@cpp-linter/clang-tidy-win32-x64 21.1.0 → 23.1.3

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

@@ -0,0 +1,242 @@
+//===--- __clang_gpu_intrinsics.h - Device-side GPU intrinsic wrappers ------===
+//
+// 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_GPU_INTRINSICS_H__
+#define __CLANG_GPU_INTRINSICS_H__
+
+#if defined(__HIP__) || defined(__CUDA__)
+
+#ifndef __GPU_DEVICE__
+#error                                                                         \
+    "__clang_gpu_intrinsics.h must be included via __clang_gpu_device_functions.h"
+#endif
+
+//===----------------------------------------------------------------------===//
+// Wavefront shuffles
+//===----------------------------------------------------------------------===//
+
+template <typename __T>
+__GPU_DEVICE__ __T __gpu_shuffle_idx_impl(__T __v, unsigned int __idx,
+                                          int __w) {
+  if constexpr (sizeof(__T) == sizeof(unsigned long long)) {
+    return __builtin_bit_cast(
+        __T, __gpu_shuffle_idx_u64(__gpu_lane_mask(), __idx,
+                                   __builtin_bit_cast(unsigned long long, __v),
+                                   (unsigned int)__w));
+  } else {
+    return __builtin_bit_cast(
+        __T, __gpu_shuffle_idx_u32(__gpu_lane_mask(), __idx,
+                                   __builtin_bit_cast(unsigned int, __v),
+                                   (unsigned int)__w));
+  }
+}
+
+template <typename __T>
+__GPU_DEVICE__ __T __shfl(MAYBE_UNDEF __T __var, int __src_lane,
+                          int __width = warpSize) {
+  return __gpu_shuffle_idx_impl(
+      __var, (unsigned int)(__src_lane & (__width - 1)), __width);
+}
+template <typename __T>
+__GPU_DEVICE__ __T __shfl_up(MAYBE_UNDEF __T __var, unsigned int __delta,
+                             int __width = warpSize) {
+  int __rel = int(__gpu_lane_id() & (unsigned int)(__width - 1));
+  int __tgt = __rel - int(__delta);
+  return __gpu_shuffle_idx_impl(
+      __var, (unsigned int)(__tgt < 0 ? __rel : __tgt), __width);
+}
+template <typename __T>
+__GPU_DEVICE__ __T __shfl_down(MAYBE_UNDEF __T __var, unsigned int __delta,
+                               int __width = warpSize) {
+  int __rel = int(__gpu_lane_id() & (unsigned int)(__width - 1));
+  int __tgt = __rel + int(__delta);
+  return __gpu_shuffle_idx_impl(
+      __var, (unsigned int)(__tgt >= __width ? __rel : __tgt), __width);
+}
+template <typename __T>
+__GPU_DEVICE__ __T __shfl_xor(MAYBE_UNDEF __T __var, int __lane_mask,
+                              int __width = warpSize) {
+  int __rel = int(__gpu_lane_id() & (unsigned int)(__width - 1));
+  int __tgt = __rel ^ __lane_mask;
+  return __gpu_shuffle_idx_impl(
+      __var, (unsigned int)(__tgt >= __width ? __rel : __tgt), __width);
+}
+
+//===----------------------------------------------------------------------===//
+// Warp synchronization
+//===----------------------------------------------------------------------===//
+
+__GPU_DEVICE__ void __syncwarp(unsigned long long __mask = -1) {
+  __scoped_atomic_thread_fence(__ATOMIC_RELEASE, __MEMORY_SCOPE_WVFRNT);
+  __gpu_sync_lane(__mask);
+  __scoped_atomic_thread_fence(__ATOMIC_ACQUIRE, __MEMORY_SCOPE_WVFRNT);
+}
+
+//===----------------------------------------------------------------------===//
+// Wave syncrhonization sync aliases.
+//===----------------------------------------------------------------------===//
+
+template <typename __MaskT>
+__GPU_DEVICE__ unsigned long long __ballot_sync(__MaskT __mask, int __pred) {
+  return __ballot(__pred) & (unsigned long long)__mask;
+}
+template <typename __MaskT>
+__GPU_DEVICE__ int __all_sync(__MaskT __mask, int __pred) {
+  return __ballot_sync(__mask, __pred) == (unsigned long long)__mask;
+}
+template <typename __MaskT>
+__GPU_DEVICE__ int __any_sync(__MaskT __mask, int __pred) {
+  return __ballot_sync(__mask, __pred) != 0ull;
+}
+
+template <typename __MaskT, typename __T>
+__GPU_DEVICE__ __T __shfl_sync(__MaskT __mask, MAYBE_UNDEF __T __var,
+                               int __src_lane, int __width = warpSize) {
+  (void)__mask;
+  return __shfl(__var, __src_lane, __width);
+}
+template <typename __MaskT, typename __T>
+__GPU_DEVICE__ __T __shfl_up_sync(__MaskT __mask, MAYBE_UNDEF __T __var,
+                                  unsigned int __delta,
+                                  int __width = warpSize) {
+  (void)__mask;
+  return __shfl_up(__var, __delta, __width);
+}
+template <typename __MaskT, typename __T>
+__GPU_DEVICE__ __T __shfl_down_sync(__MaskT __mask, MAYBE_UNDEF __T __var,
+                                    unsigned int __delta,
+                                    int __width = warpSize) {
+  (void)__mask;
+  return __shfl_down(__var, __delta, __width);
+}
+template <typename __MaskT, typename __T>
+__GPU_DEVICE__ __T __shfl_xor_sync(__MaskT __mask, MAYBE_UNDEF __T __var,
+                                   int __lane_mask, int __width = warpSize) {
+  (void)__mask;
+  return __shfl_xor(__var, __lane_mask, __width);
+}
+
+//===----------------------------------------------------------------------===//
+// Match primitives.
+//===----------------------------------------------------------------------===//
+
+template <typename __T>
+__GPU_DEVICE__ unsigned long long __match_any(__T __value) {
+  if constexpr (sizeof(__T) == sizeof(unsigned long long)) {
+    return __gpu_match_any_u64(__gpu_lane_mask(),
+                               __builtin_bit_cast(unsigned long long, __value));
+  } else {
+    return __gpu_match_any_u32(__gpu_lane_mask(),
+                               __builtin_bit_cast(unsigned int, __value));
+  }
+}
+template <typename __MaskT, typename __T>
+__GPU_DEVICE__ unsigned long long __match_any_sync(__MaskT __mask,
+                                                   __T __value) {
+  return __match_any(__value) & (unsigned long long)__mask;
+}
+
+template <typename __T>
+__GPU_DEVICE__ unsigned long long __match_all(__T __value, int *__pred) {
+  unsigned long long __r;
+  if constexpr (sizeof(__T) == sizeof(unsigned long long)) {
+    __r = __gpu_match_all_u64(__gpu_lane_mask(),
+                              __builtin_bit_cast(unsigned long long, __value));
+  } else {
+    __r = __gpu_match_all_u32(__gpu_lane_mask(),
+                              __builtin_bit_cast(unsigned int, __value));
+  }
+  *__pred = __r != 0;
+  return __r;
+}
+template <typename __MaskT, typename __T>
+__GPU_DEVICE__ unsigned long long __match_all_sync(__MaskT __mask, __T __value,
+                                                   int *__pred) {
+  (void)__mask;
+  return __match_all(__value, __pred);
+}
+
+//===----------------------------------------------------------------------===//
+// Wave reductions.
+//===----------------------------------------------------------------------===//
+
+template <typename __MaskT>
+__GPU_DEVICE__ unsigned int __reduce_add_sync(__MaskT __mask,
+                                              unsigned int __val) {
+  return __gpu_lane_add_u32((unsigned long long)__mask, __val);
+}
+template <typename __MaskT>
+__GPU_DEVICE__ int __reduce_add_sync(__MaskT __mask, int __val) {
+  return int(
+      __gpu_lane_add_u32((unsigned long long)__mask, (unsigned int)__val));
+}
+template <typename __MaskT>
+__GPU_DEVICE__ unsigned int __reduce_min_sync(__MaskT __mask,
+                                              unsigned int __val) {
+  return __gpu_lane_min_u32((unsigned long long)__mask, __val);
+}
+template <typename __MaskT>
+__GPU_DEVICE__ int __reduce_min_sync(__MaskT __mask, int __val) {
+  unsigned int __r = __gpu_lane_min_u32((unsigned long long)__mask,
+                                        (unsigned int)__val ^ 0x80000000u);
+  return int(__r ^ 0x80000000u);
+}
+template <typename __MaskT>
+__GPU_DEVICE__ unsigned int __reduce_max_sync(__MaskT __mask,
+                                              unsigned int __val) {
+  return __gpu_lane_max_u32((unsigned long long)__mask, __val);
+}
+template <typename __MaskT>
+__GPU_DEVICE__ int __reduce_max_sync(__MaskT __mask, int __val) {
+  unsigned int __r = __gpu_lane_max_u32((unsigned long long)__mask,
+                                        (unsigned int)__val ^ 0x80000000u);
+  return int(__r ^ 0x80000000u);
+}
+template <typename __MaskT>
+__GPU_DEVICE__ unsigned int __reduce_and_sync(__MaskT __mask,
+                                              unsigned int __val) {
+  return __gpu_lane_and_u32((unsigned long long)__mask, __val);
+}
+template <typename __MaskT>
+__GPU_DEVICE__ unsigned int __reduce_or_sync(__MaskT __mask,
+                                             unsigned int __val) {
+  return __gpu_lane_or_u32((unsigned long long)__mask, __val);
+}
+template <typename __MaskT>
+__GPU_DEVICE__ unsigned int __reduce_xor_sync(__MaskT __mask,
+                                              unsigned int __val) {
+  return __gpu_lane_xor_u32((unsigned long long)__mask, __val);
+}
+
+//===----------------------------------------------------------------------===//
+// Funnel shifts.
+//===----------------------------------------------------------------------===//
+
+__GPU_DEVICE__ unsigned int
+__funnelshift_l(unsigned int __lo, unsigned int __hi, unsigned int __shift) {
+  unsigned int __s = __shift & 31u;
+  return (unsigned int)((((unsigned long long)__hi << 32 | __lo) << __s) >> 32);
+}
+__GPU_DEVICE__ unsigned int
+__funnelshift_lc(unsigned int __lo, unsigned int __hi, unsigned int __shift) {
+  unsigned int __s = __shift >= 32u ? 32u : __shift;
+  return (unsigned int)((((unsigned long long)__hi << 32 | __lo) << __s) >> 32);
+}
+__GPU_DEVICE__ unsigned int
+__funnelshift_r(unsigned int __lo, unsigned int __hi, unsigned int __shift) {
+  unsigned int __s = __shift & 31u;
+  return (unsigned int)(((unsigned long long)__hi << 32 | __lo) >> __s);
+}
+__GPU_DEVICE__ unsigned int
+__funnelshift_rc(unsigned int __lo, unsigned int __hi, unsigned int __shift) {
+  unsigned int __s = __shift >= 32u ? 32u : __shift;
+  return (unsigned int)(((unsigned long long)__hi << 32 | __lo) >> __s);
+}
+
+#endif // device compile
+#endif // __CLANG_GPU_INTRINSICS_H__

package/lib/clang/{21 → 23}/include/__clang_hip_libdevice_declares.h

@@ -29,9 +29,6 @@ __device__ __attribute__((const)) float __ocml_atan2_f32(float, float);
 __device__ __attribute__((const)) float __ocml_atan_f32(float);
 __device__ __attribute__((pure)) float __ocml_atanh_f32(float);
 __device__ __attribute__((pure)) float __ocml_cbrt_f32(float);
-__device__ __attribute__((const)) float __ocml_ceil_f32(float);
-__device__ __attribute__((const)) __device__ float __ocml_copysign_f32(float,
-                                                                       float);
 __device__ float __ocml_cos_f32(float);
 __device__ float __ocml_native_cos_f32(float);
 __device__ __attribute__((pure)) __device__ float __ocml_cosh_f32(float);
@@ -43,40 +40,17 @@ __device__ __attribute__((pure)) float __ocml_erfcinv_f32(float);
 __device__ __attribute__((pure)) float __ocml_erfcx_f32(float);
 __device__ __attribute__((pure)) float __ocml_erf_f32(float);
 __device__ __attribute__((pure)) float __ocml_erfinv_f32(float);
-__device__ __attribute__((pure)) float __ocml_exp10_f32(float);
-__device__ __attribute__((pure)) float __ocml_native_exp10_f32(float);
-__device__ __attribute__((pure)) float __ocml_exp2_f32(float);
-__device__ __attribute__((pure)) float __ocml_exp_f32(float);
-__device__ __attribute__((pure)) float __ocml_native_exp_f32(float);
 __device__ __attribute__((pure)) float __ocml_expm1_f32(float);
-__device__ __attribute__((const)) float __ocml_fabs_f32(float);
 __device__ __attribute__((const)) float __ocml_fdim_f32(float, float);
-__device__ __attribute__((const)) float __ocml_floor_f32(float);
-__device__ __attribute__((const)) float __ocml_fma_f32(float, float, float);
-__device__ __attribute__((const)) float __ocml_fmax_f32(float, float);
-__device__ __attribute__((const)) float __ocml_fmin_f32(float, float);
 __device__ __attribute__((const)) __device__ float __ocml_fmod_f32(float,
                                                                    float);
-__device__ float __ocml_frexp_f32(float, __PRIVATE_AS int *);
 __device__ __attribute__((const)) float __ocml_hypot_f32(float, float);
 __device__ __attribute__((const)) int __ocml_ilogb_f32(float);
-__device__ __attribute__((const)) int __ocml_isfinite_f32(float);
-__device__ __attribute__((const)) int __ocml_isinf_f32(float);
-__device__ __attribute__((const)) int __ocml_isnan_f32(float);
 __device__ float __ocml_j0_f32(float);
 __device__ float __ocml_j1_f32(float);
-__device__ __attribute__((const)) float __ocml_ldexp_f32(float, int);
 __device__ float __ocml_lgamma_f32(float);
-__device__ __attribute__((pure)) float __ocml_log10_f32(float);
-__device__ __attribute__((pure)) float __ocml_native_log10_f32(float);
 __device__ __attribute__((pure)) float __ocml_log1p_f32(float);
-__device__ __attribute__((pure)) float __ocml_log2_f32(float);
-__device__ __attribute__((pure)) float __ocml_native_log2_f32(float);
-__device__ __attribute__((const)) float __ocml_logb_f32(float);
-__device__ __attribute__((pure)) float __ocml_log_f32(float);
-__device__ __attribute__((pure)) float __ocml_native_log_f32(float);
 __device__ float __ocml_modf_f32(float, __PRIVATE_AS float *);
-__device__ __attribute__((const)) float __ocml_nearbyint_f32(float);
 __device__ __attribute__((const)) float __ocml_nextafter_f32(float, float);
 __device__ __attribute__((const)) float __ocml_len3_f32(float, float, float);
 __device__ __attribute__((const)) float __ocml_len4_f32(float, float, float,
@@ -89,15 +63,10 @@ __device__ __attribute__((pure)) float __ocml_rcbrt_f32(float);
 __device__ __attribute__((const)) float __ocml_remainder_f32(float, float);
 __device__ float __ocml_remquo_f32(float, float, __PRIVATE_AS int *);
 __device__ __attribute__((const)) float __ocml_rhypot_f32(float, float);
-__device__ __attribute__((const)) float __ocml_rint_f32(float);
 __device__ __attribute__((const)) float __ocml_rlen3_f32(float, float, float);
 __device__ __attribute__((const)) float __ocml_rlen4_f32(float, float, float,
                                                          float);
-__device__ __attribute__((const)) float __ocml_round_f32(float);
 __device__ __attribute__((pure)) float __ocml_rsqrt_f32(float);
-__device__ __attribute__((const)) float __ocml_scalb_f32(float, float);
-__device__ __attribute__((const)) float __ocml_scalbn_f32(float, int);
-__device__ __attribute__((const)) int __ocml_signbit_f32(float);
 __device__ float __ocml_sincos_f32(float, __PRIVATE_AS float *);
 __device__ float __ocml_sincospi_f32(float, __PRIVATE_AS float *);
 __device__ float __ocml_sin_f32(float);
@@ -109,7 +78,6 @@ __device__ __attribute__((const)) float __ocml_native_sqrt_f32(float);
 __device__ float __ocml_tan_f32(float);
 __device__ __attribute__((pure)) float __ocml_tanh_f32(float);
 __device__ float __ocml_tgamma_f32(float);
-__device__ __attribute__((const)) float __ocml_trunc_f32(float);
 __device__ float __ocml_y0_f32(float);
 __device__ float __ocml_y1_f32(float);
 
@@ -150,8 +118,6 @@ __device__ __attribute__((const)) double __ocml_atan2_f64(double, double);
 __device__ __attribute__((const)) double __ocml_atan_f64(double);
 __device__ __attribute__((pure)) double __ocml_atanh_f64(double);
 __device__ __attribute__((pure)) double __ocml_cbrt_f64(double);
-__device__ __attribute__((const)) double __ocml_ceil_f64(double);
-__device__ __attribute__((const)) double __ocml_copysign_f64(double, double);
 __device__ double __ocml_cos_f64(double);
 __device__ __attribute__((pure)) double __ocml_cosh_f64(double);
 __device__ double __ocml_cospi_f64(double);
@@ -166,30 +132,18 @@ __device__ __attribute__((pure)) double __ocml_exp10_f64(double);
 __device__ __attribute__((pure)) double __ocml_exp2_f64(double);
 __device__ __attribute__((pure)) double __ocml_exp_f64(double);
 __device__ __attribute__((pure)) double __ocml_expm1_f64(double);
-__device__ __attribute__((const)) double __ocml_fabs_f64(double);
 __device__ __attribute__((const)) double __ocml_fdim_f64(double, double);
-__device__ __attribute__((const)) double __ocml_floor_f64(double);
-__device__ __attribute__((const)) double __ocml_fma_f64(double, double, double);
-__device__ __attribute__((const)) double __ocml_fmax_f64(double, double);
-__device__ __attribute__((const)) double __ocml_fmin_f64(double, double);
 __device__ __attribute__((const)) double __ocml_fmod_f64(double, double);
-__device__ double __ocml_frexp_f64(double, __PRIVATE_AS int *);
 __device__ __attribute__((const)) double __ocml_hypot_f64(double, double);
 __device__ __attribute__((const)) int __ocml_ilogb_f64(double);
-__device__ __attribute__((const)) int __ocml_isfinite_f64(double);
-__device__ __attribute__((const)) int __ocml_isinf_f64(double);
-__device__ __attribute__((const)) int __ocml_isnan_f64(double);
 __device__ double __ocml_j0_f64(double);
 __device__ double __ocml_j1_f64(double);
-__device__ __attribute__((const)) double __ocml_ldexp_f64(double, int);
 __device__ double __ocml_lgamma_f64(double);
 __device__ __attribute__((pure)) double __ocml_log10_f64(double);
 __device__ __attribute__((pure)) double __ocml_log1p_f64(double);
 __device__ __attribute__((pure)) double __ocml_log2_f64(double);
-__device__ __attribute__((const)) double __ocml_logb_f64(double);
 __device__ __attribute__((pure)) double __ocml_log_f64(double);
 __device__ double __ocml_modf_f64(double, __PRIVATE_AS double *);
-__device__ __attribute__((const)) double __ocml_nearbyint_f64(double);
 __device__ __attribute__((const)) double __ocml_nextafter_f64(double, double);
 __device__ __attribute__((const)) double __ocml_len3_f64(double, double,
                                                          double);
@@ -203,26 +157,19 @@ __device__ __attribute__((pure)) double __ocml_rcbrt_f64(double);
 __device__ __attribute__((const)) double __ocml_remainder_f64(double, double);
 __device__ double __ocml_remquo_f64(double, double, __PRIVATE_AS int *);
 __device__ __attribute__((const)) double __ocml_rhypot_f64(double, double);
-__device__ __attribute__((const)) double __ocml_rint_f64(double);
 __device__ __attribute__((const)) double __ocml_rlen3_f64(double, double,
                                                           double);
 __device__ __attribute__((const)) double __ocml_rlen4_f64(double, double,
                                                           double, double);
-__device__ __attribute__((const)) double __ocml_round_f64(double);
 __device__ __attribute__((pure)) double __ocml_rsqrt_f64(double);
-__device__ __attribute__((const)) double __ocml_scalb_f64(double, double);
-__device__ __attribute__((const)) double __ocml_scalbn_f64(double, int);
-__device__ __attribute__((const)) int __ocml_signbit_f64(double);
 __device__ double __ocml_sincos_f64(double, __PRIVATE_AS double *);
 __device__ double __ocml_sincospi_f64(double, __PRIVATE_AS double *);
 __device__ double __ocml_sin_f64(double);
 __device__ __attribute__((pure)) double __ocml_sinh_f64(double);
 __device__ double __ocml_sinpi_f64(double);
-__device__ __attribute__((const)) double __ocml_sqrt_f64(double);
 __device__ double __ocml_tan_f64(double);
 __device__ __attribute__((pure)) double __ocml_tanh_f64(double);
 __device__ double __ocml_tgamma_f64(double);
-__device__ __attribute__((const)) double __ocml_trunc_f64(double);
 __device__ double __ocml_y0_f64(double);
 __device__ double __ocml_y1_f64(double);
 
@@ -256,88 +203,34 @@ __device__ __attribute__((const)) double __ocml_fma_rtp_f64(double, double,
 __device__ __attribute__((const)) double __ocml_fma_rtz_f64(double, double,
                                                             double);
 
-__device__ __attribute__((const)) _Float16 __ocml_ceil_f16(_Float16);
 __device__ _Float16 __ocml_cos_f16(_Float16);
 __device__ __attribute__((const)) _Float16 __ocml_cvtrtn_f16_f32(float);
 __device__ __attribute__((const)) _Float16 __ocml_cvtrtp_f16_f32(float);
 __device__ __attribute__((const)) _Float16 __ocml_cvtrtz_f16_f32(float);
-__device__ __attribute__((pure)) _Float16 __ocml_exp_f16(_Float16);
-__device__ __attribute__((pure)) _Float16 __ocml_exp10_f16(_Float16);
-__device__ __attribute__((pure)) _Float16 __ocml_exp2_f16(_Float16);
-__device__ __attribute__((const)) _Float16 __ocml_floor_f16(_Float16);
-__device__ __attribute__((const)) _Float16 __ocml_fma_f16(_Float16, _Float16,
-                                                          _Float16);
-__device__ __attribute__((const)) _Float16 __ocml_fmax_f16(_Float16, _Float16);
-__device__ __attribute__((const)) _Float16 __ocml_fmin_f16(_Float16, _Float16);
-__device__ __attribute__((const)) _Float16 __ocml_fabs_f16(_Float16);
-__device__ __attribute__((const)) int __ocml_isinf_f16(_Float16);
-__device__ __attribute__((const)) int __ocml_isnan_f16(_Float16);
-__device__ __attribute__((pure)) _Float16 __ocml_log_f16(_Float16);
-__device__ __attribute__((pure)) _Float16 __ocml_log10_f16(_Float16);
-__device__ __attribute__((pure)) _Float16 __ocml_log2_f16(_Float16);
-__device__ __attribute__((const)) _Float16 __ocml_rint_f16(_Float16);
 __device__ __attribute__((const)) _Float16 __ocml_rsqrt_f16(_Float16);
 __device__ _Float16 __ocml_sin_f16(_Float16);
-__device__ __attribute__((const)) _Float16 __ocml_sqrt_f16(_Float16);
-__device__ __attribute__((const)) _Float16 __ocml_trunc_f16(_Float16);
 __device__ __attribute__((pure)) _Float16 __ocml_pown_f16(_Float16, int);
 
 typedef _Float16 __2f16 __attribute__((ext_vector_type(2)));
 typedef short __2i16 __attribute__((ext_vector_type(2)));
 
-// We need to match C99's bool and get an i1 in the IR.
-#ifdef __cplusplus
-typedef bool __ockl_bool;
-#else
-typedef _Bool __ockl_bool;
-#endif
-
-__device__ __attribute__((const)) float __ockl_fdot2(__2f16 a, __2f16 b,
-                                                     float c, __ockl_bool s);
-__device__ __attribute__((const)) __2f16 __ocml_ceil_2f16(__2f16);
-__device__ __attribute__((const)) __2f16 __ocml_fabs_2f16(__2f16);
-__device__ __2f16 __ocml_cos_2f16(__2f16);
-__device__ __attribute__((pure)) __2f16 __ocml_exp_2f16(__2f16);
-__device__ __attribute__((pure)) __2f16 __ocml_exp10_2f16(__2f16);
-__device__ __attribute__((pure)) __2f16 __ocml_exp2_2f16(__2f16);
-__device__ __attribute__((const)) __2f16 __ocml_floor_2f16(__2f16);
-__device__ __attribute__((const))
-__2f16 __ocml_fma_2f16(__2f16, __2f16, __2f16);
-__device__ __attribute__((const)) __2i16 __ocml_isinf_2f16(__2f16);
-__device__ __attribute__((const)) __2i16 __ocml_isnan_2f16(__2f16);
-__device__ __attribute__((pure)) __2f16 __ocml_log_2f16(__2f16);
-__device__ __attribute__((pure)) __2f16 __ocml_log10_2f16(__2f16);
-__device__ __attribute__((pure)) __2f16 __ocml_log2_2f16(__2f16);
-
-#if HIP_VERSION_MAJOR * 100 + HIP_VERSION_MINOR >= 560
-#define __DEPRECATED_SINCE_HIP_560(X) __attribute__((deprecated(X)))
+__device__ void __asan_poison_memory_region(const void *addr,
+                                            __SIZE_TYPE__ size);
+__device__ void __asan_unpoison_memory_region(const void *addr,
+                                              __SIZE_TYPE__ size);
+__device__ int __asan_address_is_poisoned(const void *addr);
+__device__ void *__asan_region_is_poisoned(void *beg, __SIZE_TYPE__ size);
+
+#if __has_feature(address_sanitizer)
+#define ASAN_POISON_MEMORY_REGION(addr, size)                                  \
+  __asan_poison_memory_region((addr), (size))
+#define ASAN_UNPOISON_MEMORY_REGION(addr, size)                                \
+  __asan_unpoison_memory_region((addr), (size))
 #else
-#define __DEPRECATED_SINCE_HIP_560(X)
+#define ASAN_POISON_MEMORY_REGION(addr, size) ((void)(addr), (void)(size))
+#define ASAN_UNPOISON_MEMORY_REGION(addr, size) ((void)(addr), (void)(size))
 #endif
 
-// Deprecated, should be removed when rocm releases using it are no longer
-// relevant.
-__DEPRECATED_SINCE_HIP_560("use ((_Float16)1.0) / ")
-__device__ inline _Float16 __llvm_amdgcn_rcp_f16(_Float16 x) {
-  return ((_Float16)1.0f) / x;
-}
-
-__DEPRECATED_SINCE_HIP_560("use ((__2f16)1.0) / ")
-__device__ inline __2f16
-__llvm_amdgcn_rcp_2f16(__2f16 __x)
-{
-  return ((__2f16)1.0f) / __x;
-}
-
-#undef __DEPRECATED_SINCE_HIP_560
-
-__device__ __attribute__((const)) __2f16 __ocml_rint_2f16(__2f16);
-__device__ __attribute__((const)) __2f16 __ocml_rsqrt_2f16(__2f16);
-__device__ __2f16 __ocml_sin_2f16(__2f16);
-__device__ __attribute__((const)) __2f16 __ocml_sqrt_2f16(__2f16);
-__device__ __attribute__((const)) __2f16 __ocml_trunc_2f16(__2f16);
-__device__ __attribute__((const)) __2f16 __ocml_pown_2f16(__2f16, __2i16);
-
 #ifdef __cplusplus
 } // extern "C"
 #endif

package/lib/clang/{21 → 23}/include/__clang_hip_math.h

@@ -498,7 +498,7 @@ __DEVICE__
 float log2f(float __x) { return __FAST_OR_SLOW(__log2f, __builtin_log2f)(__x); }
 
 __DEVICE__
-float logbf(float __x) { return __ocml_logb_f32(__x); }
+float logbf(float __x) { return __builtin_logbf(__x); }
 
 __DEVICE__
 float logf(float __x) { return __FAST_OR_SLOW(__logf, __builtin_logf)(__x); }
@@ -795,13 +795,13 @@ __DEVICE__
 double erfinv(double __x) { return __ocml_erfinv_f64(__x); }
 
 __DEVICE__
-double exp(double __x) { return __ocml_exp_f64(__x); }
+double exp(double __x) { return __builtin_exp(__x); }
 
 __DEVICE__
-double exp10(double __x) { return __ocml_exp10_f64(__x); }
+double exp10(double __x) { return __builtin_exp10(__x); }
 
 __DEVICE__
-double exp2(double __x) { return __ocml_exp2_f64(__x); }
+double exp2(double __x) { return __builtin_exp2(__x); }
 
 __DEVICE__
 double expm1(double __x) { return __ocml_expm1_f64(__x); }
@@ -901,7 +901,7 @@ __DEVICE__
 double log2(double __x) { return __ocml_log2_f64(__x); }
 
 __DEVICE__
-double logb(double __x) { return __ocml_logb_f64(__x); }
+double logb(double __x) { return __builtin_logb(__x); }
 
 __DEVICE__
 long int lrint(double __x) { return __builtin_rint(__x); }

package/lib/clang/{21 → 23}/include/__clang_hip_runtime_wrapper.h

@@ -25,6 +25,9 @@
 #define __constant__ __attribute__((constant))
 #define __managed__ __attribute__((managed))
 
+#define __cluster_dims__(...) __attribute__((cluster_dims(__VA_ARGS__)))
+#define __no_cluster__ __attribute__((no_cluster))
+
 #if !defined(__cplusplus) || __cplusplus < 201103L
   #define nullptr NULL;
 #endif
@@ -108,6 +111,12 @@ __attribute__((weak)) inline __device__ void free(void *__ptr) {
 #endif //__cplusplus
 
 #if !defined(__HIPCC_RTC__)
+// We must include the forward declarations before cmath is included.
+// Otherwise `constexpr` functions in <cmath> would be implicitely __host__
+// __device__. Declaring the __device__ verison before allows us to overload
+// them with __device__ versions (this behavour is only valid for system
+// headers).
+#include <__clang_cuda_math_forward_declares.h>
 #include <cmath>
 #include <cstdlib>
 #include <stdlib.h>
@@ -141,7 +150,6 @@ typedef __SIZE_TYPE__ size_t;
 #if defined(__HIPCC_RTC__)
 #include <__clang_hip_cmath.h>
 #else
-#include <__clang_cuda_math_forward_declares.h>
 #include <__clang_hip_cmath.h>
 #include <__clang_cuda_complex_builtins.h>
 #include <algorithm>

package/lib/clang/{21 → 23}/include/__clang_spirv_builtins.h

@@ -52,30 +52,30 @@
 // Builtin IDs and sizes
 
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_num_workgroups) __size_t
-    __spirv_NumWorkgroups(int);
+    __spirv_BuiltInNumWorkgroups(int);
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_workgroup_size) __size_t
-    __spirv_WorkgroupSize(int);
+    __spirv_BuiltInWorkgroupSize(int);
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_workgroup_id) __size_t
-    __spirv_WorkgroupId(int);
+    __spirv_BuiltInWorkgroupId(int);
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_local_invocation_id) __size_t
-    __spirv_LocalInvocationId(int);
+    __spirv_BuiltInLocalInvocationId(int);
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_global_invocation_id) __size_t
-    __spirv_GlobalInvocationId(int);
+    __spirv_BuiltInGlobalInvocationId(int);
 
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_global_size) __size_t
-    __spirv_GlobalSize(int);
+    __spirv_BuiltInGlobalSize(int);
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_global_offset) __size_t
-    __spirv_GlobalOffset(int);
+    __spirv_BuiltInGlobalOffset(int);
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_subgroup_size) __uint32_t
-    __spirv_SubgroupSize();
+    __spirv_BuiltInSubgroupSize();
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_subgroup_max_size) __uint32_t
-    __spirv_SubgroupMaxSize();
+    __spirv_BuiltInSubgroupMaxSize();
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_num_subgroups) __uint32_t
-    __spirv_NumSubgroups();
+    __spirv_BuiltInNumSubgroups();
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_subgroup_id) __uint32_t
-    __spirv_SubgroupId();
+    __spirv_BuiltInSubgroupId();
 extern __SPIRV_BUILTIN_ALIAS(__builtin_spirv_subgroup_local_invocation_id)
-    __uint32_t __spirv_SubgroupLocalInvocationId();
+    __uint32_t __spirv_BuiltInSubgroupLocalInvocationId();
 
 // OpGenericCastToPtrExplicit