Topic: [XPM] Primecoin Built-in Miner Sieve Performance Issue (Read 69150 times)
I remember Sunny King had already disappeared by the time I first started posting on this forum. Maybe he assumed a different identity. Its kind of hard to top his accomplishments as a dev, IMHO.
They must have moved to their own forum, don't hear a lot from these guys anymore.... Like I said, too bad. This is a good coin that actually _does_ something.
Two years after the last post? Yep, don't hear a lot from these guys anymore either.
They must have moved to their own forum, don't hear a lot from these guys anymore.... Like I said, too bad. This is a good coin that actually _does_ something.
Primecoin at 0.011/btc ratio. Honestly only the btc 1 o 1 psicology can hold it down or a major unknow bug in the concept it self.
as most in the troll box like to yell, primecoins to the moon!
0.016* as most in the troll box like to yell, primecoins to the moon!
Primecoin at 0.011/btc ratio. Honestly only the btc 1 o 1 psicology can hold it down or a major unknow bug in the concept it self.
as most in the troll box like to yell, primecoins to the moon!
as most in the troll box like to yell, primecoins to the moon!
How do you manage to run it on Raspberry Pi? I'm stuck with make, can't go further. Do you know any tutorial or discussions about this topic?
you need to make a swap file as the pi doesnt have enough ram.. run this
Code:
sudo dd if=/dev/zero of=/swapfile bs=64M count=16
sudo mkswap /swapfile
sudo swapon /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
then run the make command then delete sawp file
Code:
sudo swapoff /swapfile
sudo rm /swapfile
sudo rm /swapfile
How do you manage to run it on Raspberry Pi? I'm stuck with make, can't go further. Do you know any tutorial or discussions about this topic?
With this last commit I went from circa 1K pps to 400-500 with about the same load I had before and stayed there for 15-20'.
Then went up to ~600 for the next 20' and now it's at ~700 but fluctuating back and forth. So that's about half the pps compared to mikaelh's hp commit.
I know pps isn't an indication of speed/success but it's the only one to test against other builds for the time being.
Alas no blocks yet
and just 2 in last 1.5 days
Then went up to ~600 for the next 20' and now it's at ~700 but fluctuating back and forth. So that's about half the pps compared to mikaelh's hp commit.
I know pps isn't an indication of speed/success but it's the only one to test against other builds for the time being.
Alas no blocks yet
and just 2 in last 1.5 daysTest builds against each other on the test net to see how many blocks each one generated in the sake amount of time on the same machine. That is a far better efficacy measure than pps.
I pushed a new commit with sieve optimizations. It appears to generate larger performance improvement on systems not heavily crippled by v0.1.0 miner.
Give it a try. Please note this version has a slower warmup period.
https://github.com/primecoin/primecoin/commit/ab9fe91d711c92527fcde2fa34563b0aefd16742
Give it a try. Please note this version has a slower warmup period.
https://github.com/primecoin/primecoin/commit/ab9fe91d711c92527fcde2fa34563b0aefd16742
I had a noticeable decrease in PPS from this update ( ~1/10th of PPS from original release). I'm running a Core2 Quad. I realize PPS isn't a great measure of efficacy but the decrease seems pretty consistent and dramatic (15min thus far).
Is the "slower warmup period" longer than this, Sunny King??
Maybe wait a bit longer, what's your pps on core2quad?
It was 70-150pps with the original setup, then went down to 1-10pps with your updated .cpp and .h files. I switched to the mikaelh qt and d files now (retaining your .cpp and .h files) and I'm up above 1000pps.
odd thing happening to me a few time was that 18 instances found nothing for 12 hours, I updated clients from some of these homegrown fork, 10 minutes later boom 2 blocks, then nothing for the rest of the day, and the cycle continues
I switched to mikaelh client this morning which almost doubled my pps and I just found a block. Let's see if I get to keep it.
With this last commit I went from circa 1K pps to 400-500 with about the same load I had before and stayed there for 15-20'.
Then went up to ~600 for the next 20' and now it's at ~700 but fluctuating back and forth. So that's about half the pps compared to mikaelh's hp commit.
I know pps isn't an indication of speed/success but it's the only one to test against other builds for the time being.
Alas no blocks yet and just 2 in last 1.5 days
Then went up to ~600 for the next 20' and now it's at ~700 but fluctuating back and forth. So that's about half the pps compared to mikaelh's hp commit.
I know pps isn't an indication of speed/success but it's the only one to test against other builds for the time being.
Alas no blocks yet
and just 2 in last 1.5 days I did some block generation testing on testnet of whether using more threads than cores helps the mining speed. On a core 2 duo (2 cores) running on Chemisist's code, standard build (O3):
4 threads: 47 blocks found in 32.67 minutes = 1.44 blocks/minute
2 threads: 116 blocks in 76.43 minutes = 1.52 blocks/minute
On the stock 0.1.1 from Sunny:
2 threads: 69 blocks in 64.6 minutes = 1.07 blocks/minute
Using 2x threads basically made no difference and is possibly worse. Running for a few hours would give better data but I've got what I wanted.
This somewhat moot now with mikaelh's new code. His looks to be over 3x faster than Chemisist's.
https://bitcointalksearch.org/topic/xpm-ann-primecoin-high-performance-hp14-released-255782
4 threads: 47 blocks found in 32.67 minutes = 1.44 blocks/minute
2 threads: 116 blocks in 76.43 minutes = 1.52 blocks/minute
On the stock 0.1.1 from Sunny:
2 threads: 69 blocks in 64.6 minutes = 1.07 blocks/minute
Using 2x threads basically made no difference and is possibly worse. Running for a few hours would give better data but I've got what I wanted.
This somewhat moot now with mikaelh's new code. His looks to be over 3x faster than Chemisist's.
https://bitcointalksearch.org/topic/xpm-ann-primecoin-high-performance-hp14-released-255782
Could you run a test with larger number, like 8 thread per core.
Maybe the result will be better
I dont think modifying "setgenerate" value would get better performance.
I pushed a new commit with sieve optimizations. It appears to generate larger performance improvement on systems not heavily crippled by v0.1.0 miner.
Give it a try. Please note this version has a slower warmup period.
https://github.com/primecoin/primecoin/commit/ab9fe91d711c92527fcde2fa34563b0aefd16742
Give it a try. Please note this version has a slower warmup period.
https://github.com/primecoin/primecoin/commit/ab9fe91d711c92527fcde2fa34563b0aefd16742
I had a noticeable decrease in PPS from this update ( ~1/10th of PPS from original release). I'm running a Core2 Quad. I realize PPS isn't a great measure of efficacy but the decrease seems pretty consistent and dramatic (15min thus far).
Is the "slower warmup period" longer than this, Sunny King??
Maybe wait a bit longer, what's your pps on core2quad?
I did some block generation testing on testnet of whether using more threads than cores helps the mining speed. On a core 2 duo (2 cores) running on Chemisist's code, standard build (O3):
4 threads: 47 blocks found in 32.67 minutes = 1.44 blocks/minute
2 threads: 116 blocks in 76.43 minutes = 1.52 blocks/minute
On the stock 0.1.1 from Sunny:
2 threads: 69 blocks in 64.6 minutes = 1.07 blocks/minute
Using 2x threads basically made no difference and is possibly worse. Running for a few hours would give better data but I've got what I wanted.
This somewhat moot now with mikaelh's new code. His looks to be over 3x faster than Chemisist's.
https://bitcointalksearch.org/topic/xpm-ann-primecoin-high-performance-hp14-released-255782
4 threads: 47 blocks found in 32.67 minutes = 1.44 blocks/minute
2 threads: 116 blocks in 76.43 minutes = 1.52 blocks/minute
On the stock 0.1.1 from Sunny:
2 threads: 69 blocks in 64.6 minutes = 1.07 blocks/minute
Using 2x threads basically made no difference and is possibly worse. Running for a few hours would give better data but I've got what I wanted.
This somewhat moot now with mikaelh's new code. His looks to be over 3x faster than Chemisist's.
https://bitcointalksearch.org/topic/xpm-ann-primecoin-high-performance-hp14-released-255782
Could you run a test with larger number, like 8 thread per core.
Maybe the result will be better
Sunny about how long should it be taking us to find a block with a i7 for example getting about 4000ppc?
pps is not a reliable measure of performance, so honestly I can't give any such estimate. Hopefully later when the nature of prime chain mining is better understood a more reliable metric can be developed.
Sunny about how long should it be taking us to find a block with a i7 for example getting about 4000ppc?
I pushed a new commit with sieve optimizations. It appears to generate larger performance improvement on systems not heavily crippled by v0.1.0 miner.
Give it a try. Please note this version has a slower warmup period.
https://github.com/primecoin/primecoin/commit/ab9fe91d711c92527fcde2fa34563b0aefd16742
Give it a try. Please note this version has a slower warmup period.
https://github.com/primecoin/primecoin/commit/ab9fe91d711c92527fcde2fa34563b0aefd16742
I had a noticeable decrease in PPS from this update ( ~1/10th of PPS from original release). I'm running a Core2 Quad. I realize PPS isn't a great measure of efficacy but the decrease seems pretty consistent and dramatic (15min thus far).
Is the "slower warmup period" longer than this, Sunny King??
Those links show a code of some kind.
How do i use that text code to mine?
How do i use that text code to mine?
Guys, HOW do I mine this coin?
Where can I download the client from? And how do i start mining then?
Read first post. Where can I download the client from? And how do i start mining then?
Jump to: