> If you're on a compatible host, why not just build in 32-bit mode? If > the app is really not portable and meant to be on 32-bit platforms, then > your options seem obvious. Either build in 32-bit mode "e.g. -m32", or > fix the application to not depend on the size of given data types (that > is the max size). We build the model with -m32 option on Opteron. They work fine; almost the same as in P4/P5 CPU. At another level of simulation, which used up 4GB 32-bit address space, we have to use pure 64-bit build. We see segfault when running some test. > In reality, in proper portable code it shouldn't matter. So long as the > data types are big enough you can always mask off upper bits. In hardware implementation we simulate is strictly a 32-bit platform, the rounding of final results and intermediate computation are cut-off, not like the way of round-half-up. This might contribute to different results in 64-bit and 32-bit binary. -- View this message in context: http://www.nabble.com/Size-of-C-C%2B%2B-data-type-from-GNU-GCC-g%2B%2B-compiled-ELF-64-bit-LSB-executable%2C-AMD-x86-64-vs.-ELF-32-bit-LSB-executable%2C-Intel-80386-tf3942845.html#a11194094 Sent from the gcc - Help mailing list archive at Nabble.com.