Optimized fused-multiply-add method and system

A method for performing a fused multiply/add (FMA) operation comprising:

• multiplying a first operand and a second operand and providing at least one product of the first operand and the second operand;
• aligning a third operand with the at least one product to provide an aligned third operand;
• performing, using the at least one product and the aligned third operand, addition and subtraction in parallel paths utilizing an adder generating an adder result and a subtractor generating a subtractor result;
• counting a number of leading zero bits of at least one of the adder result and the subtractor result to provide at least one normalization shift amount;
• and
• selecting, as output of the fused multiply/add operation, between at least (i) a first result based on the adder result of the adder and (ii) a second result based on the subtractor result of the subtractor.

A non-transitory computer-readable storage medium embodying instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

• multiplying a first operand and a second operand and providing a plurality of partial products of the first operand and the second operand;
• aligning a third operand with the plurality of partial products to provide an aligned third operand;
• performing addition and subtraction in parallel paths utilizing an adder generating an adder result and a subtractor generating a subtractor result;
• counting a number of leading zero bits to provide at least one normalization shift amount of at least one of the adder result and the subtractor result;
• and
• selecting, as output of a fused multiply/add operation, between at least (i) a first result based on the adder result of the adder and (ii) a second result based on the subtractor result of the subtractor.