No, Anthony, you cannot answer how much performance is gained by using the MEMSET macro simply by looking at the assembler output. Especially not with such a contrived example.
What you *can* do, however, is use "make perf" and look at the performance comparison between gcc/glibc memset() and libast's MEMSET() given various CFLAGS. I've tried it many times on multiple architectures and with numerous combinations of flags, and the macro is indeed faster in many situations. If you think the macro is too "weird" (when, in fact, it is quite straight-forward) or that it exhibits poorer performance than the glibc/gcc version, simply build with -DMEMSET=memset or comment it out or something. A potential fix for the 64-bit issue with MEMSET() was first reported to me by Mike Frysinger back in July of 2004 (see http://bugs.gentoo.org/show_bug.cgi?id=52634), and the issue has been corrected in newer versions. Michael -- Michael Jennings (a.k.a. KainX) http://www.kainx.org/ <[EMAIL PROTECTED]> n + 1, Inc., http://www.nplus1.net/ Author, Eterm (www.eterm.org) ----------------------------------------------------------------------- "I'd like to live just long enough to be there when they cut off your head and stick it on a pike as a warning to the next ten generations that some favors come with too high a price. I want to look up into your lifeless eyes and wave.... Can you and your associates arrange that for me, Mr. Morden?" -- Vir Cotto, Babylon Five -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
