Fast normalization in a mixed precision floating-point unit

A hardware circuit architecture for returning scalar single precision denormal results to double precision normal results in architected format with limited latency impact on performance, the hardware circuit comprising:

• a first hardware circuit component configured to perform a first normalization on a first unrounded single precision denormal result of a floating-point operation, wherein the first normalization includes performing fraction leading zero removal and exponent adjustment up to a minimum exponent threshold on the single precision denormal result to yield a second unrounded single precision denormal result, wherein the first hardware circuit component is a floating-point unit normalizer circuit;
• the first hardware circuit component configured to count leading zeros of an the first unrounded single precision denormal result, wherein counting leading zeros happens in parallel to fraction rounding of the second unrounded single precision denormal result;
• a second hardware circuit component configured to pre-compute a first exponent and a second exponent for the second unrounded single precision denormal result in parallel with a second normalization a rounded single precision denormal result, wherein the first exponent equals the minimum exponent threshold of the first unrounded single precision denormal result minus the a number of leading zeros of the second unrounded single precision denormal result, and the second exponent equals the minimum exponent threshold of the first unrounded single precision denormal result minus the number of leading zeros of the second unrounded single precision denormal result plus one, wherein the second hardware circuit component is a floating-point unit subtractor circuit;
• a third hardware circuit component configured to perform fraction rounding of the second unrounded single precision denormal result to yield the rounded single precision denormal result, wherein the third hardware circuit component is a floating-point unit rounder circuit;
• a fourth hardware circuit component configured to determine whether a number of leading zeros of the rounded single precision denormal result does not equal the number of leading zeros of the second unrounded single precision denormal result and to correct one of the first exponent and the second exponent in response to the determination, wherein the fourth hardware circuit component is a floating-point unit normalizer circuit;
• and
• the fourth hardware circuit component configured to output a final fraction result by performing a second normalization of the rounded single precision denormal result back to architected format, wherein the second normalization includes selecting the first exponent or the second exponent, and normalizing the selected exponent in parallel to fraction normalization;
• wherein the architected format includes the final fraction result and one of the first exponent and the second exponent.