Holy new Mersenne prime, Batman! (M47 related)

[rgiltrap]

Originally Posted by davieddy

0.0224 sec/iter means ten days for 40M iterations.

Please keep in mind that we were running on lower performing code prior to today so that figure has limited bearing on the DC run as a whole.

[T.Rex]

Originally Posted by T.Rex

I'll see tomorrow which FFT length is used now by Glucas. I remember I've seen several messages at the beginning saying that Glucas has increased the FFT length in order to reduce the level of error under the required limit.

As Ernst said, no way to say the FFT length. However, yes, Glucas had 4 "roundoff" errors and at the end it decided to change the FFT length:
Restarting. Changed FFT length.
T.

[davieddy]

Originally Posted by rgiltrap

Please keep in mind that we were running on lower performing code prior to today so that figure has limited bearing on the DC run as a whole.

Since 13 days have passed since the initial discovery, you can
rest assured I am bearing that in mind.

[davieddy]

Originally Posted by T.Rex

As Ernst said, no way to say the FFT length. However, yes, Glucas had 4 "roundoff" errors and at the end it decided to change the FFT length:
Restarting. Changed FFT length.
T.

That must be disappointing for you.
More power to my suggestion that 2048K is inadequate!

[ixfd64]

Well, I hope you get that 30% speed increase working!

[davieddy]

Originally Posted by ixfd64

Well, I hope you get that 30% speed increase working!

???

[ixfd64]

See post #273.

[T.Rex]

30% done on my side. Still late...

[davieddy]

Originally Posted by ixfd64

See post #273.

Are you having difficult keeping up to the pace of posting?
I think the rate of growth of this thread is probably record-breaking.

[ewmayer]

Originally Posted by T.Rex

As Ernst said, no way to say the FFT length. However, yes, Glucas had 4 "roundoff" errors and at the end it decided to change the FFT length

I must say, I find that very surprising - apparently the accuracy loss from Glucas' fast-multiply-up-of-roots-of-unity-and-DWT-weights scheme is significantly more than I expected.

"And the specially modified Lamborghini blows the doors off the garage..."

Here are timings for the same 16-thread 4096K FFT running on 4 quad-cores of the Sparc VII:

[Sep 04 08:11:23] Mxxxxxxxx Iter# = 2**00000 clocks = 00:33:51.000 [ 0.0203 sec/iter] Res64: CC42FF0C3E6FDD1E AvgMaxErr = 0.000002711. MaxErr = 0.000003815

[Jeff Gilchrist]

Originally Posted by rgiltrap

As of today we have the nice new Mlucas code release working which means we have a bit more speed and a lot more reliability in terms of restart capability (just in case if anything goes pear shaped).

So Ernst, are you going to make this shiny new Mlucas code available to everyone when this is all over?

I don't have any Ferrari's or Lamborghini's in my garage, and I can't even use any special Bull Linux tricks. I'm just hoping to squeak in a finish before Tony even though I had quite a head start...