FLASH-D: FlashAttention with Hidden Softmax Division

--The transformer's attention mechanism has revolutionized AI and machine learning, with its efficient computation being crucial to its performance. However, calculating attention involves matrix operations interspersed with softmax rescaling, which inherently slows down computation and requires processing the entire input sequence. While optimized for GPUs, FlashAttention's simplicity makes it amenable to direct hardware acceleration. This work re-evaluates the core FlashAttention kernel, presenting FLASH-D a mathematically equivalent, yet simplified, formulation that achieves: (a) hiding softmax division within other nonlinear function evaluations; (b) inherently numerically stable computation of exponentials, eliminating the need for maximum value subtraction; and (c) a reduction in computational cost without introducing numerical approximations to the FlashAttention kernel. Importantly, the essential FlashAttention properties that facilitate efficient tiled implementation are fully preserved. Hardware implementation results at 28nm demonstrate that this proposed formulation achieves a 22.8% reduction in area and a 20.3% reduction in power, on average, compared to state-of-the-art parallel hardware architectures without any performance penalty. Current state-of-the-art ML and AI systems are characterized by deep learning models with billions of parameters, achieving human-level performance in tasks like image recognition [1] and natural language processing [2]. A key innovation driving this progress is the attention mechanism [3], which allows models to focus on relevant parts of the input data.