Christopher Brown - Lapedo(...)
summary
A framework called Lapedo semi‑automatically turns sequential Erlang into hybrid CPU+GPU code using the Skel skeleton library, OpenCL bindings and the Wrangler refactoring tool; it generates OpenCL marshalling and offload boilerplate, profiles inputs to choose device scheduling, and lets users swap in hybrid farms/pipelines. Reported results show ~20× speedup on 24 CPU cores after program shaping and cases exceeding ~200× when a GPU is added, with other examples moving from ~7–10× to ~12–18×.
AI-written summary · about
chapters (18)
- 0:00Start
- 2:18Ultra Stratix
- 5:24A Parallel Pattern
- 7:39Parallel Pattern Library
- 8:13General Api
- 10:54Refactoring
- 11:50Refactoring Tool in Erlang
- 12:12Refactoring Tool
- 14:53Refactorings
- 15:41Hybrid Skeletons
- 16:01Hybrid Component
- 17:23Patent Discovery
- 19:36Program Shaping Step
- 23:03Program Shaping
- 25:06Opencl Marshalling Code
- 26:24Scale Pattern
- 35:26Contact Us
- 37:48Why Are We Comparing Opencl to Airline
related talks
42mGame of Threads - You Spawn or You Die
45mBack to Front and Beyond
30mBuilding Single Page Web Applications with Purescript and Erlang
01h 03mWhat I Talk About When I Talk About Cross Platform Development
19mBridging the Gap with Gleam & Lustre
46mBuilding a Graphical IDE in Elm/Purescript (...)
13mDemo: RISC-V as an Enabler of Heterogeneous Compute
36mOptimizing ClojureScript Apps For Speed