Ediolot

#include <torch/extension.h>
#include <cstdint>

__global__ void kernel_vector_add(uint32_t N, float* a, float* b, float* c) {
    uint32_t idx = blockIdx.x * blockDim.x + threadIdx.x;
    uint32_t max_threads = gridDim.x * blockDim.x;

    for (uint32_t i = idx; i < N; i += max_threads) {
        c[i] = a[i] + b[i];
    }

Ediolot

#include <iostream>
#include <chrono>

__global__ void vectorAdd(const float *A, const float *B, float *C, uint32_t N) {
    const uint32_t threads_per_block = blockDim.x;
    const uint32_t total_blocks = gridDim.x;
    const uint32_t block_id = blockIdx.x;
    const uint32_t thread_id = threadIdx.x;

    const uint32_t total_threads = total_blocks * threads_per_block;

Ediolot

#include <iostream>

__global__ void vectorAdd(const float *A, const float *B, float *C, uint32_t N) {
    const uint32_t threads_per_block = blockDim.x;
    const uint32_t total_blocks = gridDim.x;
    const uint32_t block_id = blockIdx.x;
    const uint32_t thread_id = threadIdx.x;

    const uint32_t total_threads = total_blocks * threads_per_block;
    const uint32_t idx = block_id * threads_per_block + thread_id;

Ediolot

import numpy as np


class Vector:
    def __init__(self, w, h, d, n=None):
        self.w = w
        self.h = h
        self.d = d
        self.n = n
        self.half_w = w // 2