too hard on intel

[Guest]

oops sorry... no i mean't i wouldn't flame even if you were wildly off base... well wildly maybe...

have a few more things to say (of course... never runs out), but won't be back till the weekend. too tired to post right now...

thanks for keeping this clean fella's

i had a drink the other day... opinions were like kittens i was givin' away

 

[Guest]

Some thoughts on why a company may not recompile their code.

It may not be as simple as it seems. My company develops in java and we are planning to change to the latest version of java soon. This is being seen as a major release since lots of code needs to be changed plus a full round of testing. I don't know if this is true of Intel's compiler. Does any one know if they have released this compiler? (I'm guessing they have but I haven't heard anything)

There must be some reason that companies haven't been pushing P4 optimized code. If it was simply a matter of switching compilers we would have seen more optimized software. People are lazy but companies aren't, they will jump at a chance to make their product look better.

 

[Raystonn]

"Does any one know if they have released this compiler?"

Yes, the Intel C/C++ Compiler 5.0 has been out for quite a while. In fact, a Linux version is going into beta in May.

"There must be some reason that companies haven't been pushing P4 optimized code. If it was simply a matter of switching compilers we would have seen more optimized software. People are lazy but companies aren't, they will jump at a chance to make their product look better."

Would that were so... The fact is, it raelly is just a matter of switching compilers. Since noone really believes it, noone is doing it. When more benchmarks of recompiled code are released, the case will become more convincing.

-Raystonn

-- The center of your digital world --

 

[girish]

agreed!
mfc sucks! how do i gurantee my program when 75% of it is written by somebody else?



<font color=blue>die-hard fans don't have heat-sinks!</font color=blue>

 

[Guest]

you have a decent point and has valuable things to kept in it. however i have a few problems with it (even though some of the things i may attack still have merit). 1st you are the first person to bring up the x86 architechure. my original post made no claims about it so why are "we supposed" to be talking about it. i didn't see any signs that said x86 forum anywhere. also the none of the benchmarks i posted were from intel. as i mentioned spec is not perfect but i don't believe it as flawed as you say. if you can post any links of technical experts who back up what you say with solid reasoning i would love to read it and listen to get a better understanding. i don't think you want to get into an argument about parts of computers of amd vs. intel considering via makes most of the the mobo's for amd systems, which do not perform as well, have greater difficulty getting them stable and have problems with 4x agp settings... the best thing amd has going for it is the CPU it seems and it gets trounced is SPEC. problems with spec or not it is a fact, but as stated before you have not given any reasons to believe that specCPU is not a decent benchmark.

regardless or not whether intel adheres strictly to x86 your argument that developers shouldn't have to go back to school isn't very convincing. if you know how one complex system works a similar one is not very hard to learn. it does take time and effort, but not school. just a bit of reading and practice. with this type of logic i can hardly see where any improvements can be made. shouldn't people always be learning new ways to make things better. there are tons of optimizing algorithms that aren't architechure specific why bother learning these?

i don't see what your example is supposed to show. 3x is very arbitrary and i may be inclinded to say no but a slightly larger value i might say yes. also i don't know the specifics between their differences but it seems that standard code will work on both the PIV and the Athlon but additional code is required to take full advantage of the PIV thus making it a superset. what is wrong with this? there are certainly many deficiencies in the x86 architechure why not tweek it a little and the tweeking they have done does not seem to be monumental. clearly there is a line, but whether they have crossed it is another question. don't know why you brought up the itanium and x86 has nothing to do with anything. running code that forces mispredictions on the PIV i think would not show as poor result as the athlon though since the athlon has no branch prediction. could be wrong don't know a whole lot about this and how other components factor in. would love to have some info on it and would gladly listen if you provided sources.

finally i never, nor do i think anyone else here, tried to convince an average user to purchase a PIV over an Athlon or PIII.

i had a drink the other day... opinions were like kittens i was givin' away