The CPU Is Back in LLM Inference—Rethink the Split
artificial intelligence Aug 08, 2026 7 min read

The CPU Is Back in LLM Inference—Rethink the Split

LLM inference no longer looks like “CPU submits, GPU generates.” As assistants become agentic, tool calls and orchestration move onto the critical path, making CPU-heavy serving practical. This post breaks down prefill/decode, KV cache, and the software techniques (continuous batching, prefix caching, chunked prefill) that make CPU inference viable.

by ahsan
When grep crashes: tracing ripgrep + MUSL mallocng segfaults in huge directory walks
systems programming Aug 01, 2026 6 min read

When grep crashes: tracing ripgrep + MUSL mallocng segfaults in huge directory walks

A reported ripgrep crash affects x86_64 MUSL builds during massive directory walks. The SIGSEGV comes from musl mallocng integrity checks triggered via calloc→opendir while Rust threads iterate directories. The takeaway: allocator invariants can fail under extreme size + concurrency, and the stack trace points straight at that boundary.

by ahsan