Comments for Memkite http://memkite.com Deep Learning for iOS Mon, 28 Dec 2015 19:16:17 +0000 hourly 1 http://wordpress.org/?v=4.1.19 Comment on Swift and Metal GPU Programming on tvOS for the new Apple TV by DeepLearningKit – Deep Learning for iOS (tested on iPhone 6S), tvOS and OS X developed in Metal and Swift | Open Source Deep Learning Framework for iOS, OS X and tvOS http://memkite.com/blog/2015/09/09/swift-and-metal-gpu-programming-on-tvos-for-the-new-apple-tv/#comment-16377 Mon, 28 Dec 2015 19:16:17 +0000 http://memkite.com/?p=933#comment-16377 […] interesting feature on iOS (and most likely on tvOS, but not yet tested in our case) is that one can share memory between GPU and CPU (less copying of […]

]]>
Comment on DeepLearning.University – An Annotated Deep Learning Bibliography by Neural network paper list | Eniod's Blog http://memkite.com/deep-learning-bibliography/#comment-16334 Wed, 23 Dec 2015 15:44:07 +0000 http://memkite.com/?page_id=572#comment-16334 […] DeepLearning.University – An Annotated Deep Learning Bibliography | Memkite(github.com/memkite/DeepLearningBibliography) […]

]]>
Comment on Memkite’s DeepLearningKit by Amund Tveit (@atveit - amund@memkite.com) http://memkite.com/#comment-16301 Mon, 21 Dec 2015 12:07:02 +0000 http://188.226.216.64/?page_id=195#comment-16301 Hi Sun, Memkite is planned to be opensourced (very soon) as an iOS framework called DeepLearningKit at http://deeplearningkit.org

B.R.,
Amund

]]>
Comment on Memkite’s DeepLearningKit by Amund Tveit (@atveit - amund@memkite.com) http://memkite.com/#comment-16300 Mon, 21 Dec 2015 12:06:47 +0000 http://188.226.216.64/?page_id=195#comment-16300 Hi Christopher, Memkite is planned to be opensourced (very soon) as an iOS framework called DeepLearningKit at http://deeplearningkit.org

B.R.,
Amund

]]>
Comment on DeepLearning.University – An Annotated Deep Learning Bibliography by Machine Learning Resources | alokanand09 http://memkite.com/deep-learning-bibliography/#comment-16101 Tue, 01 Dec 2015 17:32:57 +0000 http://memkite.com/?page_id=572#comment-16101 […] Topic-wise Deep Learning Bibliography by memkite (new) […]

]]>
Comment on Memkite’s DeepLearningKit by Christopher Wittlinger http://memkite.com/#comment-16094 Mon, 30 Nov 2015 10:24:14 +0000 http://188.226.216.64/?page_id=195#comment-16094 When will Memkite be available for download?

]]>
Comment on Memkite’s DeepLearningKit by Sun Donghui http://memkite.com/#comment-16074 Sat, 28 Nov 2015 03:35:14 +0000 http://188.226.216.64/?page_id=195#comment-16074 The memkits is library? Is it ready for use in app development?

]]>
Comment on DeepLearning.University – An Annotated Deep Learning Bibliography by robertsdionne/neural-network-papers | GITROOM http://memkite.com/deep-learning-bibliography/#comment-15997 Sat, 21 Nov 2015 05:47:07 +0000 http://memkite.com/?page_id=572#comment-15997 […] DeepLearning.University – An Annotated Deep Learning Bibliography | Memkite (github.com/memkite/DeepLearningBibliography) […]

]]>
Comment on Example of Sharing Memory between GPU and CPU with Swift and Metal for iOS8 by Memkite – Deep Learning for iOS (tested on iPhone 6S), tvOS and OS X developed in Metal and Swift | Memkite http://memkite.com/blog/2014/12/30/example-of-sharing-memory-between-gpu-and-cpu-with-swift-and-metal-for-ios8/#comment-15797 Tue, 10 Nov 2015 08:17:11 +0000 http://memkite.com/?p=690#comment-15797 […] interesting feature on iOS (and most likely on tvOS, but not yet tested in our case) is that one can share memory between GPU and CPU (less copying of […]

]]>
Comment on Swift and Metal GPU Programming on tvOS for the new Apple TV by Memkite – Deep Learning for iOS (tested on iPhone 6S), tvOS and OS X developed in Metal and Swift | Memkite http://memkite.com/blog/2015/09/09/swift-and-metal-gpu-programming-on-tvos-for-the-new-apple-tv/#comment-15395 Wed, 14 Oct 2015 11:19:49 +0000 http://memkite.com/?p=933#comment-15395 […] interesting feature on iOS (and most likely on tvOS, but not yet tested in our case) is that one can share memory between GPU and CPU (less copying of […]

]]>
Comment on Contact by Swift and Metal GPU Programming on tvOS for the new Apple TV | Memkite http://memkite.com/contact/#comment-14922 Wed, 09 Sep 2015 21:54:44 +0000 http://memkite.com/?page_id=910#comment-14922 […] Contact  […]

]]>
Comment on Example of Sharing Memory between GPU and CPU with Swift and Metal for iOS8 by Swift and Metal GPU Programming on OSX 10.11 / El Capitan | Memkite http://memkite.com/blog/2014/12/30/example-of-sharing-memory-between-gpu-and-cpu-with-swift-and-metal-for-ios8/#comment-14917 Wed, 09 Sep 2015 10:31:28 +0000 http://memkite.com/?p=690#comment-14917 […] Example of Sharing Memory between GPU and CPU with Swift and Metal for iOS8 […]

]]>
Comment on Swift and Metal GPU Programming on OSX 10.11 / El Capitan by Amund Tveit (@atveit - amund@memkite.com) http://memkite.com/blog/2015/06/10/swift-and-metal-gpu-programming-on-osx-10-11-el-capitan/#comment-14893 Mon, 31 Aug 2015 08:11:57 +0000 http://memkite.com/?p=875#comment-14893 Hi, the API syntax is evolving from version to version in Xcode, I’ll probably post an update to this during the fall.

Best,
Amund

]]>
Comment on Swift and Metal GPU Programming on OSX 10.11 / El Capitan by Jim Witte http://memkite.com/blog/2015/06/10/swift-and-metal-gpu-programming-on-osx-10-11-el-capitan/#comment-14846 Sat, 22 Aug 2015 06:05:44 +0000 http://memkite.com/?p=875#comment-14846 When I try to compile on xCode (7 beta 5; 7A176x) for El Capitan beta (15A262e), I get an error on the line:

computePipelineState = try metalDevice.newComputePipelineStateWithDescriptor(computePipeLineDescriptor

of:

cannot invoke ‘newComputePipelineStateWithDescriptor’ with an argument list of type ‘(MTLComputePipelineDescriptor)’

Does this mean it wants more arguments? (and what would they be) Or has the Metal API and/or Swift syntax changed since this was written?

Thanks,
Jim

]]>
Comment on Swift and Metal GPU Programming on OSX 10.11 / El Capitan by Arjun Jain http://memkite.com/blog/2015/06/10/swift-and-metal-gpu-programming-on-osx-10-11-el-capitan/#comment-14738 Sun, 09 Aug 2015 02:45:09 +0000 http://memkite.com/?p=875#comment-14738 I get an error at metalDevice.newComputePipelineStateWithDescriptor

]]>
Comment on GPGPU Performance of Swift/Metal vs Accelerate on iPhone 6 & 5S, iPad Air and iPad Mini by Nevin Brackett-Rozinsky http://memkite.com/blog/2014/12/18/gpgpu-performance-of-swiftmetal-vs-accelerate-on-iphone-6-5s-ipad-air-and-ipad-mini/#comment-14557 Thu, 23 Jul 2015 19:28:06 +0000 http://memkite.com/?p=676#comment-14557 It looks like your Swift+Accelerate implementation will blow through the cache twice on large arrays. Have you tried processing batches of, say, 1024 elements at a time? I’m not sure how easy that is to do in Swift, but with C-style pointer arithmetic it is trivial.

Also, you could use vvrecf to take the reciprocal, rather than vvpowf.

And for that matter, did you check whether vDSP_vsadd is faster than cblas_saxpy?

So for example, you might have something like this, only Swiftier (or in a wrapper):

const int batchSize = 1024;
const int r = n % batchSize;
const float *endingPoint = startingPoint + n - r;
for (float *p = startingPoint; p < endingPoint; p += batchSize) {
vvexpf(…); // batchSize elements of p
vDSP_vsadd(…); // batchSize elements of p
vvrecf(…); // batchSize elements of p
}
// process final r < 1024 elements of p

And finally, it may be worth mentioning that 1/(1+exp(-x)) == exp(x)/(1+exp(x)), so the negation can be eliminated at the cost of replacing a reciprocal with a division (and needing to store a second array). Testing should reveal which is faster.

]]>
Comment on Continuous Integration for testing Metal GPGPU shaders for iOS with Swig, Numpy and Python by Swift and Metal GPU Programming on OSX 10.11 / El Capitan | Memkite http://memkite.com/blog/2015/05/20/continuous-integration-for-testing-metal-gpgpu-shaders-for-ios-with-swig-numpy-and-python/#comment-14300 Wed, 10 Jun 2015 13:17:45 +0000 http://memkite.com/?p=863#comment-14300 […] Continuous Integration for Testing Metal GPGPU Shaders for iOS with Swig, Numpy and Python […]

]]>
Comment on Deep Learning for Speech Recognition by Threadmill Desks – the Future Software Engineer Office Rig? | Memkite http://memkite.com/blog/2015/02/11/deep-learning-for-speech-recognition/#comment-14044 Thu, 14 May 2015 18:03:42 +0000 http://memkite.com/?p=714#comment-14044 […] the rapid advances in Deep Learning for Speech Recognition on might even get rid of the keyboard (and the desk!) eventually, even though I believe using […]

]]>
Comment on Update with 13 new papers to Deeplearning.University by Update with 29 new publications to Deeplearning.University Bibliography | Memkite http://memkite.com/blog/2015/04/07/update-with-13-new-papers-to-deeplearning-university/#comment-13581 Mon, 13 Apr 2015 06:23:34 +0000 http://memkite.com/?p=814#comment-13581 […] to more frequent updates (1-2 times per week). Underneath are 29 new deep learning papers since last update (8 days prior to this). There are many highly interesting papers, and I would in particular like to […]

]]>
Comment on Data-Parallel Programming with Metal and Swift for iPhone/iPad GPU by Dipanjan Das http://memkite.com/blog/2014/12/15/data-parallel-programming-with-metal-and-swift-for-iphoneipad-gpu/#comment-13304 Thu, 26 Mar 2015 18:17:02 +0000 http://memkite.com/?p=671#comment-13304 Hello,
Thanks for sharing a wonderful example. It really helps to understand the basic building blocks of data parallel programming with metal. However when I tried it to run the example it gives a gpu run time exception telling the accessing invalid memory address of the kernel. By modify the kernel program only coping the input array data into output array, program started working. The observation is that any arithmetic computation in the kernel program giving the above exception (even outputVector[id] = 5.0 .//or any hard code assignment is not working.)

It would be great help if you able to give any solution around the above problem.

]]>