Re: Mathematica 8.0.4 now available
- To: mathgroup at smc.vnet.net
- Subject: [mg122626] Re: Mathematica 8.0.4 now available
- From: Yves Klett <yves.klett at googlemail.com>
- Date: Fri, 4 Nov 2011 05:58:15 -0500 (EST)
- Delivered-to: l-mathgroup@mail-archive0.wolfram.com
- References: <j89uqj$60$1@smc.vnet.net> <j8tkif$evr$1@smc.vnet.net>
Whether or not a certain CUDA implementation is faster than regular code
is totally dependent on the example. For small toy examples the CUDA
overhead is huge and the use inefficient.
Consider CUDASort on my not so slow machine:
Needs["CUDALink`"]
$HistoryLength = 0;
l = Reverse@Range[10^4];
AbsoluteTiming[Do[Sort[l], {100}]]
AbsoluteTiming[Do[CUDASort[l], {100}]]
{0.0230046, Null}
{0.5071014, Null}
l = Reverse@Range[10^7];
AbsoluteTiming[a = Sort[l];]
AbsoluteTiming[b = CUDASort[l];]
a == b
{0.4420884, Null}
{0.2640528, Null}
True
For huge lists CUDASort can be quite a bit faster out of the box. Makes
a lot of noise, though, but keeps you warm in wintertime.
Regards,
Yves
Am 03.11.2011 09:49, schrieb DmitryG:
> On Oct 26, 2:41 pm, Wolfram Research <i... at wolfram.com> wrote:
>> We are pleased to announce the release of Mathematica8.0.4. This
>> new version streamlines deployment to the Computable Document
>> Format (CDF) with a new interactive wizard and new functions for
>> programmatic deployment. Key usability, stability, and security
>> enhancements include:
>>
>> * New menu options and an interactive wizard for deploying to CDF
>> * New functions CDFInformation and CDFDeploy for programmatically
>> deploying CDF files
>> * New download progress indicator for the CDF browser plugin
>> * Improved stability, speed, and security of the CDF browser
>> plugin
>> * Tested under Mac OS X Lion
>> * Improved memory use of the Share function when storing large
>> sets of data
>> * Updated MKL library to improve performance
>> * New syntax coloring to warn about use of Module variable in
>> Dynamic
>> * Performance improvements to citation functions
>> * Improved behavior of optimized variables in compiled
>> conditional statements
>> * Improved front end behavior when saving untitled notebooks
>> under new versions of Linux
>>
>> Users who have Mathematica 8.0.0 or 8.0.1 with active Premier
>> Service will receive email from us in the next few days with
>> instructions on how to download their free updates.
>>
>> Sincerely,
>>
>> Wolfram Research
>> Customer Supporthttp://www.wolfram.com/support
>
> I have tested Mathematica 8.0.4 for CUDA because now I have a new
> laptop with a new nVidia Geforce 555M card. As before, CUDADot is
> slower than regular Dot on reals, so I do not know how I could benefit
> from GPU calculations
>
> Another disappointment is that 8.0.4 does not recognize any C compiles
> except of the Intel C compiler and a Generic C compiler that are
> listed by CCompilers[Full]. There are no other compilers in the list.
> I have upgraded to Mathematica 8.0.4 under Windows7 with the new
> laptop and now I cannot use Miscosoft Visual C++ compiler. Also I have
> upgraded on my Mac Pro and now I cannot use the GCC compiler that I
> was using with Mathematica 8.0.1.
>
> I do not know whether a better CDF support and maybe a better MKL
> library compensate for the loss of C compilers and thus whether an
> upgrade to 8.0.4 makes sense.
>
> As for GPU computations, it seems we have to wait for Mathematica 9.
>
> Dmitry
>