So how is the strategy performance on EPIC?


#1

I was wondering where in the line-up EPICs execution performance is? I am imagining it is somewhere between an S2 and SoftPAC, but that is a huge range.

I don’t have an EPIC box, so would some kind person be able to run these sorting subroutines on an EPIC (and an S2 if you got a spare) and report back their findings?

The strategy is set by default to sort 400 random integers using three different sort routines.

Here were the results on an R1 and SoftPAC (older laptop i7). The i7 performs the bubble in 32ms and the insertion in 55ms at the 400 length.

I’m interested in the EPIC performance at the 400 item level since that is the last value I have for the R1 for comparison. As a bonus, it would be great if someone can test on an S2 as well - all mine are in use.

Strategy with a watch window is attached.

Watch window looks like this:

The three execution times in the float variables is the data I am looking for and hopefully will be helpful to all.

Strategy is 10.0 basic.

TestSort.R10.0b.Basic.Download.D09202018.T111631.zip (12.3 KB)

Thanks in advanced!


#2

All done at the 400 level.
Done three times and averaged.

i7 softpac.
Bubble = 0.4157
Insert = 1.0939
Sort = 0.0341

EPIC
Bubble = 2.4268
Insert = 0.9363
Sort = 0.0956

PAC S
Bubble = 7.248
Insert = 3.042
Sort = 0.356

PAC R
Bubble = 23.12
Insert = 10.054
Sort = 1.115


#3

Looks good, EPIC is faster than I expected. The insertion sort speed is interesting on SoftPAC, and your i7 is a lot faster than my i7 :wink:

Edit: nevermind, those times are in seconds and the conversion didn’t go right in my head - so looks like my i7 edged yours out a bit (qsort in .018s vs your .0341)


#4

EPIC is ~3x faster than PAC S
PAC-S is ~3x faster than PAC-R

And SoftPAC on a decent CPU is still the go to for the most performance.
(And something makes the insertion sort slow on x86)