Processor for fine-grain sparse integer and floating-point operations

A system for performing computations by a neural network, the system comprising:

• a processing circuit comprising:
• a first multiplier;
• a second multiplier;
• and
• a third multiplier,
• the processing circuit being configured to:
• identify a first plurality of weights having a first arrangement;
• generate a first set of products, each product of the first set of products being an integer product of a first activation value and a respective weight of a first plurality of weights;
• generate a second set of products, each product of the second set of products being a floating-point product of a second activation value and a respective weight of a second plurality of weights;
• and
• output at least one of the first set of products or the second set of products for use by the neural network,
• wherein in generating the first set of products the processing circuit is configured to:
• process the first plurality of weights to have a second arrangement different from the first arrangement;
• store the first plurality of weights arranged according to the second arrangement into a first memory space;
• multiply, in the first multiplier, the first activation value by a least significant sub-word of a first weight of the first plurality of weights stored in the first memory space to form a first partial product;
• multiply, in the second multiplier, the first activation value by a most significant sub-word of the first weight to form a second partial product;
• and
• add the first partial product and the second partial product,
• wherein in generating the second set of products the processing circuit is configured to:
• multiply, in the first multiplier, the second activation value by a first sub-word of a mantissa of a first weight of the second plurality of weights to form a third partial product;
• multiply, in the second multiplier, the second activation value by a second sub-word of the mantissa to form a fourth partial product;
• and
• add the third partial product and the fourth partial product.

A system for performing computations by a neural network, the system comprising:

• means for processing, the means for processing comprising:
• a first multiplier;
• a second multiplier;
• and
• a third multiplier,
• the means for processing being configured to:
• identify a first plurality of weights having a first arrangement;
• generate a first set of products, each product of the first set of products being an integer product of a first activation value and a respective weight of a first plurality of weights;
• generate a second set of products, each product of the second set of products being a floating-point product of a second activation value and a respective weight of a second plurality of weights;
• and
• output at least one of the first set of products or the second set of products for use by the neural network,
• wherein in generating the first set of products the means for processing is configured to:
• process the first plurality of weights to have a second arrangement different from the first arrangement;
• store the first plurality of weights arranged according to the second arrangement into a first memory space;
• multiply, in the first multiplier, the first activation value by a least significant sub-word of a first weight of the first plurality of weights stored in the first memory space to form a first partial product;
• multiply, in the second multiplier, the first activation value by a most significant sub-word of the first weight to form a second partial product;
• and
• add the first partial product and the second partial product,
• wherein in generating the second set of products the means for processing is configured to:
• multiply, in the first multiplier, the second activation value by a first sub-word of a mantissa of a first weight of the second plurality of weights to form a third partial product;
• multiply, in the second multiplier, the second activation value by a second sub-word of the mantissa to form a fourth partial product;
• and
• add the third partial product and the fourth partial product.