The meanings of the settings of optimization parameters are not
standardized, somewhat complicating matters. I think the
Allegro CL people share your view that except for very small
well-tested pieces, used intensively,
the safety=0 setting is not recommended. It makes life
hazardous for probably modest speed improvements.
On the other hand, I use it all the time when running
benchmarks. Also looking at the assembly code is much
easier with speed=3, safety=0, since there are no compiled-in
checks, and stuff is done in-line.
RJF
Raymond Toy wrote:
>>>>>>"Stavros" == Stavros Macrakis <stavros.macrakis@verizon.net> writes:
>>>>>>
>>>>>>
>
> Stavros> By the way, I am surprised that GCL doesn't have enough error-checking
> Stavros> to avoid fatal errors like this. (Or can it be turned on somehow?) For
>
>I believe that some parts of the code are compiled with safety 0[1].
>That may have made sense back in the early days of maxima, but
>computers today are probably 100-1000 times faster with 100-1000 times
>more memory, so I see no reason why anything is compiled with safety
>0, except, perhaps is some very small parts of the code.
>
>Ray
>
>Footnotes:
>[1] I did not check to see if the current code base still has this,
>though.
>
>
>_______________________________________________
>Maxima mailing list
>Maxima@www.math.utexas.edu
>http://www.math.utexas.edu/mailman/listinfo/maxima
>
>