On 6/19/11 1:01 AM, Lorenzo Fiorini wrote: > I have started leaving the xvnc4viewer but it didn't work at full color. > Then I installed the tigerviewer without any new parameter but I got > some "artifacts" > like white backgrounds with some gray bands.
What version of TigerVNC Viewer produced those artifacts? We use a high-speed JPEG codec called libjpeg-turbo to compress and decompress JPEG tiles. It can use either MMX or SSE2 instructions to greatly accelerate compression/decompression, but the MMX code paths are not often used, since most chips these days have SSE2. Thus, it wouldn't surprise me if you uncovered a bug in the MMX code. > I tried to add -NoJPEG: it solved the artifacts. > Then the users started to complain about lost keys ( that is the UI > can't catch up with the typing speed ) and slow screen updates. > I read that using -NoJPEG could have an impact on performance and > guessed that Tight protocol requires > faster CPUs so I removed -NoJPEG and tried hextile. No idea what would be causing the lost keys or the slow screen updates, but we have fixed a number of bugs in TigerVNC, so make sure you're using the latest version. >> I guess the next question would be: what exactly is your workload, and >> how are you measuring performance? > > This is a typical GDM/xinetd/Vnc configuration used for ERP/Office > applications. > Once entered in the standard Gnome desktop the main apps are Firefox > 5.0, Thunderbird 5.0, Libreoffice 3.3.2 > and our ERP which uses an extra light X11 interface. > > Actually the system performs very well at server side. Every tool I > use tells me that it's faster than previous. > > Could you point me to a tool or a method to measure the vnc speed at > client side? > > Monday I'll be there and I'll do some tests following your suggestions. I don't know of a way to measure the performance of 2D apps. What I typically do is measure video frame rates-- either playing a full-screen movie or running an animation in VirtualGL and measuring the frame rate from the client side using TCBench (a program in the VirtualGL source tree which monitors a small area at the center of the client window and counts the changes within that area over a 30-second period.) TCBench is only meaningful if the client is double buffered (which TigerVNC is.) Otherwise, it would measure multiple changes within the same frame. These benchmarks may or may not have any real meaning for your workload. ------------------------------------------------------------------------------ EditLive Enterprise is the world's most technically advanced content authoring tool. Experience the power of Track Changes, Inline Image Editing and ensure content is compliant with Accessibility Checking. http://p.sf.net/sfu/ephox-dev2dev _______________________________________________ Tigervnc-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/tigervnc-users
