| 摘要 |
PURPOSE: To quickly and exactly perform the calculations of integers and floating point amount by single precision and double precision by generating specified responses to instructed exceptions including the exceptions of operations to be zero with code and infinity within a numerical processor and selectively masking the generated response. CONSTITUTION: All the numerical exceptions, that is, errors are detected. The natures of the detected exceptions, that is, the errors are instructed in a proper state register 44. The responses specific to instructed exceptions to be zero with code and infinity, that is, errors, including the exceptions of the adopted calculations are generated in a processor 22. Lastly, what response the user of software performs can not be made to be determined by generating an interruption, one of the generated responses or instructed errors is selectively masked and a calculation can be continued in accordance with an instructed, specified and reliable response. Thus, the calculations of integers and floating point amount by single precision and double precision can be quickly and exactly performed. |
