On 6 jun 2007, at 16.15, Bennett Helm wrote:
On Jun 6, 2007, at 9:59 AM, Johannes Knaus wrote:
Well, I can confirm that Gerards PPC-Version runs just fine
(flawlessly and speedy - no lags anymore) on my PPCs, while the
universal binary just behaves like driving through small city
streets with a truck that can't brake ;-)
Thank you very much Gerard.
Gerard: What configure settings did you use?
Anders: Does Gerard's version perform the same as yours?
No, it is faster than the universal, but not as fast as the one
compiled on my machine (iBook G4). See figures below.
Note that the "benchmark" does not reflect "real world use", but is
more of a (very) rough measure of screen updating, etc. I haven't
used the universal binary in any production work yet, will try tomorrow.
===
Benchmark with LyX User's guide (time in seconds).
Instant preview on.
Date Open Scroll Page Comments
down down
(arrow)
07-05-21 6 172 41 Skipping lines.
RC1(uni) 7 275 53
RC1(PPC) 5 220 28
07-06-03 6 222 28
RC-Ger 6 235 44
===
Another question (for everyone): would it be preferable to produce
a universal binary (assuming we can work out the performance
problem), or would you prefer separate versions for Intel and PPC
Macs?
It is of course nice with ONE Mac version. On the other hand it will
be a larger binary, and there is no major problem for a user to
select version.
So, I guess my opinion is: If you can fix the performance issue, a
universal binary would be nice, but in my opinion it is not a major
priority.
But as you I think we need to understand what causes the performance
lag to ensure 1.5 binary(ies) with good performance.
/Anders
