finished convolution TIL

Author: secondspass

extras/tiled_convolution_visualization.ipynb

@@ -1,5 +1,23 @@
 {
  "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "348c3cea-e7e6-459d-ab18-c76f13fa651f",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Requirement already satisfied: drawsvg in ./vizenv/lib/python3.10/site-packages (2.3.0)\n"
+     ]
+    }
+   ],
+   "source": [
+    "!pip install drawsvg"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 1,

img/TILs/TIL_parallel_convolution_with_tiling/TIL_parallel_convolution_block_overlap.jpg

@@ -19,7 +19,7 @@
 
 
 
-<meta property="og:updated_time" content="2024-03-10T00:00:00&#43;00:00"/>
+<meta property="og:updated_time" content="2024-05-29T00:00:00&#43;00:00"/>
 
 
 

img/TILs/TIL_parallel_convolution_with_tiling/TIL_parallel_convolution_calculating_output_border.jpg

@@ -7,11 +7,22 @@
     <generator>Hugo -- gohugo.io</generator>
     <language>en-US</language>
     <copyright>Copyright (c) 2016 - 2019. Licensed under CC BY-NC-SA 4.0</copyright>
-    <lastBuildDate>Sun, 10 Mar 2024 00:00:00 +0000</lastBuildDate>
+    <lastBuildDate>Wed, 29 May 2024 00:00:00 +0000</lastBuildDate>
 
 	<atom:link href="/index.xml" rel="self" type="application/rss+xml" />
 
 
+    <item>
+      <title>TIL Parallel Convolution with Tiling in GPUs</title>
+      <link>/tils/til_parallel_convolution_with_tiling/</link>
+      <pubDate>Wed, 29 May 2024 00:00:00 +0000</pubDate>
+      
+      <guid>/tils/til_parallel_convolution_with_tiling/</guid>
+      <description>In the continued saga of &amp;ldquo;Subil learns how to program GPUs&amp;rdquo;, today I learned about convolution and how to program it for a GPU.
+What is convolution? Well, I learned just enough so that I&amp;rsquo;m satisfied with how it works, but definitely not enough to explain it with any depth in this post. See this article or this video for explanations.
+For my purposes, convolution is this: For each input element in a 2D matrix of size NxN, apply a function on it that takes that input element and some of its neighboring elements (so your input to the function is a smaller nxn matrix with your input element at the center), multiplies this smaller matrix by a another known nxn matrix called a convolution kernel or convolution filter and sums these multiplied elements together to produce one output element.</description>
+    </item>
+    
     <item>
       <title>TIL corner turning in GPUs</title>
       <link>/tils/til_corner_turning/</link>

img/TILs/TIL_parallel_convolution_with_tiling/TIL_parallel_convolution_convolution_example.jpg

@@ -2,6 +2,11 @@
 <urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"
   xmlns:xhtml="http://www.w3.org/1999/xhtml">
 
+  <url>
+    <loc>/tils/til_parallel_convolution_with_tiling/</loc>
+    <lastmod>2024-05-29T00:00:00+00:00</lastmod>
+  </url>
+  
   <url>
     <loc>/tils/til_corner_turning/</loc>
     <lastmod>2024-03-10T00:00:00+00:00</lastmod>
@@ -166,7 +171,7 @@
 
   <url>
     <loc>/tags/gpus/</loc>
-    <lastmod>2024-03-10T00:00:00+00:00</lastmod>
+    <lastmod>2024-05-29T00:00:00+00:00</lastmod>
     <priority>0</priority>
   </url>
 
@@ -232,13 +237,13 @@
 
   <url>
     <loc>/</loc>
-    <lastmod>2024-03-10T00:00:00+00:00</lastmod>
+    <lastmod>2024-05-29T00:00:00+00:00</lastmod>
     <priority>0</priority>
   </url>
 
   <url>
     <loc>/tils/</loc>
-    <lastmod>2024-03-10T00:00:00+00:00</lastmod>
+    <lastmod>2024-05-29T00:00:00+00:00</lastmod>
     <priority>0</priority>
   </url>
 
@@ -250,7 +255,7 @@
 
   <url>
     <loc>/tags/til/</loc>
-    <lastmod>2024-03-10T00:00:00+00:00</lastmod>
+    <lastmod>2024-05-29T00:00:00+00:00</lastmod>
     <priority>0</priority>
   </url>