EFFICIENT PARALLEL FLOATING POINT EXCEPTION HANDLING IN PROCESSOR
PROBLEM TO BE SOLVED: To provide a method of efficiently handling floating point exceptions in a processor that executes SIMD instructions.SOLUTION: A method comprises the steps of: identifying a numerical exception for a SIMD floating point operation; initiating a first SIMD micro-operation to gene...
Saved in:
Main Authors | , , , , , |
---|---|
Format | Patent |
Language | English Japanese |
Published |
17.12.2015
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | PROBLEM TO BE SOLVED: To provide a method of efficiently handling floating point exceptions in a processor that executes SIMD instructions.SOLUTION: A method comprises the steps of: identifying a numerical exception for a SIMD floating point operation; initiating a first SIMD micro-operation to generate a first packed partial result for the SIMD floating point operation; initiating a second SIMD micro-operation to generate a second packed partial result for the SIMD floating point operation; initiating a SIMD denormalization micro-operation to combine the first and second packed partial results, denormalize a first element of the combined first and second packed partial results, and generate a third packed result having a denormalized element; storing the third packed result for the SIMD floating point operation; and setting a flag identifying the denormalized element of the third packed result in the first packed partial result.
【課題】SIMD命令を実行するプロセッサにおいて、浮動小数点例外を効率的に処理する。【解決手段】SIMD浮動小数点演算の数値例外を特定するステップと、SIMD浮動小数点演算の第1Packed部分結果を生成するため、第1SIMDマイクロ演算を開始するステップと、SIMD浮動小数点演算の第2Packed部分結果を生成するため、第2SIMDマイクロ演算を開始するステップと、第1及び第2Packed部分結果を合成し、合成された第1及び第2Packed部分結果の第1要素を非正規化して非正規化要素を有する第3Packed結果を生成するため、SIMD非正規化マイクロ演算を開始するステップと、SIMD浮動小数点演算の第3Packed結果を格納するステップと、第3Packed結果の非正規化要素を特定するフラグを第1Packed部分結果に設定するステップと、を有する。【選択図】図8 |
---|---|
Bibliography: | Application Number: JP20150131282 |