Mcl on PowerPc Benchmarks

[poeck@informatik.uni-wuerzburg.de (Karsten Poeck)]

We did today some benchmarks of mcl 2.0.1 on a powerpc 7100/66.
Emulation is faster than a vx and has between 50 and 80 % of a quadra 700
Overall it is probably like an fx, but I can't verify this at the moment


Most of our stuff worked, but one application crashed the compiler (he
claimed that nil was a variable and had some bogus value)


Karsten Poeck

Results for some benmarkings in seconds


Scheduling speed:
        powerpc 7100/66  Quadra 700
N-queens-1     4.5610 1.7290
N-queens-2     6.1290 2.1650
Further N-Queens did not work on the powerpc, but fine on the Quadra.

timetabling 
powerpc 7100/66  Quadra 700
139.5970         77.5000


Some of the Gabriel Benchmarks:
non opt means speed 0, safety 3, opt means speed3 safety 0
	q700 non opt;	q700 opt;	7100/66 non opt;	7100/66 opt;	vx non opt;	vx opt
BOYER  	4.027	2.174	6.448	3.923	10.018	6.792
BROWSE 	6.915	4.616	14.266	10.444	15.178	11.723
CTAK   	0.546	0.440	1.299	0.966	1.839	1.627
DDERIV 	1.084	0.688	1.663	1.198	2.596	2.191
DERIV  	0.859	0.567	1.657	0.849	2.537	1.636
DESTRU-M     	1.036	0.521	2.385	1.429	2.034	1.211
DESTRU 	1.030	0.616	2.345	1.431	2.078	1.202
DIV2   	1.383	0.944	1.857	1.210	3.022	2.208
FFT-MOD      	6.251	4.763	90.461	69.667	18.878	14.859
FFT    	6.239	4.773	90.278	69.774	17.930	14.896
FPRINT 	2.774	2.990	4.771	4.775	4.465	4.454
FREAD  	0.362	0.343	0.888	0.865	0.983	0.945
FRPOLY 	11.775	8.236	31.361	29.329	22.108	19.601
PUZZLE-M     	12.735	9.327	28.384	22.844	25.089	22.640
PUZZLE 	11.149	9.468	28.027	21.377	24.723	22.430
STAK   	0.983	0.383	1.520	0.735	2.692	1.903
TAK-MOD      	1.657	0.900	3.035	1.954	3.163	2.233
TAK    	1.657	0.900	3.029	1.953	3.164	2.233
TAKL   	1.859	0.746	3.262	1.514	4.365	2.119
TAKR   	0.269	0.178	0.345	0.215	0.454	0.309


surprisingly slow is the FFT-MOD, but this is probably due to the not
existing fpu



We also tested a bit of quickdraw 
displaying ten hierarchies took
21.209 seconds on the Quadra 700 and 35.477 on the 7100/66