Architecture & Systems Labs

I use diagrams, simulations and recorded traces to study how computers execute work. Pick a lab below to explore a GPU, a cluster, a multicore CPU or the runtime behind its simulator.

One Xe-HPG render slice contains four Xe-cores; a detail view shows the vector and matrix engines in one core.

Xe-Diagram →

Explore my Intel Xe GPU knowledge portfolio: start with a render slice, then look inside a Xe-core to see its vector and matrix engines.

A message travels from a POWER2 node through the SP2 switch to another node.

IBM SP2 →

Start inside a POWER2 processor, zoom out to an SP2 cluster, and replay the journey of a message between compute nodes.

Two modeled cores each have private caches and connect to a shared last-level cache.

Coffee Lake →

Step through a recorded CPU simulation inspired by the Intel Core i7-8700K to see how cores execute work and coordinate access to shared data.

An idle worker steals a ready task from another worker’s queue.

Inside oneTBB →

Step through a work-stealing example, then inspect the C++ and recorded task timelines to see how oneTBB runs the Coffee Lake simulator in parallel.