> > I hate to even post my results (all with -DFORCE_REG) since > the PowerPC > > doesn't really shine here, but then again, I do believe in truth in > > advertising: ;) > > > > siev bubble matrix fib machine and configuration > > 2.05 2.78 3.26 3.27 500Mhz i686-pc-cygwin > DIRECT_THREADED > > 2.24 2.93 3.51 3.65 500Mhz i686-pc-cygwin > > 6.32 6.56 5.05 8.07 667Mhz powerpc-apple-darwin6.0 > > DIRECT_THREADED > > 6.56 6.64 5.35 7.93 667Mhz powerpc-apple-darwin6.0
Here's another: 4.20 4.55 3.46 5.47 1Ghz powerpc-apple-darwin6.0 gforth-0.5.0 -DFORCE_REG -DDIRECT_THREADED > This is a little strange. My benchmarks on a relatively slow machine > (an accelerated 7600--note the slow bus) were quite a bit faster: > > GFORTH BENCHMARK SUITE RESULTS > All with PowerMac G3/400, 50 MHz bus, 128 MB RAM, 2MB, 200 > MHz backside > cache, Mac OS X Public Beta (Darwin 1.3). > All with gcc -O3, version 2.95.2. > Pfe-0.30.96 with regs, next unrolled, static libraries. > Gforth-0.5.0 with regs, direct threaded. > > ./gforth-fast --die-on-signal -p siev.fs -e main bye > 2.47s user 0.07s system 96% cpu 2.626 total > ./gforth-fast --die-on-signal -p bubble.fs -e main bye > 2.81s user 0.07s system 96% cpu 2.995 total > ./gforth-fast --die-on-signal -p -m 160000 matrix.fs -e main bye > 1.93s user 0.05s system 97% cpu 2.022 total > ./gforth-fast --die-on-signal -p fib.fs -e main bye > 3.16s user 0.06s system 99% cpu 3.231 total > > I did compile with -O3. You can see these and pfe results, plus > automatic Forth to C translation results, at: > > http://www-personal.umich.edu/~williams/archive/forth/hatforth /gfbench-results.html As Joel said, Jaguar is not faster. The switch to gcc 3.1 was pretty painful around here, and obviously, the debates have not even begun to settle down yet. My cygwin is using gcc version 2.95.3-5 and if I were to use -O3, I'd turn it on there too, so what's the point? > > It's interesting to compare this to the times to build our BootROM: > > > > cygwin direct: build-time was 93 seconds > > cygwin indirect: build-time was 89 seconds > > darwin direct: build-time was 129 seconds > > darwin indirect: build-time was 130 seconds > > > > And just for yucks: ;) > > > > SwiftForth on Win2K: build-time was 45 seconds > > Are you using bash as sh instead of zsh as sh on Darwin? That > speeds up builds quite a bit. It's supposed to become the > default for sh at some point--maybe it already has with 10.2? It's not the default yet, and I try to use as much "out of the box" as I can. I've got a whole bunch of people using it already, and convincing Mac people to use the Terminal (tcsh) is hard enough. Usually, people use a shell window in BBEdit, which adds yet another level of overhead. So, trying to convince any but the true unix gurus (which I am not) that they need to setup bash usually ends in people going back to MPW in Classic. ;( DaR --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
