LVO tables in exec libraries

[bloodline]

Does anyone know why the lvo tables are constructed as a list of JMP <address> instructions rather than just a simple list of pointers?

That would have saved 2bytes per entry, is there any advantage to the JMP approach?

[Leffmann]

The advantage is that the function call becomes shorter, a single instruction on 68000 and 68010. If you had just stored an array of pointers you would've needed one more instruction of 2 bytes and one temporary address register per call on 68000 on 68010.

I don't know how many library and device functions are active on a typical Workbench installation, a thousand maybe, so the savings would've been just a couple of kilobytes, minus the above cost for all the software.

[bloodline]

Originally Posted by Leffmann

The advantage is that the function call becomes shorter, a single instruction on 68000 and 68010. If you had just stored an array of pointers you would've needed one more instruction of 2 bytes and one temporary address register per call on 68000 on 68010.

I don't know how many library and device functions are active on a typical Workbench installation, a thousand maybe, so the savings would've been just a couple of kilobytes, minus the above cost for all the software.

Good call. Many thanks