Fire 4 is availabe

@kranium wrote:

@nabildanial93 wrote:

Can I ask why the binary is too big in size, albeit it is already packed with mpress? The uncompressed version is even bigger (50mb++)!

Truthfully, I don't understand exactly why it's a big exe...
it might have something to do with compiler optimizations, etc.
I tried VC2010, 2012, and 2013, result was the same,
but as long as it runs and is stable, that's all I care about!

(Packing it only helped 1-2%)

Hi no there's no book hidden in there. Its just the way Kranium has compiled it. Basically the actual engine is only 150Kb but the EXE file has been blown up to 25Mb during the compile. The rest of it is just empty so 99% of the EXE files contains a simple block of 0-bytes with no code or data. It doesn't do any harm or affect the perfromance of the engine, but may use slightly more RAM as a result.

[pgn][Event "3 Minutes/Game"]
[Site "Fire 4 x64 Gauntlet"]
[Date "2014.12.11"]
[Round "20.4"]
[White "Fire 4 x64"]
[Black "Houdini 4 Pro x64"]
[Result "1-0"]
[PlyCount "67"]
[EventDate "2014.??.??"]

1. d4 {book 0s} d5 {book 0s} 2. c4 {book 0s} c6 {book 0s} 3. Qc2 {book 0s} Nf6
{book 0s} 4. Nf3 {book 0s} dxc4 {book 0s} 5. e4 {book 0s} b5 {book 0s} 6. b3 {
book 0s} Bg4 {0.00/15 2} 7. bxc4 {0.36/13 2 (Nbd2)} Bxf3 {-0.05/16 2} 8. gxf3 {
0.26/14 4} e6 {0.44/17 22 (Qxd4)} 9. Be3 {0.52/14 10} c5 {0.50/16 4 (bxc4)} 10.
d5 {0.60/13 3} b4 {0.41/16 4 (Bd6)} 11. e5 {1.04/12 1 (Qb2)} Nbd7 {0.66/15 3}
12. Qa4 {1.01/15 2 (dxe6)} Be7 {1.43/15 15 (exd5)} 13. d6 {1.24/15 3} O-O {1.
89/15 2} 14. dxe7 {1.27/16 1} Qxe7 {1.84/14 0} 15. exf6 {1.24/16 3} Qxf6 {1.80/
13 0} 16. Qxd7 {1.31/16 2} Qxa1 {1.37/16 1} 17. Qd1 {1.38/16 6} Qxa2 {1.51/16 2
} 18. Bd3 {1.42/15 3 (Bxc5)} Rfd8 {1.41/16 13 (a5)} 19. Qc2 {1.43/13 1} Qa1 {
1.49/16 1 (b3)} 20. Bxh7+ {1.59/15 2 (O-O)} Kh8 {1.48/15 0} 21. Be4 {1.57/14 1
(Bd3)} Rac8 {1.56/15 2} 22. O-O {1.62/14 1 (Qb3)} f5 {1.88/14 1 (a5)} 23. Bd3 {
1.48/15 2} a5 {1.48/15 1} 24. Bg5 {1.64/17 13} Rd6 {1.86/17 2} 25. Qe2 {1.72/
16 1} Qa4 {2.11/17 3} 26. Bf4 {1.72/15 1} Ra6 {2.33/15 2} 27. Kh1 {2.06/15 2
(Be5)} Qe8 {2.33/16 1 (Qc6)} 28. Rg1 {2.11/16 1 (Be5)} Ra7 {3.24/17 14} 29. Be5
{2.09/15 2 (Qe3)} a4 {3.81/15 12} 30. Qe3 {2.18/16 6} Qh5 {3.82/16 10 (Kg8)}
31. Be2 {2.31/13 1 (Nd2)} Rcc7 {3.91/15 5 (a3)} 32. f4 {3.54/14 1 (Bxc7)} Qe8 {
7.22/16 7} 33. Qh3+ {3.93/15 1} Kg8 {6.50/12 0} 34. Bh5 {3.98/14 0 (Nd2)} 1-0

[/pgn]

#short_and_sweet #EnginesWar

Fire 4 x64 Gauntlet (finished)
https://www.chess2u.com/t9183-fire-4-x64-gauntlet#62786


https://www.chess2u.com/t6886p45-engineswar-rating-list#62788

#EnginesWar #Fire

41 elo !

41 ELO is a great achievement for Fire - congrat. May be Fire will replace Gull 3 in season 8 TCEC

tks.. Vladimir for you testing...really was a big difference .... on fire 4 ... 41 Elo ... congratulations to Kranium ... good work friend ... much success in the future .. best

@kranium wrote:

Nicolas wrote:

Fire 4 uses more memory (RAM) than it should. This can make Fire 4 slower ?!?!

Hi Nicholas-

I tried some different things concerning memory management, the foremost is utilizing thread local storage SMP...,
which minimizes the 'shared' resources among threads...ie each thread maintains data local to the thread.
This increases MP speed by reducing the need for synchronization...thread locking/waiting/ etc.

https://software.intel.com/en-us/articles/use-thread-local-storage-to-reduce-synchronization

so from a memory management perspective, Fire is somewhat of a unique implementation.
(I'm not aware of any other engine doing this)

Commit and Private seem unaffected, but Working Set bumps up a bit.

Don't forget: RMs Syzygy TB code automatically caches TB hits via the OS...it's normal to see RAM usage climb while utilizing Syzygy bases.

here's a Resource Monitor comparison with SF (both 6 cores - 128 Mb hash) to illustrate the point
http://www.chesslogik.com/images/Fire-SF-mem.jpg

Regards-
Norm

Interesting, thanks for the detailed explanation!

@kranium wrote:

Nicolas wrote:

Fire 4 uses more memory (RAM) than it should. This can make Fire 4 slower ?!?!

Hi Nicholas-

I tried some different things concerning memory management, the foremost is utilizing thread local storage SMP...,
which minimizes the 'shared' resources among threads...ie each thread maintains data local to the thread.
This increases MP speed by reducing the need for synchronization...thread locking/waiting/ etc.

https://software.intel.com/en-us/articles/use-thread-local-storage-to-reduce-synchronization

so from a memory management perspective, Fire is somewhat of a unique implementation.
(I'm not aware of any other engine doing this)

Commit and Private seem unaffected, but Working Set bumps up a bit.

Don't forget: RMs Syzygy TB code automatically caches TB hits via the OS...it's normal to see RAM usage climb while utilizing Syzygy bases.

here's a Resource Monitor comparison with SF (both 6 cores - 128 Mb hash) to illustrate the point
http://www.chesslogik.com/images/Fire-SF-mem.jpg

Regards-
Norm

No, I don't think this is the case here. The original binary size is huge, and Norman packed it using Mpress to compress it into a 24.6 MB binary. While it may not give any impact to performance, the downside of using Mpress is Mpress has to decompress the binary in memory before execution. The memory usage is bigger when using more threads, as binary decompression is done per thread.

You can verify what I was saying by doing this in Arena:

1) Set the hash size at 4MB, and number of threads at 1.

Fire 4 screenshot


2) Now set the number of threads at 8.

Fire 4 screenshot

Most compressed executables decompress the original code in memory and most require slightly more memory to run (because they need to store the decompressor code, the compressed data and the decompressed code). Moreover, some compressed executables have additional requirements, such as those that write the decompressed executable to the file system before executing it.

http://en.wikipedia.org/wiki/Executable_compression

Last edited by nabildanial93 on Sat Dec 13, 2014 6:26 am; edited 1 time in total (Reason for editing : Add wiki reference)

@kranium wrote:

http://users.telenet.be/chesslogik/downloads/Fire_4.rar

Is it based on Stockfish 4?

Mark Briggs wrote:

@kranium wrote:

http://users.telenet.be/chesslogik/downloads/Fire_4.rar

Is it based on Stockfish 4?

of course not, don't be ridiculous
totally uncalled for
pure maliciousness, and quite rude of you to even suggest it

Hmmm I didn't attack you, it was just a question!

Fire_4.rtf file:


Can you please tell us more about the engine?

Kranium,

will there be a version for old processors (w32 SSE2)?

Mark Briggs wrote:

Hmmm I didn't attack you, it was just a question!

Fire_4.rtf file:

Code:

Fire 4 has been completely re-engineered from top to bottom,
and does not use any source code from (or closely related to) Ippolit.

Can you please tell us more about the engine?

It's not a difficult question to answer, is it?

It looks like I should have said a gain of 40 ELO's from Fire 4. It's early in my 3/3 tournament but so far it is IMPRESSIVE!!!
GOOD JOB, Mr. Schmidt!!!

Fire 4 64-bit 4CPU Gauntlet [Long Time Control]
https://www.chess2u.com/t9190p15-fire-4-64-bit-4cpu-gauntlet-long-time-control#63406

#Beautiful_Game :

[pgn][Event "Fire 4 x64 Gauntlet"]
[Site "Philadelphia"]
[Date "2014.12.23"]
[Round "36.2"]
[White "Fire 4 x64"]
[Black "Houdini 4 x64 B syz"]
[Result "1-0"]
[ECO "B80"]
[WhiteElo "3300"]
[BlackElo "3300"]
[Annotator "0.14;-0.02"]
[PlyCount "59"]
[EventDate "2014.12.14"]
[EventType "simul"]
[Source "Russell"]
[TimeControl "40/1140:20/1140:1140"]

{Intel(R) Core(TM) i7-4940MX CPU @ 3.10GHz 3292 MHz  W=18.9 plies; 7,037kN/s;
Perfect2014.ctg  B=21.8 plies; 9,164kN/s; 1,661 TBAs; Perfect2014.ctg} 1. e4 {
B 0} c5 {B 0} 2. Nf3 {B 0} d6 {B 0} 3. d4 {B 0} cxd4 {B 0} 4. Nxd4 {B 0} Nf6 {
B 0} 5. Nc3 {B 0} a6 {B 0} 6. Be3 {B 0} e6 {B 0} 7. f3 {B 0} b5 {B 0} 8. Qd2 {
B 0} Nbd7 {B 0} 9. O-O-O {0.14/17 17} Bb7 {-0.02/24 39} 10. Kb1 {0.20/19 46
(a2-a3)} Rc8 {-0.03/23 126 (b5-b4)} 11. g4 {0.14/19 40} h6 {-0.01/22 54
(Nd7-e5)} 12. a3 {0.22/19 31} Nb6 {0.15/21 24 (Nd7-e5)} 13. Bf2 {0.15/19 121
(h2-h4)} Nfd7 {-0.03/22 48} 14. Qe1 {0.12/19 47 (h2-h4)} Qc7 {-0.01/22 28
(h6-h5)} 15. Rg1 {0.12/19 43 (h2-h4)} Nc4 {0.04/23 153} 16. Bxc4 {0.06/18 24}
Qxc4 {0.08/23 28} 17. Nb3 {0.09/19 48 (h2-h4)} Qc7 {-0.02/23 21} 18. h4 {0.08/
20 41 (Nb3-d4)} Ne5 {0.00/24 49} 19. Rg3 {0.09/20 19 (Qe1-e2)} Be7 {0.16/23 63
(h6-h5)} 20. f4 {0.18/21 58} Nc4 {0.17/22 37} 21. g5 {0.20/21 191} g6 {0.11/22
14 (h6xg5)} 22. Rgd3 {0.21/19 17 (Nb3-d2)} hxg5 {0.14/21 22} 23. hxg5 {0.20/18
7} Rh5 {0.14/22 112 (Ke8-f8)} 24. Qe2 {0.21/18 23 (Bf2-g3)} Kf8 {0.18/21 34
(Bb7-c6)} 25. Qf3 {0.44/17 14 (Bf2-d4)} Kg8 {0.18/20 11 (Kf8-e8)} 26. Bd4 {0.
47/20 23} Bf8 {0.29/21 17} 27. Bf6 {1.09/19 16} Bg7 {0.29/19 0 (d6-d5)} 28. Rh1
{1.61/18 12 (Qf3-e2)} d5 {5.92/19 69} 29. Rxh5 {3.12/17 9} dxe4 {7.06/20 24}
30. Rh7 {3.60/19 13} 1-0

[/pgn]

[pgn][Event "3 Minutes/Game"]
[Site "Stockfish 02-01-2015 64bit"]
[Date "2015.01.03"]
[Round "39.5"]
[White "Stockfish 02-01-2015 64bit"]
[Black "Fire 4 x64"]
[Result "0-1"]
[PlyCount "52"]
[EventDate "2015.??.??"]

1. d4 {book 0s} Nf6 {book 0s} 2. Nf3 {book 0s} d5 {book 0s} 3. c4 {book 0s} c6
{book 0s} 4. Qc2 {book 0s} dxc4 {book 0s} 5. e4 {book 0s} b5 {book 0s} 6. b3 {
book 0s} Qa5+ {0.20/13 2} 7. Bd2 {0.67/17 2} cxb3 {0.20/15 2} 8. Qxb3 {0.62/19
4} Qd8 {0.22/15 6 (Qb6)} 9. Bd3 {0.57/17 3 (Qc2)} Bg4 {0.32/14 3 (e6)} 10. Ne5
{0.63/16 2 (a4)} Qxd4 {0.03/15 4 (e6)} 11. Qxf7+ {1.06/19 3} Kd8 {-0.46/6 0}
12. Bc3 {1.15/21 3} Qd6 {0.02/15 2} 13. Nxg4 {1.07/21 2 (Ba5+)} Qxd3 {0.08/15 1
} 14. Nxf6 {1.34/22 3 (Ba5+)} gxf6 {0.30/15 8} 15. Ba5+ {1.27/21 3} Kc8 {0.24/
16 3} 16. Nc3 {1.30/23 5 (Qe6+)} Qc4 {0.08/16 4 (Kb7)} 17. Qe8+ {1.36/19 2
(Qxc4)} Kb7 {0.86/6 0} 18. Qd8 {0.00/19 3 (a4)} e5 {0.00/14 2} 19. a4 {0.00/21
4} b4 {-0.08/15 2} 20. Qc7+ {0.00/22 1 (Ne2)} Ka6 {-0.23/6 0} 21. Ne2 {0.00/24
1 (Qc8+)} Qe6 {-1.14/14 2 (b3)} 22. Rc1 {-1.25/20 6 (Rd1)} Be7 {-1.82/15 1
(Qd7)} 23. Nc3 {-3.52/21 4} Rc8 {-2.42/15 1} 24. Nb5 {-4.53/23 11} Nd7 {-2.54/
14 1} 25. O-O {-5.05/21 3} cxb5 {-2.85/14 2} 26. axb5+ {-5.13/20 2} Kxb5 {-0.
95/6 0} 0-1

[/pgn]

#short_and_sweet

Dear Kranium...how do you do? I hope all is well and has passed happy holidays (Christmas, New Year) ... something new for Fire 4.1 ... hug and have a great year, very successful ..

Hello.

Well, i downloaded the engine to test it, but my antivirus detect the 32bits of fire4.exe as a virus. However not the 64bits version.

Sounds like a false positive to me, never had any problems with Fire.

All Norton products are bug themself ! Remove them quickly !  

Nicolas wrote:

Sounds like a false positive to me, never had any problems with Fire.

All Norton products are bug themself ! Remove them quickly !  

I have Norton. I do sometimes get false positives but it does keep me safe. I do seem to recall an instance as described above.

Many one knows when will out fire 5?

Who share link on Fire 5?