Advanced execution of extended floating-point add operations in a narrow dataflow

An apparatus for performing additive arithmetic operations on operands in a floating-point arithmetic unit, the operands including a first operand and a second operand, the first operand being a first floating-point number having a first mantissa, the second operand being a second floating-point number having a second mantissa, the second operand being greater than the first operand, the precision of the floating-point arithmetic unit exceeding the precision enabled by a bit length of data input to the floating-point arithmetic unit, said apparatus comprising:

• means for separating the first mantissa into a high portion and a low portion and separating the second mantissa into a high portion and a low portion;
• means for loading the high portion of the first mantissa into a first operand register and loading the high portion of the second mantissa into a second operand register, the first and second operand registers each having a data width of N bits;
• means for aligning the high portion of the first mantissa with the high portion of the second mantissa;
• means for moving the aligned high portion of the first mantissa into a sum register and moving the high portion of the second mantissa into a carry register, said sum register and said carry register each having a data width of 2N bits;
• means for concatenating in the carry register the low portion of the second mantissa with the high portion of the second mantissa;
• means for aligning the low portion of the first mantissa in accordance with said aligning of the high portion of the first mantissa;
• a hold-function multiplexing circuit, said circuit concatenating in the sum register the aligned low portion of the first mantissa with the aligned high portion of the first mantissa;
• an adder, said adder performing an additive arithmetic operation on the aligned first mantissa and the second mantissa, the adder having a data width of 2N bits.

A floating point arithmetic unit for performing additive arithmetic operations on operands, the operands including a first operand and a second operand, the first operand being a first floating-point number having a first mantissa, the second operand being a second floating-point number having a second mantissa, the second operand being greater than the first operand, the first and second mantissas each being separated into a low portion and a high portion, the precision of the floating-point arithmetic unit exceeding the precision enabled by a bit length of data input to the floating-point arithmetic unit, said floating point arithmetic unit comprising:

• first and second operand registers, said operand registers each having a data width of N bits, said first operand register initially loaded with the high portion of the first mantissa, said second operand register initially loaded with high portion of the second mantissa;
• an aligner for aligning the high portion of the first mantissa with the high portion of the second mantissa and for aligning the low portion of the first mantissa in accordance with the aligning of the high portion of the first mantissa;
• a carry register for containing a concatenation of the high portion of the second mantissa with the low portion of the second mantissa;
• a sum register, the sum register first loaded with the aligned high portion of the first mantissa;
• a hold-function circuit, said circuit concatenating in the sum register the aligned low portion of the first mantissa with the aligned high portion of the first mantissa;
• an adder, said adder performing an additive arithmetic operation on the aligned first mantissa and the second mantissa, wherein the sum register, carry register, and adder each have a data width of 2N bits.