Adriaan van Os wrote:
Roland Turcan wrote:
Hello FPC-Pascal users discussions!
What is the best way to keep the binary compatibility among different
CPU architectures.
I need to make our own database engine compatible and other binary
files, that user still can use our data from i386 or PowerPC computer.
I really don't like the idea to change the order of all numeric
variables on PowerPC, because it costs time of CPU.
There must be easier way how to keep the compatibility.
Could you give me some hints for better solution.
Whether you "like the idea" or not, the best way is as follows:
- add a byte to the header of source data that indicates the endianness,
high-endian for PowerPC, low-endian for i386
- perform byte swapping if the endianness of source data is different
from the endianness of the current CPU
With "source data" I mean
- reading files
- reading data over a network
This is best built-into a streaming toolbox, so that it happens
automatically.
I forgot to mention:
- write data in the endianness of the current CPU
To me, this is obvious, but for example Postgres always transfers data high-endian over a network,
which is ridicuous. So, for example if they send binary data from an i386 machine to another i386
machine, they perform byte-swapping twice, unnecessary,
Regards,
Adriaan van Os
_______________________________________________
fpc-pascal maillist - fpc-pascal@lists.freepascal.org
http://lists.freepascal.org/mailman/listinfo/fpc-pascal
