Size of C/C++ data type from GNU GCC/g++ compiled ELF 64-bit LSB executable, AMD x86-64 vs. ELF 32-bit LSB executable, Intel 80386

[Andrew Haley]

tom peng writes:

 > Thanks for correcting my typo on "signed long int 4 8".
 > 
 > The problem I have here is about simulating a hardware model in C/C++. The
 > hardware model development was based on 32-bit CPU originally. We had
 > presumed "signed long int" as 32-bit/4 byte length -- as specified in C/C++
 > code "typedef signed long int int32".
 > 
 > It is apparent that this "int32" in 64-bit ELF executable is of length 8
 > bytes, not what we expect. It "might" cause the hw simulation inconsistency
 > when running in 64 vs. 32 ELF format.
 > 
 > I hope to get a generic picture about the size difference of C/C++ data type
 > and their modifiers in 32/64 ELF format compiled by GCC.
 > 
 > Are those 32-bit / 64-bit difference stipulated in C/C++ standard or ruled
 > by the GNU/GCC compiler?

Neither of those.  They are defined by the ABI, at
http://refspecs.freestandards.org/elf/x86_64-abi-0.95.pdf Page 12.

GCC, as the system compiler, must conform to this specification.  It
isn't something that we compiler writers get to decide for ourseleves.

Andrew.