With the discussion yesterday on M. performance, I wrote a little test program to see if there was significant differences (basically the test program is a for / endfor loop setting 5 memvars to datetime() x number of cycles - with and without using m. for variables). What I found in my simple test was that there is very little performance difference using either convention. However, I found in running this same test program on different computers with different processors / operating systems, the variable convention (m. or not to m.) performance was not significantly different, but some processing time differences were rather surprising.

Test machine 1 - Pentium 1 133MHz. 48 MB VFP 6.0 SP5 Runtime (old test machine)

Using m. for all variables in 1 to 10000 loop 26.375 Seconds
NOT using m. for all variables in 1 to  10000  loop 26.212 Seconds
Using m. for all variables in 1 to 1000000 loop  2635.311 Seconds
NOT using m. for all variables in 1 to  1000000  loop  2618.036 Seconds

Test machine 2 - HP Pentium 4 1.4GHz 512MB WinME VFP6.0 SP5

Using m. for all variables in 1 to 10000 loop  6.641 Seconds
NOT using m. for all variables in 1 to  10000  loop  6.639 Seconds
Using m. for all variables in 1 to 1000000 loop  668.769 Seconds
NOT using m. for all variables in 1 to  1000000  loop  691.564 Seconds

Test machine 3 - Gateway Pentium 4 1.8GHz 256MB Win2000 Pro VFP6.0 SP5

Using m. for all variables in 1 to 10000 loop  0.311 Seconds
NOT using m. for all variables in 1 to  10000  loop  0.300 Seconds
Using m. for all variables in 1 to 1000000 loop  31.014 Seconds
NOT using m. for all variables in 1 to  1000000  loop  30.544 Seconds

2 : 1 = 10.5 X faster processor, 10.6 X RAM = 3.94 X faster processing
3 : 1 = 13.5 X faster processor, 5.3 X RAM = 874.35 X faster processing
3 : 2 = 1.3 X faster processor, 0.5 X RAM = 21.5 X faster processing

I realize that this is like comparing a Corvette, Camero, and Yugo, but the time differences were very surprising. I thought that the 1.4Ghz P4 on Win ME would do better that this.

Both P4 computers had only VFP running, both had virus detection. The major difference was the OS. The Win ME box has a few more programs loaded at startup, but the Win2000 box has SQL Server and Oracle plus some other programs loaded at start up and half the memory, so the background load should not be that much different.

I expected that a faster processor would yield faster processing times. I was expecting maybe 2 to 3 times faster on the 1.8GHz box compared to 1.4Ghz. box. What I found was more like 21 times faster. I wasn't anticipating this much difference in 2 computers that I had thought were somewhat closer in performance than this. I realize that WinME is practically worthless OS, and the result of this will make me upgrade OS of the 1.4GH box. Does this make sense?

Elmer