[XPM] [ANN] Primecoin High Performance | HP14 released!

[Tuck Fheman]

Does reducing the # of threads increase chainspermin for anyone else?

On my i7 I started using 6 threads and saw 8-12 CPMs.

When I switched to -1 it dropped to 3-6 CPMs.

I changed back to 6 and it increased again.

[1714162867]

1714162867

[1714162867]

1714162867

[1714162867]

1714162867

[1714162867]

1714162867

[GSnak]

It's looking like mining is getting significantly less profitable without a gpu miner.

says the guy with 400-500k pps. lulz

If everyone had a GPU miner... you know how that would go.

[mikaelh]

difficulty 9 already, that's the reason

What's the reason it is only 9?
I'm trying to get my head around the idea that Quarkcoin launched a few hours ago and is already 57 difficultly.*

Is that an apples to oranges comparison? Surely Quarkcoin can't have 6 times the network strength of Primecoin.

As Primecoin difficulty rises, the supply of coins really gets choked off - at difficultly of 32 . . . that's 999/(32*32)= .975 coins per block.

*Oh, it's gone up to 63 since I wrote those two sentences.

It is very much like comparing apples and oranges. For Primecoin, the integer part of difficulty determines the required probable prime chain length. E.g. if difficulty is 9.07233632, a probable prime chain of length 9 is required. The fractional part is used to determine how many length-9 chains are accepted. It scales logarithmically so that going from 9.8 to 9.9 is a lot harder than going from 9.1 to 9.2. The details for the fractional part are in Sunny King's paper.

The current algorithm also does more work for longer target chain lengths. That's why PPS numbers are also dropping.

[rethaw]

Does reducing the # of threads increase chainspermin for anyone else?

On my i7 I started using 6 threads and saw 8-12 CPMs.

When I switched to -1 it dropped to 3-6 CPMs.

I changed back to 6 and it increased again.

I wasn't able to replicate this. Maybe this has something to do with hyperthreading on your processor. Is it a quad-core?

[ivanlabrie]

Does reducing the # of threads increase chainspermin for anyone else?

On my i7 I started using 6 threads and saw 8-12 CPMs.

When I switched to -1 it dropped to 3-6 CPMs.

I changed back to 6 and it increased again.

I wasn't able to replicate this. Maybe this has something to do with hyperthreading on your processor. Is it a quad-core?

It's a well known fact that you see a boost in flops when disabling HT...but I'm not sure if that would be a good idea or not in this case. (intel burn test surely likes that)

[redphlegm]

I still don't understand how I overcame this error few days ago and it keeps coming back -- any ideas?  I tried swap commands but I guess the server I am has it locked and does not allow swapping, but it has enough memory to processor the make command: make -f makefile.unix USE_UPNP=-
I even tried make clean command and still no progress when I launch this. Any updates or ideas?

g++: internal compiler error: Killed (program cc1plus)
Please submit a full bug report,
with preprocessed source if appropriate.
See <file:///usr/share/doc/gcc-4.6/README.Bugs> for instructions.
make: *** [obj/alert.o] Error 4

RAM.

[FreeTrade]

It is very much like comparing apples and oranges. For Primecoin, the integer part of difficulty determines the required probable prime chain length. E.g. if difficulty is 9.07233632, a probable prime chain of length 9 is required. The fractional part is used to determine how many length-9 chains are accepted. It scales logarithmically so that going from 9.8 to 9.9 is a lot harder than going from 9.1 to 9.2. The details for the fractional part are in Sunny King's paper.

The current algorithm also does more work for longer target chain lengths. That's why PPS numbers are also dropping.

Thanks Mikael, I read that it takes a 30 fold increase in processing power to increase difficulty by 1 . . . if so difficult to imagine difficultly ever climbing beyond 12 or 13.

999/(13*13) =  5.9 per block

[Tamis]

difficulty is rising fast ! gpu miner out ?

[maco]

HP6 dropped me to 800 pps instead of 3k pps normally.

[jubalix]

did we all see the PPS drop

I went from 14k to 8k....or is that hp6 v hp5 effect?

[redphlegm]

I feel like I'm sort of flying blind on sievepercentage. I don't know what it's supposed to be changing so I don't know whether to tweak it up or down. Comparatively speaking, I don't know if running default hp5 is better / worse than default hp6 for me since pps isn't a great metric to use.

Any suggestions / insight?

[jubalix]

just tried to compile hp6, go error

Code:

:~$ cd ~/primecoin/src 
:~/primecoin/src$ sudo make -f makefile.unix USE_UPNP=-
Building LevelDB ...
make[1]: Entering directory `/home/ubuntu/primecoin/src/leveldb'
g++ -I. -I./include -fno-builtin-memcmp -pthread -DOS_LINUX -DLEVELDB_PLATFORM_POSIX -O2 -pthread -Wall -Wextra -Wformat -Wformat-security -Wno-unused-parameter -g -DBOOST_SPIRIT_THREADSAFE -D_FILE_OFFSET_BITS=64 -I/home/ubuntu/primecoin/src -I/home/ubuntu/primecoin/src/obj -DUSE_IPV6=1 -I/home/ubuntu/primecoin/src/leveldb/include -I/home/ubuntu/primecoin/src/leveldb/helpers -DHAVE_BUILD_INFO -fno-stack-protector -fstack-protector-all -Wstack-protector -D_FORTIFY_SOURCE=2  -c db/._builder.cc -o db/._builder.o
db/._builder.cc:1:1: warning: null character(s) ignored [enabled by default]
db/._builder.cc:1:2: error: stray ‘\5’ in program
db/._builder.cc:1:2: error: stray ‘\26’ in program
db/._builder.cc:1:2: error: stray ‘\7’ in program
.....

db/._builder.cc:1:9: error: ‘Mac’ does not name a type
db/._builder.cc:1:158: error: expected unqualified-id before numeric constant
db/._builder.cc:1:167: error: ‘Firefox’ does not name a type
db/._builder.cc:1:3827: error: ‘This’ does not name a type
make[1]: *** [db/._builder.o] Error 1
make[1]: Leaving directory `/home/ubuntu/primecoin/src/leveldb'
make: *** [leveldb/libleveldb.a] Error 2

any ideas on how to fix this???

skay fixed

[dudeguy]

-hp6 released!

Lots of changes in this release. Here's a list of notable changes:
 * Added fast divisibility tests before doing the expensive Fermat's test
 * Lots of other small optimizations
 * Introduced a new experimental tuning parameter "sievepercentage". Default value is 10, minimum is 1, maximum is 100.
 * Added new RPC command "getchainspermin"
 * Added the following information to "getmininginfo": chainspermin, difficulty, sievepercentage, and sievesize
 * Attempt to fix the random crash while mining

My PPS on mainnet went up 23% with this release with difficulty >= 9 on mainnet. This release attempts to fix the random crash while mining. I'm not sure if it's actually fixed because I have heard both positive and negative reports. The crash is more likely with high number of cores. 16 and 32 core machines are suffering the most.

The "sievepercentage" parameter is there for people to experiment with. It limits how many base primes are used to filter out candidate multipliers in the sieve. 8% seems optimal for testnet but I have no idea about mainnet.

This build is doing horrible for me. Dropped down to 9 blocks last hour, and 5 this one (almost at the end of the hour). Was getting between 8 and 20 the last build, even after diff 9. Gonna give it one more hour of variation, then switch back. My servers restart primecoind upon crashing, so could the reduction be from multiple crashes? Any explanation?

Can we change "sievepercentage" from debug console? My primecoin.conf is not working no matter what I try. I've literally tried everything.

[LlamaMaster]

My pps went down after I changed the sievesize from 1M to 2M.  Should I be worried, or is this "normal"?

[reb0rn21]

My pps went down after I changed the sievesize from 1M to 2M.  Should I be worried, or is this "normal"?

its normal

[notlist3d]

-hp6 released!

Lots of changes in this release. Here's a list of notable changes:
 * Added fast divisibility tests before doing the expensive Fermat's test
 * Lots of other small optimizations
 * Introduced a new experimental tuning parameter "sievepercentage". Default value is 10, minimum is 1, maximum is 100.
 * Added new RPC command "getchainspermin"
 * Added the following information to "getmininginfo": chainspermin, difficulty, sievepercentage, and sievesize
 * Attempt to fix the random crash while mining

My PPS on mainnet went up 23% with this release with difficulty >= 9 on mainnet. This release attempts to fix the random crash while mining. I'm not sure if it's actually fixed because I have heard both positive and negative reports. The crash is more likely with high number of cores. 16 and 32 core machines are suffering the most.

The "sievepercentage" parameter is there for people to experiment with. It limits how many base primes are used to filter out candidate multipliers in the sieve. 8% seems optimal for testnet but I have no idea about mainnet.

This build is doing horrible for me. Dropped down to 9 blocks last hour, and 5 this one (almost at the end of the hour). Was getting between 8 and 20 the last build, even after diff 9. Gonna give it one more hour of variation, then switch back. My servers restart primecoind upon crashing, so could the reduction be from multiple crashes? Any explanation?

Can we change "sievepercentage" from debug console? My primecoin.conf is not working no matter what I try. I've literally tried everything.

Same here on not getting primecoin.conf to work.  But it's probley me.

[jubalix]

It is very much like comparing apples and oranges. For Primecoin, the integer part of difficulty determines the required probable prime chain length. E.g. if difficulty is 9.07233632, a probable prime chain of length 9 is required. The fractional part is used to determine how many length-9 chains are accepted. It scales logarithmically so that going from 9.8 to 9.9 is a lot harder than going from 9.1 to 9.2. The details for the fractional part are in Sunny King's paper.

The current algorithm also does more work for longer target chain lengths. That's why PPS numbers are also dropping.

Thanks Mikael, I read that it takes a 30 fold increase in processing power to increase difficulty by 1 . . . if so difficult to imagine difficultly ever climbing beyond 12 or 13.

999/(13*13) =  5.9 per block

i thing there are about 1000x optimization left in the code....at least

[Tamis]

My pps went down after I changed the sievesize from 1M to 2M.  Should I be worried, or is this "normal"?

it's normal !

[OnlyC]

sievepercentage=1
sievesize=1000000

with sievepercentage=1,  I got biggest PPS and more chainspermin. Is it good?

[FreeTrade]

Thanks Mikael, I read that it takes a 30 fold increase in processing power to increase difficulty by 1 . . . if so difficult to imagine difficultly ever climbing beyond 12 or 13.

999/(13*13) =  5.9 per block

i thing there are about 1000x optimization left in the code....at least

so . . . 999/(15*15) =  4.44 per block