Main Menu
Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - Frz

#1
Quote from: cybe on November 13, 2013, 04:46:15 PM
Doesn't seem faster for me. Or should I have downloaded the remotes/origin/linux-bigmemversion?
Sad to hear, on what system are you running this and how did you compile it? Bigmem is only for using >512mb ram per thread, I'm not yet sure if that's a good idea.
#2
Quote from: tyeken8 on November 13, 2013, 03:49:20 PM
Quote from: Frz on November 13, 2013, 03:18:43 PM
I tried switching to the sphlib library for better sha512 performance, I seem to be getting about 50% more collisions per minute on my test system

I would be happy if someone could test how it affects their performance.
https://github.com/FrzMe/jhProtominer/tree/linux-port

Great work!!
It boost my E5620 from 91 to ~96. Not very fast, don't know why..

I've merged your changeset (-b sph), add memory of 1024 ~4096MB as well.

Thank you!

I've also experimented with bigger memory sizes (https://github.com/FrzMe/jhProtominer/commit/ce443328095264700f8eb3bd3cbea4187e9e115c ) not sure of the results yet. I've moved the worker function to a new file that can be tweaked with macros to avoid code duplication in the source hope I've done that correctly :D I've moved the worker function to a new file that can be tweaked with macros to avoid code duplication in the source hope I've done that correctly :D

My test system is rather slow that's probably the reason why I get so much more performance with the new hash library (8virtual cores with 4500 mhz host cpu capped -> ~ 1/4 of a more or less modern xeon => I originally had about 30cps, now around 42)

QuoteThe last version has 76 c/m in my system, and this new one about 84c/m. BUT, this new version doesn't find any hash in several minutes I have been running it.
Maybe you are just unlucky? :( Of course there might be something wrong but it works fine for me.

QuoteHow are you compiling this?
make CFLAGS="-Ofast -march=native"
Haven't tried it on Windows.
#3
I tried switching to the sphlib library for better sha512 performance, I seem to be getting about 50% more collisions per minute on my test system

I would be happy if someone could test how it affects their performance.
https://github.com/FrzMe/jhProtominer/tree/linux-port