{"681884":{"#nid":"681884","#data":{"type":"event","title":"MS Defense by Julia Startsev","body":[{"value":"\u003Cp\u003EJulia Startsev\u003Cbr\u003E(Advisor: Prof. Santosh Pande) will defend a master\u2019s thesis entitled,\u003Cbr\u003ELoopOracle : Speculative HotSpot Detection and Optimization using Machine Learning in Web Browsing\u0026nbsp;\u003Cbr\u003EEnvironments\u003Cbr\u003EOn\u003Cbr\u003ETuesday, April 22nd at 10:00 a.m.\u003Cbr\u003ERemote (Zoom Link)\u003C\/p\u003E\u003Cp\u003EAbstract\u003Cbr\u003EJust-in-Time (JIT) compilation in web browsers is critical to the performance of modern web\u0026nbsp;\u003Cbr\u003Eapplications. These compilers typically rely on runtime profiling to identify frequently executed\u0026nbsp;\u003Cbr\u003E(hot) functions before applying optimizations, yet this process introduces overhead especially on\u0026nbsp;\u003Cbr\u003Efirst execution. Although widely used techniques\u2014such as per-application caching and simple\u0026nbsp;\u003Cbr\u003Ethreshold-based recompilation\u2014partially mitigate these issues, they often fall short for\u003Cbr\u003Eloop-intensive workloads with unpredictable runtime characteristics. To address these challenges,\u0026nbsp;\u003Cbr\u003Ethis thesis proposes a lightweight speculative optimization framework\u2014LoopOracle. The framework\u0026nbsp;\u003Cbr\u003Eembeds lightweight machine learning models and survival analysis directly into inline caches (ICs).\u0026nbsp;\u003Cbr\u003EThe thesis considers two optimization targets: JIT recompilation and inlining decisions. Similar to\u0026nbsp;\u003Cbr\u003Ehow traditional techniques sample code execution to guide compilation, LoopOracle observes loop\u0026nbsp;\u003Cbr\u003Epatterns and probabilistically infers whether a function is likely to run often enough to warrant\u0026nbsp;\u003Cbr\u003Ean alternative strategy compared to the baseline compiler. By integrating these predictive models\u0026nbsp;\u003Cbr\u003Ewithin the existing IC mechanism, the framework integrates training, monitoring, inference and\u0026nbsp;\u003Cbr\u003Emodel switching seamlessly into the compiler pipeline. This thesis presents a concrete\u0026nbsp;\u003Cbr\u003Eimplementation in Mozilla\u2019s SpiderMonkey compiler. To evaluate the framework, empirical evaluations\u0026nbsp;\u003Cbr\u003Eon two industry-standard benchmarks\u2014Jetstream 2.2 (emphasizing computationally expensive web\u0026nbsp;\u003Cbr\u003Eapplications) and Speedometer 3.0 (focusing on DOM interaction)\u2014are presented in contrast to\u0026nbsp;\u003Cbr\u003Eindustry established approaches such as JitHints.\u003Cbr\u003ECommittee:\u003C\/p\u003E\u003Cp\u003E\u25cf Prof. Santosh Pande \u2013 School of Computing (advisor)\u003Cbr\u003E\u25cf Prof. Ada Gravilovska Habl \u2013 School of Computing\u003Cbr\u003E\u25cf Matthew Gaudet \u2013 Staff Engineer, SpiderMonkey Team, Mozilla C\u003Cbr\u003E\u0026nbsp;\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003ELoopOracle : Speculative HotSpot Detection and Optimization using Machine Learning in Web Browsing Environments\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"LoopOracle : Speculative HotSpot Detection and Optimization using Machine Learning in Web Browsing Environments"}],"uid":"27707","created_gmt":"2025-04-17 19:10:45","changed_gmt":"2025-04-17 19:11:15","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2025-04-22T10:00:00-04:00","event_time_end":"2025-04-22T11:00:00-04:00","event_time_end_last":"2025-04-22T11:00:00-04:00","gmt_time_start":"2025-04-22 14:00:00","gmt_time_end":"2025-04-22 15:00:00","gmt_time_end_last":"2025-04-22 15:00:00","rrule":null,"timezone":"America\/New_York"},"location":"Remote (Zoom Link)","extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"111531","name":"ms defense"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}