Mma 2.0 slower than 1.0?
- To: mathgroup at yoda.physics.unc.edu
- Subject: Mma 2.0 slower than 1.0?
- From: barry at math.ucla.edu (Barry Merriman)
- Date: Tue, 27 Aug 91 15:15:23 PDT
It appears to me that Mma 2.0 is slower than Mma 1.0 on numerical
calculations.
For example, consider this trivial benchmark (100,000 real
multiplies)
Timing[ Block[{x = 1.00001,y = 1.0000}, Do[ y *= x, {100000} ]; y] ]
On my NeXTStation, I get, using Mma kernel 1.2:
{54.1833 Second, 2.718268237192294}
While using Mma kernel 2.0 I get
{59.4667 Second, 2.71827}
Thus it appears that 2.0 is 10% slower than 1.0 on the trivial
benchmark of doing 100,000 multiplies. Hmm. Has anyone else
benchmarked 1.0 vs 2.0? I thought 2.0 was supposed to be faster!
(I have experimented with Compile-ing the above program, but it
always hits an error and says
CompiledFunction::cfn:
Numerical error encountered at instruction 23
; proceeding with uncompiled evaluation.
and so end up taking about 1 second longer. Probably I need to learn
more about how to convert the benchmark to a suitable form for the
compiler to enhance...I'm waiting for my docs...)