[XPM] Primecoin Built-in Miner Sieve Performance Issue

[Boing7898]

Just tested the 3 new windows releases from anonppcoin, shinkicker and zalfrin. Zalfrins is the fastest but shinkickers didn't crash on exit.

 

Heh... that's a new development. It also appears to be broken again, not pulling in new blocks as they come. Last time that was a UPNP problem.

All right, here's a fixed version. Includes all dlls, shouldn't be crashing on exit, and should be following the blockchain properly. 

https://www.dropbox.com/s/6dwcc5p1koj86vz/release5.zip

Hmm, something's wrong. It doesn't open :/
Win 8 x64 AMD Phenom X4 965

[1713538441]

1713538441

[1713538441]

1713538441

[1713538441]

1713538441

[Zalfrin]

Just tested the 3 new windows releases from anonppcoin, shinkicker and zalfrin. Zalfrins is the fastest but shinkickers didn't crash on exit.

 

Heh... that's a new development. It also appears to be broken again, not pulling in new blocks as they come. Last time that was a UPNP problem.

All right, here's a fixed version. Includes all dlls, shouldn't be crashing on exit, and should be following the blockchain properly. 

https://www.dropbox.com/s/6dwcc5p1koj86vz/release5.zip

Hmm, something's wrong. It doesn't open :/
Win 8 x64 AMD Phenom X4 965

Yes, that version was specifically targeted at a sandy bridge i5 processor architecture, so it's unlikely to work on other processors like your AMD. release6 MIGHT work for you, but there hasn't been a whole lot of luck with my builds if you aren't using a similar processor.

[Boing7898]

Just tested the 3 new windows releases from anonppcoin, shinkicker and zalfrin. Zalfrins is the fastest but shinkickers didn't crash on exit.

 

Heh... that's a new development. It also appears to be broken again, not pulling in new blocks as they come. Last time that was a UPNP problem.

All right, here's a fixed version. Includes all dlls, shouldn't be crashing on exit, and should be following the blockchain properly. 

https://www.dropbox.com/s/6dwcc5p1koj86vz/release5.zip

Hmm, something's wrong. It doesn't open :/
Win 8 x64 AMD Phenom X4 965

Yes, that version was specifically targeted at a sandy bridge i5 processor architecture, so it's unlikely to work on other processors like your AMD. release6 MIGHT work for you, but there hasn't been a whole lot of luck with my builds if you aren't using a similar processor.

Tried release 6 and 7, both crash :/

[altsay]

Just pushed another commit on github.

https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675

sieve round limit further reduced to 1 second.
added mike's suggestion.

Give it a try and let me know how it goes.

This nearly tripled my PPS from the original release.

Now getting 2000-2400 PPS on an i7-980X @ 4 GHz. 7 blocks found so far (of which 1 was orphaned).

Compiled with -O3 -march=native.

Do you mean:

make -f makefile.unix -O3 -march=native

Something like that. More specifically:

Code:

make CXXFLAGS="-O3 -march=native" -j6 -f makefile.unix

These extra flags speed things up? As I did not use any flag when compiling

[Zalfrin]

Just pushed another commit on github.

https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675

sieve round limit further reduced to 1 second.
added mike's suggestion.

Give it a try and let me know how it goes.

This nearly tripled my PPS from the original release.

Now getting 2000-2400 PPS on an i7-980X @ 4 GHz. 7 blocks found so far (of which 1 was orphaned).

Compiled with -O3 -march=native.

Do you mean:

make -f makefile.unix -O3 -march=native

Something like that. More specifically:

Code:

make CXXFLAGS="-O3 -march=native" -j6 -f makefile.unix

These extra flags speed things up? As I did not use any flag when compiling

Yes, they should. -O3 is a higher level of optimization than the default -O2, so runtime performance should be better but compile will take longer. -march=native optimizes the instructions specifically for your processor.

[Zalfrin]

Here's another binary I've made which targets simply x86-64 processors. This SHOULD work over a wider range of processors than my earlier releases. It may not work on 32-bit OSes (not sure).

https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip

[Boing7898]

Here's another binary I've made which targets simply x86-64 processors. This SHOULD work over a wider range of processors than my earlier releases. It may not work on 32-bit OSes (not sure).

https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip

Thanks! That works. I'll try mining and see the speed I get.

EDIT: It opens up, starts downloading the blocks and then it crashes.. ;__;

[captainfuture]

Here's another binary I've made which targets simply x86-64 processors. This SHOULD work over a wider range of processors than my earlier releases. It may not work on 32-bit OSes (not sure).

https://www.dropbox.com/s/q7w4fosqjyqydff/release8_x86-64.zip

i will give it a try

[Zalfrin]

EDIT: It opens up, starts downloading the blocks and then it crashes.. ;__;

Geeze... just can't win. Works fine on MY computer.  

[captainfuture]

pps now higher. hope to found a block with this

[UNOE]

Could be the 32-bit windows, as I am running x64. Maybe a -m32 switch is needed during the compile to force it into 32-bit mode.

right. my i5 and zeon boxes are 64bit win7

Here's one more you can try. Last one from me tonight. This one has -mtune=i686 -m32, so hopefully it will cross compile to 32-bit OSes. Still runs on my 64-bit one.

https://www.dropbox.com/s/uvgn1bcso6sppd0/release7_tuned_for_i686_32-bit.zip

Updated client works only on 4core cpu's.
Were tested all versions of updated client. APPCRASH on all versions of windows (W7 embedded standart x32, w7 home basic x32, w7 proffessional x64, w8 proffesional x64) with 2 core processors celeron, sempron, athlon x2 and 6core phenom x6. Works only on quad core: i5 and athlon X4 works well.

I'm running in a Windows VM and maxing out 6 cores 12 threads at 4.7ghz 1950 PPS.  Not sure why you only have 4 cores.
Found 6 blocks in 8 hours

[fabrizziop]

This is my build, I used -O2 -march=native, I'm running a FX 8350, it gives me ~1250 PPS on that. I placed all the dlls in the rar. Please do your virus check and everything, I compiled that on a VM but still take precautions running a file made by another person.

https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar

[Remember remember the 5th of November]

I know my Phenom II x4 955 is quite old by today's standards, but is it supposed to be getting just ~120(not constant) with three cores and even less with 4(I have used all the optimizations provided in the threads)? People reporting over 1000 on a single CPU machine is just ridiculous.

[Ethera]

double xeons 6 cores 3.07 showing this

18:02:44

{
"blocks" : 13970,
"currentblocksize" : 1000,
"currentblocktx" : 0,
"errors" : "",
"generate" : true,
"genproclimit" : -1,
"primespersec" : 3235,
"pooledtx" : 0,
"testnet" : false
}


all shows around the same primespersec. thats wtih the release8 (releases before crashed).

Now lets see if it changes the things (5 servers havent found for 14 hours a single block).

[Zalfrin]

all shows around the same primespersec. thats wtih the release8 (releases before crashed).

Now lets see if it changes the things (5 servers havent found for 14 hours a single block).

Cool, glad to hear this version is working better for some folks anyways.

[fabrizziop]

I know my Phenom II x4 955 is quite old by today's standards, but is it supposed to be getting just ~120(not constant) with three cores and even less with 4(I have used all the optimizations provided in the threads)? People reporting over 1000 on a single CPU machine is just ridiculous.

It isn't normal. With the original build released by sunny king I got 10 pps on 8 cores of my FX 8350, and 80 pps on 4 cores (using 1 core of each module). With the recent commits and the build I made, I get 600 pps on 4 cores (1 per module) and 1200 pps on the 8 cores. And I'm a noob at compiling, so it could be made better.

[tocket]

Just pushed another commit on github.

https://github.com/primecoin/primecoin/commit/eda38d62ad571be02095ce44469c0cc26a2b3675

sieve round limit further reduced to 1 second.
added mike's suggestion.

Give it a try and let me know how it goes.

This nearly tripled my PPS from the original release.

Now getting 2000-2400 PPS on an i7-980X @ 4 GHz. 7 blocks found so far (of which 1 was orphaned).

Compiled with -O3 -march=native.

Do you mean:

make -f makefile.unix -O3 -march=native

Something like that. More specifically:

Code:

make CXXFLAGS="-O3 -march=native" -j6 -f makefile.unix

These extra flags speed things up? As I did not use any flag when compiling

Supposedly, yes. However, I tried without and I'm not seeing much difference in the PPS. It is difficult to measure though, so the difference could still be there but it's probably small.

[Ethera]

all shows around the same primespersec. thats wtih the release8 (releases before crashed).

Now lets see if it changes the things (5 servers havent found for 14 hours a single block).

Cool, glad to hear this version is working better for some folks anyways.

waiting to see if anything will be better (meaning finding actual blocks). If yes, then I will definately drop some xpm's for your great work!

[dudeguy]

This is my build, I used -O2 -march=native, I'm running a FX 8350, it gives me ~1250 PPS on that. I placed all the dlls in the rar. Please do your virus check and everything, I compiled that on a VM but still take precautions running a file made by another person.

https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar

So this seems to work better for Older AMD processors? My Girlfriend has a Turion X2 laptop laying around. I'll give it a try (in VMWARE) and let everyone know

[fabrizziop]

This is my build, I used -O2 -march=native, I'm running a FX 8350, it gives me ~1250 PPS on that. I placed all the dlls in the rar. Please do your virus check and everything, I compiled that on a VM but still take precautions running a file made by another person.

https://www.dropbox.com/s/6hycwi0i4dzwu1i/Primecoin.rar

So this seems to work better for Older AMD processors? My Girlfriend has a Turion X2 laptop laying around. I'll give it a try (in VMWARE) and let everyone know

I don't think it will work there, if you want I can make you a build for that ASAP.