I guess I will need to try and clean up my system again, cause even using
./configure and make I'm still seeing a speed decrease for some reason.
Perhaps now, it is all the fink stuff that I needed with 0.6.1.  However, I
did try it on a clean system and everything appears to be fine.  I even took
the opportunity to try it on a Dual 2G G5 with 10.2.7 and 10.3 (Panther).

Here's the results:

siev    bubble  matrix    fib   machine and configuration
 1.07    1.32    0.78    1.32   1Ghz G4 0.6.1 -DFORCE_REG -DDIRECT_THREADED
 1.50    1.92    1.17    1.80   1Ghz G4 OSX 10.2.6 GCC 3.1
 0.78    0.88    0.47    0.89   2Ghz Dual G5 OSX 10.2.7 GCC 3.1
 0.76    0.90    0.47    0.92   2Ghz Dual G5 OSX 10.3 GCC 3.3

Sorry for the false alarm!

You all did great! ;)

DaR

On 8/26/03 12:05 PM, "Anton Ertl" <[EMAIL PROTECTED]> wrote:

> Dennis Ruffer wrote:
>> 
>> Initial report from OSX land shows this update to be extremely slow on a G4:
>> 
>> siev    0.6.1        0.6.2
>> real    0m2.415s    0m15.860s
>> user    0m1.600s    0m9.460s
>> sys     0m0.050s    0m0.020s
>> 
>> bubble
>> real    0m3.719s    0m17.632s
>> user    0m1.920s    0m10.360s
>> sys     0m0.040s    0m0.060s
>> 
>> matrix
>> real    0m2.328s    0m12.047s
>> user    0m0.890s    0m6.910s
>> sys     0m0.020s    0m0.050s
>> 
>> fib
>> real    0m6.302s    0m17.245s
>> user    0m1.980s    0m10.340s
>> sys     0m0.030s    0m0.040s
> 
> On the only MacOS X system I have for testing (a PowerBook with AFAIK
> a 500MHz PPC 750 (G3) running Darwin 5.4 (AFAIK MacOS X 10.1)), I get
> the following results:
> 
> gforth 0.6.1:
> 1.89s user 0.05s system 84% cpu 2.296 total
> 2.18s user 0.09s system 96% cpu 2.363 total
> 1.19s user 0.01s system 93% cpu 1.285 total
> 2.48s user 0.06s system 97% cpu 2.598 total
> 
> gforth 0.6.2:
> 1.69s user 0.06s system 93% cpu 1.878 total
> 2.19s user 0.03s system 94% cpu 2.348 total
> 1.18s user 0.04s system 94% cpu 1.293 total
> 2.28s user 0.12s system 96% cpu 2.497 total
> 
> Built with just
> 
> ./configure; make
> 
> On a 450MHz PPC 7400 (G4) under Linux I get the following results for
> gforth 0.6.2, which are in line with the results on the PowerBook:
> 
> siev   1.84
> bubble 2.28
> matrix 1.29
> fib    2.27
> 
>> In both cases, I compile with the following line:
>> env CFLAGS="-I/sw/include -no-cpp-precomp" LDFLAGS="-L/sw/lib"
>> GFORTH="./gforth -i ./kernl32b.fi -m4M" ./BUILD-FROM-SCRATCH
>> --enable-force-reg --enable-direct-threaded
>> 
>> Please let me know if I should be doing something different with this new
>> version.
> 
> I don't see anything that would hurt the performance, but maybe you
> should try a simple "./configure; make" first and see if it makes a
> difference.  If so, we can look more closely at what causes the
> slowdown.
> 
> - anton
> 
> ---------------------------------------------------------------------
> 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