On 9/8/14 8:40 PM, Nathan Kidd wrote: > TL;DR: VGL_READBACK=sync is always required for VP 9, 10 and 11.
Docs updated accordingly. > I'd like to eventually understand why the SEGV in 3 occurs (presumably > bad buffer interaction between VGL and VP), but this isn't high on my > list since a) VGL_READBACK=sync works, and b) I'm quite used by now, to > saying "ViewPerf is a GPU benchmark, it doesn't make much sense to run > it in any kind of remote setting." Yes and no. If frame spoiling is enabled, then no OpenGL benchmark is useful in a remote setting, but if you disable spoiling, then you can use the benchmark to get some sense of how the remote solution might perform under a real application workload. The thing that most casual users don't understand, however, is that what they're really testing is the remote solution's ability to compress and transmit the images generated by the benchmark. The GPU performance is probably not factoring into the results. So in a sense, yes, the benchmark is not being used for its intended purpose, but it is still useful-- as long as it is being used to compare different remote solutions, as opposed to trying to compare remote vs. local. When I worked for Sun, it seemed like I had to spend 1/3 of my time explaining to people why comparing benchmarks in a remote vs. local setting is kind of like trying to compare the performance of a sports car and a telephone. ------------------------------------------------------------------------------ Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS Reports Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk _______________________________________________ VirtualGL-Devel mailing list VirtualGL-Devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/virtualgl-devel
