Re: Justifying Optimization (VBDis)
21 Jan 2003 00:13:32 -0500

          From comp.compilers

Related articles
Justifying Optimization (MICHAEL DILLON) (2003-01-17)
Re: Justifying Optimization (Joachim Durchholz) (2003-01-20)
Re: Justifying Optimization (srikanth) (2003-01-21)
Re: Justifying Optimization (Christian Bau) (2003-01-21)
Re: Justifying Optimization (2003-01-21)
Re: Justifying Optimization (2003-01-21)
Re: Justifying Optimization (Sid TOUATI) (2003-01-25)
Re: Justifying Optimization (Conor O'Neill) (2003-01-25)
Re: Justifying Optimization (Jan C. =?iso-8859-1?Q?Vorbr=FCggen?=) (2003-01-25)
Re: Justifying Optimization (Jan C.=?iso-8859-1?Q?Vorbr=FCggen?=) (2003-01-25)
Re: Justifying Optimization (Jan C.=?iso-8859-1?Q?Vorbr=FCggen?=) (2003-01-25)
[15 later articles]
| List of all articles for this month |

From: (VBDis)
Newsgroups: comp.compilers
Date: 21 Jan 2003 00:13:32 -0500
Organization: AOL Bertelsmann Online GmbH & Co. KG
References: 03-01-088
Keywords: optimize
Posted-Date: 21 Jan 2003 00:13:32 EST

In general I agree with John, our famous mod :-)

MICHAEL DILLON <> schreibt:
>Developers are stating that optimized code produces errors

This statement is incorrect, since buggy or misleaded optimizers /can/ produce
errors, as well as buggy or misleaded compilers can produce errors, even if
they do no optimization.

>and makes debugging more difficult.

This is (almost ;-) true. So why do you insist in debugging optimized code,
when OTOH you suppose that both the optimized and non-optimized code will
produce exactly the same results???

IMO the validation of the source code should be separated from the validation
of the compiler specific optimization.

And finally, what would you do if the outcome of the optimized and
non-optimized versions differ? Would you introduce workarounds into your source
code, to outwit bugs in the compiler optimization??? Is it this task which you
are paid for? What if these workarounds will no more work with the next
compiler version?


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.