home *** CD-ROM | disk | FTP | other *** search
/ developer.apple.com / developer.apple.com.tar / developer.apple.com / tools / images / optimizn-help-window.gif < prev    next >
Graphics Interchange Format  |  2009-11-09  |  32KB  |  571x294  |  8-bit (202 colors)
Labels: dialog box | sky | window
OCR: uns gned char br tmp1 obr tmp1 br tmp1 brainptr[] c[ii] AltiVec Unal gn This loop contains 8-bit integer computa tion. you spend large amount of time in this loop (i.e T shows up hotspot ir time profile) and it does significant amount of computation, you may be able to use AltiVec to vastly improve the performance of f this code This instruction appears tc be the start of a loop that is not aligned to a 32- byte address boundary For optimal performance on PPC970 you should align the start of hot loop using compiler directive. ur Code eWarrior 8 use #pragma function 32 In CodeWarrior also possible to use 'asm {align 32} (broken ir Code eWarrior 8 With gcc, use the -falign- -loops=32 compiler flag (broken in gcc 3.1 gcc 3.3 or later should be used) Tm BRAIN CONT Cobr tmp1 BRAIN change clate ...