<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>cuda on Suriya's site</title><link>https://subygan.com/tech/algos/cuda/</link><description>Recent content in cuda on Suriya's site</description><generator>Hugo</generator><language>en</language><copyright>This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.</copyright><item><title>1d convolution</title><link>https://subygan.com/tech/algos/cuda/06_1d_convolution/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/06_1d_convolution/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-cpp" data-lang="cpp">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;iostream&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;stdio.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">convolution_1d_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> kernel, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output,
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> input_size, &lt;span style="color:#66d9ef">int&lt;/span> kernel_size) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> i &lt;span style="color:#f92672">=&lt;/span> blockDim.x &lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> output_size &lt;span style="color:#f92672">=&lt;/span> input_size &lt;span style="color:#f92672">-&lt;/span> kernel_size &lt;span style="color:#f92672">+&lt;/span>&lt;span style="color:#ae81ff">1&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> (i &lt;span style="color:#f92672">&amp;lt;&lt;/span> output_size ) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">float&lt;/span> sum &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">0.0f&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">for&lt;/span> ( &lt;span style="color:#66d9ef">int&lt;/span> j &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">0&lt;/span>; j&lt;span style="color:#f92672">&amp;lt;&lt;/span>kernel_size; j&lt;span style="color:#f92672">++&lt;/span>) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> sum &lt;span style="color:#f92672">+=&lt;/span> input[i&lt;span style="color:#f92672">+&lt;/span>j]&lt;span style="color:#f92672">*&lt;/span>kernel[j];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> }
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> output[i] &lt;span style="color:#f92672">=&lt;/span> sum;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> }
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input, kernel, output are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> kernel, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> input_size, &lt;span style="color:#66d9ef">int&lt;/span> kernel_size) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> output_size &lt;span style="color:#f92672">=&lt;/span> input_size &lt;span style="color:#f92672">-&lt;/span> kernel_size &lt;span style="color:#f92672">+&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (output_size &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> convolution_1d_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, kernel, output, input_size, kernel_size);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>color inversion</title><link>https://subygan.com/tech/algos/cuda/04_color_inversion/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/04_color_inversion/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-cpp" data-lang="cpp">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">invert_kernel&lt;/span>(&lt;span style="color:#66d9ef">unsigned&lt;/span> &lt;span style="color:#66d9ef">char&lt;/span>&lt;span style="color:#f92672">*&lt;/span> image, &lt;span style="color:#66d9ef">int&lt;/span> width, &lt;span style="color:#66d9ef">int&lt;/span> height) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> i &lt;span style="color:#f92672">=&lt;/span> blockIdx.x &lt;span style="color:#f92672">*&lt;/span> blockDim.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> image[i&lt;span style="color:#f92672">*&lt;/span>&lt;span style="color:#ae81ff">4&lt;/span>] &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">255&lt;/span> &lt;span style="color:#f92672">-&lt;/span> image[i&lt;span style="color:#f92672">*&lt;/span>&lt;span style="color:#ae81ff">4&lt;/span>];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> image[i&lt;span style="color:#f92672">*&lt;/span>&lt;span style="color:#ae81ff">4&lt;/span>&lt;span style="color:#f92672">+&lt;/span>&lt;span style="color:#ae81ff">1&lt;/span>] &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">255&lt;/span> &lt;span style="color:#f92672">-&lt;/span> image[i&lt;span style="color:#f92672">*&lt;/span>&lt;span style="color:#ae81ff">4&lt;/span>&lt;span style="color:#f92672">+&lt;/span>&lt;span style="color:#ae81ff">1&lt;/span>];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> image[i&lt;span style="color:#f92672">*&lt;/span>&lt;span style="color:#ae81ff">4&lt;/span>&lt;span style="color:#f92672">+&lt;/span>&lt;span style="color:#ae81ff">2&lt;/span>] &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">255&lt;/span> &lt;span style="color:#f92672">-&lt;/span> image[i&lt;span style="color:#f92672">*&lt;/span>&lt;span style="color:#ae81ff">4&lt;/span>&lt;span style="color:#f92672">+&lt;/span>&lt;span style="color:#ae81ff">2&lt;/span>];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// image_input, image_output are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">unsigned&lt;/span> &lt;span style="color:#66d9ef">char&lt;/span>&lt;span style="color:#f92672">*&lt;/span> image, &lt;span style="color:#66d9ef">int&lt;/span> width, &lt;span style="color:#66d9ef">int&lt;/span> height) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (width &lt;span style="color:#f92672">*&lt;/span> height &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> invert_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(image, width, height);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>count 2d array element</title><link>https://subygan.com/tech/algos/cuda/14_count_2d_array/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/14_count_2d_array/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-c++" data-lang="c++">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">count_2d_equal_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">int&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">int&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N, &lt;span style="color:#66d9ef">int&lt;/span> M, &lt;span style="color:#66d9ef">int&lt;/span> K) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> idx &lt;span style="color:#f92672">=&lt;/span> blockDim.x&lt;span style="color:#f92672">*&lt;/span>blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> idy &lt;span style="color:#f92672">=&lt;/span> blockDim.y&lt;span style="color:#f92672">*&lt;/span>blockIdx.y &lt;span style="color:#f92672">+&lt;/span> threadIdx.y;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> ((idx&lt;span style="color:#f92672">&amp;lt;&lt;/span>N &lt;span style="color:#f92672">&amp;amp;&amp;amp;&lt;/span> idy&lt;span style="color:#f92672">&amp;lt;&lt;/span>M) &lt;span style="color:#f92672">&amp;amp;&amp;amp;&lt;/span> input[idx&lt;span style="color:#f92672">*&lt;/span>M&lt;span style="color:#f92672">+&lt;/span>idy]&lt;span style="color:#f92672">==&lt;/span>K) atomicAdd(output,&lt;span style="color:#ae81ff">1&lt;/span>);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input, output are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">int&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">int&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N, &lt;span style="color:#66d9ef">int&lt;/span> M, &lt;span style="color:#66d9ef">int&lt;/span> K) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> dim3 &lt;span style="color:#a6e22e">threadsPerBlock&lt;/span>(&lt;span style="color:#ae81ff">16&lt;/span>, &lt;span style="color:#ae81ff">16&lt;/span>);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> dim3 &lt;span style="color:#a6e22e">blocksPerGrid&lt;/span>((N &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock.x &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock.x,
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> (M &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock.y &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock.y);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> count_2d_equal_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, output, N, M, K);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>count array element</title><link>https://subygan.com/tech/algos/cuda/13_count_array/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/13_count_array/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-c++" data-lang="c++">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">count_equal_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">int&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">int&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N, &lt;span style="color:#66d9ef">int&lt;/span> K) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> id&lt;span style="color:#f92672">=&lt;/span>blockDim.x &lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span>(id&lt;span style="color:#f92672">&amp;lt;&lt;/span>N &lt;span style="color:#f92672">&amp;amp;&amp;amp;&lt;/span> input[id]&lt;span style="color:#f92672">==&lt;/span>K) atomicAdd(output, &lt;span style="color:#ae81ff">1&lt;/span>);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input, output are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">int&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">int&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N, &lt;span style="color:#66d9ef">int&lt;/span> K) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (N &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> count_equal_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, output, N, K);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>leaky relu</title><link>https://subygan.com/tech/algos/cuda/09_leaky_relu/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/09_leaky_relu/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-c++" data-lang="c++">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">leaky_relu_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> id &lt;span style="color:#f92672">=&lt;/span> blockDim.x&lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> (id&lt;span style="color:#f92672">&amp;lt;&lt;/span>N) output[id] &lt;span style="color:#f92672">=&lt;/span> input[id]&lt;span style="color:#f92672">&amp;gt;&lt;/span>&lt;span style="color:#ae81ff">0&lt;/span>&lt;span style="color:#f92672">?&lt;/span> input[id] &lt;span style="color:#f92672">:&lt;/span> &lt;span style="color:#ae81ff">0.01&lt;/span>&lt;span style="color:#f92672">*&lt;/span>input[id];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input, output are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (N &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> 
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> leaky_relu_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, output, N);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>matrix addition</title><link>https://subygan.com/tech/algos/cuda/05_matrix_addition/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/05_matrix_addition/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-c++" data-lang="c++">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">matrix_add&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> A, &lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> B, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> C, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> id &lt;span style="color:#f92672">=&lt;/span> blockDim.x &lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> (id&lt;span style="color:#f92672">&amp;lt;&lt;/span>N) C[id] &lt;span style="color:#f92672">=&lt;/span> A[id]&lt;span style="color:#f92672">+&lt;/span>B[id];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// A, B, C are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> A, &lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> B, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> C, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (N &lt;span style="color:#f92672">*&lt;/span> N &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> matrix_add&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(A, B, C, N);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>matrix copy</title><link>https://subygan.com/tech/algos/cuda/11_matrix_copy/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/11_matrix_copy/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-c++" data-lang="c++">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;stdio.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;iostream&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">copy_matrix_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> A, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> B, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> idx &lt;span style="color:#f92672">=&lt;/span> blockDim.x&lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> (idx&lt;span style="color:#f92672">&amp;lt;&lt;/span>N){
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> B[idx] &lt;span style="color:#f92672">=&lt;/span> A[idx];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> } 
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// A, B are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> A, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> B, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> total &lt;span style="color:#f92672">=&lt;/span> N &lt;span style="color:#f92672">*&lt;/span> N;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (total &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> copy_matrix_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(A, B, total);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>} 
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>matrix multiply</title><link>https://subygan.com/tech/algos/cuda/02_matrix_multipy/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/02_matrix_multipy/</guid><description>&lt;p>The tricky part in this was that, I took some time to intuit that each thread should associate to each value in the output matrix.&lt;/p></description></item><item><title>matrix transpose</title><link>https://subygan.com/tech/algos/cuda/03_matrix_transpose/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/03_matrix_transpose/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-cpp" data-lang="cpp">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;iostream&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;stdio.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">matrix_transpose_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> rows, &lt;span style="color:#66d9ef">int&lt;/span> cols) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> i &lt;span style="color:#f92672">=&lt;/span> blockIdx.x &lt;span style="color:#f92672">*&lt;/span> blockDim.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> j &lt;span style="color:#f92672">=&lt;/span> blockIdx.y &lt;span style="color:#f92672">*&lt;/span> blockDim.y &lt;span style="color:#f92672">+&lt;/span> threadIdx.y;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> printf(&lt;span style="color:#e6db74">&amp;#34;%d&amp;#34;&lt;/span>,i);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> (i&lt;span style="color:#f92672">&amp;lt;&lt;/span>rows &lt;span style="color:#f92672">&amp;amp;&amp;amp;&lt;/span> j&lt;span style="color:#f92672">&amp;lt;&lt;/span>cols){
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> output[j&lt;span style="color:#f92672">*&lt;/span>rows&lt;span style="color:#f92672">+&lt;/span>i] &lt;span style="color:#f92672">=&lt;/span> input[i&lt;span style="color:#f92672">*&lt;/span>cols&lt;span style="color:#f92672">+&lt;/span>j];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> }
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input, output are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> rows, &lt;span style="color:#66d9ef">int&lt;/span> cols) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> dim3 &lt;span style="color:#a6e22e">threadsPerBlock&lt;/span>(&lt;span style="color:#ae81ff">16&lt;/span>, &lt;span style="color:#ae81ff">16&lt;/span>);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> dim3 &lt;span style="color:#a6e22e">blocksPerGrid&lt;/span>((cols &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock.x &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock.x,
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> (rows &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock.y &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock.y);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> matrix_transpose_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, output, rows, cols);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>rainbow table</title><link>https://subygan.com/tech/algos/cuda/10_rainbow_table/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/10_rainbow_table/</guid><description/></item><item><title>relu</title><link>https://subygan.com/tech/algos/cuda/08_ReLu/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/08_ReLu/</guid><description>&lt;p>&lt;code>ReLU(x) = max(0,x)&lt;/code>&lt;/p>
&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-c++" data-lang="c++">&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">relu_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> id &lt;span style="color:#f92672">=&lt;/span> blockDim.x&lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> (id&lt;span style="color:#f92672">&amp;lt;&lt;/span>N) output[id] &lt;span style="color:#f92672">=&lt;/span> std&lt;span style="color:#f92672">::&lt;/span>max(&lt;span style="color:#ae81ff">0.0f&lt;/span>,input[id]);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input, output are device pointers (i.e. pointers to memory on the GPU)
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (N &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> relu_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, output, N);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>reverse array</title><link>https://subygan.com/tech/algos/cuda/07_reverse_array/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/07_reverse_array/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-c++" data-lang="c++">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;iostream&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;stdio.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">reverse_array&lt;/span>(&lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> i &lt;span style="color:#f92672">=&lt;/span> blockDim.x &lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> (i&lt;span style="color:#f92672">&amp;lt;&lt;/span>N&lt;span style="color:#f92672">/&lt;/span>&lt;span style="color:#ae81ff">2&lt;/span>) { 
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> std&lt;span style="color:#f92672">::&lt;/span>swap(input[i],input[N&lt;span style="color:#f92672">-&lt;/span>i&lt;span style="color:#f92672">-&lt;/span>&lt;span style="color:#ae81ff">1&lt;/span>]);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> }
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input is device pointer
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (N &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> reverse_array&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, N);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>silu</title><link>https://subygan.com/tech/algos/cuda/15_sigmoid_linear_unit/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/15_sigmoid_linear_unit/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-cpp" data-lang="cpp">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">silu_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> id &lt;span style="color:#f92672">=&lt;/span> blockDim.x &lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span> (id&lt;span style="color:#f92672">&amp;lt;&lt;/span>N) output[id] &lt;span style="color:#f92672">=&lt;/span> input[id]&lt;span style="color:#f92672">*&lt;/span>(&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">/&lt;/span>(&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">+&lt;/span>std&lt;span style="color:#f92672">::&lt;/span>exp(&lt;span style="color:#f92672">-&lt;/span>&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">*&lt;/span>(input[id]))));
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input, output are device pointers
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (N &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> silu_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, output, N);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>simple inference</title><link>https://subygan.com/tech/algos/cuda/12_simple_inference/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/12_simple_inference/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-python" data-lang="python">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#f92672">import&lt;/span> torch
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#f92672">import&lt;/span> torch.nn &lt;span style="color:#66d9ef">as&lt;/span> nn
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e"># input, model, and output are on the GPU&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#66d9ef">def&lt;/span> &lt;span style="color:#a6e22e">solve&lt;/span>(input: torch&lt;span style="color:#f92672">.&lt;/span>Tensor, model: nn&lt;span style="color:#f92672">.&lt;/span>Module, output: torch&lt;span style="color:#f92672">.&lt;/span>Tensor):
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> model&lt;span style="color:#f92672">.&lt;/span>eval()
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">with&lt;/span> torch&lt;span style="color:#f92672">.&lt;/span>no_grad():
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> output&lt;span style="color:#f92672">.&lt;/span>copy_(model(input))
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>swilu</title><link>https://subygan.com/tech/algos/cuda/16_swish_gated_linear_unit/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/16_swish_gated_linear_unit/</guid><description>&lt;div class="highlight">&lt;pre tabindex="0" style="color:#f8f8f2;background-color:#272822;-moz-tab-size:4;-o-tab-size:4;tab-size:4;">&lt;code class="language-c++" data-lang="c++">&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">#include&lt;/span> &lt;span style="color:#75715e">&amp;lt;cuda_runtime.h&amp;gt;&lt;/span>&lt;span style="color:#75715e">
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#66d9ef">float&lt;/span> &lt;span style="color:#a6e22e">silu&lt;/span>(&lt;span style="color:#66d9ef">float&lt;/span> x){
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">return&lt;/span> x&lt;span style="color:#f92672">*&lt;/span>(&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">/&lt;/span>(&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">+&lt;/span>std&lt;span style="color:#f92672">::&lt;/span>exp(&lt;span style="color:#f92672">-&lt;/span>&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">*&lt;/span>(x))));
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>__global__ &lt;span style="color:#66d9ef">void&lt;/span> &lt;span style="color:#a6e22e">swiglu_kernel&lt;/span>(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> halfN) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> id &lt;span style="color:#f92672">=&lt;/span> blockDim.x &lt;span style="color:#f92672">*&lt;/span> blockIdx.x &lt;span style="color:#f92672">+&lt;/span> threadIdx.x;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">if&lt;/span>(id&lt;span style="color:#f92672">&amp;lt;&lt;/span>halfN) output[id] &lt;span style="color:#f92672">=&lt;/span> input[id&lt;span style="color:#f92672">%&lt;/span>halfN]&lt;span style="color:#f92672">*&lt;/span>(&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">/&lt;/span>(&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">+&lt;/span>std&lt;span style="color:#f92672">::&lt;/span>exp(&lt;span style="color:#f92672">-&lt;/span>&lt;span style="color:#ae81ff">1&lt;/span>&lt;span style="color:#f92672">*&lt;/span>(input[id&lt;span style="color:#f92672">%&lt;/span>halfN])))) &lt;span style="color:#f92672">*&lt;/span> input[id&lt;span style="color:#f92672">%&lt;/span>halfN &lt;span style="color:#f92672">+&lt;/span> halfN];
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">// input, output are device pointers
&lt;/span>&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>&lt;span style="color:#75715e">&lt;/span>&lt;span style="color:#66d9ef">extern&lt;/span> &lt;span style="color:#e6db74">&amp;#34;C&amp;#34;&lt;/span> &lt;span style="color:#66d9ef">void&lt;/span> solve(&lt;span style="color:#66d9ef">const&lt;/span> &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> input, &lt;span style="color:#66d9ef">float&lt;/span>&lt;span style="color:#f92672">*&lt;/span> output, &lt;span style="color:#66d9ef">int&lt;/span> N) {
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> halfN &lt;span style="color:#f92672">=&lt;/span> N &lt;span style="color:#f92672">/&lt;/span> &lt;span style="color:#ae81ff">2&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> threadsPerBlock &lt;span style="color:#f92672">=&lt;/span> &lt;span style="color:#ae81ff">256&lt;/span>;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> &lt;span style="color:#66d9ef">int&lt;/span> blocksPerGrid &lt;span style="color:#f92672">=&lt;/span> (halfN &lt;span style="color:#f92672">+&lt;/span> threadsPerBlock &lt;span style="color:#f92672">-&lt;/span> &lt;span style="color:#ae81ff">1&lt;/span>) &lt;span style="color:#f92672">/&lt;/span> threadsPerBlock;
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> swiglu_kernel&lt;span style="color:#f92672">&amp;lt;&amp;lt;&amp;lt;&lt;/span>blocksPerGrid, threadsPerBlock&lt;span style="color:#f92672">&amp;gt;&amp;gt;&amp;gt;&lt;/span>(input, output, halfN);
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span> cudaDeviceSynchronize();
&lt;/span>&lt;/span>&lt;span style="display:flex;">&lt;span>}
&lt;/span>&lt;/span>&lt;/code>&lt;/pre>&lt;/div></description></item><item><title>vector addition</title><link>https://subygan.com/tech/algos/cuda/01_vector_addition/</link><pubDate>Wed, 12 Nov 2025 00:00:00 +0000</pubDate><guid>https://subygan.com/tech/algos/cuda/01_vector_addition/</guid><description>&lt;p>The question was&lt;/p>
&lt;pre tabindex="0">&lt;code>Implement a program that performs element-wise addition of two vectors containing 32-bit floating point numbers on a GPU. The program should take two input vectors of equal length and produce a single output vector containing their sum.

Implementation Requirements
External libraries are not permitted
The solve function signature must remain unchanged
The final result must be stored in vector C
&lt;/code>&lt;/pre>&lt;p>Identifying the thread uniquely is always done like using
&lt;code>blockIdx.x * blockDim.x + threadIdx.x&lt;/code>&lt;/p></description></item></channel></rss>