GPUs perform most efficiently when all threads in a warp execute the same sequence of instructions convergently. However, when threads in a warp encounter a divergent branch, the hardware serializes the execution of diverged paths. We consider a class of convergence opportunities wherein multiple threads are expected to eventually execute a given segment of code, but not all threads arrive at the same time, resulting in serialized duplicate execution of common code subsequences such as function calls and loop bodies. Our goal is to promote convergence by helping threads that execute common code arrive together before allowing execution to proceed. We propose a new user-guided compiler mechanism, Speculative Reconvergence, to help identify and exploit previously untapped convergence opportunities that increase SIMT efficiency and improve performance. For the set of workloads we study, we see improvements ranging from 10% to 3× in both SIMT efficiency and in performance.

We introduce a learning-based framework to optimize tensor programs for deep learning workloads. Efficient implementations of tensor operators, such as matrix multiplication and high dimensional convolution are key enablers of effective deep learning systems. However, existing systems rely on manually optimized libraries such as cuDNN where only a narrow range of server class GPUs are well-supported. The reliance on hardware specific operator libraries limits the applicability of high-level graph optimizations and incurs significant engineering costs when deploying to new hardware targets. We use learning to remove this engineering burden. We learn domain specific statistical cost models to guide the search of tensor operator implementations over billions of possible program variants. We further accelerate the search by effective model transfer across workloads. Experimental results show that our framework delivers performance competitive with state-of-the-art hand-tuned libraries for low-power CPU, mobile GPU, and server-class GPU.

Most user interfaces require the user to focus on one element at a time, but StarCraft 2 is a game where players often control more than a hundred units simultaneously. The game interface provides an optional mechanism called "control groups" that allows players to select multiple units and assign them to a group in order to quickly recall previous selections of units. From an analysis of over 3,000 replays, we show that the usage of control groups is a key differentiator of individual players as well as players of different skill levels---novice users rarely use control groups while experts nearly always do. But players also behave differently in how they use their control groups, especially in time-pressured situations. While certain control group behaviors are common across all skill levels, expert players appear to be better at remaining composed and sustaining control group use in battle. We also qualitatively analyze discussions on web forums from players about how they use control groups to provide context about how such a simple interface mechanic has produced numerous ways of optimizing unit control.