difference between commercial and opensource TBB

difference between commercial and opensource TBB

Hi All, I have good experience in boost, posix and windows threads. I am focusing on Intel TBB. What is the difference between Open Source TBB and Commercial TBB. I want to create a signature for a binary data ( like SHA1). Will TBB improves the performance here on Linux/Win/Mac? Thanks in advance.dpn

6 post / 0 nuovi
Ultimo contenuto
Per informazioni complete sulle ottimizzazioni del compilatore, consultare l'Avviso sull'ottimizzazione

"What is the difference between Open Source TBB and Commercial TBB."
It's a mystery. :-)

"I want to create a signature for a binary data ( like SHA1). Will TBB improves the performance here on Linux/Win/Mac?"
You cannot exploit parallelism if the algorithm expressly requires sequential operation (unlike a block cypher, where TBB is an excellent choice).

(Corrected.) Removed a whole sentence because of a spelling error.

Thank you Raf Schietekat for your valuable information.I am going to drop to use TBB.Reasons:1. No clearity on TBB products.2. It doesn't improve performance for SHA1 calc.Best Regards,dpn.

Best Reply

Well,
of course there are differences:)

the biggest difference between oss and
comm versions areinstallation andcommercial technical support. oss
version is distributed via archives and support chanel is this forum and
bugzilla on oss site.

But binaries and sources are the same
(except copyright comments: oss version includes GPLv2 part).

dpn, there are questions to you:

1. what clearity do you mean? we are about 6 years on the market.

2. you can parallel on file level i.e. make signatures for several files at once. The advantage of this that threading algorithm will be the same on all platforms regardless of OS threading model.

thanks
--Vladimir

Now, I am clear about OpenSource TBB and Commercial TBB.I expected block level parallelism, not file level.Seems to be, it might suitable for my requirements.Let me do feasibility study.Thanks alot Vladimir.Regards,dpn

"You cannot exploit parallelism if the algorithm expressly requires
sequential operation (unlike a block cypher, where TBB is an excellent
choice)."
I just realised that I neglected to point out that tasks may still be a good way to schedule the total workload, using TBB enqueued tasks (rather than spawned ones), and a separate I/O thread. Each task would process a number of blocks, fed to the TBB scheduler from a plain I/O thread. This way TBB is never blocked for I/O, and any number of input files can be processed without oversubscription. Priority could be given to blocks further away from the end of their respective files (for an earlier finish), or closest (for earlier partial results), or some other strategy.

Lascia un commento

Eseguire l'accesso per aggiungere un commento. Non siete membri? Iscriviti oggi