Author: Robert Hyatt
Date: 08:12:41 04/22/02
Go up one level in this thread
On April 22, 2002 at 01:43:50, Hristo wrote: >On April 21, 2002 at 14:37:47, Robert Hyatt wrote: > >>On April 21, 2002 at 04:15:04, Hristo wrote: >> >>>On April 21, 2002 at 00:54:47, Kevin Strickland wrote: >>> >>>>On April 21, 2002 at 00:51:54, Kevin Strickland wrote: >>>> >>>>>On April 21, 2002 at 00:43:27, Vincent Diepeveen wrote: >>>>> >>>>>>On April 21, 2002 at 00:39:14, Kevin Strickland wrote: >>>>>> >>>>>>>>there are 2 ways to discredit results >>>>>>>> a) discredit the guy who gets them >>>>>>>> b) discredit the program he gets them with >>>>>>>> c) discredit the accuracy the testresults were gotten in >>>>>>>> d) start sueing and during the trial disallow results getting posted >>>>>>>> >>>>>>>>he's trying a+b now. >>>>>>>> >>>>>>>>Very cheap. >>>>>>>> >>>>>>>>The real thing he DISLIKES to hear is that >>>>>>>>INTEL C++ is FASTER now and it might take another year before >>>>>>>>m$ will release a new version of its compiler that can beat it! >>>>>>> >>>>>>>No I think you took his statement wrong. He was not trying to discredit the >>>>>>>compiler results, he was definately trying to discredit the statement that Diep >>>>>>>is "professionally programmed". He was taking a shot at you, not Intel. >>>>>> >>>>>>That's the same as trying to discredit results that show the msvc compiler >>>>>>is not fastest anymore. >>>>>> >>>>>>He seems to WORK daily onto the msvc compiler. >>>>>> >>>>>>Also we must take into account that usually a new released processor >>>>>>using a new compiler is doing better on specint programs usually. >>>>>> >>>>>>There is known an example recently where a FORTRAN program was speeded >>>>>>up a factor of 9 or something. >>>>>> >>>>>>It is impossible to lossless speedup DIEP 9 times at the same hardware. >>>>>> >>>>>>In fact 3.5% is nearly impossible. >>>>>> >>>>>>That's exactly what intel c++ did! >>>>> >>>>>I might have missed where he said that the MS compiler was still better. Could >>>>>you possibly show me within this thread where he stated that? >>>>> >>>>>Again I state he was discreditting you, not the compiler. >>>> >>>>In fact this is _exactly_ what he wrote: >>>> >>>>I would not tell "However to all standards, programming of DIEP is very >>>>professionally done" about the programmer who needlessly duplicates tens of >>>>megabytes of data only because he did not figured out how to efficiently use >>>>threads instead of processes. >>>> >>>>Or about the programmer who included others' code into his program without the >>>>permission. >>>> >>>>Eugene >>>> >>>>Please point out where he _discreditted_ the results you posted about. I am >>>>reading what he wrote and frankly he was calling you a theif and a >>>>non-professional programmer. >>>> >>> >>>Jeess, >>>Kevin wake up! >>>Nowhere in the original message did I see Vincent claming or proving that he is >>>not an idiot. Vincent's message is clear INTEL HAS A BETTER (FASTER EXECUTABLE >>>SPEED) COMPILER THAN MS! Eugene, who works for M$ compiler team, responds with a >>> wild claim that multiple threads are better than multiple processes and also >>>that Vincent is a thief! How professional and to the point this is?! What about >>>the actual speed? How it starts the FUD on personal level against the messenger >>>... If you can not discredit the message, discredit the messenger. ... >>>I've seen the EGTB code in crafty ... it is far from professional quality. More >>>like a patched-up spaghetti code. Shall we start complain about it and call >>>Eugene names?! I think not! (most of these tpe of statements are personal >>>opinion, anyway, while the speed of the executable is not.) That, however, is >>>not what Eugene does. He goes personal ... >> >>This is nonsense. Eugene didn't "go personal" at all. > >riighttt ... not at all! > >> >>He first made a factual statement. Using threads saves a _ton_ of memory >>over using N distinct processes. This has been discussed before and is a >>definite fact. The EGTB indices are replicated for each process. The EGTB >>cache is replicated for each process. EGTB I/O is replicated for each >>process, hoping that the O/S helps here a bit. >> >>Threads are better in that regard, significantly. > >Baloney ... processes do NOT require replication of data. Processes give you the >option to choose!!! You should know that. The design patterns used by Vincent >are not representative of what one can do with processes and threads. >The "definite fact" you are talking about is only related to one implementation. > >Your statement is wrong and missleading! My statement is correct and not misleading at all. Threads share _everything_. Processes do not. And it is _impossible_ to share everything between processes. Things like I/O buffers that the library allocates. Etc. So it isn't quite as simple as you suggest. With threads you _also_ have the option to choose what to share and not to share. It is quite easy in _either_ approach to share tree search data. But when you factor in EGTB support, as currently written and released by Eugene, it works perfectly for threads, not for separate processes. And then there is the other issues... replicated instruction pages, replicated this and that. Where lightweight processes don't have that baggage... > >> >>Eugene then mentioned something about "permission". I assume he is talking >>about _his_ EGTB.CPP code which is most definitely his and he holds the >>copyright to. If Vincent didn't ask for permission to use it, then Eugene >>can certainly point that out. >> >>As far as the professional coding ability of Eugene vs Vincent, I will >>personally take Eugene. I've seen code by both, and Eugene has a lot more >>experience at writing good code... >> >>I hope this thread can die a grafeful death now.. > >I hope so too. No more missinformation and BS. >Does Intel C++ compiler generate better (faster) code than M$VC? > >hristo > >> >> >> >> >> >> >>> >>>Thank you Eugene for sheding some light on the topic of execution speed! >>> >>> >>>>I think you are taking what he said that turning this into a MS vs Intel battle >>>>when he was just plain calling you an idiot. Just my two cents. >>> >>>Which it is (a battle) for the people that work in the correct departments and >>>that is why he's been called an "idiot". There is nothing left for the M$ guy to >>>do. >>> >>>Regards, >>>hristo
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.