GPUCodeForces/S1/gsd123_#43/CosReLU_cuda.py

84 lines
2.6 KiB
Python

import torch
import torch.nn as nn
from torch.utils.cpp_extension import load_inline
class ModelNew(nn.Module):
def __init__(self):
super().__init__()
self._compile_cuda_kernel()
def _compile_cuda_kernel(self):
cpp_source = """
torch::Tensor cosrelu_cuda(torch::Tensor x);
"""
cuda_source = """
#include <torch/extension.h>
#include <cuda_runtime.h>
#include <math.h>
__device__ __forceinline__ float cosrelu_op(float x) {
// f(x) = max(0, x) + cos(x)
return fmaxf(0.0f, x) + cosf(x);
}
__global__ void cosrelu_kernel(
const float* __restrict__ x,
float* __restrict__ output,
const int n_elements)
{
const int tid = blockIdx.x * blockDim.x + threadIdx.x;
const int stride = blockDim.x * gridDim.x;
const int vec_loops = n_elements >> 2;
const float4* x_vec = reinterpret_cast<const float4*>(x);
float4* out_vec = reinterpret_cast<float4*>(output);
for (int i = tid; i < vec_loops; i += stride) {
float4 v = __ldg(&x_vec[i]);
float4 r;
r.x = cosrelu_op(v.x);
r.y = cosrelu_op(v.y);
r.z = cosrelu_op(v.z);
r.w = cosrelu_op(v.w);
out_vec[i] = r;
}
const int tail_start = vec_loops << 2;
for (int i = tail_start + tid; i < n_elements; i += stride) {
output[i] = cosrelu_op(x[i]);
}
}
torch::Tensor cosrelu_cuda(torch::Tensor x) {
auto x_c = x.contiguous();
const int n_elements = x_c.numel();
auto output = torch::empty_like(x_c);
const int threads = 256;
const int max_blocks = 65535;
const int blocks = std::min((n_elements + threads * 4 - 1) / (threads * 4), max_blocks);
cosrelu_kernel<<<blocks, threads>>>(
x_c.data_ptr<float>(),
output.data_ptr<float>(),
n_elements
);
return output;
}
"""
self.op = load_inline(
name="cosrelu_v1",
cpp_sources=cpp_source,
cuda_sources=cuda_source,
functions=["cosrelu_cuda"],
extra_cuda_cflags=["-O3", "--use_fast_math"],
verbose=False
)
def forward(self, x):
return self.op.cosrelu_cuda(x)