topic Rate dsytrd Intel MKL in Intel® oneAPI Math Kernel Library

Rate dsytrd Intel MKL

yuriisig — Thu, 12 Aug 2010 20:49:58 GMT

I had a message, where the comparison of my dsptrd and thedsytrd Intl MKL took place. But it was deleted. Why? This information is very important.

Rate dsytrd Intel MKL

Aubrey_W_ — Thu, 12 Aug 2010 21:42:36 GMT

Hello,

I deleted your thread because it was too similar to your other thread at http://software.intel.com/en-us/forums/showthread.php?t=76595. We welcome you to share your results, but please understand that too many similar threads hurts the value of the forums. Please feel free to add more details to your existing thread.

Best regards,

==
Aubrey W.
Intel Software Network Support

Rate dsytrd Intel MKL

yuriisig — Thu, 12 Aug 2010 23:58:19 GMT

You have removed important information where it was compared my dsptrd with dsytrd Intel MKL. This comparison shows that dsytrd Intel MKL requires of fundamental changes.Unfortunately you're not very good in this area.

Rate dsytrd Intel MKL

mecej4 — Fri, 13 Aug 2010 03:12:23 GMT

I saw that post before it was removed, and came away musing as to what useful purpose it served, since it simply showed execution times for an unseen version of a routine performing much faster than an MKL routine of the same name.

I am just an MKL user, and I speak only for myself.

If the faster version were genuine, you could market it to interested parties. Once you have made a claim for your version of one routine being faster, posts showing timing details for other routines are just as believable as the first claim, and more timing data will draw little interest until they are verified and the faster routines are made generally available.

Using Intel's forums to make derogatory comments without supporting evidence could be seen as an abuse of hospitality. I am surprised that the post in question was not truncated or removed by the moderators even earlier.

I feel that most users of MKL would have similar opinions, and would agree that this forum is not the place for aggressive marketing, and that the tactics that have been used are unlikely to advance your cause.

Rate dsytrd Intel MKL

yuriisig — Fri, 13 Aug 2010 04:08:02 GMT

I read the fable "The fox and grapes" by Krylov when a pupil. For example, I wrote that the function dgeem Intel MKL for 65 nm (IA32) and P4 processors was slow. But you did not believe me. Professor Granovsky proved it. And you took advantage of this proof. My program of diagonalisation is used in quantum chemical calculations (I have published articles and Intel have read them). If Intel wants tobe sure I'm right, let them look up my computer. But as I understand they do not want it. Sorry for my bad English. When I went to work in the U.S., I had been studying it for 3 months.

Rate dsytrd Intel MKL

yuriisig — Fri, 13 Aug 2010 06:19:12 GMT

Quoting mecej4

I feel that most users of MKL would have similar opinions

Youmake amistake. The truth is important for users. And you do not need it.I offered to look at my computer. But you do not want to do it. Why?

Rate dsytrd Intel MKL

mecej4 — Fri, 13 Aug 2010 15:33:48 GMT

Please do not confuse me ("mecej4") with anyone at Intel or elsewhere. Until this month (August 2010) I had not heard about you ("yuriisig") or about your previous interactions with Intel.

Rate dsytrd Intel MKL

yuriisig — Sat, 14 Aug 2010 17:18:01 GMT

My web page (it is not currently available) and publications, which used my diagonalization, can be downloaded here: http://depositfiles.com/files/fmy2ueaad