@cpp-linter/clang-tidy-win32-x64 20.1.0 → 22.1.8

package/README.md

@@ -1,7 +1,7 @@
 # @cpp-linter/clang-tidy-win32-x64
 
-The Windows x64 binary of clang-tidy from LLVM 20.1.0, for [`@cpp-linter/clang-tidy`](https://www.npmjs.com/package/@cpp-linter/clang-tidy).
+The Windows x64 binary of clang-tidy from LLVM 22.1.8, 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-20.1.0-3`](https://github.com/cpp-linter/clang-tools-static-binaries/releases/tag/llvm-20.1.0-3) 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-22.1.8-3`](https://github.com/cpp-linter/clang-tools-static-binaries/releases/tag/llvm-22.1.8-3) 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: 65729536 bytes
-sha256: 3820f1bc6c7b64b09d282aa933e4fe6b332ff33433cadcf9474e8a104435edd3
+size: 71897600 bytes
+sha256: 2d57d16ed4740b1399e70a17dad8f7b8f786cfeffc7756de2d0b5525b7676365

package/lib/clang/22/include/__clang_cuda_complex_builtins.h

@@ -0,0 +1,229 @@
+/*===-- __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__)
+#pragma omp declare target
+#define __DEVICE__ __attribute__((noinline, nothrow, cold, weak))
+#else
+#define __DEVICE__ __device__ inline
+#endif
+
+#ifdef __NVPTX__
+// FIXME: NVPTX should use generic builtins.
+#define _SCALBNd __nv_scalbn
+#define _SCALBNf __nv_scalbnf
+#define _LOGBd __nv_logb
+#define _LOGBf __nv_logbf
+#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__)
+#pragma omp end declare target
+#endif
+
+#pragma pop_macro("__DEVICE__")
+
+#endif // __CLANG_CUDA_COMPLEX_BUILTINS

package/lib/clang/{20 → 22}/include/__clang_cuda_device_functions.h

@@ -528,7 +528,7 @@ __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);

package/lib/clang/{20 → 22}/include/__clang_cuda_intrinsics.h

@@ -479,6 +479,293 @@ inline __device__ unsigned __funnelshift_rc(unsigned low32, unsigned high32,
   return ret;
 }
 
+#if defined(__cplusplus) && (__cplusplus >= 201103L)
+
+#pragma push_macro("__INTRINSIC_LOAD")
+#define __INTRINSIC_LOAD(__FnName, __AsmOp, __DeclType, __TmpType, __AsmType,  \
+                         __Clobber)                                            \
+  inline __device__ __DeclType __FnName(const __DeclType *__ptr) {             \
+    __TmpType __ret;                                                           \
+    asm(__AsmOp " %0, [%1];" : __AsmType(__ret) : "l"(__ptr)__Clobber);        \
+    return (__DeclType)__ret;                                                  \
+  }
+
+#pragma push_macro("__INTRINSIC_LOAD2")
+#define __INTRINSIC_LOAD2(__FnName, __AsmOp, __DeclType, __TmpType, __AsmType, \
+                          __Clobber)                                           \
+  inline __device__ __DeclType __FnName(const __DeclType *__ptr) {             \
+    __DeclType __ret;                                                          \
+    __TmpType __tmp;                                                           \
+    asm(__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, \
+                          __Clobber)                                           \
+  inline __device__ __DeclType __FnName(const __DeclType *__ptr) {             \
+    __DeclType __ret;                                                          \
+    __TmpType __tmp;                                                           \
+    asm(__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;                                                              \
+  }
+
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.s8", char, unsigned int, "=r", );
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.s8", signed char, unsigned int, "=r", );
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.s16", short, unsigned short, "=h", );
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.s32", int, unsigned int, "=r", );
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.s64", long long, unsigned long long,
+                 "=l", );
+
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.s8", char2, int2, "=r", );
+__INTRINSIC_LOAD4(__ldcg, "ld.global.cg.v4.s8", char4, int4, "=r", );
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.s16", short2, short2, "=h", );
+__INTRINSIC_LOAD4(__ldcg, "ld.global.cg.v4.s16", short4, short4, "=h", );
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.s32", int2, int2, "=r", );
+__INTRINSIC_LOAD4(__ldcg, "ld.global.cg.v4.s32", int4, int4, "=r", );
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.s64 ", longlong2, longlong2, "=l", );
+
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.u8", unsigned char, unsigned int,
+                 "=r", );
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.u16", unsigned short, unsigned short,
+                 "=h", );
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.u32", unsigned int, unsigned int,
+                 "=r", );
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.u64", unsigned long long,
+                 unsigned long long, "=l", );
+
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.u8", uchar2, int2, "=r", );
+__INTRINSIC_LOAD4(__ldcg, "ld.global.cg.v4.u8", uchar4, int4, "=r", );
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.u16", ushort2, ushort2, "=h", );
+__INTRINSIC_LOAD4(__ldcg, "ld.global.cg.v4.u16", ushort4, ushort4, "=h", );
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.u32", uint2, uint2, "=r", );
+__INTRINSIC_LOAD4(__ldcg, "ld.global.cg.v4.u32", uint4, uint4, "=r", );
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.u64", ulonglong2, ulonglong2,
+                  "=l", );
+
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.f32", float, float, "=f", );
+__INTRINSIC_LOAD(__ldcg, "ld.global.cg.f64", double, double, "=d", );
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.f32", float2, float2, "=f", );
+__INTRINSIC_LOAD4(__ldcg, "ld.global.cg.v4.f32", float4, float4, "=f", );
+__INTRINSIC_LOAD2(__ldcg, "ld.global.cg.v2.f64", double2, double2, "=d", );
+
+inline __device__ long __ldcg(const long *__ptr) {
+  unsigned long __ret;
+  if (sizeof(long) == 8) {
+    asm("ld.global.cg.s64 %0, [%1];" : "=l"(__ret) : "l"(__ptr));
+  } else {
+    asm("ld.global.cg.s32 %0, [%1];" : "=r"(__ret) : "l"(__ptr));
+  }
+  return (long)__ret;
+}
+
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.u8", unsigned char, unsigned int,
+                 "=r", : "memory");
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.u16", unsigned short, unsigned short,
+                 "=h", : "memory");
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.u32", unsigned int, unsigned int,
+                 "=r", : "memory");
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.u64", unsigned long long,
+                 unsigned long long, "=l", : "memory");
+
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.s8", char, unsigned int,
+                 "=r", : "memory");
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.s8", signed char, unsigned int,
+                 "=r", : "memory");
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.s16", short, unsigned short,
+                 "=h", : "memory");
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.s32", int, unsigned int,
+                 "=r", : "memory");
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.s64", long long, unsigned long long,
+                 "=l", : "memory");
+
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.u8", uchar2, uint2,
+                  "=r", : "memory");
+__INTRINSIC_LOAD4(__ldcv, "ld.global.cv.v4.u8", uchar4, uint4,
+                  "=r", : "memory");
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.u16", ushort2, ushort2,
+                  "=h", : "memory");
+__INTRINSIC_LOAD4(__ldcv, "ld.global.cv.v4.u16", ushort4, ushort4,
+                  "=h", : "memory");
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.u32", uint2, uint2,
+                  "=r", : "memory");
+__INTRINSIC_LOAD4(__ldcv, "ld.global.cv.v4.u32", uint4, uint4,
+                  "=r", : "memory");
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.u64", ulonglong2, ulonglong2,
+                  "=l", : "memory");
+
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.s8", char2, int2, "=r", : "memory");
+__INTRINSIC_LOAD4(__ldcv, "ld.global.cv.v4.s8", char4, int4, "=r", : "memory");
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.s16", short2, short2,
+                  "=h", : "memory");
+__INTRINSIC_LOAD4(__ldcv, "ld.global.cv.v4.s16", short4, short4,
+                  "=h", : "memory");
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.s32", int2, int2, "=r", : "memory");
+__INTRINSIC_LOAD4(__ldcv, "ld.global.cv.v4.s32", int4, int4, "=r", : "memory");
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.s64", longlong2, longlong2,
+                  "=l", : "memory");
+
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.f32", float, float, "=f", : "memory");
+__INTRINSIC_LOAD(__ldcv, "ld.global.cv.f64", double, double, "=d", : "memory");
+
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.f32", float2, float2,
+                  "=f", : "memory");
+__INTRINSIC_LOAD4(__ldcv, "ld.global.cv.v4.f32", float4, float4,
+                  "=f", : "memory");
+__INTRINSIC_LOAD2(__ldcv, "ld.global.cv.v2.f64", double2, double2,
+                  "=d", : "memory");
+
+inline __device__ long __ldcv(const long *__ptr) {
+  unsigned long __ret;
+  if (sizeof(long) == 8) {
+    asm("ld.global.cv.s64 %0, [%1];" : "=l"(__ret) : "l"(__ptr));
+  } else {
+    asm("ld.global.cv.s32 %0, [%1];" : "=r"(__ret) : "l"(__ptr));
+  }
+  return (long)__ret;
+}
+
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.s8", char, unsigned int, "=r", );
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.s8", signed char, signed int, "=r", );
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.s16", short, unsigned short, "=h", );
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.s32", int, unsigned int, "=r", );
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.s64", long long, unsigned long long,
+                 "=l", );
+
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.s8", char2, int2, "=r", );
+__INTRINSIC_LOAD4(__ldcs, "ld.global.cs.v4.s8", char4, int4, "=r", );
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.s16", short2, short2, "=h", );
+__INTRINSIC_LOAD4(__ldcs, "ld.global.cs.v4.s16", short4, short4, "=h", );
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.s32", int2, int2, "=r", );
+__INTRINSIC_LOAD4(__ldcs, "ld.global.cs.v4.s32", int4, int4, "=r", );
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.s64", longlong2, longlong2, "=l", );
+
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.u8", unsigned char, unsigned int,
+                 "=r", );
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.u16", unsigned short, unsigned short,
+                 "=h", );
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.u32", unsigned int, unsigned int,
+                 "=r", );
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.u64", unsigned long long,
+                 unsigned long long, "=l", );
+
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.u8", uchar2, uint2, "=r", );
+__INTRINSIC_LOAD4(__ldcs, "ld.global.cs.v4.u8", uchar4, uint4, "=r", );
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.u16", ushort2, ushort2, "=h", );
+__INTRINSIC_LOAD4(__ldcs, "ld.global.cs.v4.u16", ushort4, ushort4, "=h", );
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.u32", uint2, uint2, "=r", );
+__INTRINSIC_LOAD4(__ldcs, "ld.global.cs.v4.u32", uint4, uint4, "=r", );
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.u64", ulonglong2, ulonglong2,
+                  "=l", );
+
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.f32", float, float, "=f", );
+__INTRINSIC_LOAD(__ldcs, "ld.global.cs.f64", double, double, "=d", );
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.f32", float2, float2, "=f", );
+__INTRINSIC_LOAD4(__ldcs, "ld.global.cs.v4.f32", float4, float4, "=f", );
+__INTRINSIC_LOAD2(__ldcs, "ld.global.cs.v2.f64", double2, double2, "=d", );
+
+#pragma pop_macro("__INTRINSIC_LOAD")
+#pragma pop_macro("__INTRINSIC_LOAD2")
+#pragma pop_macro("__INTRINSIC_LOAD4")
+
+inline __device__ long __ldcs(const long *__ptr) {
+  unsigned long __ret;
+  if (sizeof(long) == 8) {
+    asm("ld.global.cs.s64 %0, [%1];" : "=l"(__ret) : "l"(__ptr));
+  } else {
+    asm("ld.global.cs.s32 %0, [%1];" : "=r"(__ret) : "l"(__ptr));
+  }
+  return (long)__ret;
+}
+
+#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");                                                           \
+  }
+
+__INTRINSIC_STORE(__stwt, "st.global.wt.s8", char, int, "r");
+__INTRINSIC_STORE(__stwt, "st.global.wt.s8", signed char, int, "r");
+__INTRINSIC_STORE(__stwt, "st.global.wt.s16", short, short, "h");
+__INTRINSIC_STORE(__stwt, "st.global.wt.s32", int, int, "r");
+__INTRINSIC_STORE(__stwt, "st.global.wt.s64", long long, long long, "l");
+
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.s8", char2, int2, "r");
+__INTRINSIC_STORE4(__stwt, "st.global.wt.v4.s8", char4, int4, "r");
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.s16", short2, short2, "h");
+__INTRINSIC_STORE4(__stwt, "st.global.wt.v4.s16", short4, short4, "h");
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.s32", int2, int2, "r");
+__INTRINSIC_STORE4(__stwt, "st.global.wt.v4.s32", int4, int4, "r");
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.s64", longlong2, longlong2, "l");
+
+__INTRINSIC_STORE(__stwt, "st.global.wt.u8", unsigned char, int, "r");
+__INTRINSIC_STORE(__stwt, "st.global.wt.u16", unsigned short, unsigned short,
+                  "h");
+__INTRINSIC_STORE(__stwt, "st.global.wt.u32", unsigned int, unsigned int, "r");
+__INTRINSIC_STORE(__stwt, "st.global.wt.u64", unsigned long long,
+                  unsigned long long, "l");
+
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.u8", uchar2, uchar2, "r");
+__INTRINSIC_STORE4(__stwt, "st.global.wt.v4.u8", uchar4, uint4, "r");
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.u16", ushort2, ushort2, "h");
+__INTRINSIC_STORE4(__stwt, "st.global.wt.v4.u16", ushort4, ushort4, "h");
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.u32", uint2, uint2, "r");
+__INTRINSIC_STORE4(__stwt, "st.global.wt.v4.u32", uint4, uint4, "r");
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.u64", ulonglong2, ulonglong2, "l");
+
+__INTRINSIC_STORE(__stwt, "st.global.wt.f32", float, float, "f");
+__INTRINSIC_STORE(__stwt, "st.global.wt.f64", double, double, "d");
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.f32", float2, float2, "f");
+__INTRINSIC_STORE4(__stwt, "st.global.wt.v4.f32", float4, float4, "f");
+__INTRINSIC_STORE2(__stwt, "st.global.wt.v2.f64", double2, double2, "d");
+
+#pragma pop_macro("__INTRINSIC_STORE")
+#pragma pop_macro("__INTRINSIC_STORE2")
+#pragma pop_macro("__INTRINSIC_STORE4")
+
+#endif // defined(__cplusplus) && (__cplusplus >= 201103L)
 #endif // !defined(__CUDA_ARCH__) || __CUDA_ARCH__ >= 320
 
 #if CUDA_VERSION >= 11000
@@ -515,32 +802,32 @@ __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/{20 → 22}/include/__clang_cuda_runtime_wrapper.h

@@ -38,12 +38,25 @@
 
 // Include some standard headers to avoid CUDA headers including them
 // while some required macros (like __THROW) are in a weird state.
+#include <climits>
 #include <cmath>
 #include <cstdlib>
 #include <stdlib.h>
 #include <string.h>
 #undef __CUDACC__
 
+// math_functions.h from CUDA 13.2+ defines _NV_RSQRT_SPECIFIER.
+// Clang does not include it, so we need to define it ourselves.
+#if defined(__GNUC__) && defined(__GLIBC_PREREQ)
+#if __GLIBC_PREREQ(2, 42)
+#define _NV_RSQRT_SPECIFIER noexcept(true)
+#endif
+#endif
+
+#ifndef _NV_RSQRT_SPECIFIER
+#define _NV_RSQRT_SPECIFIER
+#endif
+
 // Preserve common macros that will be changed below by us or by CUDA
 // headers.
 #pragma push_macro("__THROW")
@@ -384,7 +397,13 @@ __host__ __device__ void __nv_tex_surf_handler(const char *name, T *ptr,
 // will continue to fail as it does now.
 #endif // CUDA_VERSION
 #endif // __cplusplus >= 201103L && CUDA_VERSION >= 9000
+#include "surface_indirect_functions.h"
+#if CUDA_VERSION < 13000
+// Direct texture fetch functions had been deprecated since CUDA-11.
+// The file in CUDA-12 only carried unused texture types, and is no longer
+// needed.
 #include "texture_fetch_functions.h"
+#endif // CUDA_VERSION < 13000
 #include "texture_indirect_functions.h"
 
 // Restore state of __CUDA_ARCH__ and __THROW we had on entry.