There are some more fine-grained controls but the basic one is just -O2,
which is a combination of optimization techniques that should work on most
source code.  I'd rather not test out each separate optimization to see
which one is causing the problem, but I may have to.  It may also turn out
that no individual optimization is to blame but that a combination of them
work together to cause the problem.  Or it may happen that -O2 just can't
be supported and the API ends up being slow, which would be unfortunate. 
It would be nice if one of the developers could figure this problem out,
since GCC 3 is now pretty standard on most new linux distros.  I'm using
Fedora Core 1.

Thanks,
Carl

> Hi Carl,
>
> Does gcc allow you to selectively turn optimizations on and off? Possibly
> through the use of #pragmas?
>
> Jeff
>
>> -----Original Message-----
>> From: [EMAIL PROTECTED]
>> [mailto:[EMAIL PROTECTED] On Behalf Of Carl Youngblood
>> Sent: Friday, March 19, 2004 4:44 PM
>> To: [EMAIL PROTECTED]
>> Subject: Re: Trouble linking Crypto++ 5.1 on Debian unstable
>>
>>
>> >On February 5, 2004 05:17 pm, Wei Dai wrote:
>> >> On Thu, Feb 05, 2004 at 11:12:38AM -0500, J Smith wrote:
>> >> > If you're interested, the exception is as follows, and happens as
>> soon as
>> >> > the RSA testing begins.
>> >> >
>> >> > CryptoPP::Exception caught: InvertibleRSAFunction: computational
>> >> > error during private key operation
>> >>
>> >> Try turning off optimization?
>> >
>> >Yeah, I had tried that after my last post and it seemed to
>> work. Just
>> >forgot to post back the results. Methinks gcc 3.3.3 might be
>> a tad bit
>> >unstable, perhaps? Or perhaps this issue is simply limited
>> to Debian.
>> >Either way, it seems to be working now.
>> >
>> >Thanks again, Wei.
>> >
>> >J
>>
>> I'm getting these same compile/exception problems, and I'm
>> trying to find a better way to fix them.  Without
>> optimization the API is considerably slower than some of the
>> other ones I'm evaluating.  Does anyone have a better idea?
>>
>> Thanks,
>> Carl Youngblood
>>
>

Reply via email to