Peaked at #3 Now #4
LLM Optimizes ffmpeg Blur 50x Faster
Developer uses an LLM to optimize ffmpeg's dblur, cutting per-frame time from 2452ms to 49ms.
Key points
- @RomainHedouin says ffmpeg blurring copied the same chunk of memory as many times as the frame's pixel width, moving 95 GB of data per frame.
- The human-written code took 2452 ms per frame, while the human+LLM version took 49 ms.
- The improvement was made in a project called "dblur".
Key points and the reaction summary are written by AI from the posts on this page. Check the original post. How we use AI
Original post
The ffmpeg saga continues with "dblur" & an improvement I found with an LLM
Blurring was a bit slow
That's because it copied the same chunk of memory as many times as the frame's pixel width... for no reason 💀
95 GB of data per frame
Human code: 2452 ms
Human+LLM: 49 ms https://t.co/B4w3KyP2PF
Top replies on X
Reaction on X: The top replies, including two from the original poster, largely agree: @RomainHedouin argues the compiler excuse doesn't apply here and says 2.5 seconds per frame makes the code go from unusable to very fast, while other replies joke about humans writing "slop" too and about copying repeatedly so the computer doesn't forget.
Sign in to see 5 top replies from X
From @RomainHedouin, @RomainHedouin, @xdNiBoR and others. Spam removed, with English and Chinese translations.
Discussion 0
Sign up Sign in to join the discussion
No comments yet. Start the conversation.