#performance

Microsoft announced TypeScript 7.0, featuring a major compiler rewrite that delivers up to 12x faster build times across large codebases like VS Code, Sentry, and Playwright. This release dramatically reduces compilation time for TypeScript projects, which is critical for developer productivity in large-scale JavaScript/TypeScript applications. It also signals a new era of compiler performance for the ecosystem. Based on benchmarks, TypeScript 7.0 achieved 11.9x speedup on the VS Code codebase (from 125.7s to 10.6s), 8.9x on Sentry, and 8.7x on Playwright. The release also includes new language features, though details were not specified in the brief.

ClickHouse published a blog post describing how they scaled PgBouncer, a popular PostgreSQL connection pooler, to 4x its original throughput by enabling SOREUSEPORT and implementing peering between processes. PgBouncer is widely used in PostgreSQL deployments, and this performance improvement benefits many users who rely on it for connection pooling. The techniques used can inspire similar optimizations in other network services. SOREUSEPORT allows multiple processes to bind to the same TCP port, distributing incoming connections evenly. Peering forwards cancellation requests to the correct process, eliminating bottlenecks and improving scalability.

QBE 1.3 was released on June 2, 2026, adding roughly 7,000 lines of code while removing 1,500, introducing metaprogramming capabilities, performance improvements to close the CoreMark gap, and expanded cross-platform support. This release brings QBE closer to its goal of achieving 70% of GCC -O2 performance, making it a more viable compact backend for systems programming and compiler projects like Hare and cproc. The performance improvements focus on optimizing hot functions in the CoreMark benchmark, such as eeisdigit() and crcu8(), through classic optimization passes. Metaprogramming simplifies backend development by generating code automatically.

Puranjay Mohan presented work on improving RCU performance by allowing normal RCU callbacks to be executed after expedited grace periods, and a new kmallocnolock() function for lockless memory allocation from any kernel context was discussed at LSFMM+BPF 2026. These developments significantly enhance kernel scalability by reducing memory allocation latency and speeding up RCU grace period completion, benefiting high-performance workloads under memory pressure. The RCU improvement involves tracking both non-expedited and expedited grace-period numbers to allow callbacks to run when either completes, while kmallocnolock() enables lockless allocation without needing to hold locks.

A project called WATaBoy demonstrates that JIT-compiling Game Boy CPU instructions to WebAssembly can achieve faster performance than native interpreters. This approach leverages the web browser's existing JIT infrastructure for WebAssembly. This is significant because it shows a novel way to overcome platform restrictions (e.g., iOS blocking JIT) by using WebAssembly as a portable, high-performance target. It could enable smoother emulation on constrained platforms and inspire similar approaches for other retro systems. The project notes that while native interpreters can have roughly 1000% overhead, WebAssembly JIT overhead is around 20%, making the JIT approach faster. However, Firefox was observed to be 25% slower than Chrome/Safari for this workload.