I have 100K SLOC app that compiles/links fine on Dec Alpha. When I compile/link the code on Intel platform, I find that a few global class objects are larger in the compilation (per watch window "sizeof"
than what the link load map shows. Now the difference I infer to be due to alignment padding. Nonetheless, in one case the compiler generates object with 40 byte size, and load map shows 36 bytes! I haven't produced this in a small app yet. The compilation settings have 8 byte alignment. Any ideas on how to force linker to give 8 byte doubleword alignment as default for Intel? (Alpha Dev Studio uses double word boundaries for ALL globals.) Appreciate your experiences learned...
than what the link load map shows. Now the difference I infer to be due to alignment padding. Nonetheless, in one case the compiler generates object with 40 byte size, and load map shows 36 bytes! I haven't produced this in a small app yet. The compilation settings have 8 byte alignment. Any ideas on how to force linker to give 8 byte doubleword alignment as default for Intel? (Alpha Dev Studio uses double word boundaries for ALL globals.) Appreciate your experiences learned...