1) Last Year in DFSan - The proposed structure of the discussion is to start from a brief requirement list from each project that requires SPIR-V support followed by the design proposals/overviews of each project that integrates SPIR-V in their flow. The panel will be concluded by the open discussions at the end. 2) Pure sparse tensor computation in COMET compiler - MLIR is considered to be the emerging IR framework for doing high level optimizations like Vectorization. LLVM’s VPlan based vectorizer provides accurate cost modelling by building explicit model and selecting the best from various candidates. In this work, we are trying to propose a VPlan like, dialect-agnostic vectorization framework for MLIR. We plan to build or use the LLVM’s TargetTransformInfo (TTI) like information at MLIR level for better target related details in the cost analysis. Our current focus is on languages like Fortran / C / C++ on x86 / AArch64 targets. 3) VPlan like vectorizer for MLIR - MLIR is considered to be the emerging IR framework for doing high level optimizations like Vectorization. LLVM’s VPlan based vectorizer provides accurate cost modelling by building explicit model and selecting the best from various candidates. In this work, we are trying to propose a VPlan like, dialect-agnostic vectorization framework for MLIR. We plan to build or use the LLVM’s TargetTransformInfo (TTI) like information at MLIR level for better target related details in the cost analysis. Our current focus is on languages like Fortran / C / C++ on x86 / AArch64 targets.