Jorge, here's the benchmark results.  To me, it looks like the difference is
down in the mud of non-repeatable variability.

For CC=gcc2
Each benchmark takes about 30s on a 486-66 (gcc-2.6.3 -DFORCE_REG)
time ./gforth-fast --die-on-signal -p
.:/usr/local/lib/gforth/site-forth:/usr/local/share/gforth/site-forth:/usr/l
ocal/lib/gforth/0.6.1:/usr/local/share/gforth/0.6.1:. siev.fs -e "main bye"

real    0m1.200s
user    0m1.020s
sys     0m0.020s
time ./gforth-fast --die-on-signal -p
.:/usr/local/lib/gforth/site-forth:/usr/local/share/gforth/site-forth:/usr/l
ocal/lib/gforth/0.6.1:/usr/local/share/gforth/0.6.1:. bubble.fs -e "main
bye"
redefined cell  redefined list
real    0m1.404s
user    0m1.020s
sys     0m0.030s
time ./gforth-fast --die-on-signal -p
.:/usr/local/lib/gforth/site-forth:/usr/local/share/gforth/site-forth:/usr/l
ocal/lib/gforth/0.6.1:/usr/local/share/gforth/0.6.1:. matrix.fs -e "main
bye"

real    0m0.809s
user    0m0.720s
sys     0m0.000s
time ./gforth-fast --die-on-signal -p
.:/usr/local/lib/gforth/site-forth:/usr/local/share/gforth/site-forth:/usr/l
ocal/lib/gforth/0.6.1:/usr/local/share/gforth/0.6.1:. fib.fs -e "main bye"

real    0m1.513s
user    0m1.300s
sys     0m0.040s

Without overriding gcc:
Each benchmark takes about 30s on a 486-66 (gcc-2.6.3 -DFORCE_REG)
time ./gforth-fast --die-on-signal -p
.:/usr/local/lib/gforth/site-forth:/usr/local/share/gforth/site-forth:/usr/l
ocal/lib/gforth/0.6.1:/usr/local/share/gforth/0.6.1:. siev.fs -e "main bye"

real    0m1.130s
user    0m1.080s
sys     0m0.000s
time ./gforth-fast --die-on-signal -p
.:/usr/local/lib/gforth/site-forth:/usr/local/share/gforth/site-forth:/usr/l
ocal/lib/gforth/0.6.1:/usr/local/share/gforth/0.6.1:. bubble.fs -e "main
bye"
redefined cell  redefined list
real    0m1.417s
user    0m1.320s
sys     0m0.000s
time ./gforth-fast --die-on-signal -p
.:/usr/local/lib/gforth/site-forth:/usr/local/share/gforth/site-forth:/usr/l
ocal/lib/gforth/0.6.1:/usr/local/share/gforth/0.6.1:. matrix.fs -e "main
bye"

real    0m0.759s
user    0m0.650s
sys     0m0.020s
time ./gforth-fast --die-on-signal -p
.:/usr/local/lib/gforth/site-forth:/usr/local/share/gforth/site-forth:/usr/l
ocal/lib/gforth/0.6.1:/usr/local/share/gforth/0.6.1:. fib.fs -e "main bye"

real    0m1.483s
user    0m1.370s
sys     0m0.010s

DaR

On 3/23/03 2:12 PM, "Jorge Acereda Maci�" <[EMAIL PROTECTED]> wrote:

> 
> On Sunday, March 23, 2003, at 08:47 PM, Dennis Ruffer wrote:
> 
>> I see it as yet another environmental dependency.  If gcc 3.3 solves
>> these
>> (?) problems, how are you going to make the Fink version switch?  If
>> you
>> want to go to the effort to optimize for each OS version, fine, but I
>> think
>> that's a waste of your time.
> 
> Put it the other way: Why should I switch to gcc3?
> 
>> The benchmarks you refer to are not for 0.6.1, but for 0.5.0.  This new
>> version does not have near as much (if any) penalty for using gcc 3.1.
> 
> Why? Superinstructions? Do you have benchmark results?
> 
> Greetings,
>  Jorge
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to