[3376] in java-interest
Re: XShm and solaris version
daemon@ATHENA.MIT.EDU (Jim Graham)
Thu Nov 9 19:51:24 1995
Date: Thu, 9 Nov 1995 12:21:26 -0800
From: flar@bendenweyr.Eng.Sun.COM (Jim Graham)
To: java-interest@java.Eng.Sun.COM, dgourley@CS.Berkeley.EDU
Hi David,
> This may be a silly question, but is there any reason that the solaris(X11
> based versions) of java do not use the XShm extensions?
It's not a silly question and for a while I was looking forward to making
this supposedly trivial change and expect some performance. But...
As it turns out, the XShm extension is not always a win on Solaris. In
fact it sometimes loses more than it gains, especially for small images
(about the size of the single scanlines that we use for the majority of
our PutImage requests). On some depths on some Solaris framebuffers,
it even induces an X11 error (asynchronously) which makes it even more
of a problem to detect and rely on. It does win on the most common
framebuffer - the TGX - and even for small images, but the support
considerations for the other framebuffers are a serious drawback.
Since we don't rely heavily on the PutImage request - it is only used
when the image is first created, then CopyArea is used to draw the bits
on the window - it isn't an important area to optimize. And given the
time that would have to be spent to discover and work around the
problems I've seen with it, I would rather spend time finding a faster
way to draw transparent gif images or optimizing the pixel conversion
loops, both of which are currently affecting us more than the PutImage
performance.
...jim
-
This message was sent to the java-interest mailing list
Info: send 'help' to java-interest-request@java.sun.com
