Worked on performance optimization in the intel/sycl-tla repository, focusing on enhancing tensor operation throughput for GPU workloads. Developed and integrated support for kFactor=8 in the MmaTensorOpMultiplicandTileIterator, which required careful adjustment of index calculations to ensure correctness for both contiguous and strided memory accesses. This change aligned the code path with tensor core usage, paving the way for higher-throughput tensor operations. The work was implemented in C++ and leveraged low-level programming techniques to maximize efficiency. Emphasis was placed on performance optimization and GPU computing, with the update designed to maintain accuracy while enabling more advanced tensor workloads.