Intel's "cripple AMD" function:Will Intel be forced to remove the "cripple AMD" funct

[poke349]

The Intel compiler will instead do stuff like use x87 instead of SSE when it's an AMD chip even if SSE is supported.

Interesting...

I'm using version 11.0.061 and I when I scroll through the assembly it generates, I don't see this.

When I tell it to use SSE, it uses SSE, and it doesn't generate an x87 version so that it runs slower on AMD.
I've done a full search for x87 instructions in the assembly and I find very little x87.


EDIT:
On second thought, I tell it use SSE using intrinsics. So it kinda has to.
But as for auto-vectorization, that may be fuzzy area which is what I think you're talking about.

[alfaunits]

Sorry sir, but not-use-best IS intentionally crippling performance.

No it's not. Crippling performance on AMD CPUs is providing lower performance than any other compiler. Not providing the best performance is simply not giving a f*** about your competition. And I'd be seriously worried if a company did give a f*** about helping its competition


As Sierra noted, Intel knows what works on their CPUs - it's not just the instruction set, it is also memory alighment of data, prefetching optimization (done differently for different Intel CPUs even), etc. and also the fact that AMD CPUs (just like Intel of course) have errors that can cause the different optimizations to not only perform slower but not work at all. It's not Intel's work to optimize for AMD CPUs. The only thing they should be expected to do is not provide the worst performance of all compilers.

If you don't see a difference, please tell me what you do so I can make a competitor to your business and ask you to optimize my business because you think it should be this way

[keithlm]

Intels compiler must be extremely rigorously tested, it's not simply "Do you have SSE2? Great, we'll use it."

Actually it is that simple; that is EXACTLY how it is supposed to work.

Period. End of discussion.

You are attempting to remove the responsibility from the manufacturer and put it onto the consumer or end user. (Which in the case of this discussion would be the compiler developer.)

This would be similar to claiming that the CPU manufacturer needs to get and test every motherboard that anybody might possibly make anywhere in the world. To claim that would just as wrong as what you are attempting to claim. There is no doubt that they will probably test SOME. Maybe even a lot. But to claim they need to test all would be naive.

It's not Intel's work to optimize for AMD CPUs.

It is also not their job to CERTIFY what the AMD CPU can and can not do.

It is however Intel's job to use what the CPU has been certified by AMD to do.

People seem to be attempting to make this into a much more complicated issue than merely using a certain optimization or not.

[safan80]

Intel made the compiler for it's own cpus it should not be forced to support AMD or VIA products. AMD and VIA can make their own compilers. It's one thing for AMD and VIA to have an x86 license it's quite another for intel to have to write software for companies who can't do it for themselves.

[highoctane]

It is also not their job to CERTIFY what the AMD CPU can and can not do.

It is however Intel's job to use what the CPU has been certified by AMD to do.

Why is it Intels responsibility to optimize their compiler for AMD cpu's in any way shape or form. I recon developers can choose how they want to compile or optimize their code freely unless they are saying a software developer doesn't know anything about compiling their code.

Maybe AMD's compiler's are not optimized for Intel cpu's, should we be lodging an investigation into AMD's compilers.

Seriously in what businesses do you expect full support from the competition when using a competitors own product.

Maybe Apple can get its engineers working on optimizing their iphone apps for Palm or Google Droid.

[Eastcoasthandle]

If SSE 2 (for example) is basically the same regardless of what CPU it's found in why are you of the opinion that Intel has the right to flag it for their CPU only? SSE 2 (in this example) is shared (albeit how they function in each CPU may not be exactly the same) so there should be some co-operation between companies in this endeavor. This makes no sense to me.

[quintus]

It is also not their job to CERTIFY what the AMD CPU can and can not do.

+1

[highoctane]

If SSE 2 (for example) is basically the same regardless of what CPU it's found in why are you of the opinion that Intel has the right to flag it for their CPU only?

Well for one its regarding an Intel compiler from what I gather which is developed by Intel for use and optimized for their processors.

Should Ford worry about how their diagnostic software interfaces with a Kia even though OBDII is a standard.

[Eastcoasthandle]

Well for one its regarding an Intel compiler from what I gather which is developed by Intel for use and optimized for their processors.

Should Ford worry about how their diagnostic software interfaces with a Kia even though OBDII is a standard.

But isn't the entire line of SSE from Intel to begin with? Where others license it's use? If so, in order for that to apply from my point of view is that Intel should be on the only ones capable of SSE to begin with. AMD would have to come up with some other derivative. However, the problem here is that's not how the market developed. Both AMD and Intel have agreed to share it's use. That should include any other enhancement as outlined in the OP from the very beginning.

[safan80]

If SSE 2 (for example) is basically the same regardless of what CPU it's found in why are you of the opinion that Intel has the right to flag it for their CPU only? SSE 2 (in this example) is shared (albeit how they function in each CPU may not be exactly the same) so there should be some co-operation between companies in this endeavor. This makes no sense to me.

I take your never made something yourself and had others copy it, and use it for things your did not intend it to do then had people complain it did not work "right".

Intel made the compiler for intel cpus only there's no reason for them to be supporting AMD cpus.

AMD should start hiring their own devs for their cpu division. AMD can make their compilers they've just been too cheap in the past to do so.

You don't see AMD telling Nvidia that they should be allowed to use Cuda or PhsyX do you?

edit: share use applies to the hardware side of things not the software side of things.

[highoctane]

But isn't the entire line of SSE from Intel to begin with?

I think AMD added something of their own with sse 4a or something like that so I don't think the entire line of sse if from Intel.

While they are shared I really don't see how its Intel's responsibility to implement these functions via their own compiler.

[Eastcoasthandle]

I take your never made something yourself and had others copy it, and use it for things your did not intend it to do then had people complain it did not work "right".

Intel made the compiler for intel cpus only there's no reason for them to be supporting AMD cpus.

AMD should start hiring their own devs for their cpu division. AMD can make their compilers they've just been too cheap in the past to do so.

You don't see AMD telling Nvidia that they should be allowed to use Cuda or PhsyX do you?

edit: share use applies to the hardware side of things not the software side of things.

But in this case it applies to both as one doesn't work without the other. And there is no such thing as someone copying something when the technology (both hardware and the software aspect) is being shared. There is no exclusivity here. Which for whatever reason you believe exists. The information in the OP should have been something enforced a long, long time ago. This is hardly a black and white situation but a gray area where Intel should have made available all enhancements made do to the lack of exclusivity here.

[safan80]

But in this case it applies to both as one doesn't work without the other. And there is no such thing as someone copying something when the technology (both hardware and the software aspect) is being shared. There is no exclusivity here. Which for whatever reason you believe exists.

The sharing only applies to the X86 license which is for hardware only!

Anybody could write a compiler.

Intel hires it's own devs AMD could do the same.

Just because AMD is too cheap to hire it's own devs does mean people should go to intel crying about it. if AMD can not properly support it's own product you can't blame that on intel.

[Eastcoasthandle]

So AMD doesn't use SSE now? This is hopeless.

[highoctane]

This is hardly a black and white situation but a gray area where Intel should have made available all enhancements made do to the lack of exclusivity here.

I feel the burden is on the shoulders of the software developer to ensure his product is optimized and his customers are satisfied with his product not Intel or AMD. I don't see exclusivity here because the developer can freely choose to optimize his product to suit his clients needs or to the developers own satisfaction.

[gosh]

I feel the burden is on the shoulders of the software developer to ensure his product is optimized and his customers are satisfied with his product not Intel or AMD. I don't see exclusivity here because the developer can freely choose to optimize his product to suit his clients needs or to the developers own satisfaction.

Intel has 80+ % market share
It is not a healthy market

[Final8ty]

So AMD doesn't use SSE now? This is hopeless.

It looks to me like another BatmanAA scenario. Disable a function that the competitor is quite capable of running when the competing hardware is detected.

[Serpentarius]

The sharing only applies to the X86 license which is for hardware only!

Anybody could write a compiler.

Intel hires it's own devs AMD could do the same.

Just because AMD is too cheap to hire it's own devs does mean people should go to intel crying about it. if AMD can not properly support it's own product you can't blame that on intel.

Bill Gates said something similar on his Internet Explorer
but this is beyond one-sided

this could've impacted all softwares .... including synthetic benchmarks

what will the world be ..... if somehow this stint is removed, and benchmarks revised their results ... and AMD is not actually an underdog processor as they are initially thought are

[safan80]

Bill Gates said something similar on his Internet Explorer
but this is beyond one-sided

this could've impacted all softwares .... including synthetic benchmarks

How so? AMD makes it's own compilers

http://developer.amd.com/Pages/default.aspx

[STEvil]

Simple solution: Find a program that easily lets us change out cpuid string and re-bench everything to see who took the time to set things right and who went with the flow!

Whats the easiest program to use for this?

edit - technically its the vendor string I guess that needs to be change/intercepted/whatever..

[Chumbucket843]

hmmm... this has already been done. nano is an example where the cpu supports every intel sse instruction but the code path is chosen based on cpuid, not the actual capabilities.
http://arstechnica.com/hardware/revi...o-review.ars/6

[demonkevy666]

Simple solution: Find a program that easily lets us change out cpuid string and re-bench everything to see who took the time to set things right and who went with the flow!

Whats the easiest program to use for this?

edit - technically its the vendor string I guess that needs to be change/intercepted/whatever..

it IS possible to change cpu string to show intel in cpu-z, though no benches where ran after this it show it is possible to do.

It looks to me like another BatmanAA scenario. Disable a function that the competitor is quite capable of running when the competing hardware is detected.

it's would seem that way but then it could go the same way physic went on ageia to nvidia (but it would be a compiler)

[hollo]

if you re-read the OP it's pretty clear the FTC didn't complain that Intel's compiler wasn't supporting AMD CPUs properly, it's complaining that Intel misrepresented whether their compiler was supporting AMD CPUs, which is a different kettle of fish.

[...]
To the public, OEMs, ISVs, and benchmarking organizations, the slower performance of non-Intel CPUs on Intel-compiled software applications appeared to be caused by the non-Intel CPUs rather than the Intel software. Intel failed to disclose the effects of the changes it made to its software in or about 2003 and later to its customers or the public. Intel also disseminated false or misleading documentation about its compiler and libraries. Intel represented to ISVs, OEMs, benchmarking organizations, and the public that programs inherently performed better on Intel CPUs than on competing CPUs. In truth and in fact, many differences were due largely or entirely to the Intel software. Intel’s misleading or false statements and omissions about the performance of its software were material to ISVs, OEMs, benchmarking organizations, and the public in their purchase or use of CPUs. Therefore, Intel’s representations that programs inherently performed better on Intel CPUs than on competing CPUs were, and are, false or misleading. Intel’s failure to disclose that the differences were due largely to the Intel software, in light of the representations made, was, and is, a deceptive practice. Moreover, those misrepresentations and omissions were likely to harm the reputation of other x86 CPUs companies, and harmed competition.
[...]

there is also this part, which i disagree with, because as people have mentioned software requires testing and it's natural that Intel wouldn't test AMD hardware..

[...]
Intel’s software design changes slowed the performance of non-Intel x86 CPUs and had no sufficiently justifiable technological benefit.
[...]

..and although it's sneaky but not illegal of Intel to fail to disclose the issue, it's unethical of them to spread false information when asked about it directly:

[...]
Some ISVs requested information from Intel concerning the apparent variation in performance of identical software run on Intel and non-Intel CPUs. In response to such requests, on numerous occasions, Intel misrepresented, expressly or by implication, the source of the problem and whether it could be solved.

this part..

Requiring that, with respect to those Intel customers that purchased from Intel a software compiler that had or has the design or effect of impairing the actual or apparent performance of microprocessors not manufactured by Intel ("Defective Compiler"), as described in the Complaint:

1. Intel provide them, at no additional charge, a substitute compiler that is not a Defective Compiler;
2. Intel compensate them for the cost of recompiling the software they had compiled on the Defective Compiler and of substituting, and distributing to their own customers, the recompiled software for software compiled on a Defective Compiler; and
3. Intel give public notice and warning, in a manner likely to be communicated to persons that have purchased software compiled on Defective Compilers purchased from Intel, of the possible need to replace that software.

..is the slap on the wrist that would have never happened if Intel had been honest from the start. it would be xtreeeeemely ironic if Intel end up needing to test a lot of AMD hardware to make sure it is compatible with their compiler, or help AMD develop their own compiler, to comply with 1.

[STEvil]

it IS possible to change cpu string to show intel in cpu-z, though no benches where ran after this it show it is possible to do.

Yeah but that doesnt help anything, need to change the vendorid or intercept the call..

[Serpentarius]

How so? AMD makes it's own compilers

http://developer.amd.com/Pages/default.aspx

it's not about who makes better compilers or who invest heavily on compiler dev
it's about Intel's underhand methods ... if they're had superior products, they've more than just winning but to actually sneak at one's back and shot at their leg ... is not something to be proud of

hmmm... this has already been done. nano is an example where the cpu supports every intel sse instruction but the code path is chosen based on cpuid, not the actual capabilities.
http://arstechnica.com/hardware/revi...o-review.ars/6

i doubt that ... "Many software programmers consider Intel's compiler the best optimizing compiler on the market, and it is often the preferred compiler for the most critical applications" ...

and with such a large mainstream using Intel, it is obvious that any dumbass would develop software would run best on the mainstream don't you?
heck, even an illiterate beggar would choose the well-dressed person to beg from

critical applications - how critical is relative, but most project developer considers every project is crucial to them (in fact, we all should be serious in developments, do we?) ...

but for some reason, i doubt the "fairness" PCMark mentioned, they can claim anything as long as they're close source