DS2 speed compared to 4gl.

[ciaran2]

It's not about theory, it's about real application.
Does anyone have experience that code written in DS2 actually executed faster than similar code written in 4gl?
Examples found on the Internet are artificial and not useful in practical use. Has anyone actually had faster code that executes multi-threaded?

[yabwon]

Did you read this one:

https://www.lexjansen.com/mwsug/2019/SP/MWSUG-2019-SP-031.pdf

Bart

[SASKiwi]

What is your actual use case? There are many more ways to speed up programs than just to look at multi-threading. In my experience most SAS processing is IO constrained unless you are using computationally intensive operations. Tuning IO can have a major impact on performance.

[ciaran2]

You write about your experience. So you had examples where multithreading gave good results? What was the problem you managed to solve?

[SASKiwi]

@ciaran2  - No I don't have any experiences to share regarding multi-threading. Have you read the paper linked above? Most DS2 multi-threaded examples tried were slower than a single-threaded version.

In my experience minimising and tuning IO results in significant performance improvements. Also if you really want to speed up a SAS application, parallel processing, if appropriate, can reduce processing time massively. I have one parallel processing job that originally took over 14 hours to run and now takes under 4 hours. The nice thing about parallel processing is you don't need to recode you just split it into separate SAS sessions.