Casey Destroys Optimization Myths | TheStandup
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Related videos
Go’s Big Mistake
The PrimeTime
37.9k views
Recovering from AI Psychosis | TheStandup
The PrimeTime
38.3k views
Mythos unleashed on Opensource
The PrimeTime
331.6k views
Microsoft doesn't understand FPS
The PrimeTime
47.2k views
The Mythos Situation | TheStandup
The PrimeTime
63.3k views
Linus Lays down the Law
The PrimeTime
118.5k views
What's really going on with AI, Expert weighs in | TheStandup
The PrimeTime
59.7k views
Why is Microsoft updating their text editors!? | TheStandup
The PrimeTime
59.9k views
OpenAI's New Era
The PrimeTime
217.3k views
AGI Achieved?! | TheStandup
The PrimeTime
85.3k views
Top Comments (10)
Casey's patience is legendary.
the only "always do x" i can think of that works for optimization is "always make sure it's slow before trying to optimize it"
Casey: "To be fair, I am the wrong one to explain it" *proceeds to give a university-level lesson about floating point*
I'm amazed Casey manages to keep it together with all the clowning and interruptions at TheStandup
The author of the tweet neglected to mention that they were measuring all of this on a Redstone CPU in Minecraft.
Awlays here for Casey taking a wrecking ball to sacred cows
always here for some more mergecop
If it was "always do X", the compiler would've done it for you
I was recently working on some CPU image processing stuff where the inverse-multiply thing actually *did* matter. But like Casey said, It was only once I verified that I wasn't memory-bound/everything was in L1, AND I was getting a lot of serialization. Even then for my algorithm it was only like a 5% speedup. Other stuff like making my loops play nice with the cache were like 50% speedups.
"Two web devs are bullying low level backend dev for 25 minutes"
Unlock the Data Inside
Turn Videos into Knowledge
- Get FREE 10/day: transcripts, summaries, chats
- Chat with videos, export text & PDF
- $1 free API credit for RAG, chatbots & research
Free forever plan • All features unlocked
Top Comments (10)
Casey's patience is legendary.
the only "always do x" i can think of that works for optimization is "always make sure it's slow before trying to optimize it"
Casey: "To be fair, I am the wrong one to explain it" *proceeds to give a university-level lesson about floating point*
I'm amazed Casey manages to keep it together with all the clowning and interruptions at TheStandup
The author of the tweet neglected to mention that they were measuring all of this on a Redstone CPU in Minecraft.
Awlays here for Casey taking a wrecking ball to sacred cows
always here for some more mergecop
If it was "always do X", the compiler would've done it for you
I was recently working on some CPU image processing stuff where the inverse-multiply thing actually *did* matter. But like Casey said, It was only once I verified that I wasn't memory-bound/everything was in L1, AND I was getting a lot of serialization. Even then for my algorithm it was only like a 5% speedup. Other stuff like making my loops play nice with the cache were like 50% speedups.
"Two web devs are bullying low level backend dev for 25 minutes"