Intel AVX and GIMPS
 

As some of you probably know, many future chips will support the AVX instruction set, starting with Intel's "Sandy Bridge" in January 2011.

According to Wikipedia, AVX will allow higher more GFLOPS, and future versions will supposedly handle 1024-bit registers. Does anyone know how this will affect GIMPS?

I also have a somewhat unrelated question: there are two other instruction sets on the horizon, FMA3 and FMA4. Does anyone know if they will be of any benefit?

I have high hopes for AVX, but it will be of no value until I rewrite all the assembly code to use the AVX istructions.

FMA3 and FMA4 are not instruction sets, they are individual instructions -- fused multiply add. They could be quite useful depending on how Intel and AMD implement them

Hmm, Wikipedia is [url=http://en.wikipedia.org/wiki/FMA_instruction_set]listing them as instruction sets[/url]. Perhaps the article needs to be updated.

[QUOTE=ixfd64;241510]Hmm, Wikipedia is [URL="http://en.wikipedia.org/wiki/FMA_instruction_set"]listing them as instruction sets[/URL]. Perhaps the article needs to be updated.[/QUOTE]Perhaps it's just a question of whether one considers a group of instructions that all perform the same basic operation, differing only in operand length and rotation of operand order, to be an "instruction set".

Maybe "FMA instruction set" clues one in, like "ADD instruction set" would.

[QUOTE=Prime95;241271]I have high hopes for AVX, but it will be of no value until I rewrite all the assembly code to use the AVX istructions.

FMA3 and FMA4 are not instruction sets, they are individual instructions -- fused multiply add. They could be quite useful depending on how Intel and AMD implement them[/QUOTE]

Sounds great but I hope you won't drop the optimizations for older cpus.

Wow! Nice speed for non-optimized for AVX and mobile CPU :w00t: :
[url]http://www.mersenne.org/report_benchmarks/?exover=1&exbad=1&specific_cpu=4369746[/url]

[QUOTE=Lorenzo;244678]Wow! Nice speed for non-optimized for AVX and mobile CPU :w00t: :
[url]http://www.mersenne.org/report_benchmarks/?exover=1&exbad=1&specific_cpu=4369746[/url][/QUOTE]

They're available already!!!!!!!!

Yes, among the fastest of all CPUs benchmarked to date.

And another one here:
[url]http://www.mersenne.org/report_benchmarks/?speed_lo=2000&speed_hi=9900&exbad=1&specific_cpu=4369510[/url]

[QUOTE=Lorenzo;244678]Wow! Nice speed for non-optimized for AVX and mobile CPU :w00t: :
[url]http://www.mersenne.org/report_benchmarks/?exover=1&exbad=1&specific_cpu=4369746[/url][/QUOTE]

And that's not even getting to the i7-2920XM!

Very nice speed and medium price (205$) for Intel Core i5-2500 @ 3.30GHz :tu: It's good choose for Prime95.

Waiting for support AVX instruction to Prime95!

For Prime95 I recommend the i5-2500[B]K[/B]. The unlocked multiplier, specially if coupled with a good mobo, fast memory and a decent cooler, will allow us to overclock the machine so as to make Prime95 fly. Throw in the AVX otpimizations, if and when GW provides them, and we´ll be in for some real awesome benchies (and of course a hefty amount of real work...)

Has Mr. Woltman actually stated that he believes this instruction set will be helpful? I don't know anything about the math or this particular set of instructions(or any set for that matter), I just think it's a good question to ask.

I can just picture a bunch of rumors flying for weeks, and the Mr. Woltman comes on and bursts everybody's bubble by saying the instruction set won't help and he isn't going to use it.

[QUOTE=jasong;245501]Has Mr. Woltman actually stated that he believes this instruction set will be helpful? I don't know anything about the math or this particular set of instructions(or any set for that matter), I just think it's a good question to ask.

I can just picture a bunch of rumors flying for weeks, and the Mr. Woltman comes on and bursts everybody's bubble by saying the instruction set won't help and he isn't going to use it.[/QUOTE]
Just look at the second post. He thinks they will probably useful.

[QUOTE=Prime95;241271]I have high hopes for AVX, but it will be of no value until I rewrite all the assembly code to use the AVX istructions.
[/QUOTE]
Anything new with regard to AVX?

[QUOTE=Brain;257789]Anything new with regard to AVX?[/QUOTE]

Patience :)

I'm still working on another project before beginning the AVX rewrite.

[QUOTE=Prime95;257798]Patience :)

I'm still working on another project before beginning the AVX rewrite.[/QUOTE]

GPU! GPU! GPU! :devil:

Rumors
 

[QUOTE=Prime95;257798]Patience :)

I'm still working on another project before beginning the AVX rewrite.[/QUOTE]

Just curious. Thanks for all your work.

After being correct with my LL-result-positive-test-email-test on the 200... exponent I would bet it is GPU P-1 coding. If the project wouldn't deal with GIMPS he wouldn't have posted it here. ;-) And 26.6 wouldn't be called "another project". Or it's simply the spring work in his garden. We'll see. Looking forward to it.