On Thu, Sep 03, 2009 at 07:56:26AM +0300, Karri Kaksonen wrote: > I also compiled the solitaire with the compiler from the subversion > trunk. The binary was 61 bytes bigger than what the previous compiler > produced. Some runtime functions have been changed. The multiplication and division routines are now a lot faster in many cases, unfortunately at the expense of code size. And, the compiler will inline more code. While this can be roughly controlled with --codesize, in some places choosing a bigger --codesize argument will actually make the code smaller! This is because inlining the code will make it visible to the optimizer, which is then able to remove part of it or remove register loads in other places, because it can now see that they're no longer necessary. So size changes depend a lot on your code. Especially in such a big application, I would say 61 bytes are almost nothing:-) > I left all the intermediate *.s files from the compilation and > removed all the files that were equal. If you are interested download > http://sipo.fi/diffs.tar.gz There are two trees "new" for the current > svn compiler "old" for the 2.11.something compiler. The generated code with -O is so different, that it's probably not worth the trouble to compare the two as long as there is no problem. If you have a place where you think that the current compiler generates extremely large or ineffective code, please let me know. > But the code seems to work with both compilers. Thanks! Regards Uz -- Ullrich von Bassewitz uz@musoftware.de ---------------------------------------------------------------------- To unsubscribe from the list send mail to majordomo@musoftware.de with the string "unsubscribe cc65" in the body(!) of the mail.Received on Thu Sep 3 11:19:08 2009
This archive was generated by hypermail 2.1.8 : 2009-09-03 11:19:10 CEST