編譯器參數(shù)調(diào)優(yōu)方法

《編譯器參數(shù)調(diào)優(yōu)方法》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫。

1、www.intel.com/software/productsIntel?CompilersForXeon?Processor*Allotherbrandsandnamesarethepropertyoftheirrespectiveowners1IntelConfidentialIA64_Tools_Overview2.pptAgendaGeneralXeon?processoroptimizationsLoopleveloptimizationsMulti-passoptimizationsOtherAgendaGeneralXeon?pr

2、ocessoroptimizationsLoopleveloptimizationsMulti-passoptimizationsOtherGeneralOptimizations/Od,-O0:disableoptimizations/Zi,-g:CreateSymbols/O1,-O1:Optimizesforspeedwithoutincreasingcodesize–i.e.disableslibraryfunctioninlining/O2,-O2–default–Optimizeforspeed/O3,-O3–High-levelo

3、ptimizationsAgendaGeneralXeon?processoroptimizationsLoopleveloptimizationsMulti-passoptimizationsOtherInstructionSchedulingScheduleinstructionstobeoptimalforspecificprocessorinstructionlatenciesandcachesizesWindowsLinuxPentium?processorsandPentiumprocessorswithMMX?technology

4、-G5-tpp5PentiumPro,PentiumIIandPentiumIIIprocessors-G6(Default)-tpp6(Default)Pentium4processor-G7-tpp7Note:defaultmaychangeinfuturecompilersShift/MultiplyLatencyPentiumShifthas~1xlatencyofaddsMultiplyhas~10xlatencyofaddsPentiumPro,II,andIIIShifthas~1xlatencyofaddsMultiplyha

5、s~3xlatencyofaddsPentium4(maychangeinfuturereleases)Shifthas~8xlatencyofaddsMultiplyhas~26xlatencyofaddsUndertheCovers:P4Compileraccountsforthesedifferencesforyou!for(inti=0;i

6、edxcmpl%ecx,%edxjl.B1.7.B1.7:#-tpp7movl(%ebx,%edx,4),%eaxaddl%eax,%eaxaddl%eax,%eaxaddl%eax,%eaxaddl%eax,%eaxaddl%eax,%eaxmovl%eax,(%esi,%edx,4)addl$1,%edxcmpl%ecx,%edxjl.B1.7UndertheCovers:XeonWhichProcessor:[a]x?Torequireatleast...UseWindows*Linux*PentiumProandPentiumIIpro

7、cessorswithCMOVandFCMOVinstructionsiQaxiaxiPentiumprocessorswithMMXinstructionsMQaxMaxMPentiumIIIprocessorwithStreamingSIMDExtensions(impliesiandMabove)KQaxKaxKPentium4processorwithStreamingSIMDExtensions2(impliesi,MandKabove)WQaxWaxWAutomaticProcessorDispatchSingleexecutabl

8、ePentium4targetthatrunsonallx86processors.ForTargetProcessorituses:Processo