PSI, mpg, GHz P4-days and P90 years

[ATH]

Originally Posted by cheesehead

If we said that an L-L test was credited with, say, 900 MFLOPS, an appropriate response would be, "Yes, but for how long did that speed have to be maintained in order to finish the L-L test?" So then we'd have to multiply that by time needed to complete the test (e.g., 900 MFLOPS * 24 days = 21600 MFLOPS-day or 21.6 GFLOPS-day), to quantify the amount of computation accomplished.

http://en.wikipedia.org/wiki/FLOPS

"Alternatively, the singular FLOP (or flop) is used as an abbreviation for "FLoating-point OPeration", and a flop count is a count of these operations (e.g., required by a given algorithm or computer program)."

This is the way I used it, so 900 MFLOPS * 24 days * 86400 sec/day = 1866.24 TeraFlop.

[cheesehead]

Originally Posted by ATH

This is the way I used it, so 900 MFLOPS

... which is speed, in units of floating-point operations per second, as I wrote ...

* 24 days * 86400 sec/day

I consciously abbreviated my posts above by leaving out the seconds/day and days/year conversion factors.

I regret that decision.

[ATH]

What I meant was, that we could use retina's idea but with FLOP instead of FLOPS. So P90-years and P4 Ghz-days could be converted to TeraFlop or Petaflop, just have to be careful not to add "s" for plural so you confuse them with the speed unit :)

From the status page: http://www.mersenne.org/primenet/summary.txt

2756.91 P90 CPU yr/day = 33186.934 GFLOP/s
So
1 P90 CPU yr = 33186.934 GFLOP/s * 86400 sec/day / 2756.91 P90 CPU yr/day = 1040 Teraflop (or 1.04 Petaflop)

1 GHz P4-day = 0.16434 P90years = 170.91 TeraFlop

[xilman]

What's wrong with MIPS-year? Worked fine back in the good old days. We might want to use GIPS-years now but that's a minor detail.

Paul

[ET_]

Originally Posted by xilman

What's wrong with MIPS-year? Worked fine back in the good old days. We might want to use GIPS-years now but that's a minor detail.

Paul

I may be wrong, but I read something about the word "instruction" related to MIPS acronym. A same CISC instruction may be worth 10 RISC instructions, or be implemented in a different fashion on different architectures, taking different timings. MIPS-yeat does not, in my opinion, indicate a unique (rigogous) value.

Luigi

[RMAC9.5]

Originally posted by Cheesehead.
See? That's one reason I continue to defend using the P90-year as an appropriate GIMPS unit of work measurement -- it specifies a specific CPU (original Pentium) at a specific clock rate (90MHz) maintained for a specific length of time (year), thus together denoting a specific amount of computing accomplished.

I would also like to continue to use P90 years for the following reason. I have successfully "borged" a few computers over the years that I have run Prime95. Each time I have been successful, I have had to answer the following, usually unasked, question. What do you get for participating in this hobby? My answer is that "Our team gets P90 years credit and we get to compete with all of these other teams on the Team Prime Rib web site". Explaining credit in P90 years to the owners of my "borged" PCs is infinitely easier than any of the other proposed solutions that I have seen in this thread.

[xilman]

Originally Posted by ET_

I may be wrong, but I read something about the word "instruction" related to MIPS acronym. A same CISC instruction may be worth 10 RISC instructions, or be implemented in a different fashion on different architectures, taking different timings. MIPS-yeat does not, in my opinion, indicate a unique (rigogous) value.

Luigi

Ah, but a MIPS was defined as the speed of a VAX 11/780.

So a MIPS year is defined to be the amount of computation performed by such a machine in one year.

Paul

[ET_]

Originally Posted by xilman

Ah, but a MIPS was defined as the speed of a VAX 11/780.

So a MIPS year is defined to be the amount of computation performed by such a machine in one year.

Paul

Right!

But wasn't Vax MIPS used to check integer instructions, like for Dhrystones? When it comes to Whetstones, Vax was much slower, and actual Prime95 FFTs are written in floating point instructions. Under this perspective, FLOPs could achieve a better approximation. But,again, FLOPs are not standard like MIPS, as you can see when you benchmark a Cray...

Luigi

[RMAC9.5]

Best definition of MIPS that I know of:
Meaningless Indicator of Processor Speed