Christoph Ruegg
a97604768c
Docs: simplify home page, fix tooltips
11 years ago
Christoph Ruegg
3f7b955a19
Merge pull request #303 from MattHeffron/SparseVector.AbsoluteMaximumIndex
Add (restore?) the missing SparseVector.AbsoluteMaximumIndex() override implementations
11 years ago
Matt Heffron
31a7927062
Add (restore?) the missing SparseVector.AbsoluteMaximumIndex() override implementations.
11 years ago
Christoph Ruegg
850a7af28e
Docs: NuGet & Binaries, simplified home page
11 years ago
Christoph Ruegg
eab4eae0b7
Docs: Platform Compatibility
11 years ago
Christoph Ruegg
20f44564cf
Docs: Build & Tools
11 years ago
Christoph Ruegg
431b727ebe
Dependencies
11 years ago
Christoph Ruegg
f18e96e260
Docs: rewording MKL section
11 years ago
Christoph Ruegg
9771f1bdd1
Release v3.6.0
11 years ago
Christoph Ruegg
a6878cc7d1
Distributions: ChiSquared.CDF test fix for more robust implementation #288
11 years ago
Christoph Ruegg
4814200541
Merge remote-tracking branch 'bbieniek/master'
11 years ago
Christoph Ruegg
9dbf23a928
Tests: port vector storage combinator tests to matrices
11 years ago
Christoph Ruegg
36b568c5f1
Tests: cleanup, simplifications
11 years ago
Christoph Ruegg
54b48383e6
Tests: merge dense and sparse vector arithmetic tests, cover mixed cases
11 years ago
Christoph Ruegg
b244510f0e
Tests: refactor LA theory tests to use enum datasets (much more tool friendly)
11 years ago
Christoph Ruegg
14ec5a392b
Buildfix: disable Fibonacci tests where not available (pcl, 3.5).
11 years ago
Christoph Ruegg
44ea89a983
Generate: Unfold, Fibonacci
11 years ago
Christoph Ruegg
004d8155e3
Generate: Normal and Standard, deprecate Stable and Gaussian
11 years ago
Christoph Ruegg
fbfe077552
LA: migrate mixed-storage fallback implementations to use higher order functions
11 years ago
Christoph Ruegg
e591f86e2f
LA: Vector.Find/2, Exists/2, ForAll/2
11 years ago
Christoph Ruegg
73140be313
Docs: document simpler MKL loading with native provider loader
11 years ago
Christoph Ruegg
4967e057b4
Native Provider Loader: code simplification
11 years ago
Christoph Ruegg
a112a2a380
Native Provider Loader: optionally set directory from the outside via Control.NativeProviderPath
11 years ago
Christoph Ruegg
241797fc20
Release Notes: cosmetics: highlight bug fixes
11 years ago
Christoph Ruegg
fa3dae304f
Merge pull request #298 from kjbartel/NativeProviderLoaderFixes
Native provider loader fixes
11 years ago
Kuan Bartel
5a01cd91bf
Changed architecture directory name from "amd64" to "x64" in NativeProviderLoader.
Changed UnitTests-MKL project to only have "Any CPU" platform configuration. Changed output path to "out\MKL\Windows\" so that the tests could be run using the NativeProviderLoader without needing any manual copying.
11 years ago
Kuan Bartel
67642e9c76
Changed to use LoadLibraryEx rather than LoadLibrary to properly find library dependencies and added searching for the architecture specific folder in the current domain's base directory which fixes a problem with running unit tests with MSTest / Xunit in Visual Studio (assemblies seem to be copied to the temp directory so fails to find the native library).
11 years ago
Christoph Ruegg
43c361dd87
Native: replace NATIVEMKL build constant with NATIVE as it can be used by other native providers as well #296
11 years ago
Christoph Ruegg
60beb0eca6
Providers: disable native provider loader on portable builds #296
11 years ago
Christoph Ruegg
e0d62b721e
Tooling update
11 years ago
Christoph Ruegg
6525a4b923
Merge pull request #296 from kjbartel/NativePlatformDetect
Detect processor architecture and process bitness for loading native providers
11 years ago
Kuan Bartel
b8de1e563a
Changed behaviour to only attempt to load the library if it is found within an architecture specific subfolder. If not found then LoadLibrary will return false allowing the normal P/Invoke load behaviour as P/Invoke searches in a slightly different order to the dlopen() and LoadLibrary() functions.
For Unix based OS changed to only supporting x86 and amd64 as I haven't found a reliable way to detect the architecture. MKL only supports x86 and amd64 anyway so it isn't currently a problem but could be for other native providers which support other architectures.
11 years ago
Kuan Bartel
4ea75fad55
Added support for loading native libraries on Linux and MacOSX. I'm not sure how to detect the processor on Unix systems though as Environment.GetEnvironmentVariable("PROCESSOR_ARCHITECTURE") won't work. x86 and AMD64 are fine as I could just use Environment.Is64BitProcess but there's no way to detect ARM and IA64 in the BCL. So currently for Unix I'm not using an architecture directory.
Other small changes rearranging some code and changing some property names.
11 years ago
Kuan Bartel
a75e4f6060
Added NativeProviderLoader, a helper class for loading native provider libraries depending on the processor architecture and the process bitness. MklLinearAlgebraProvider modified to call the helper class in InitializeVerify.
Not sure exactly how to automate building AnyCPU build though as both x86 and x64 builds need to be built first and then the files copied into subdirectories (x86 and amd64) of the AnyCPU output directory.
11 years ago
Christoph Ruegg
1f26138b59
LA: proper exception when trying to create a matrix of an empty vector array #294
11 years ago
Christoph Ruegg
bd64c8e612
LA: Matrix.Fold2, Find/2, Exists/2, ForAll/2
11 years ago
Christoph Ruegg
d1651ebb4f
LA: Storage should not make Zeros/ExistingData arguments optional (internal only until v4). Minor optimization.
11 years ago
Christoph Ruegg
70dda56f85
LA: MatrixStorage.Find/Find2; leverage Find2 in Equals #291
11 years ago
Christoph Ruegg
c4c0ee3955
LA: inline xml docs typo
11 years ago
Christoph Ruegg
aa79ec7a97
Merge pull request #290 from cuda/simplify_native
Native Tweaks
11 years ago
Marcus Cuda
b7cd353d58
added no convergence check
11 years ago
Marcus Cuda
341dec231d
fixed tests for complex evd
11 years ago
Marcus Cuda
296ff72b4b
added native Complex EVD support
11 years ago
bbieniek
e0e7292324
When DF parameter was high, the CDF returned NaN value. This was resolved by using the GammaLowerRegularized function straight from the CDF instead of division and multipliciation by Gamma function.
11 years ago
Marcus Cuda
129702d49e
simplified lapack templates
11 years ago
Christoph Ruegg
78a7aca144
Docs: tweak build & tools docs, merge into website
11 years ago
Christoph Ruegg
bc5165ab43
Build: dependencies & tools upgrade, docs build updated
11 years ago
Christoph Ruegg
810b50cd21
Special Functions: stronger unit tests for Gamma/GammaLn #196
12 years ago
Christoph Ruegg
94a37cb1c1
LA: More consistent range checking in MatrixStorage.Clear and related routines #284
12 years ago
Christoph Ruegg
4d96f30889
LA: fix loop range in MatrixStorage.ClearColumns (built-in storage backends not affected) #284
12 years ago