[XPM] [ANN] Primecoin High Performance | HP14 released! - page 116.

redphlegm

sr. member

Activity: 246

Merit: 250

My spoon is too big!

Quote from: dreamwatcher on July 14, 2013, 07:14:42 PM

Quote from: mikaelh on July 14, 2013, 06:14:14 PM

Quote from: dreamwatcher on July 14, 2013, 06:04:06 PM

I found that installing libmpfr-dev which it required for Freicoin, also satisfies the gmp requirement.

Is there a difference in speed between the mpfr and gmp libraries?

libmpfr-dev pulls libgmp as a dependency. The actual mpfr does floating point calculations. Comparing mpfr and gmp is like comparing apples and bananas.

Thank you.

I did not know that it used gmp as a dependency.

I see so many libraries and packages that appear to do the same things. Some better at certain applications while worse at others.

I guess the next major step would be to rewrite the mining routine in OpenCL. Though I would not doubt somebody has already did it and is not sharing.

Can't say I would blame them at this point.

Yeah, I was thinking that earlier. It seems a likely explanation to the skyrocketing difficulty (that surprises even Sunny).

I'm curious - how / why were OCL miners even publicly released anyway? What caused the person / group to release it/them and not horde anymore?

dudeguy

member

Activity: 182

Merit: 10

Quote from: dreamwatcher on July 14, 2013, 07:14:42 PM

Quote from: mikaelh on July 14, 2013, 06:14:14 PM

Quote from: dreamwatcher on July 14, 2013, 06:04:06 PM

I found that installing libmpfr-dev which it required for Freicoin, also satisfies the gmp requirement.

Is there a difference in speed between the mpfr and gmp libraries?

libmpfr-dev pulls libgmp as a dependency. The actual mpfr does floating point calculations. Comparing mpfr and gmp is like comparing apples and bananas.

Thank you.

I did not know that it used gmp as a dependency.

I see so many libraries and packages that appear to do the same things. Some better at certain applications while worse at others.

I guess the next major step would be to rewrite the mining routine in OpenCL. Though I would not doubt somebody has already did it and is not sharing. Cheesy

Can't say I would blame them at this point.

Luke Jr we are talking to you! Tongue

fabrizziop

hero member

Activity: 506

Merit: 500

Quote from: dudeguy on July 14, 2013, 07:05:52 PM

Quote from: fabrizziop on July 14, 2013, 06:49:42 PM

Finally managed to build it on Ubuntu 13.04 amd64, it gives me around 4500 pps on my FX 8350 Cheesy

Congrats!

Next project; improve HP3??? Wink

In my case it's weird, if I compile with march=native the pps rate is lower.

dreamwatcher

legendary

Activity: 1064

Merit: 1000

Quote from: mikaelh on July 14, 2013, 06:14:14 PM

Quote from: dreamwatcher on July 14, 2013, 06:04:06 PM

I found that installing libmpfr-dev which it required for Freicoin, also satisfies the gmp requirement.

Is there a difference in speed between the mpfr and gmp libraries?

libmpfr-dev pulls libgmp as a dependency. The actual mpfr does floating point calculations. Comparing mpfr and gmp is like comparing apples and bananas.

Thank you.

I did not know that it used gmp as a dependency.

I see so many libraries and packages that appear to do the same things. Some better at certain applications while worse at others.

I guess the next major step would be to rewrite the mining routine in OpenCL. Though I would not doubt somebody has already did it and is not sharing. Cheesy

Can't say I would blame them at this point.

dudeguy

member

Activity: 182

Merit: 10

Quote from: fabrizziop on July 14, 2013, 06:49:42 PM

Finally managed to build it on Ubuntu 13.04 amd64, it gives me around 4500 pps on my FX 8350 Cheesy

Congrats!

Next project; improve HP3??? Wink

dudeguy

member

Activity: 182

Merit: 10

Quote from: forsetifox on July 14, 2013, 06:56:58 PM

Just found another block. 4 blocks in 24 hours @ 2500 PPS.

Shocked

I'm so jelly. Hold onto those precious coins.

I've been well over 3000 for the last 48 hours. Nothing. I made sure do re-download the blockchain before switching builds. No excuses. Just bad luck?

forsetifox

sr. member

Activity: 266

Merit: 250

Just found another block. 4 blocks in 24 hours @ 2500 PPS.

Shocked

fabrizziop

hero member

Activity: 506

Merit: 500

Finally managed to build it on Ubuntu 13.04 amd64, it gives me around 4500 pps on my FX 8350 Cheesy

redphlegm

sr. member

Activity: 246

Merit: 250

My spoon is too big!

Quote from: mikaelh on July 14, 2013, 06:14:14 PM

Quote from: dreamwatcher on July 14, 2013, 06:04:06 PM

I found that installing libmpfr-dev which it required for Freicoin, also satisfies the gmp requirement.

Is there a difference in speed between the mpfr and gmp libraries?

libmpfr-dev pulls libgmp as a dependency. The actual mpfr does floating point calculations. Comparing mpfr and gmp is like comparing apples and bananas.

Well I tried "sudo port uninstall gmp" and then "sudo port install mpfr". Then I recompiled. It compiled without errors but the performance was even worse than not having mpfr installed (with macports).

I'm at a loss now... Sad

matt4054

legendary

Activity: 1946

Merit: 1035

Quote from: mikaelh on July 14, 2013, 05:53:19 PM

Quote from: redphlegm on July 14, 2013, 05:38:56 PM

I'm getting pretty abysmal performance with this build on osx (using the osx makefile). 1player's build seems to be working better on osx while all other machines are better using mikhael's hp3. Any suggestions / optimizations to the osx makefile would be much appreciated.

I'm tempted to VM an ubuntu build and see how that goes but I'd rather run it natively in osx with this high performance build.

Where did you get your libgmp? That's what's doing most of the heavy lifting. I would probably look for optimization opportunities in how libgmp is compiled.

As far as I am concerned, I got it fresh from the current MacPorts

Greydon Isis

newbie

Activity: 14

Merit: 0

Quote from: ?? on ??

So... if things continue like that, reward is 1 XPM in two weeks? Wink

is XPM going outta control many good thing may come of this added attention, think I need to check the exchange price again~~~

dudeguy

member

Activity: 182

Merit: 10

Quote from: itod on July 14, 2013, 06:26:41 PM

Quote from: jubalix on July 14, 2013, 06:23:19 PM

ok this is officially crazy

on a E2 8cc

build 1 = original hp2 compiled = found 7 blocks 2 orphans [17 - 11K pps]/24 hrs
build 2 = original hp1 compiled = found 5 blocks 2 orphans [20k - 16K pps]/24 hrs
build 3 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/12 hrs
build 4 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 5 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 6 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 7 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 8 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]

there is no way that the blocks would all would be found in just the first 2....

is statistically impossible

there is something going on here

Are all these started in parallel? Can it be explained by the change of difficulty working in different timeframes?

Actually Sunny King was just turning his server room on and off. There's your reason. Cool

altsay

sr. member

Activity: 359

Merit: 250

Quote from: jubalix on July 14, 2013, 06:23:19 PM

ok this is officially crazy

on a E2 8cc

build 1 = original hp2 compiled = found 7 blocks 2 orphans [17 - 11K pps]/24 hrs
build 2 = original hp1 compiled = found 5 blocks 2 orphans [20k - 16K pps]/24 hrs
build 3 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/12 hrs
build 4 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 5 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 6 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 7 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 8 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]

there is no way that the blocks would all would be found in just the first 2....

is statistically impossible

there is something going on here

These numbers for 8 core instances are great actually

dudeguy

member

Activity: 182

Merit: 10

Quote from: jubalix on July 14, 2013, 06:23:19 PM

ok this is officially crazy

on a E2 8cc

build 1 = original hp2 compiled = found 7 blocks 2 orphans [17 - 11K pps]/24 hrs
build 2 = original hp1 compiled = found 5 blocks 2 orphans [20k - 16K pps]/24 hrs
build 3 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/12 hrs
build 4 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 5 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 6 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 7 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 8 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]

there is no way that the blocks would all would be found in just the first 2....

is statistically impossible

there is something going on here

What peer IPs are you connected to? Yourself? ARe you talking to your own servers indirectly working together to solve with them? Someone explain is this is kind of possible given the Peer to Peer technology implemented in mining.

jubalix

legendary

Activity: 2632

Merit: 1023

Quote from: ?? on ??

Quote from: jubalix on July 14, 2013, 06:23:19 PM

is statistically impossible

there is something going on here

What's going on is that you do not understand statistics at all.

IIts statistically unlikley that none of the other instances find zero, while the first two find blocks, it's geting to the point where p<0.05 so statistically significant.

forsetifox

sr. member

Activity: 266

Merit: 250

Man.. this coin is strange. 3 days and no coins. Just got 3 blocks today. O.o

hp3 client @ 2500PPS

itod

legendary

Activity: 1974

Merit: 1077

^ Will code for Bitcoins

Quote from: jubalix on July 14, 2013, 06:23:19 PM

ok this is officially crazy

on a E2 8cc

build 1 = original hp2 compiled = found 7 blocks 2 orphans [17 - 11K pps]/24 hrs
build 2 = original hp1 compiled = found 5 blocks 2 orphans [20k - 16K pps]/24 hrs
build 3 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/12 hrs
build 4 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 5 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 6 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 7 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 8 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]

there is no way that the blocks would all would be found in just the first 2....

is statistically impossible

there is something going on here

Are all these started in parallel? Can it be explained by the change of difficulty working in different timeframes?

bidji29

sr. member

Activity: 392

Merit: 250

Hp1 report higher pps, but he find less block.
Still, bad luck

redphlegm

sr. member

Activity: 246

Merit: 250

My spoon is too big!

Quote from: mikaelh on July 14, 2013, 05:53:19 PM

Quote from: redphlegm on July 14, 2013, 05:38:56 PM

I'm getting pretty abysmal performance with this build on osx (using the osx makefile). 1player's build seems to be working better on osx while all other machines are better using mikhael's hp3. Any suggestions / optimizations to the osx makefile would be much appreciated.

I'm tempted to VM an ubuntu build and see how that goes but I'd rather run it natively in osx with this high performance build.

Where did you get your libgmp? That's what's doing most of the heavy lifting. I would probably look for optimization opportunities in how libgmp is compiled.

sudo port install gmp

jubalix

legendary

Activity: 2632

Merit: 1023

ok this is officially crazy

on a E2 8cc

build 1 = original hp2 compiled = found 7 blocks 2 orphans [17 - 11K pps]/24 hrs
build 2 = original hp1 compiled = found 5 blocks 2 orphans [20k - 16K pps]/24 hrs
build 3 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/12 hrs
build 4 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 5 = original hp2 compiled = found 0 [17 - 11K pps]/12 hrs
build 6 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 7 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]
build 8 = AWS image of hp1, compiled = found 0 [20 - 16K pps]/3 hrs [now deleted instance]

there is no way that the blocks would all would be found in just the first 2....

is statistically impossible

there is something going on here

Topic: [XPM] [ANN] Primecoin High Performance | HP14 released! - page 116. (Read 397647 times)