[INFRA] Import NVIDIA/CCCL upstream as optimization reference library
CCCL (CUDA C++ Core Libraries) provides: - CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk) - Thrust: high-level parallel algorithms (transform_reduce, sort, scan) - libcudacxx: CUDA C++ standard library (atomics, barriers, memory) - cudax: experimental features (memory resources, allocators) - Tuning policies: per-SM hardware-specific algorithm parameters Competition optimization vectors mapped to CCCL: - Output TPS (83% weight): warp_reduce, block_reduce, device_topk - Input TPS (14% weight): device_scan, block_load, prefetch - Cache TPS (3% weight): prefix caching strategy patterns - Memory (0.9 util): pooled/cached/buddy allocators Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only) License: Apache-2.0
This commit is contained in:
100
cccl_upstream/thrust/examples/include/timer.h
Normal file
100
cccl_upstream/thrust/examples/include/timer.h
Normal file
@@ -0,0 +1,100 @@
|
||||
// SPDX-FileCopyrightText: Copyright (c) 2008-2009, NVIDIA Corporation. All rights reserved.
|
||||
// SPDX-License-Identifier: Apache-2.0
|
||||
|
||||
#pragma once
|
||||
|
||||
// A simple timer class
|
||||
|
||||
#ifdef __CUDACC__
|
||||
|
||||
// use CUDA's high-resolution timers when possible
|
||||
# include <thrust/system/cuda/error.h>
|
||||
# include <thrust/system_error.h>
|
||||
|
||||
# include <string>
|
||||
|
||||
# include <cuda_runtime_api.h>
|
||||
|
||||
void cuda_safe_call(cudaError_t error, const std::string& message = "")
|
||||
{
|
||||
if (error)
|
||||
{
|
||||
throw thrust::system_error(error, thrust::cuda_category(), message);
|
||||
}
|
||||
}
|
||||
|
||||
struct timer
|
||||
{
|
||||
cudaEvent_t start;
|
||||
cudaEvent_t end;
|
||||
|
||||
timer()
|
||||
{
|
||||
cuda_safe_call(cudaEventCreate(&start));
|
||||
cuda_safe_call(cudaEventCreate(&end));
|
||||
restart();
|
||||
}
|
||||
|
||||
~timer()
|
||||
{
|
||||
static_cast<void>(cudaEventDestroy(start));
|
||||
static_cast<void>(cudaEventDestroy(end));
|
||||
}
|
||||
|
||||
void restart()
|
||||
{
|
||||
cuda_safe_call(cudaEventRecord(start, nullptr));
|
||||
}
|
||||
|
||||
double elapsed()
|
||||
{
|
||||
cuda_safe_call(cudaEventRecord(end, nullptr));
|
||||
cuda_safe_call(cudaEventSynchronize(end));
|
||||
|
||||
float ms_elapsed;
|
||||
cuda_safe_call(cudaEventElapsedTime(&ms_elapsed, start, end));
|
||||
return ms_elapsed / 1e3;
|
||||
}
|
||||
|
||||
double epsilon()
|
||||
{
|
||||
return 0.5e-6;
|
||||
}
|
||||
};
|
||||
|
||||
#else
|
||||
|
||||
// fallback to clock()
|
||||
# include <ctime>
|
||||
|
||||
struct timer
|
||||
{
|
||||
clock_t start;
|
||||
clock_t end;
|
||||
|
||||
timer()
|
||||
{
|
||||
restart();
|
||||
}
|
||||
|
||||
~timer() = default;
|
||||
|
||||
void restart()
|
||||
{
|
||||
start = clock();
|
||||
}
|
||||
|
||||
double elapsed()
|
||||
{
|
||||
end = clock();
|
||||
|
||||
return static_cast<double>(end - start) / static_cast<double>(CLOCKS_PER_SEC);
|
||||
}
|
||||
|
||||
double epsilon()
|
||||
{
|
||||
return 1.0 / static_cast<double>(CLOCKS_PER_SEC);
|
||||
}
|
||||
};
|
||||
|
||||
#endif
|
||||
Reference in New Issue
Block a user