Intel® Manycore Testing Lab (Archived)

Upgraded systems on the MTL - now with 40-cores

So you thought that your application or lab scaled effectively with 32-cores. Now the MTL has available 40-core systems for your use - why not see ifyour codescales to 40-cores.

Thesenew systems consist of10-core CPUs (Intel Xeon E7-4860)on a 4 socketserver thus giving 40-cores on all the Linux systems.

Note: all batch nodes are currently configured with Hyperthreading off.

Preliminary experiments with SMP parallelism in GHC 7.0.2 (Haskell)

As per the agreement with Intel, I am reporting my experiences with the Intel Manycore Testing Lab (Linux). This was my first time in the lab, and I wanted to test GHC's [1] SMP parallelism [2] features.

The first challenge was to actually get GHC to work on the lab. There was a working version of ghc under /opt/ghc6.13/bin/ghc, but I really needed GHC 7. So first I built GHC 7.0.2-rc2, which worked without much trouble.

Anomalous performance on batch nodes


I've been running experiments on MTL over the last week, and I'm seeing some strange results. I ran similar experiments on MTL last summer and was able to obtain consistent performance measurements for operations on four concurrent data structures. This time, taking performance measurements on six concurrent data structures, I'm not having much luck obtaining good results (i.e., small standard deviation, few anomalies). I'm starting to wonder if something is happening systemically.

Seg fault with -O0 or -O1 flag but not with -O2 or -O3

I have a strange issue with an embarrasingly parallel code.When I run it with -O0 or -O1 optimization flags my code throws segfault. However, with -O2 or -O3 flag it runs just fine. The debugger (gdb) says the error is at a OpenMP parallel for directive. I would appreciate if someone has an idea as to what could be the possible reason. I am also debugging in the meantime and will post if I find something. Thank you.


This question may have been asked before, but here goes. The Intel mult-core CPUs offer an option which I believe is called hyperhtreading. Hyperthreadingdoubles the number of physical cores on an Intel processor; thus a quad core becomes an eight core procssor, etc.

How does Cilk++ square with this?


Preliminary results on a parallelizing runtime for an implicitly parallel language


Hello everyone, my name is Dan Amelang and I'm a PhD student at the University of California, San Diego. As part of my dissertation research I'm developing a declarative, implicitly parallel programming language. The language is similar to previous work in dataflow, stream processing and functional languages. Closely related work includes SISAL, Kahn Process Networks, Dataflow Process Networks, and StreamIt, among others. It also borrows from Haskell somewhat.

Haskell: forkOS vs forkIO

First of all I want to boast I've successfully compiled and used cisco vpnclient version for Linux for the first time since few years ;)I know I'm not haskell guru or something but I wrote simple Monte-Carlo pi computing multi-threaded implementation. I wrote one version using forkIO function and the second one using forkOS. I did simple benchmark too:

Acknowledging the Intel Manycore Testing Lab in your publications

Dear Intel Manycore Testing Lab users,

When you are reporting your parallelism and concurrency results in white papers, presentations, documents, books, blogs, forums, capstones, or any other published method of sharing your findings, we prefer that you include two specific URLs with your acknowledgment.

In some cases, including the URLs in the references sections of your monograph might be more appropriate than in the thank you section itself. We encourage you to use this acknowledgement:

The Intel Manycore Testing Lab is open for business but new accounts are being postponed until JAN 3 2011

Dear Intel Manycore Testing Lab Users,

Due to the demands of our current year-end schedule, we will not be able to create any new accounts on the Lab systems starting from noon PDT, Wednesday 12/15, 2010, through Sunday, January 2nd, 2011. This means all requests for new professor or student accounts (or both) received during this time will be processed on Monday, January 3rd, 2011, when account creation will resume as normal.

We apologize in advance for any inconvenience this temporary suspension of new account creation may cause.

Suscribirse a Intel® Manycore Testing Lab (Archived)