Nvidia Corporation

Adam Dreyer

Rui Manuel Bastos

Narayan Kulshrestha

Chad D Walker

Charles Tsen

A method for correcting a shift error in a fused multiply add operation. The method comprises adjusting a normalized floating-point number before performing a shift error correction to produce an adjusted normalized floating-point number, and correcting a shift error in the adjusted normalized floating-point number. The correcting the shift error comprises shifting a mantissa of the adjusted normalized floating-point number in one direction. A fused multiply add module comprising a normalizer module, a compensation logic, and a round. The normalizer module is operable to normalize a floating-point number to produce a normalized floating-point number. The floating-point number is normalized based upon an estimated quantity of leading zeros. The compensation logic is operable to manage a correction of a shift error in the normalized floating-point number. The rounder is operable to correct the shift error with a mantissa shift in only one direction.

Efficient correction of normalizer shift amount errors in fused multiply add operations

One embodiment of the invention sets forth a CROP configured to perform both color raster operations and atomic transactions. Upon receiving an atomic transaction, the distribution unit within the CROP transmits a read request to the L2 cache for retrieving the destination operand. The distribution unit also transmits the source operands and the operation code to the latency buffer for storage until the destination operand is retrieved from the L2 cache. The processing pipeline transmits the operation code, the source and destination operands and an atomic flag to the blend unit for processing. The blend unit performs the atomic transaction on the source and destination operands based on the operation code and returns the result of the atomic transaction to the processing pipeline for storage in the internal cache. The processing pipeline writes the result of the atomic transaction to the L2 cache for storage at the memory location associated with the atomic transaction.

Growing community of inventors

Portland, OR, United States of America

Adam Dreyer