld##__Mode, "ld.global." #__Mode ".v4.s32", int4, int4,  \
+                    "=r", __Volatile, __Clobber)                               \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.s64", longlong2,   \
+                    longlong2, "=l", __Volatile, __Clobber)                    \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".u8", unsigned char,    \
+                   unsigned int, "=r", __Volatile, __Clobber)                  \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".u16", unsigned short,  \
+                   unsigned short, "=h", __Volatile, __Clobber)                \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".u32", unsigned int,    \
+                   unsigned int, "=r", __Volatile, __Clobber)                  \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".u64",                  \
+                   unsigned long long, unsigned long long, "=l", __Volatile,   \
+                   __Clobber)                                                  \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.u8", uchar2,       \
+                    uint2, "=r", __Volatile, __Clobber)                        \
+  __INTRINSIC_LOAD4(__ld##__Mode, "ld.global." #__Mode ".v4.u8", uchar4,       \
+                    uint4, "=r", __Volatile, __Clobber)                        \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.u16", ushort2,     \
+                    ushort2, "=h", __Volatile, __Clobber)                      \
+  __INTRINSIC_LOAD4(__ld##__Mode, "ld.global." #__Mode ".v4.u16", ushort4,     \
+                    ushort4, "=h", __Volatile, __Clobber)                      \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.u32", uint2,       \
+                    uint2, "=r", __Volatile, __Clobber)                        \
+  __INTRINSIC_LOAD4(__ld##__Mode, "ld.global." #__Mode ".v4.u32", uint4,       \
+                    uint4, "=r", __Volatile, __Clobber)                        \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.u64", ulonglong2,  \
+                    ulonglong2, "=l", __Volatile, __Clobber)                   \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".f32", float, float,    \
+                   "=f", __Volatile, __Clobber)                                \
+  __INTRINSIC_LOAD(__ld##__Mode, "ld.global." #__Mode ".f64", double, double,  \
+                   "=d", __Volatile, __Clobber)                                \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.f32", float2,      \
+                    float2, "=f", __Volatile, __Clobber)                       \
+  __INTRINSIC_LOAD4(__ld##__Mode, "ld.global." #__Mode ".v4.f32", float4,      \
+                    float4, "=f", __Volatile, __Clobber)                       \
+  __INTRINSIC_LOAD2(__ld##__Mode, "ld.global." #__Mode ".v2.f64", double2,     \
+                    double2, "=d", __Volatile, __Clobber)                      \
+  __INTRINSIC_LOAD_LONG(__Mode)                                                \
+  __INTRINSIC_LOAD_ULONG(__Mode)
+
+__INTRINSIC_LOAD_FAMILY(ca, __volatile__, /* no clobber */)
+__INTRINSIC_LOAD_FAMILY(cg, __volatile__, /* no clobber */)
+__INTRINSIC_LOAD_FAMILY(cs, __volatile__, /* no clobber */)
+__INTRINSIC_LOAD_FAMILY(cv, /* not volatile */, : "memory")
+__INTRINSIC_LOAD_FAMILY(lu, /* not volatile */, : "memory")
+
+#pragma pop_macro("__INTRINSIC_LOAD")
+#pragma pop_macro("__INTRINSIC_LOAD2")
+#pragma pop_macro("__INTRINSIC_LOAD4")
+#pragma pop_macro("__INTRINSIC_LOAD_LONG")
+#pragma pop_macro("__INTRINSIC_LOAD_ULONG")
+#pragma pop_macro("__INTRINSIC_LOAD_FAMILY")
+
+#pragma push_macro("__INTRINSIC_STORE")
+#define __INTRINSIC_STORE(__FnName, __AsmOp, __DeclType, __TmpType, __AsmType) \
+  inline __device__ void __FnName(__DeclType *__ptr, __DeclType __value) {     \
+    __TmpType __tmp = (__TmpType)__value;                                      \
+    __asm__(__AsmOp " [%0], %1;" ::"l"(__ptr), __AsmType(__tmp) : "memory");   \
+  }
+
+#pragma push_macro("__INTRINSIC_STORE2")
+#define __INTRINSIC_STORE2(__FnName, __AsmOp, __DeclType, __TmpType,           \
+                           __AsmType)                                          \
+  inline __device__ void __FnName(__DeclType *__ptr, __DeclType __value) {     \
+    __TmpType __tmp;                                                           \
+    using __ElementType = decltype(__tmp.x);                                   \
+    __tmp.x = (__ElementType)(__value.x);                                      \
+    __tmp.y = (__ElementType)(__value.y);                                      \
+    __asm__(__AsmOp " [%0], {%1,%2};" ::"l"(__ptr), __AsmType(__tmp.x),        \
+            __AsmType(__tmp.y)                                                 \
+            : "memory");                                                       \
+  }
+
+#pragma push_macro("__INTRINSIC_STORE4")
+#define __INTRINSIC_STORE4(__FnName, __AsmOp, __DeclType, __TmpType,           \
+                           __AsmType)                                          \
+  inline __device__ void __FnName(__DeclType *__ptr, __DeclType __value) {     \
+    __TmpType __tmp;                                                           \
+    using __ElementType = decltype(__tmp.x);                                   \
+    __tmp.x = (__ElementType)(__value.x);                                      \
+    __tmp.y = (__ElementType)(__value.y);                                      \
+    __tmp.z = (__ElementType)(__value.z);                                      \
+    __tmp.w = (__ElementType)(__value.w);                                      \
+    __asm__(__AsmOp " [%0], {%1,%2,%3,%4};" ::"l"(__ptr), __AsmType(__tmp.x),  \
+            __AsmType(__tmp.y), __AsmType(__tmp.z), __AsmType(__tmp.w)         \
+            : "memory");                                                       \
+  }
+
+#pragma push_macro("__INTRINSIC_STORE_LONG")
+#define __INTRINSIC_STORE_LONG(__Mode)                                         \
+  inline __device__ void __st##__Mode(long *__ptr, long __value) {             \
+    if (__SIZEOF_LONG__ == __SIZEOF_LONG_LONG__) {                             \
+      __st##__Mode((long long *)__ptr, (long long)__value);                    \
+    } else {                                                                   \
+      __st##__Mode((int *)__ptr, (int)__value);                                \
+    }                                                                          \
+  }
+
+#pragma push_macro("__INTRINSIC_STORE_ULONG")
+#define __INTRINSIC_STORE_ULONG(__Mode)                                        \
+  inline __device__ void __st##__Mode(unsigned long *__ptr,                    \
+                                      unsigned long __value) {                 \
+    if (__SIZEOF_LONG__ == __SIZEOF_LONG_LONG__) {                             \
+      __st##__Mode((unsigned long long *)__ptr, (unsigned long long)__value);  \
+    } else {                                                                   \
+      __st##__Mode((unsigned int *)__ptr, (unsigned int)__value);              \
+    }                                                                          \
+  }
+
+#pragma push_macro("__INTRINSIC_STORE_FAMILY")
+#define __INTRINSIC_STORE_FAMILY(__Mode)                                       \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".s8", char, int, "r")  \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".s8", signed char,     \
+                    int, "r")                                                  \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".s16", short, short,   \
+                    "h")                                                       \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".s32", int, int, "r")  \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".s64", long long,      \
+                    long long, "l")                                            \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.s8", char2, int2, \
+                     "r")                                                      \
+  __INTRINSIC_STORE4(__st##__Mode, "st.global." #__Mode ".v4.s8", char4, int4, \
+                     "r")                                                      \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.s16", short2,     \
+                     short2, "h")                                              \
+  __INTRINSIC_STORE4(__st##__Mode, "st.global." #__Mode ".v4.s16", short4,     \
+                     short4, "h")                                              \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.s32", int2, int2, \
+                     "r")                                                      \
+  __INTRINSIC_STORE4(__st##__Mode, "st.global." #__Mode ".v4.s32", int4, int4, \
+                     "r")                                                      \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.s64", longlong2,  \
+                     longlong2, "l")                                           \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".u8", unsigned char,   \
+                    int, "r")                                                  \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".u16", unsigned short, \
+                    unsigned short, "h")                                       \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".u32", unsigned int,   \
+                    unsigned int, "r")                                         \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".u64",                 \
+                    unsigned long long, unsigned long long, "l")               \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.u8", uchar2,      \
+                     uchar2, "r")                                              \
+  __INTRINSIC_STORE4(__st##__Mode, "st.global." #__Mode ".v4.u8", uchar4,      \
+                     uint4, "r")                                               \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.u16", ushort2,    \
+                     ushort2, "h")                                             \
+  __INTRINSIC_STORE4(__st##__Mode, "st.global." #__Mode ".v4.u16", ushort4,    \
+                     ushort4, "h")                                             \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.u32", uint2,      \
+                     uint2, "r")                                               \
+  __INTRINSIC_STORE4(__st##__Mode, "st.global." #__Mode ".v4.u32", uint4,      \
+                     uint4, "r")                                               \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.u64", ulonglong2, \
+                     ulonglong2, "l")                                          \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".f32", float, float,   \
+                    "f")                                                       \
+  __INTRINSIC_STORE(__st##__Mode, "st.global." #__Mode ".f64", double, double, \
+                    "d")                                                       \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.f32", float2,     \
+                     float2, "f")                                              \
+  __INTRINSIC_STORE4(__st##__Mode, "st.global." #__Mode ".v4.f32", float4,     \
+                     float4, "f")                                              \
+  __INTRINSIC_STORE2(__st##__Mode, "st.global." #__Mode ".v2.f64", double2,    \
+                     double2, "d")                                             \
+  __INTRINSIC_STORE_LONG(__Mode)                                               \
+  __INTRINSIC_STORE_ULONG(__Mode)
+
+__INTRINSIC_STORE_FAMILY(cg)
+__INTRINSIC_STORE_FAMILY(cs)
+__INTRINSIC_STORE_FAMILY(wb)
+__INTRINSIC_STORE_FAMILY(wt)
+
+#pragma pop_macro("__INTRINSIC_STORE")
+#pragma pop_macro("__INTRINSIC_STORE2")
+#pragma pop_macro("__INTRINSIC_STORE4")
+#pragma pop_macro("__INTRINSIC_STORE_LONG")
+#pragma pop_macro("__INTRINSIC_STORE_ULONG")
+#pragma pop_macro("__INTRINSIC_STORE_FAMILY")
+
+#endif // defined(__cplusplus) && (__cplusplus >= 201103L)
 #endif // !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 320
 
 #if CUDA_VERSION >= 11000
@@ -515,32 +783,32 @@ __device__ inline cuuint32_t __nvvm_get_smem_pointer(void *__ptr) {
 #if !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 800
 __device__ inline unsigned __reduce_add_sync(unsigned __mask,
                                              unsigned __value) {
-  return __nvvm_redux_sync_add(__mask, __value);
+  return __nvvm_redux_sync_add(__value, __mask);
 }
 __device__ inline unsigned __reduce_min_sync(unsigned __mask,
                                              unsigned __value) {
-  return __nvvm_redux_sync_umin(__mask, __value);
+  return __nvvm_redux_sync_umin(__value, __mask);
 }
 __device__ inline unsigned __reduce_max_sync(unsigned __mask,
                                              unsigned __value) {
-  return __nvvm_redux_sync_umax(__mask, __value);
+  return __nvvm_redux_sync_umax(__value, __mask);
 }
 __device__ inline int __reduce_min_sync(unsigned __mask, int __value) {
-  return __nvvm_redux_sync_min(__mask, __value);
+  return __nvvm_redux_sync_min(__value, __mask);
 }
 __device__ inline int __reduce_max_sync(unsigned __mask, int __value) {
-  return __nvvm_redux_sync_max(__mask, __value);
+  return __nvvm_redux_sync_max(__value, __mask);
 }
 __device__ inline unsigned __reduce_or_sync(unsigned __mask, unsigned __value) {
-  return __nvvm_redux_sync_or(__mask, __value);
+  return __nvvm_redux_sync_or(__value, __mask);
 }
 __device__ inline unsigned __reduce_and_sync(unsigned __mask,
                                              unsigned __value) {
-  return __nvvm_redux_sync_and(__mask, __value);
+  return __nvvm_redux_sync_and(__value, __mask);
 }
 __device__ inline unsigned __reduce_xor_sync(unsigned __mask,
                                              unsigned __value) {
-  return __nvvm_redux_sync_xor(__mask, __value);
+  return __nvvm_redux_sync_xor(__value, __mask);
 }
 
 __device__ inline void __nv_memcpy_async_shared_global_4(void *__dst,

package/lib/clang/{19 → 23}/include/__clang_cuda_math.h

@@ -12,6 +12,10 @@
 #error "This file is for CUDA compilation only."
 #endif
 
+// The __CLANG_GPU_DISABLE_MATH_WRAPPERS macro provides a way to let standard
+// libcalls reach the link step instead of being eagerly replaced.
+#ifndef __CLANG_GPU_DISABLE_MATH_WRAPPERS
+
 #ifndef __OPENMP_NVPTX__
 #if CUDA_VERSION < 9000
 #error This file is intended to be used with CUDA-9+ only.
@@ -345,4 +349,5 @@ __DEVICE__ float ynf(int __a, float __b) { return __nv_ynf(__a, __b); }
 #pragma pop_macro("__DEVICE_VOID__")
 #pragma pop_macro("__FAST_OR_SLOW")
 
+#endif // __CLANG_GPU_DISABLE_MATH_WRAPPERS
 #endif // __CLANG_CUDA_MATH_H__