TalkChess.com

abrok.eu is not an official SF site. Official SF9 binaries will be available from the official site once the best compiles are selected:
https://stockfishchess.org/download/

The abrok POPCNT compile is at least 5% slower than that by Ipman, BTW.

A Distel wrote:Official release version of Stockfish 9
!! latest version !!
http://abrok.eu/stockfish/

Windows x64 for Haswell CPUs
Windows x64 for modern computers
Windows x64
Windows 32
Linux x64 for Haswell CPUs
Linux x64 for modern computers
Linux x64
Author: mstembera
Date: Wed Jan 31 11:41:09 2018 +0100
Timestamp: 1517395269

Bench 5023629

Some macOS binaries - feel free to post on abrok, just let me know and I will take down the dropbox link.

https://www.dropbox.com/s/l52oiye0ts079 ... S.zip?dl=1
source: https://github.com/official-stockfish/S ... 802294bbad
use whichever one work on your machine. For me , Stockfish-9-bmi2-O3 is fastest on a 2015 iMac (2.7M nps), Stockfish-9-mPro-O2_mo is fastest on a 2010 Mac Pro (2.0M nps). Will work on pulling together some cFish binaries, link will be posted on the cFish thread.

On a 2014 Apple Mac Pro, when running Stockfish 9, I am seeing a nps of 4.7m! This is with two cores running in analysis mode (middle game phase) within Hiarcs Explorer. This is more than double the speed quoted by Michael B at 2.0m nps? Perhaps Michael is quoting for 1 core only?

kk wrote:On a 2014 Apple Mac Pro, when running Stockfish 9, I am seeing a nps of 4.7m! This is with two cores running in analysis mode (middle game phase) within Hiarcs Explorer. This is more than double the speed quoted by Michael B at 2.0m nps? Perhaps Michael is quoting for 1 core only?

I assume the speeds he reports are from the built-in Stockfish benchmarking routine, which by default uses one thread.

kk wrote:On a 2014 Apple Mac Pro, when running Stockfish 9, I am seeing a nps of 4.7m! This is with two cores running in analysis mode (middle game phase) within Hiarcs Explorer. This is more than double the speed quoted by Michael B at 2.0m nps? Perhaps Michael is quoting for 1 core only?

Correct one core only.

tpoppins wrote:abrok.eu is not an official SF site. Official SF9 binaries will be available from the official site once the best compiles are selected:
https://stockfishchess.org/download/

The abrok POPCNT compile is at least 5% slower than that by Ipman, BTW.

I noticed ipman-compiled SF9 binaries were packed with UPX packer and was
tagged by some AV vendor as malicious. I wonder why it was packed and
reduced to around 1/4th the size of the normal compile.

MikeGL wrote:I noticed ipman-compiled SF9 binaries were packed with UPX packer and were tagged by some AV vendor as malicious. I wonder why it was packed and
reduced to around 1/4th the size of the normal compile.

I don't know why Ipman chooses to pack his builds that way but I doubt that that's the reason why they are so small. The size reduction is most likely due to them being dynamic compiles, i.e. certain libraries (like libstdc++-6) are not linked statically during compilation. The supposed benefit is a small (a couple percent) increase in speed. SugaR and SF BYO (and some other engines I forget ATM) are typically built that way.