AMD Alleges Intel Compilers Create Slower AMD Code

[imported_Frallan]

Originally posted by: AnandThenMan

AMD Alleges Intel Compilers Create Slower AMD Code

Seems Intel's compiler also creates slow code on the Intel central processing unit.

Yupp and that will probably save Intels @ss on this one. All they have to asy is that it is optimized for P4 and everything els gets shatfted. Now then they are hitting themself as well as AMD and then its ok...

 

[imported_Frallan]

Hmm I just realized something... As I stated above I believe AMD is going to loose this one. How ever in order to win this INTEL will have to explain how the compilers work and how. This will then be used by AMD to strengthen the other case. Brilliant.

The best answer to this from INTELs side would be to try to bundle it into the other case where it would hardly be noticed except as a "strong point" for INTEL. In taking this to court AMD has at least prevented damage in the other case. This is looking more and more like a game of chess. Right now AMD have made a Very strong opening however we all know that INTEL have openings and is about to make the counter attack.

 

[The Pentium Guy]

Wow. They should DIE and BURN in HELL now. I can't wait to hear intel's crappy "counter attack":
"We have just been doing what isi n our best intesrest to do"

Or somethign stupid like that.

 

[Cerb]

Originally posted by: HDTVMan
A while ago there was an guy who did programming in a binary level. As I recall he removed the check for Genuine Intel and AMD got a big performance boost. I for one thought he was full of crap but apparently it must be legit.

All hell is going to break loose.

If Intel deliberately caused instability for Genuine AMD there is going to be hell to pay.

Companies and Consumers could file suits against Intel. Tough to prove but I would imagine AMD has some binary programmers under microscope with this one.

What Applications and what else has been degraded or worse made unstable.

1. It's Authentic AMD .
2. Instability? Um, no; that would be VIA (up to the KT266A), SiS (at least until the 735), and ALi.

The trouble is that they will have to prove that doing this is anti-competitive. AMD would have been a lot better off to have taken some of the Alpha folks for a compiler team.

Saying it's optomized for Intel CPUs is not enough, if AMD can prove that Intel's compiler dispatching slower code to non-Intel CPUs is anti-competitive. It should dispatch the proper code if the CPU has the feature, regardless of brand. But legally, it could be a whole other matter.

 

[The Pentium Guy]

A while ago there was an guy who did programming in a binary level. As I recall he removed the check for Genuine Intel and AMD got a big performance boost. I for one thought he was full of crap but apparently it must be legit.

That's excellent news actually . When I make my switch over to AMD from this melting prescott, hopefully they'll have sued Intel and fixed this issue!

Unfortunately, try explaining this to a panel of jury: "What? Ghz means more! How can 3 ghz be slower than 1.8? They must be lying to us or doing some type of trick or something!"

Die Intel Die! Burn Burn Burn. I'm extremely angry with their malpractice. Normally I don't start/engage in flame wars, but if they're DOING something like this, then this should be their fate. Seriously. I see threads about "Intel vs amd" all the time and try to avoid those. But I can't forgive something like this. Unbeleivably incredible.

Or, actually, on a more positive note.... competition would be helpful. If Intel went bankrupt all of a suden then AMD would price gouge. So we need competition.

//-// The Pentium Guy //-//

 

[Duvie]

http://www.techreport.com/

read the front page there....Intel is guilty as heck and should pay...The more and more I read about what they have done the more and more I am disgusted in them....What a bunch of chicken-sh^ts...I am embarassed for them as much as I am disgusted at them....

 

[christoph83]

First of all this is part of the same lawsuit not another lawsuit. A lot of people here jumping to conclusions based off a couple of articles. And quite honestly I don't see what the big deal is. Although its a cheap move by intel to make their compiler run slower on amd chips, amd to me, seems even worse for making this part of their case.

It's intels compiler and they can do whatever they want with it. If you notice on their website it says that their compilers are designed for intel 32 bit and 64 bit processors. Heck they are even releasing a new compiler designed for their multi-core chips.

If AMD were smart, they would get up and design their own compiler and sell it for 400 bucks like intel is doing promising 20-30% boosts over intels compiler. But instead they are suing.

And since it seems there are other options for compilers out there, people have other choices. So based off all of this, Unless intel marketed balanced performance on AMD and intel chips, how exactly is this illegal?

 

[Duvie]

AMD and Untel have a cross-licensing agreement with SSE and SSE2, right??? then by having their cpu flags not enable SSE and SSE2 on the compiler even though AMD has them seems a violation of that...

I wouldn't argue with it if AMD needed something added above Intel to run more optimized. i wouldn't expect Intel to optimize or tweak for their competitor but this isn't about this...this is about a very specific cpu flag that sole purpose is to look for intel trademark cpus..Once it was removed the compiler would use the SSE and SSE2 functions and perform 10% faster on the opteron...that is just plain crookedness...If you can justify this I would question why you are...

 

[exar333]

Originally posted by: ssvegeta1010
Wow, that just sounds horrible. I can understand all the aggresive, unethical retail stuff being done, but this just sounds even more blantantly illegal.

3 Intels, No AMD as of yet. (Receiving one soon ) I have seen the errors of my ways.

illegal? whats illegal about writing a code that optimizes for your
products? didb't they design the x86 processor? you all live in a dream land where companies work together for the common good of people....get real...try to shed a little fanboyism and see the real world...i love AMD...but I think some of what they are doing is just as nefarious...what happens to Intel stockholdsers if they fold the company? those are common "joes" as well....AMD is no angel...

 

[DrMrLordX]

Christoph, whether or not AMD should sue will only be known after the court battle is over. If AMD wins, then it will be obvious(albeit in hindsight) that they should have sued. Considering the fact that AMD has been fighting, and winning, an uphill battle for years vs Intel, I am not prepared to say "don't sue, just be more competative" to AMD. They're already competative. Now they're seeking justice. Apparently, they have spent a long time gathering evidence, and now they're chosen to use it.

They've got a technological upper-hand at the moment, so this is an excellent time for AMD to land a staggering blow against Intel's strongest corporate unit: their marketting dept(which seems to be guiding their overall business strategy, and has been ever since Intel struck a deal with RAMBUS, Inc). If AMD can shake consumer and vendor confidence in Intel's overall business ethics, they will gain a foothold in OEMs that have previously been off-limits to AMD processors. It will be a huge paradigm shift. Other small players may re-emerge and try to challenge AMD.

As far as the legality of Intel's slanted compiler goes, this discussion has already taken place in other threads. Go read them and make up your own mind. Many individuals, ranging from common people to judges, will think of this as a monopolistic practice.

 

[harrysinghk]

well i am sworn of them Intel chips for good ever since i found that they wont let you overclock them and totally discourage it. If i am willing to take the heat risk and other factors i dont see why they wont let people tinker some and try to use their intellect and get more out of the chips after all all that extra effort and hw isnt cheap either.

 

[christoph83]

Originally posted by: Duvie
AMD and Untel have a cross-licensing agreement with SSE and SSE2, right??? then by having their cpu flags not enable SSE and SSE2 on the compiler even though AMD has them seems a violation of that...

I wouldn't argue with it if AMD needed something added above Intel to run more optimized. i wouldn't expect Intel to optimize or tweak for their competitor but this isn't about this...this is about a very specific cpu flag that sole purpose is to look for intel trademark cpus..Once it was removed the compiler would use the SSE and SSE2 functions and perform 10% faster on the opteron...that is just plain crookedness...If you can justify this I would question why you are...

Maybe that's something to check out. But AFAIK the liscence is just to let amd use those SSE,SSE2 instructions in their chips. Since when does intel have to optimize their own compiler to enhance AMD's chips?

Like I said this is a cheap move. But once again it's intel's compiler and they can do whatever they want with it unless they specifically said it was optimized for AMD chips. AMD deserves no money for this and has no case on this portion of the suit. The only thing this shows is how babyish they are both acting. Intel crying to themselves because their code would run faster on an AMD chip if fully optimized, and AMD crying that intel wont optimize their chips in a compiler OWNED and marketed to run on intel chips.

 

[Leper Messiah]

Originally posted by: Duvie
http://www.techreport.com/

read the front page there....Intel is guilty as heck and should pay...The more and more I read about what they have done the more and more I am disgusted in them....What a bunch of chicken-sh^ts...I am embarassed for them as much as I am disgusted at them....

Matrox, what does matrox have to do with this.....oh. You expect me to scroll down?

 

[HendrixFan]

Originally posted by: christoph83

Originally posted by: Duvie
AMD and Untel have a cross-licensing agreement with SSE and SSE2, right??? then by having their cpu flags not enable SSE and SSE2 on the compiler even though AMD has them seems a violation of that...

I wouldn't argue with it if AMD needed something added above Intel to run more optimized. i wouldn't expect Intel to optimize or tweak for their competitor but this isn't about this...this is about a very specific cpu flag that sole purpose is to look for intel trademark cpus..Once it was removed the compiler would use the SSE and SSE2 functions and perform 10% faster on the opteron...that is just plain crookedness...If you can justify this I would question why you are...

Maybe that's something to check out. But AFAIK the liscence is just to let amd use those SSE,SSE2 instructions in their chips. Since when does intel have to optimize their own compiler to enhance AMD's chips?

Like I said this is a cheap move. But once again it's intel's compiler and they can do whatever they want with it unless they specifically said it was optimized for AMD chips. AMD deserves no money for this and has no case on this portion of the suit. The only thing this shows is how babyish they are both acting. Intel crying to themselves because their code would run faster on an AMD chip if fully optimized, and AMD crying that intel wont optimize their chips in a compiler OWNED and marketed to run on intel chips.

This by itself would make your statement true. However, when pieced together with the market position Intel has, and followed with the many other examples cited in the suit; this is a damning fact for Intel.

Its simple economics. If you are a company in a competitive marketplace, making a compiler that is *intentionally* malfunctional will hurt the product and your company's reputation. In a monopoly driven marketplace, doing this will help to ensure that competitors do not threaten the monopoly. In the US, this is illegal.

 

[UNCjigga]

Those "Evil Inside" t-shirts were no joke!!!

 

[Duvie]

Originally posted by: christoph83

Originally posted by: Duvie
AMD and Untel have a cross-licensing agreement with SSE and SSE2, right??? then by having their cpu flags not enable SSE and SSE2 on the compiler even though AMD has them seems a violation of that...

I wouldn't argue with it if AMD needed something added above Intel to run more optimized. i wouldn't expect Intel to optimize or tweak for their competitor but this isn't about this...this is about a very specific cpu flag that sole purpose is to look for intel trademark cpus..Once it was removed the compiler would use the SSE and SSE2 functions and perform 10% faster on the opteron...that is just plain crookedness...If you can justify this I would question why you are...

Maybe that's something to check out. But AFAIK the liscence is just to let amd use those SSE,SSE2 instructions in their chips. Since when does intel have to optimize their own compiler to enhance AMD's chips?

Like I said this is a cheap move. But once again it's intel's compiler and they can do whatever they want with it unless they specifically said it was optimized for AMD chips. AMD deserves no money for this and has no case on this portion of the suit. The only thing this shows is how babyish they are both acting. Intel crying to themselves because their code would run faster on an AMD chip if fully optimized, and AMD crying that intel wont optimize their chips in a compiler OWNED and marketed to run on intel chips.

That is where you are going wrong....They dont have to optimize anything if you read that article...AMD is using the same SSE and SSE2 codes as Intel and doesn;t need optimization. Instead as the article showed that Intel uses a cpu flag to search for its cpu and if it is not an Intel cpu it wont enable SSE and SSE2 codes period even when they are there...What you dont get, they had to add code to the compiler to it to get it not to work on others...Instead of a simple if SSE and SSE2 are there use them...

They went out of their way to make it not work....Your logic is flawed here...


You are right about one thing...they should create thier own but sell it cheaper then Intel...heck I mean if Intel and the bunch of hackls they have become lately I am sure AMD can do it and likely do it better....

 

[Lithan]

Let's not forget they don't give their compiler away. They are selling it and PURPOSELY making it disfunctional. That is more than illegal. That is throw your ass in jail for fraud kind of illegal.

 

[HDTVMan]

Originally posted by: christoph83
First of all this is part of the same lawsuit not another lawsuit. A lot of people here jumping to conclusions based off a couple of articles. And quite honestly I don't see what the big deal is. Although its a cheap move by intel to make their compiler run slower on amd chips, amd to me, seems even worse for making this part of their case.

It's intels compiler and they can do whatever they want with it. If you notice on their website it says that their compilers are designed for intel 32 bit and 64 bit processors. Heck they are even releasing a new compiler designed for their multi-core chips.

If AMD were smart, they would get up and design their own compiler and sell it for 400 bucks like intel is doing promising 20-30% boosts over intels compiler. But instead they are suing.

And since it seems there are other options for compilers out there, people have other choices. So based off all of this, Unless intel marketed balanced performance on AMD and intel chips, how exactly is this illegal?

Sorry forcing instability into a consumers application because you didnt buy from Intel is highly illegal and unethical. This can lead to a large class action suit if it can be proven they did do intentially. Which it appears it was.

 

[crimson117]

Originally posted by: Lithan
Let's not forget they don't give their compiler away. They are selling it and PURPOSELY making it disfunctional. That is more than illegal. That is throw your ass in jail for fraud kind of illegal.

I was chatting about this on slashdot, and came to thinking that for this allegation, Intel didn't do anything illegal.

They only claim that their compiler produces code optimized for Intel processors. By optimized, they mean it uses standard optimized instruction sets like MMX and SSE when it detects Intel processors, and forces old-fashioned slow byte-by-byte operations for non-Intel processors. Sure it's a lousy way to write a compiler, and it only makes sense for developers who are compiling code that will only ever be run on Intel processors, but there's nothing actually illegal about it.

From http://www.intel.com/cd/software/products/asmo-na/eng/compilers/219754.htm
Achieve outstanding application performance on Intel® processors using Intel® C++ Compiler for Windows*.

See? They say it'll give great performance on Intel processors. Any wise developer should think, before using this compiler, "Well what does it do on non-Intel processors?", then ask the question, and if they find out it write crappy code for non-Intel processors, they should use another compiler.

 

[Fox5]

Originally posted by: Bona Fide
Hm...AMD sure seems to be busting out all these allegations out of nowhere...I'm sure someone thinks that these are starting to lose their legitimacy. Even if AMD is right, they are gonna get some disrespect from the consumer community because of all this all-out warfare with Intel.

All the allegations have been rumored about in the computer industry for years, including this latest one.
I'm not sure how many current programs have tags to keep them from running on AMD processors, but I actually encountered a few in the past.(they brought up error messages saying this program can only be run on an Intel processor)

BTW, I wouldn't be surprised if Intel used the compiler trick to also hurt the performance of their Celeron processors, and thus create more 'value' for their Pentium line.

A while ago there was an guy who did programming in a binary level. As I recall he removed the check for Genuine Intel and AMD got a big performance boost. I for one thought he was full of crap but apparently it must be legit.

I believe someone recompiled the quake 3 binary for AMD and got a huge boost in performance. Quake 3 was one of the few games were intel consistently dominated in, but after the recompile AMD and Intel were dead even.

Like I said this is a cheap move. But once again it's intel's compiler and they can do whatever they want with it unless they specifically said it was optimized for AMD chips. AMD deserves no money for this and has no case on this portion of the suit. The only thing this shows is how babyish they are both acting. Intel crying to themselves because their code would run faster on an AMD chip if fully optimized, and AMD crying that intel wont optimize their chips in a compiler OWNED and marketed to run on intel chips.

Supposendly it's not just not optimizing for AMD, it is checking to see if the cpu is an AMD and then running a slower codepath if it is. The compiler should just check for SSE support or not SSE support and then go from there, not check for Intel or AMD, and if AMD use x86, if Intel check for SSE.

You are right about one thing...they should create thier own but sell it cheaper then Intel...heck I mean if Intel and the bunch of hackls they have become lately I am sure AMD can do it and likely do it better....

AMD doesn't do much else then design CPUs, Intel is a far more capable company. Though a compiler is fairly related to a CPU, so you'd think AMD would take some involvement in it, but AMD has never really designed much good that wasn't their cpu, or directly integrated into their cpu.(hypertransport) Their chipsets always kind of sucked, stable, but not high performance.

See? They say it'll give great performance on Intel processors. Any wise developer should think, before using this compiler, "Well what does it do on non-Intel processors?", then ask the question, and if they find out it write crappy code for non-Intel processors, they should use another compiler.

I believe the Intel compiler is actually the fastest x86 compiler around, whether or not it doesn't fully support AMD, so it's not an option of finding a better product, the compiler is the best product.

 

[100proof]

Originally posted by: HDTVMan
A while ago there was an guy who did programming in a binary level. As I recall he removed the check for Genuine Intel and AMD got a big performance boost. I for one thought he was full of crap but apparently it must be legit.

All hell is going to break loose.

If Intel deliberately caused instability for Genuine AMD there is going to be hell to pay.

Companies and Consumers could file suits against Intel. Tough to prove but I would imagine AMD has some binary programmers under microscope with this one.

What Applications and what else has been degraded or worse made unstable.

http://www.swallowtail.org/naughty-intel.html

 

[JDCentral]

We should all just use gcc...



What about the other x86 processors?

Is VIA going to sue for Intel not optimizing it's compilers for the x86-Eden CPUs?

I am a partial AMD fanboy... but I think that AMD will lose this particular argument.

If it really wants full optimization, AMD should write their own compilers...

Which would increase the price of their chips, making them closer to intel's priceline.

*thinking out loud*

 

[Duvie]

<<The results for all unit tests of the software were identical between the original and patched versions. As can be seen, patching out the 'GenuineIntel' check makes no difference to the P4, but increases the performance of the Opteron by up to 10%. If the code was compiled with '-axW', then the Opteron really suffers: the 'Simplex' test is more than two times slower (it uses lots of exp() calls, and hence really benefits from vectorisation).


To Intel: there is a standard mechanism out there (invented by you!) for questioning a CPU as to its capabilities. You should be using that, not checking for the presence of your trademark. I don't expect Intel to support AMD-specific extensions, and I also don't expect Intel to have to test its compiler on AMD CPUs. However, if a CPU states that it can do SSE3 or whatever then I expect the code produced by the Intel compiler to use SSE3 instructions rather than to check first if the chip was made by Intel. It was not acceptable for Microsoft to go out and deliberately cripple Windows under DR-DOS, and likewise it isn't acceptable for you to cripple a product that you sell for not inconsequential sums of money so that it won't perform properly on competitors' hardware.>>



seems logical to me....To the ppl up above defending this, rethink it....It is exactly what I was saying...

 

[Duvie]

Originally posted by: JDCentral
We should all just use gcc...



What about the other x86 processors?

Is VIA going to sue for Intel not optimizing it's compilers for the x86-Eden CPUs?

I am a partial AMD fanboy... but I think that AMD will lose this particular argument.

If it really wants full optimization, AMD should write their own compilers...

Which would increase the price of their chips, making them closer to intel's priceline.

*thinking out loud*

What are you fvcking ppl not getting??? It does not need OPTMIZATION!!!!!!!!!! It has the same SSE, SSE2 and now SSE3....Just treat it as the same and the playing field is leveled...Untel needs to quit being a bunch of pussies and realise they are behind, plain and simple....This type of sabotage is sad...

I will be relooking up my link but i believe their is one of the articles of anittrust lwas that applies to this type of thing....I will find it....

 

[Jeff7]

It's the old analogy of comparing cars to CPUs. Hypothetical example - you go to buy Premium gas at a place owned by Ford. But you own a Honda. So you put the nozzleinto the gas tank, and the pump detects that you are using a competitor's car. It quietly switches to Regular gas - in effect, disabling the enhancements of Premium gas. Now you simply assume that it's going to give you the Premium gas you selected regardless of who made the car, just as you'd expect Intel's compiler to use SSE and SSE2 instructions if the CPU has them, regardless of who made it.


Now good luck to AMD's legal team explaining to a jury what the meaning of "hertz" is, much less how a compiler is and what it does. Intel can just show up with the Blue Man Group, say "Intel Inside," play their 5 note theme, and win the case on notoriety alone.