| 1 | //===-- GPU implementation of the nanosleep function ----------------------===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #include "src/time/nanosleep.h" |
| 10 | |
| 11 | #include "src/__support/common.h" |
| 12 | #include "src/__support/macros/config.h" |
| 13 | #include "src/__support/time/gpu/time_utils.h" |
| 14 | |
| 15 | namespace LIBC_NAMESPACE_DECL { |
| 16 | |
| 17 | LLVM_LIBC_FUNCTION(int, nanosleep, |
| 18 | (const struct timespec *req, struct timespec *rem)) { |
| 19 | if (!GPU_CLOCKS_PER_SEC || !req) |
| 20 | return -1; |
| 21 | |
| 22 | uint64_t nsecs = req->tv_nsec + req->tv_sec * TICKS_PER_SEC; |
| 23 | uint64_t tick_rate = TICKS_PER_SEC / GPU_CLOCKS_PER_SEC; |
| 24 | |
| 25 | uint64_t start = gpu::fixed_frequency_clock(); |
| 26 | #if defined(LIBC_TARGET_ARCH_IS_NVPTX) |
| 27 | uint64_t end = start + (nsecs + tick_rate - 1) / tick_rate; |
| 28 | uint64_t cur = gpu::fixed_frequency_clock(); |
| 29 | // The NVPTX architecture supports sleeping and guaruntees the actual time |
| 30 | // slept will be somewhere between zero and twice the requested amount. Here |
| 31 | // we will sleep again if we undershot the time. |
| 32 | while (cur < end) { |
| 33 | if (__nvvm_reflect("__CUDA_ARCH" ) >= 700) |
| 34 | LIBC_INLINE_ASM("nanosleep.u32 %0;" ::"r" (nsecs)); |
| 35 | cur = gpu::fixed_frequency_clock(); |
| 36 | nsecs -= nsecs > cur - start ? cur - start : 0; |
| 37 | } |
| 38 | #elif defined(LIBC_TARGET_ARCH_IS_AMDGPU) |
| 39 | uint64_t end = start + (nsecs + tick_rate - 1) / tick_rate; |
| 40 | uint64_t cur = gpu::fixed_frequency_clock(); |
| 41 | // The AMDGPU architecture does not provide a sleep implementation with a |
| 42 | // known delay so we simply repeatedly sleep with a large value of ~960 clock |
| 43 | // cycles and check until we've passed the time using the known frequency. |
| 44 | __builtin_amdgcn_s_sleep(2); |
| 45 | while (cur < end) { |
| 46 | __builtin_amdgcn_s_sleep(15); |
| 47 | cur = gpu::fixed_frequency_clock(); |
| 48 | } |
| 49 | #else |
| 50 | // Sleeping is not supported. |
| 51 | if (rem) { |
| 52 | rem->tv_sec = req->tv_sec; |
| 53 | rem->tv_nsec = req->tv_nsec; |
| 54 | } |
| 55 | return -1; |
| 56 | #endif |
| 57 | uint64_t stop = gpu::fixed_frequency_clock(); |
| 58 | |
| 59 | // Check to make sure we slept for at least the desired duration and set the |
| 60 | // remaining time if not. |
| 61 | uint64_t elapsed = (stop - start) * tick_rate; |
| 62 | if (elapsed < nsecs) { |
| 63 | if (rem) { |
| 64 | rem->tv_sec = (nsecs - elapsed) / TICKS_PER_SEC; |
| 65 | rem->tv_nsec = (nsecs - elapsed) % TICKS_PER_SEC; |
| 66 | } |
| 67 | return -1; |
| 68 | } |
| 69 | |
| 70 | return 0; |
| 71 | } |
| 72 | |
| 73 | } // namespace LIBC_NAMESPACE_DECL |
| 74 | |