Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!zephyr.ens.tek.com!gvgpsa!treehouse!andyp From: andyp@treehouse.UUCP (Andy Peterman) Newsgroups: comp.sys.mac.programmer Subject: Re: C++/MacApp questions Message-ID: <558@treehouse.UUCP> Date: 2 Oct 90 21:53:07 GMT References: <10594@pt.cs.cmu.edu> <242@farcomp.UUCP> <45290@apple.Apple.COM> Organization: The Tree House Lines: 37 In article <45290@apple.Apple.COM> marc@Apple.COM (Mark Dawson) writes: >.............. Also note that a bigger RAM cache does NOT always speed up >compiling time--if you've made it too big, it will actually slow it down (when >it flushes the cache, it has to write that stuff out to the disk--so a big >cache will have lots of stuff to write out (read:much slower)). For my work, >(8mb fx, 4mb MPW partion size) I've found that a 96K is the biggest I can use >before performance starts to degrade. I thought I'd check this out since I normally use 1024K cache and do intensive MPW compiles. The table below shows the results of compiling one C module (about 1000 lines) of an application and linking it using the time results from BuildProgram. The entire program is about 30,000 lines and generates a 160K application. Most of the time is spent linking. I did these tests on an 8 Meg IIcx with MultiFinder turned off. The second time shown is a repeat build just after the first, which is improved due to the caching action: Cache Size Time 2nd Time (in seconds) Off 50 --- 32K 33 31 64K 32 29 96K 31 28 128K 31 29 256K 30 27 384K 32 27 512K 32 28 1024K 37 29 From this, it seems that a 256K cache gives the best results. I've always been use to using a large cache since it seems to improve performance for most other tasks on the Mac, but I guess I'm going to lower it when in MPW. I sure wish the linker would take more advantage of available RAM and reduce the disk I/O! Andy Peterman treehouse!andyp@gvgpsa.gvg.tek.com