root <[EMAIL PROTECTED]> writes: | > My understanding is this: there is an Axiom image (interpsys) used at | > compile time; think of as the image the build machine needs to drive | > the compilation. One of the purpose of such image is for example, to | > build some of the algebra needed to build the installed image. Once | > that image is built, with database and all of that, a second build | > should be fired up to build the "real" Axiom image. The one that the | > user really cares about. | | The key difference between interpsys (the compile image) and AXIOMsys | (the final user image) is that "warm data" is added to AXIOMsys. A | study I did showed there was a highly skewed distribution of algebra | files which get loaded in any computation (e.g. PositiveInteger is | almost certainly loaded every time). I looked at that distribution | and collected the most frequently loaded algebra files into a list. | | interpsys + (the frequent file list loaded) => AXIOMsys | | which cuts down on the time it takes to perform computation. It's | one of the dozens of optimizations which make Axiom run reasonably.
That begs the question: why not put those warm data and files in interpsys too, so that we only have one image to care about? -- Gaby _______________________________________________ Axiom-developer mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/axiom-developer
