|In reply to Comment 34 (Anonymous):|
I perfectly understand what you're saying, and it makes total sense if
your goal is to make a technical comparison of Trance and the UAE JIT.
This is of course very interesting if you're developing JITs or if
you're a technical geek. It doesn't necessarily say much about how
useful MorphOS or UAE is though.
I'm not saying that it's totally irrelevant, but say you get the
result that MorphOS is twice as fast as UAE. What does this mean for
the actual user? You don't know.. because you don't even know if it's
CPU performance that is the problem. Maybe what the user really needs
is faster disk access, or improved GUI response, or something.
You're right that testing feature X in program Y will only be
meaningful for people using feature X in program Y. But at least
those people exist. There is no one out there who needs a
statistically fast CPU. :-)
Now personally I would like it's interesting to have all sorts of
tests, from bogomips to compiler times to very specific and
non-representative special cases. I totally agree with you that to get
a "general" picture, something like compiler times is probably the
best option of those discussed in this thread, but not everyone is
interested in that.