[MiNT] Proposal for SLB extension

[Joerg_Westheide@su.maus.de (Jörg Westheide)]

Hi Konrad!

KMK>And also waste of memory (1000 processes may have 1000 basepages,
KMK>which implies 1000 pointers 4 byte each,
I think this 4K are not the problem if you're running 1000 processes ;-)

KMK>so there is *no* need to waste resources (either CPU time or memory)
KMK>on supplying that for *each* library function.
Well, I think speeding up MiNT's trap handler will have a much greater effect
on performance than saving one single long word parameter on shared lib
calls.
So IMO this should be done first

KMK>The proposal I made makes this optional: i.e. if someone wants the
KMK>old behaviour, it can be easily done. Also there is perfect backward
KMK>compatibility.
The backdraw of your proposal is that you have to check wheter the basepage
pointer is needed at run time. So your code has to deal with both
possibilities.
And saying "Sorry, your version of lib foo requires a pointer to the basepage
on every call. This is not supported." is no good solution.
So it's easier and probably faster to provide this pointer than to
distinguish between the two versions

KMK>But if someone does NOT want to impose any overhead, he can configure
KMK>the SLB so that there will be NO overhead at all.
And if there are two different implementation of the same lib you will run
into trouble.
Dont' ask why there should be two different versions. You know that people
tend to reinvent the wheel

KMK>It is not a question of "how to supply the basepage pointer with
KMK>minimum  waste", but "how to avoid the waste". :)
I think this waste is to small for being worth the expense of removing it

Bye

                Joerg