BEGIN:VCALENDAR
VERSION:2.0
PRODID:IEEE vTools.Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Asia/Kolkata
BEGIN:STANDARD
DTSTART:19451014T230000
TZOFFSETFROM:+0630
TZOFFSETTO:+0530
TZNAME:IST
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20261006T063734Z
UID:FE0C8BC0-4B5F-46BC-9414-1A70B49461F8
DTSTART;TZID=Asia/Kolkata:20261011T150000
DTEND;TZID=Asia/Kolkata:20261011T170000
DESCRIPTION:Modern processors execute billions of operations per second\, y
 et performance is often limited by how fast data can be supplied\, not how
  fast it can be computed. This session will examine that gap\, the memory 
 wall\, and the architecture that bridges it. It covers the memory hierarch
 y and locality\, SRAM and DRAM\, building larger memories from chips\, int
 erleaving and the memory address map\, and cache design: address mapping\,
  cache types\, replacement and write policies\, and the three types of mis
 ses. Performance is quantified with Average Memory Access Time (AMAT)\, in
 cluding multi-level caches. A live demonstration shows how loop order alon
 e changes execution time. Real-world examples (LLM KV cache and prefix cac
 hing\, browser\, OS and database caches) and an anonymised edge AI case st
 udy connect the concepts to modern systems.\n\nCo-sponsored by: Vellore In
 stitute of Technology\, Chennai\n\nVirtual: https://events.vtools.ieee.org
 /m/582768
LOCATION:Virtual: https://events.vtools.ieee.org/m/582768
ORGANIZER:pritam_bhattacharjee@live.com
SEQUENCE:72
SUMMARY:Technical Talk on &quot;Memory Hierarchy and Caches: Why Memory Decides 
 System Performance&quot;
URL;VALUE=URI:https://events.vtools.ieee.org/m/582768
X-ALT-DESC:Description: &lt;br /&gt;&lt;p style=&quot;text-align: justify\;&quot;&gt;Modern proce
 ssors execute billions of operations per second\, yet performance is often
  limited by how fast data can be supplied\, not how fast it can be compute
 d. This session will examine that gap\, the memory wall\, and the architec
 ture that bridges it. It covers the memory hierarchy and locality\, SRAM a
 nd DRAM\, building larger memories from chips\, interleaving and the memor
 y address map\, and cache design: address mapping\, cache types\, replacem
 ent and write policies\, and the three types of misses. Performance is qua
 ntified with Average Memory Access Time (AMAT)\, including multi-level cac
 hes. A live demonstration shows how loop order alone changes execution tim
 e. Real-world examples (LLM KV cache and prefix caching\, browser\, OS and
  database caches) and an anonymised edge AI case study connect the concept
 s to modern systems.&lt;/p&gt;
END:VEVENT
END:VCALENDAR

