Abstract: Certain aspects of the present disclosure provide techniques and apparatus for quantized machine learning. A quantized input matrix is accessed at a layer of a neural network, and a first interim value is generated in an accumulator by performing matrix multiplication, using the accumulator, of the quantized input matrix and a quantized weight matrix associated with the layer of the neural network. The first interim value is normalized based at least in part on one or more leading sign bits of the first interim value, and the normalized first interim value is dequantized. A second interim value is generated by applying a rounded right-shift operation to the dequantized normalized first interim value, and activation data is generated by applying an activation function to the second interim value.

Abstract: Techniques for power-density-based clock cell spacing and resulting integrated circuits (ICs) are disclosed herein. In one example, the techniques determine power-usage density for different types of clock cells, as power-usage density relates to heat and IR droop. With the power-usage density for each type of clock cell determined, the techniques assign a keep-out region for each type of clock cell that is not fixed for all types of clock cells. These regions are instead based on the heat and IR droop corresponding to estimated power-usage density for each type of clock cell. Clock cells are then placed in a layout of an IC. The resulting IC has clock cells spaced sufficiently to reduce heat and IR droop while concurrently having excellent timing closure and performance.

Abstract: Techniques for power-density-based clock cell spacing and resulting integrated circuits (ICs) are disclosed herein. In one example, the techniques determine power-usage density for different types of clock cells, as power-usage density relates to heat and IR droop. With the power-usage density for each type of clock cell determined, the techniques assign a keep-out region for each type of clock cell that is not fixed for all types of clock cells. These regions are instead based on the heat and IR droop corresponding to estimated power-usage density for each type of clock cell. Clock cells are then placed in a layout of an IC. The resulting IC has clock cells spaced sufficiently to reduce heat and IR droop while concurrently having excellent timing closure and performance.