I make computer games and music plugins. Formerly lead dev @RiverCityRansom.
Brisbane, Australia
Joined January 2009
- Tweets2.7K
- Following46
- Followers700
- Likes219
Related to the recent "94ms to open a Run dialog" discussion: Win11 added a snipping utility to the PrintSc key. Today I measured it: it takes 1 second to start (and then another ~200ms of animation).
I just updated and Dual CV Source, and released it with a huge price drop. And my other Rack Extension devices are now free. @reasonstudios
Read more on my blog: andrewrussell.net/2025/02/du…
Andrew Russell retweeted
I'm tired of seeing lazy "software performance doesn't matter" excuses on forums and social media, so I devoted an entire video to the mountains of evidence debunking that ridiculous idea: computerenhance.com/p/perfor…
I just, by chance, stumbled on another one of these comparisons from Microsoft. I'll give you one guess, @cmuratori, what the new start-up time is 😂
Modern software performance is firmly in its "beyond parody" stage.
youtube.com/watch?v=CT7nnXej…
For anyone playing along at home, here's context with timestamp: youtu.be/GC-0tCy4P1U?t=1076
So, clearly they've been listening to that collected user data 🤣
The discussion between Casey and Bob on Clean Code is fascinating. I am a big fan of analysis of the form “it produces N ‘things’”. I personally like “I have to look in N places to follow X logic”.
The conversation between me and @unclebobmartin has concluded. The second (final) part is here:
github.com/cmuratori/misc/bl…
It is interesting to consider that N can grow unbounded, and so it would be wise to keep it in check. But, except for the most BASIC of spaghetti code, the complexity of ‘things’ is typically self-limiting.
Andrew Russell retweeted
By doing this you make the easy job easier. If the polymorphic, 'clean' API does what you need, and the internals are all correct, then you don't need to investigate deeper.
But when the API is wrong, when the internals have bugs, basically when things are complex and hard ...
Casey @cmuratori has put out a great video on the performance perils of "clean code" recently. I want to add some thoughts.
A common mistake I see: people measuring properties of a (maybe) successful system, e.g.: 'it is clean' and extrapolating that into a prescription.
From this perspective, the "clean code" ideas are not actually a methodology for making maintainable code.
What they are is extra work and extra complexity you have to add, to fulfil a requirement that you almost certainly don't have!
For completeness, a link to Casey's tweet: xitter.cf/cmuratori/status/16304…
Using their own example code, I made a simple demonstration of how horrible "Clean Code" ideas are for software performance:
computerenhance.com/p/clean-…
