Rounding unit for decimal floating-point division

A method for performing a decimal floating-point division operation using a decimal floating-point divider circuit, comprising:

• receiving, by the decimal floating-point divider circuit, a decimal floating-point dividend and a decimal floating-point divisor;
• obtaining, by the decimal floating-point divider circuit, a preliminary quotient having a first precision level,
• wherein the preliminary quotient is calculated from the decimal floating-point dividend and the decimal-floating point divisor, and
• wherein the preliminary quotient has an exponent not in a normal range according to a rule standard;
• receiving, by the decimal floating-point divider circuit, a rounding mode;
• obtaining a shifted quotient by shifting the preliminary quotient to the right to place the exponent in the normal range;
• obtaining a shifted truncated quotient by truncating the shifted quotient to a second precision level, wherein the first precision level is at least one digit greater than the second precision level;
• obtaining an estimated dividend by multiplying the shifted truncated quotient by the decimal floating-point divisor;
• obtaining a remainder by subtracting the estimated dividend from the decimal floating-point dividend;
• selecting a rounding action based on the preliminary quotient, the rounding mode, and the remainder;
• and
• obtaining a rounded quotient having the second precision level by rounding the shifted quotient according to the rounding action.

A decimal floating-point divider circuit, comprising:

• a rounding unit comprising functionality to:
• receive a decimal floating-point dividend, a decimal floating-point divisor, and a rounding mode;
• receive a preliminary quotient having a first precision level, wherein the preliminary quotient is calculated from the decimal floating-point dividend and the decimal-floating point divisor;
• select a rounding action based on the preliminary quotient and the rounding mode;
• and
• obtain a rounded quotient having a second precision level by rounding the preliminary quotient according to the rounding action, wherein the first precision level is at least one digit greater than the second precision level;
• and
• a calculation unit, operatively connected to the rounding unit, and comprising functionality to:
• receive the decimal floating-point dividend and the decimal floating-point divisor;
• calculate the preliminary quotient from the decimal floating-point dividend and the decimal-floating point divisor by performing a number of Newton-Raphson iterations with a fused-multiply-add (FMA) unit;
• obtain a truncated quotient by truncating the preliminary quotient to the second precision level;
• and
• calculate an estimated dividend by multiplying the truncated quotient by the decimal floating-point divisor.