Mersenne Twister implentation

[rprosperi]

(02-06-2017 07:58 AM)grsbanks Wrote:  On a real device with display set to FIX 2:

LISTTIME(1) yields: "91.00 / 4.00 = 22.75"
LISTTIME(10) yeilds: "791.00 / 55.0 = 14.38
LISTTIME(100) yields: "7709.00 / 561.00 = 13.74"
LISTTIME(1000) yields: "77041.00 / 5617.00 = 13.72"

On an emulated Prime, also with display set to FIX 2:

LISTTIME(1) gives a division by zero because method B completes in less than a millisecond.
LISTTIME(10): "16.00 / 1.00 = 16.00"
LISTTIME(100): "159.00 / 10.00 = 15.90"
LISTTIME(1000): "1580.00 / 98.00 = 16.12"
LISTTIME(10000): "15528.00 / 975.00 = 15.93"

Thanks for sharing the results.

While one should naturally expect that the actual times on the physical Prime and virtual Prime are quite different, one would not expect the relative times (the ratios on the device compared to the ratios on the emulator) to be this different (about 15%).

Tim/Cyril - why would the relative speed differences for these 2 types of commands vary this much between the device and virtual Prime? Internally, if they are presumably running the same code, the relative speeds should also be the same, right?

Higher ratios on the emulator implies the FOR...LOOP generated code is better optimized on the PC?

Just curious...