Modern Times wrote:It isn't an XP64 issue. It doesn't work on W7 or W8 on AMD Phenom II either. The compile clearly expects the full Intel SSE4.2 instruction set to be present, which is not the case on our CPUs.
I got an email telling me that it runs on Phenom2 with W7 so I assumed it is the OS again but you are right! It works on XP64 AND an AMD 8350 as well. So it is really a hardware issue and I can continue with the normal 64 bit compile - great
(Is there a possibility to get a pure POPcount compile? th ediffeence now between 64 and 64-SSE42 is 3%, that is usually the same difference to the POP only compile. In other words: Is there a big diffeence in speed between full SSE42 and just POPcount?)
Modern Times wrote:I'm not an expert - but my understanding is that it is just the popcount instruction that gives the speed increase.
Even Jim had a lof of difficulty in the past compiling a popcount exe for Phenom II, so I wouldn't be hopeful of a quick solution.
Up to you if you wait or not. A 3% speed increase might be a couple of Elo, I don't know how critical that might be in terms of its ranking.
There is no problem for Houdini or Komodo doing an autodetection for POP ANDPhenom2 but anyhow, I already started with the normal 64bit as hat small difference is not worth the trouble.
Martin Thoresen wrote:Time for you to move on from ancient operating systems Ingo.
I know, I know ...
But there are 3 reasons why this is difficult:
1. 8 Computers need a license (Money)
2. 8 perfectly running computers upgrade (time)
3. 1 + 2 to get no difference in the result
Basicaly I hesitate because there would be no need if the compilers do not touch the "otimize for W7" button (and there is no gain in speed from XP64 (wich is WServer2003, it is not the ancient normal XP!) to W7!)
Bye
Ingo
Move to Linux. No licence fee's. Install on as many computers as you wish and frequent upgrades. No Microsoft hassles and broken service packs.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
At least in computer chess, the power of money is now defeated. Is that good or bad? I am not a philosopher.
P.S. I am currently running a match against Komodo CCT at 0:30m+0.250s ponder=on, and under these conditions, it seems that Komodo is way superiour: Intermediate result is 70% for Komodo. Maybe it gets better, only 30 from 80 games have been played.
I think it's bad. Developing a chess program takes a lot of time. If you only get back " Thank you" that's not enough. Better give a reward in money or 'Talk Chess dollars' ?.
Don wrote:
Move to Linux. No licence fee's. Install on as many computers as you wish and frequent upgrades. No Microsoft hassles and broken service packs.
Sorry, but that is no option. I have 2 Linux devices running (one laptop and one Raspi) both give me more headache than all 8 IPON computers together!
IWB wrote:Sorry, but that is no option. I have 2 Linux devices running (one laptop and one Raspi) both give me more headache than all 8 IPON computers together!
No, I will not continue in a flamewar.
Bye
Ingo
I agree with this. I've dabbled in Linux for 10 years or more, and I always come back to Windows. You replace one Microsoft hassle with ten Linux hassles.
Yes ! Stockfish 4 ROCKS ! Using the latest developmental versions I was already beating Houdini 3 users online regularly ; now with Stockfish 4 I'm CRUSHING Houdini 3 users ; ( of course with one changed parameter which I will not reveal for obvious reasons )
)