Previously, it took approximately 8 to 17 cycles, but now we have reduced it to less than one cycle, as illustrated in the above gif. We have successfully minimized the number of progress bar cycles required for the Lightbulb scenario in a large source file. We have made substantial enhancements to the performance of both the CPU and memory, specifically when dealing with extensive source files. High-level fixes included a complete overhaul of the diagnostics pipeline to answer span-oriented questions, infer which analyzers are applicable to run on specific spans, and deprioritize slow analyzers based on historical run.Medium-level fixes included cancelling background diagnostics (squiggles) when computing Lightbulb diagnostics, optimizing background indexing, limiting analyzers that are applicable per document, addressing inefficiencies in several 1st party/2nd party analyzers like PublicAPI analyzer, and improving our caching model.
#VISUAL STUDIO FOR MAC EXTENSIONS CODE#
Low-level fixes included limiting tree walks, improving string handling, optimizing source text handling, reducing allocations, and optimizing data structure-level optimizations in the diagnostics code paths.
We have focused on optimizing all components involved in bringing up a lightbulb through a combination of low/medium/high level fixes. The lightbulb/code actions have been greatly improved through comprehensive redesign of the diagnostics and code analysis architecture. Additionally, we’ve enhanced the way Visual Studio handles calls to git for this specific case. We’ve further reduced the time it takes to switch between git branches in Visual Studio by optimizing file system changes detection and processing. We’re excited to share that our measurements indicate a 42% improvement in the Git public repository. Git repositories with longer history and more time between commits will see greater speed improvements. We are pleased to announce that the Git History feature in Visual Studio has been enhanced to provide faster loading times for file and folder scenarios. For instance, in Chromium projects, the improvements result in a 50% acceleration, saving around 20 seconds of time. These improvements provide a 20% to 50% speed boost in closing solutions, depending on their size, with the most significant performance gains observed in larger projects (1000+ projects). These enhancements greatly benefit larger projects, enhancing productivity and reducing loading times.Įnhancements have been made to the Solution-Close process, which notably accelerates closing a solution with C++ projects. Additionally, for a Solution Filters scenario with 63 projects, Chromium loads in just 20 seconds, which is 3x faster. The cached Chromium solution now loads in approximately 30 seconds. Testing with the Chromium repository, which consists of 11000 projects, showed a remarkable 8x faster cached project load time, saving 3.5 minutes in Solution Load time. Significant enhancements have been made to Solution-Open performance and Solution Filters, particularly for larger projects.
0 kommentar(er)
