• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Infinite OGR

Q

Qythyx

Member
I was thinking about OGR a bit and how/when OGR 26 and higher will start and end. I was thinking that because the OGR contests double in complexity at each increment (like OGR 25 is twice as difficult as OGR 24) they will necessarily settle down to about 18 months for each contest.

Why? Because 18 months is the historical amount of time needed to double the computing power in a given market range (Moore's law). That is, the most powerful PC you can get today is about 1.2 Ghz (not including multi-CPU). The trend says that in 18 months the most powerful PC will be about 2.4 Ghz. Now increasing bits (e.g. from 32 to 64) also makes a difference and is factored into this 18 month projection.

Anywho, because of this trend, and assuming it continues, the OGR contest must eventually settle down to that rate (** assuming that additional people are not added to the effort, this is obviously a false assumption but additional people just lower the time below 18 months **). You see, if a given contest is done faster than 18 months, then the next one (at twice the complexity) will take longer than the first. Similarly, if a given contest takes more than 18 months, then because of Moore's law, the next one will take less the first one (because after the 18 months the computing power has more than doubled). So, if a given contest takes exactly 18 months, then so should the next, and the next...

This won't work on RC5 because the contests don't double. If we ever did a RC5 65 then it would definitely take a lot less time then RC5 64 because of the increased computing power (although in this case it is greatly due to the increased number of participants too). But you see my point I hope. Of course, no one will do an RC5 65, if anything the next one will be RC5 128, a LOT more than twice as hard.

I realize there are quite a few generalizations and math "smoothing" in this idea but please forgive.

Any thoughts.
 
Where did you hear that OGR 25 would be twice as long as OGR 24? according to my calculations, OGR 25 is about 20+ times longer than OGR 24. Here is the basic reasoning behind my calculations:

According to the latest graphs, we are approximately 158% done with OGR-24 (100% done with pass 1 and 58% done with pass 2). As of the last stats run there were a total of approximately 125,000 teranodes done. dividing this number by 1.58, you get that one pass of OGR-24 is approximately 80,000 teranodes.

Doing the same with OGR 25, you get 1,200,000 teranodes done after 68% of pass 1. this extrapolates to approximately 1,790,000 tera nodes total.

Using these numbers, OGR 25 is 22 times longer than OGR 24.
 
This has probably been discussed before. The way OGRs work, isn't the number of possibilities factorial? Then OGR-25 should be exactly 25x as large a task as OGR-24. And then OGR-26 would be 26x as large as OGR-25, meaning OGR-26 would take 650x as long as OGR-24.

But I could be wrong.
 
BoberFett,

You have the basics right. OGR25 is 25 times longer than OGR24. Except that there are "pruning" algorithms that cut quite a bit of nodes out before they ever get calculated. So, sciencewhiz's guess of 22 times longer is probably pretty close. It may get a little better by the end of the node space (perhaps 20 or 21 times longer) since the longest stubs are at the beginning of the nodespace and tend to get smaller at the end, and sciencewhiz's calculations come from the first 2/3 of the nodespace.

By this logic, it is probably safe to assume that OGR26 will be somewhere around 20-25 times longer than OGR25 (which means 400 to 600 times longer than OGR24). OGR is easily going to become VERY long term in the future...

JHutch
 
That would explain d.net's desire to complete OGR-24 and get as much of OGR-25 done before they even start into OGR-26 (as expressed in bovine's last .plan). Sounds like OGR-26 could turn into a multi-year project! :Q And just think... there are rulers on the books that need checking up to OGR-150!! Eeeep! I'm thinking that AMD needs to get going with some SMP action for the T-Bird and quick! 😉

I'm a bit concerned about the stub sizes that they'll choose for OGR-26 (and I'd guess that D.net is as well 😉 ). Even when they increased the fixed lengths in OGR-25 to 6 points from 5, the WU's still increased considerably (on average) in length. It might not be a bad idea if they tried a few OGR-26 stubs with 8 points fixed to get a feel for how long they would take to process.

Any other thoughts on this?

-Brian
 
Actually we have thought about stubs for OGR-26. We think we have pretty close to the perfect size selected. Should be pretty close to the time it takes for an OGR-25 stub. Of course there always are the long ones that take alot longer than you ever expect them to...

moose
 
Awww geee Moose, do ya have to tease me like that? 😉

It's just my horridly overactive curiosity, but I'd love to know a bit about what went into the decision and what the actual stubs will be like. 🙂 Puuuuuuuuleeeeaaaze! You know... stuff like how many fixed points there will be and will they all be the same or will some have 7 and some 8, etc. etc. etc.

I give you fair warning, I have an 8-year old, so I know how to whine until information is forthcoming. 😉

Seriously, thanks for any info you provide.

-Brian
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top