miniblog.

← Back to all posts
Apr 4, 2018 at 12:59
The death of Dennard scaling, Moore's law, and the future of ISA design and performance: https://www.eejournal.com/article/fifty-or-sixty-years-of-processor-developmentfor-this/ (I loved Patterson book and he's the definitive authority here!)
Fifty (or Sixty) Years of Processor Development…for This?
Fifty (or Sixty) Years of Processor Development…for This?
“We cross our bridges when we come to them and burn them behind us, with nothing to show for our progress except a memory of the smell of smoke, and a presumption that once our eyes watered.” – Ros…

Related Posts

Jan 25, 2026 at 16:22
2
An AI benchmark website that tries to run comparable benchmarks regularly to discover when LLM performance is degrading:
AI Benchmark Tool - Best AI Models 2025 | Compare Claude vs GPT vs Gemini
AI Benchmark Tool - Best AI Models 2025 | Compare Claude vs GPT vs Gemini
Ultimate AI benchmarking tool for 2025. Test AI performance, compare AI models, and find the best AI for coding. Real-time AI benchmark results and LLM performance tests.
Apr 3, 2025 at 15:56
2
When writing long-lived programs (daemons etc) in Rust, I find myself asking *where* I should put data. In a GC'd language it's just "I have a string" but Rust forces me to find somewhere to put it. You do get a performance benefit for this work though.
Feb 4, 2025 at 16:32
I've just squeezed another 5% of performance out of difftastic by finding a few HashSet values that weren't FxHashSet. I do wonder whether hash DoS resistance is a good default. Sure, Rust programs are often pretty fast anyway, but it feels like a different threat model to the rest of Rust.