In April 2026, Nobodqwe developed new FMA correctness tests for the HLSL shader compute pipeline within the llvm/offload-test-suite repository. Focusing on GPU programming and shader development, Nobodqwe expanded test coverage for the offload compute path by designing and integrating tests that validate fused multiply-add operations under various data configurations. The work involved writing HLSL and YAML code to ensure the compute pipeline’s correctness and robustness, particularly for edge cases. By improving integration validation and data-config coverage, Nobodqwe’s contributions enhanced CI regression protection for graphics and compute workloads, providing deeper guarantees for shader operation reliability and early bug detection.