> Going to a 64-bit compile (on Athlon64/Opteron machine)
> is also known to give an interesting speedup (compared to the 32-bit 
> version ran on the same machine)

	Since I have now access to a dual Opteron machine, here are some
timings with POV-Ray 3.6.1 for Unix to support the above statement:

Machine specs:
   CPU: (dual) AMD Opteron(tm) Processor 246  (2 GHz)  with 8 GB RAM
   OS : Tao Linux (based on RedHat Enterprise Linux 3) kernel 2.4.21-15.0.3.TLsatasmp
   CXX: gcc-3.4.1 (user-installed)

Compiling specs:
   distro: source distribution of povray-3.6.1 for Unix
   flags : detected by configure, i.e.
     -pipe -Wno-multichar -O3 -msse -mfpmath=sse -msse2 -march=k8 -mtune=k8
     -minline-all-stringops
     (-m32 is passed through CXXFLAGS/CFLAGS for the 32-bit version)

Benchmark specs:
   version: Standard POV-Ray 3.6 benchmark version 1.02
   command: ./povray --benchmark  (built-in benchmark mode which is equivalent
     to running  povray benchmark.ini)

Results:

32-bit mode
   Parse Time:    0 hours  0 minutes  2 seconds (2 seconds)
   Photon Time:   0 hours  0 minutes 40 seconds (40 seconds)
   Render Time:   0 hours 29 minutes 12 seconds (1752 seconds)
   Total Time:    0 hours 29 minutes 54 seconds (1794 seconds)

64-bit mode
   Parse Time:    0 hours  0 minutes  1 seconds (1 seconds)
   Photon Time:   0 hours  0 minutes 34 seconds (34 seconds)
   Render Time:   0 hours 23 minutes 45 seconds (1425 seconds)
   Total Time:    0 hours 24 minutes 20 seconds (1460 seconds)

	So on this machine, the 64-bit binary is roughly 20% faster than
the 32-bit binary when ran on the standard benchmark.

	- NC

Post a reply to this message