AMD Alleges Intel Compilers Create Slower AMD Code

edxwelch writes "In AMD's recient anti-trust lawsuit AMD have examined the Intel compiler and found that it deliberatly runs code slower when it detects that the processor is an AMD. "To achieve this, Intel designed the compiler to compile code along several alternate code paths. ... By design, the code paths were not created equally. If the program detects a "Genuine Intel" microprocessor, it executes a fully optimized code path and operates with the maximum efficiency. However, if the program detects an "Authentic AMD" microprocessor, it executes a different code path that will degrade the program's performance or cause it to crash.""

Comments

It's true--and they know about it

[Eponymous Cowboy]

I noticed this problem back in January of 2004, with Intel C++ 8.0, and went through heck over nine months with Intel's customer support to get it fixed until I eventually had to abandon their compiler.

On any non-Intel processors, it specifically included an alternate code path for "memcpy" that actually used "rep movsb" to copy one byte at a time, instead of (for example) "rep movsd" to copy a doubleword at a time (or MMX instructions to copy quadwords). This was probably the most brain-dead memcpy I'd ever seen, and was around 4X slower than even a typical naive assembly memcpy:

push ecx
shr ecx, 2
rep movsd
pop ecx
and ecx, 3
rep movsb

They responded with completely ridiculous answers, such as:

"Our 8.0 memcpy was indeed optimized for a Pentium(r)4 Processor,when we reworked this routine we used the simplest, most robust, and straightforward implementation for older processors so that we didn't need the extra code to check for alignment, length, overlap, and other conditions."

BS. I went and added the following line to the beginning of my source code:

extern "C" int __intel_cpu_indicator;

then I added:

__intel_cpu_indicator = -512;

to the "main" function.

This forced Intel C++ to use the "Pentium 4" memcpy regardless of which processor in in the machine. It turns out that their special "Pentium 4" memcpy which I tested thoroughly in all kinds of situations, and it worked perfectly fine on an AMD Athlon and a Pentium III. I pointed this out to them.

I received the following response:

"The fast mempcy is over 2000 lines of hand coded assembly, with lots of special cases where different code paths are chosen based on relative alignment of the source and destination. ... If the performance of memcpy/memset only are improved for Pentium III will that satisfy you?"

I answered "No," saying that I needed support for AMD processors as well. I also gave them a copy of my own memcpy routine that was 50% faster than theirs--and just used MMX. They closed the support issue and did nothing to resolve it.

I switched back to Visual C++.

Re:Simply ludicrous

[RootsLINUX]

Well, I don't think they would make such a claim without at least some sort of evidence. And we all know that Intel is a corporation that doesn't like to play nice with the other kids. During my senior year of undergrad my computer architecture professor consistently refered to Intel as "the evil empire", so who knows. I'd like to see AMD present some hard evidence though.

Re:How can this be done?

[the_weasel]

Which is exactly what we did when we discovered this. And sure enough, when the Intel compiler thought our AMD was a Genuine Intel certain functions increased dramatically.

The company I work for submitted a few reports on this a few months ago, as I am sure did many others. I am very pleased to see them following up on it.

as someone who worked at a compiler company

[MatD]

I know how this can happen (and it has nothing to do with being evil).
The engineers get the specs for the next version of the compiler. They also get a slew of bug reports from the last version. They have a short amount of time to impliment the new specs, and fix the bugs.
The bug reports will be something like, "on AMD processors when doing a memcopy with optimization xyz turned on, the processors mispredicts half the time. This makes it very slow."
The engineer in that case, turns the optimization off for that generated code, thereby 'fixing' the bug (but not really). It happens all the time.
It's not a nefarious plot, it's the same time crunch issue that every software engineer has to deal with.

Re:Simply ludicrous

[yomahz]

Yes but, they knew enough to block W3C's validator from their site:

http://validator.w3.org/check?uri=http%3A%2F%2Fwww .slashdot.org%2F