Adding Stack Mutable Types To The Spiral Language

Описание: Things are coming to a head, and as we proceed in our journey, we need to optimize what can be. Concerned about the register and local memory usage, we play around with various optimizations and in the end, the stack mutable layout types that we build into the language in this video result in significant improvements. A great advantage compared to what we were doing before is that it's very easy to switch between heap and stack mutable types. Previously we'd have to bundle individual references into a record to get this kind of functionality and that involved manual labor.

And if there is one thing programmers don't like, it's manual labor.

---

#spiral #functionalprogramming #machinelearning #reinforcementlearning #programming #cpp #programminglanguage #compiler #parallelprogramming #cuda #gpu

Playlist(Staged FP in Spiral):    • Staged Functional Programming In Spiral  
Spiral: https://github.com/mrakgr/The-Spiral-...
Spiral's ML Library: https://github.com/mrakgr/Spiral-s-ML...
Github: https://github.com/mrakgr/

If you have interesting work opportunities and require an expert functional programmer, don't hesitate to get in touch. My email is on my Github profile. Put "Work" as the subject in order to avoid the spam filters.

TOC:

00:00:00 - Start
00:00:02 - (Incorrect) performance figures for the new matrix multiplication kernel on the Leduc game
00:01:49 - Analysis of the actual number of registers being used
00:07:33 - Idea 1: Getting rid of table backups
00:11:55 - Idea 2: Stack mutable layout types
00:24:22 - Adding the new feature to the Spiral language
00:46:21 - Reminiscing about the reified join points in the 2018 version of Spiral
01:56:46 - Changing the projective to a linear loop in the matmult kernel
02:00:58 - No improved performance on Leduc from the improved matmult kernel
02:09:37 - Breaking change from v2.14
02:10:57 - The testing phase
03:19:28 - Successfully going from 85 to 51ms