Differences between version 17 and predecessor to the previous major change of MacOSX.
Other diffs: Previous Revision, Previous Author, or view the Annotated Edit History
| Newer page: | version 17 | Last edited on Wednesday, July 6, 2005 10:42:00 am | by PhilMurray | Revert |
| Older page: | version 16 | Last edited on Wednesday, July 6, 2005 10:17:00 am | by AristotlePagaltzis | Revert |
@@ -6,8 +6,10 @@
Speculation has it that this is due to the very use of a MicroKernel.
''I have seen articles which state that it is due to the time to create a new thread and demonstrated it through simple C programs. Can't find the article at present but find plenty of references saying [MySQL] and [Apache] are painfully slow -- IanMcDonald''
+
+''AFAIK It only applied to pthreads. Every app written in Carbon or Cocoa (ie, every single GUI app on OSX) uses NSThreads which doesn't seem to be painfully slow from actually using MacOSX. Incidentally, my 1.4Ghz Mac Mini kicks the crap out of my Celeron 1.1GHz PC when benchmarking MySQL, which other benchmarks would have you believe that's not possible, so it can't be _that_ bad :P. Also, some have attributed the slowness in MySQL to the fact that InnoDB uses fcntl() to guarantee data has been written to disk instead of fsync(). See http://lists.apple.com/archives/darwin-dev/2005/Feb/msg00072.html -- PhilMurray''
Subsequent revisions have become much better in terms of perception of latency. While 10.0.x was quite painful on a G3, 10.2 was snappier (well, until you started swapping, at least) and apparently 10.3.x is further optimised.
Note: the Ⅹ in [MacOSX] has nothing to do with the X windowing system, it is the roman numeral for 10, which follows on from the previous version [MacOS] 9.
