[clang] [Clang] Implement resource directory headers for common GPU intrinsics (PR #110179)

[Johannes Doerfert via cfe-commits]

================
@@ -0,0 +1,184 @@
+//===-- nvptxintrin.h - NVPTX 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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __NVPTXINTRIN_H
+#define __NVPTXINTRIN_H
+
+#ifndef __NVPTX__
+#error "This file is intended for NVPTX targets or offloading to NVPTX
+#endif
+
+#include <stdbool.h>
+#include <stdint.h>
+
+#if defined(__HIP__) || defined(__CUDA__)
+#define _DEFAULT_ATTRS __attribute__((device)) __attribute__((always_inline))
+#else
+#define _DEFAULT_ATTRS __attribute__((always_inline))
+#endif
+
+#pragma omp begin declare target device_type(nohost)
+#pragma omp begin declare variant match(device = {arch(nvptx64)})
+
+// Type aliases to the address spaces used by the NVPTX backend.
+#define _private __attribute__((opencl_private))
+#define _constant __attribute__((opencl_constant))
+#define _local __attribute__((opencl_local))
+#define _global __attribute__((opencl_global))
+
+// Attribute to declare a function as a kernel.
+#define _kernel __attribute__((nvptx_kernel))
+
+// Returns the number of CUDA blocks in the 'x' dimension.
+_DEFAULT_ATTRS static inline uint32_t _get_num_blocks_x() {
+  return __nvvm_read_ptx_sreg_nctaid_x();
+}
+
+// Returns the number of CUDA blocks in the 'y' dimension.
+_DEFAULT_ATTRS static inline uint32_t _get_num_blocks_y() {
+  return __nvvm_read_ptx_sreg_nctaid_y();
+}
+
+// Returns the number of CUDA blocks in the 'z' dimension.
+_DEFAULT_ATTRS static inline uint32_t _get_num_blocks_z() {
+  return __nvvm_read_ptx_sreg_nctaid_z();
+}
+
+// Returns the total number of CUDA blocks.
+_DEFAULT_ATTRS static inline uint64_t _get_num_blocks() {
+  return _get_num_blocks_x() * _get_num_blocks_y() * _get_num_blocks_z();
+}
+
+// Returns the 'x' dimension of the current CUDA block's id.
+_DEFAULT_ATTRS static inline uint32_t _get_block_id_x() {
+  return __nvvm_read_ptx_sreg_ctaid_x();
+}
+
+// Returns the 'y' dimension of the current CUDA block's id.
+_DEFAULT_ATTRS static inline uint32_t _get_block_id_y() {
+  return __nvvm_read_ptx_sreg_ctaid_y();
+}
+
+// Returns the 'z' dimension of the current CUDA block's id.
+_DEFAULT_ATTRS static inline uint32_t _get_block_id_z() {
+  return __nvvm_read_ptx_sreg_ctaid_z();
+}
+
+// Returns the absolute id of the CUDA block.
+_DEFAULT_ATTRS static inline uint64_t _get_block_id() {
+  return _get_block_id_x() + _get_num_blocks_x() * _get_block_id_y() +
+         _get_num_blocks_x() * _get_num_blocks_y() * _get_block_id_z();
+}
+
+// Returns the number of CUDA threads in the 'x' dimension.
+_DEFAULT_ATTRS static inline uint32_t _get_num_threads_x() {
+  return __nvvm_read_ptx_sreg_ntid_x();
+}
+
+// Returns the number of CUDA threads in the 'y' dimension.
+_DEFAULT_ATTRS static inline uint32_t _get_num_threads_y() {
+  return __nvvm_read_ptx_sreg_ntid_y();
+}
+
+// Returns the number of CUDA threads in the 'z' dimension.
+_DEFAULT_ATTRS static inline uint32_t _get_num_threads_z() {
+  return __nvvm_read_ptx_sreg_ntid_z();
+}
+
+// Returns the total number of threads in the block.
+_DEFAULT_ATTRS static inline uint64_t _get_num_threads() {
+  return _get_num_threads_x() * _get_num_threads_y() * _get_num_threads_z();
+}
+
+// Returns the 'x' dimension id of the thread in the current CUDA block.
+_DEFAULT_ATTRS static inline uint32_t _get_thread_id_x() {
+  return __nvvm_read_ptx_sreg_tid_x();
+}
+
+// Returns the 'y' dimension id of the thread in the current CUDA block.
+_DEFAULT_ATTRS static inline uint32_t _get_thread_id_y() {
+  return __nvvm_read_ptx_sreg_tid_y();
+}
+
+// Returns the 'z' dimension id of the thread in the current CUDA block.
+_DEFAULT_ATTRS static inline uint32_t _get_thread_id_z() {
+  return __nvvm_read_ptx_sreg_tid_z();
+}
+
+// Returns the absolute id of the thread in the current CUDA block.
+_DEFAULT_ATTRS static inline uint64_t _get_thread_id() {
+  return _get_thread_id_x() + _get_num_threads_x() * _get_thread_id_y() +
+         _get_num_threads_x() * _get_num_threads_y() * _get_thread_id_z();
+}
+
+// Returns the size of a CUDA warp, always 32 on NVIDIA hardware.
+_DEFAULT_ATTRS static inline uint32_t _get_lane_size() { return 32; }
----------------
jdoerfert wrote:

Can we use the macro/builtin for this? 

https://github.com/llvm/llvm-project/pull/110179