================ @@ -0,0 +1,119 @@ +//===-- gpuintrin.h - Generic GPU intrinsic functions ---------------------===// +// +// 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 +// +//===----------------------------------------------------------------------===// +// +// Provides wrappers around the clang builtins for accessing GPU hardware +// features. The interface is intended to be portable between architectures, but +// some targets may provide different implementations. This header can be +// included for all the common GPU programming languages, namely OpenMP, HIP, +// CUDA, and OpenCL. +// +//===----------------------------------------------------------------------===// + +#ifndef __GPUINTRIN_H +#define __GPUINTRIN_H + +#if defined(__NVPTX__) +#include <nvptxintrin.h> +#elif defined(__AMDGPU__) +#include <amdgpuintrin.h> +#else +#error "This header is only meant to be used on GPU architectures." +#endif + +// Returns the number of blocks in the requested dimension. +_DEFAULT_FN_ATTRS static inline uint32_t __gpu_num_blocks(int __dim) { + switch (__dim) { + case 0: + return __gpu_num_blocks_x(); + case 1: + return __gpu_num_blocks_y(); + case 2: + return __gpu_num_blocks_z(); + default: + __builtin_unreachable(); + } +} + +// Returns the number of block id in the requested dimension. +_DEFAULT_FN_ATTRS static inline uint32_t __gpu_block_id(int __dim) { + switch (__dim) { + case 0: + return __gpu_block_id_x(); + case 1: + return __gpu_block_id_y(); + case 2: + return __gpu_block_id_z(); + default: + __builtin_unreachable(); + } +} + +// Returns the number of threads in the requested dimension. +_DEFAULT_FN_ATTRS static inline uint32_t __gpu_num_threads(int __dim) { + switch (__dim) { + case 0: + return __gpu_num_threads_x(); + case 1: + return __gpu_num_threads_y(); + case 2: + return __gpu_num_threads_z(); + default: + __builtin_unreachable(); + } +} + +// Returns the thread id in the requested dimension. +_DEFAULT_FN_ATTRS static inline uint32_t __gpu_thread_id(int __dim) { + switch (__dim) { + case 0: + return __gpu_thread_id_x(); + case 1: + return __gpu_thread_id_y(); + case 2: + return __gpu_thread_id_z(); + default: + __builtin_unreachable(); + } +} + +// Get the first active thread inside the lane. +_DEFAULT_FN_ATTRS static inline uint64_t +__gpu_first_lane_id(uint64_t __lane_mask) { + return __builtin_ffsll(__lane_mask) - 1; +} + +// Conditional that is only true for a single thread in a lane. +_DEFAULT_FN_ATTRS static inline bool +__gpu_is_first_in_lane(uint64_t __lane_mask) { + return __gpu_lane_id() == __gpu_first_lane_id(__lane_mask); +} + +// Gets the sum of all lanes inside the warp or wavefront. +_DEFAULT_FN_ATTRS static inline uint32_t +__gpu_lane_reduce_u32(uint64_t __lane_mask, uint32_t x) { + for (uint32_t step = __gpu_num_lanes() / 2; step > 0; step /= 2) { + uint32_t index = step + __gpu_lane_id(); + x += __gpu_shuffle_idx_u32(__lane_mask, index, x); + } + return __gpu_broadcast_u32(__lane_mask, x); +} + +// Gets the accumulator scan of the threads in the warp or wavefront. +_DEFAULT_FN_ATTRS static inline uint32_t +__gpu_lane_scan_u32(uint64_t __lane_mask, uint32_t x) { + for (uint32_t step = 1; step < __gpu_num_lanes(); step *= 2) { + uint32_t index = __gpu_lane_id() - step; + uint32_t bitmask = __gpu_lane_id() >= step; + x += -bitmask & __gpu_shuffle_idx_u32(__lane_mask, index, x); + } + return x; +} + +#undef _DEFAULT_FN_ATTRS ---------------- jhuber6 wrote:
I kind of included them transitively, but I think it'd be better to just duplicate it so people can just include the NVPTX specific ones manually. https://github.com/llvm/llvm-project/pull/110179 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
