Online Softmax Across Tiled Attention

FlashAttention Memory-Efficient SRAM Pipelinining

Mode: Auto Loop
BASELINE PATH (QUADRATIC HBM) FULL N×N MATRIX Materialized in HBM O(N²) R/W Bottleneck AVOIDED HBM TRAFFIC 0 KB ESTIMATED SPEEDUP 1.0x (Baseline) SRAM COMPUTE CHAMBER (ON-CHIP) LOAD UPDATE SCALE QUERY TILE Q_i KEY TILE K_j VALUE TILE V_j LOCAL MAP S_ij RUNNING STATS ACCUMULATOR O_i HBM OUTPUT O SRAM REGISTER HUD ACTIVE TILE: TILE 1 / 2 m_i = -inf | l_i = 0.0