I ran a similar test the other day. Here are my timings with (fibo 33) and (cFibo 33)
I'm including cFibo (since I can now run it on emu64) and ersatz. emu64: 21.632 sec emu64/cFibo: 0.111 sec pil32: 4.477 sec ersatz: 12.797 sec On Tue, Nov 6, 2012 at 11:04 AM, Alexander Burger <[email protected]>wrote: > Hi Jon, > > > I installed the latest "ongoing" (v3.1.0.12) on my iMac, and > > compared the timing results of (bench (fibo 33)). Not too useful, I > > was just curious. ;-) > > Yeah, interesting :) > > > > pil32: 0.804 sec > > emu64: 10.032 sec > > > > My EmuLisp in Safari: 5.82 sec > > My EmuLisp in Chrome: 8.102 sec > > My EmuLisp in Chromium: 8.261 sec > > > > The fibo used in all cases was this: > > (de fibo (N) (if (>= 2 N) 1 (+ (fibo (dec N)) (fibo (- N 2))))) > > > > And the returned value was 3524578. ;-) > > If I try this on an x86-64 machine having all of them installed (pil32, > pil64 and emu64), I get: > > pil32: 0.89 sec > pil64: 0.42 sec > emu64: 12.3 sec > > All quite similar (with the same result, 3524578). > > > I did also tests with the chess program, getting similar relations. You > can let play it against itself, with e.g.: > > $ time ./pil games/chess.l -main -'do 12 (msg (go))' -bye > > > A database stress, however, running 40 concurrent processes hammering > data into the database, showed a drop in speed for emu of only a factor > of three. Here the bottleneck is in file I/O, locking etc. > > Cheers, > - Alex > -- > UNSUBSCRIBE: mailto:[email protected]?subject=Unsubscribe >
