[Date Index] [Thread Index] [Author Index]
Bug on SPARC (needs recent patches for Solaris 10)
Unless I am mistaken (and others feel free to correct me), there is a bug either in some recent patches on Sun's Solaris 10, or in Mathematica. The result of this is that Mathematica (5.1 & 5.2) uses up loads of CPU time even when computing something as simple as 1+1. Since it is only recent, I assume it is a patch released by Sun that has "caused" this (using "caused" in a very loose way, as it could be a Mathematica bug, rather than a Solaris bug). How to replicate. 1) Have an affected system - see below. 2) Run 'mathematica' (i.e. the GUI interface), not the command line one 'math'. 3) Compute In:= 1+1 (or anything else for that matter) 4) Look at the CPU usage - it will keep climbing well after Mathematica has computed 1+1. *Affected Systems* Sun Ultra 60, 80 or Blade 100 that has Solaris 10 and recent Solaris patches and either Mathematica 5.1 or 5.2. (No idea about other systems, but Blade 1000 seems to have *no problems*). It appears both the first general release of Solaris 10 (03/2005) and the Update 1 (01/2006) are affected, but need recent Solaris patches. As for how recent, or what patches I don't know. Looking at newsgroup posts on comp.unix.solaris, it seems the following systems are affected: * Mathematica 5.2, Ultra 60 with Solaris 10 1/06, Kernel 118822-25. * Mathematica 5.2, Ultra 60 with Solaris 10 3/05, Kernel 118822-30. * Mathematica 5.2, Ultra 80 with Solaris 10 1/06, Kernel 118822-26. * Mathematica 5.1, Ultra 80 with Solaris 10 1/06, Kernel 118822-26. * Mathematica 5.2, Blade 100 with Solaris 10 1/06, Kernel 118833-02. These are *NOT* affected. * Mathematica 5.2, Blade 1000 with Solaris 10 3/2005, Kernel 118833-02. * Mathematica 5.2, Ultra 2 with Solaris 9 12/2003, Kernel 118833-02. * Mathematica 4.2, Ultra 80 with Solaris 10 1/06, Kernel 118822-26. * Mathematica 4.0, Ultra 80 with Solaris 10 1/06, Kernel 118822-26. Note #1 The Ultra 80 listed 4 times is the same machine. It is fine with Mathematica 4 and 4.2, but fails with 5.1 and 5.2 Note #2 The Blade 1000 and Blade 100 are interesting, as one is affected and the other is not, despite running the same Solaris version (3/05) and the same kernel patch. Both are managed by the same person (not me), and he says they both regularly patched via smpatch. Anyone else see this behavior? Anyone know of a workaround? -- Dave K MCSE. MCSE = Minefield Consultant and Solitaire Expert. Please note my email address changes periodically to avoid spam. It is always of the form: month-year@domain. Hitting reply will work for a couple of months only. Later set it manually.