| Type: | Package |
| Title: | GPU Functions for R Objects |
| Description: | Provides GPU enabled functions for 'R' objects in a simple and approachable manner. New 'gpu*' and 'vcl*' classes have been provided to wrap typical 'R' objects (e.g. vector, matrix), in both host and device spaces, to mirror typical 'R' syntax without the need to know 'OpenCL'. |
| Version: | 2.0.6 |
| Date: | 2024-05-22 |
| Maintainer: | Ruoyong Xu <ruoyong.xu@mail.utoronto.ca> |
| VignetteBuilder: | knitr |
| License: | GPL-2 | GPL-3 [expanded from: GPL (≥ 2)] |
| Encoding: | UTF-8 |
| Depends: | R (≥ 3.0.2), methods, utils |
| Imports: | Rcpp (≥ 0.12.15), RViennaCL, BH, RcppEigen |
| LinkingTo: | Rcpp (≥ 0.12.15), RcppEigen (≥ 0.3.3.4.0), RViennaCL (≥ 1.7.1.7), BH |
| NeedsCompilation: | yes |
| Suggests: | testthat, knitr |
| URL: | https://github.com/eborgnine/gpuR |
| BugReports: | https://github.com/cdeterman/gpuR/issues/new |
| SystemRequirements: | C++11 (supporting at least std=c++0x), OpenCL shared library (provided by an SDK such as AMD/NVIDIA) and OpenCL headers including the C++ header file (provided by Khronos if not by SDK) |
| OS_type: | unix |
| RoxygenNote: | 7.3.1 |
| Packaged: | 2024-05-22 13:53:23 UTC; ruoyong |
| Author: | Charles Determan Jr [aut, cph],
Patrick Brown 
[aut, ctb],
Ruoyong Xu [cre,
ctb],
The Khronos Group Inc [cph] (Copyright holder of some included header
files) |
| Repository: | CRAN |
| Date/Publication: | 2024-05-23 16:00:02 UTC |
GPU functions for R Objects
Description
This package was developed to provide simple to use R functions that leverage the power of GPU's but also retain a format familiar to the R user. There are a hand full of other R packages that provide some GPU functionality but nearly all rely on a CUDA backend thereby restricting the user to NVIDIA GPU hardware. In the spirit of being as broadly applicable as possible, this GPU code herein relies upon OpenCL via the ViennaCL library.
OpenCL, in contrast to CUDA, is open source and can be used across different graphics cards (e.g. NVIDIA, AMD, Intel). This package removes the complex code needed for GPU computing and provides easier to use functions to apply on R objects.
| Package: | gpuR |
| Type: | Package |
| Version: | 1.0.0 |
| Date: | 2015-03-31 |
| License: | GPL-3 |
| Copyright: | (c) 2015 Charles E. Determan Jr. |
| URL: | https://github.com/cdeterman/gpuR |
| LazyLoad: | yes |
Note
There are other packages that also provide wrappers for OpenCL programming including OpenCL by Simon Urbanek and ROpenCL at Open Analytics by Willem Ligtenberg. Both of these packages provide the R user an interface to directly call OpenCL functions. This package, however, hides these functions so the user does not require any knowledge of OpenCL to begin using their GPU. The idea behind this package is to provide a means to begin using existing algorithms without the need to write extensive amounts of C/C++/OpenCL code.
Author(s)
Charles Determan cdetermanjr@gmail.com
Maintainer: Charles Determan cdetermanjr@gmail.com
Methods for gpu/vcl Vector
Description
Multiply two gpuR objects, if they are conformable. If both are vectors of the same length, it will return the inner product (as a matrix).
Usage
## S4 method for signature 'gpuVector,gpuVector'
x %*% y
## S4 method for signature 'gpuVector,gpuMatrix'
x %*% y
## S4 method for signature 'vclMatrix,vclMatrix'
x %*% y
## S4 method for signature 'vclMatrix,vclVector'
x %*% y
## S4 method for signature 'vclMatrix,matrix'
x %*% y
## S4 method for signature 'matrix,vclMatrix'
x %*% y
## S4 method for signature 'vclVector,vclVector'
x %*% y
## S4 method for signature 'vclVector,vclMatrix'
x %*% y
## S4 method for signature 'gpuMatrix,gpuMatrix'
x %*% y
## S4 method for signature 'gpuMatrix,gpuVector'
x %*% y
## S4 method for signature 'gpuMatrix,matrix'
x %*% y
## S4 method for signature 'matrix,gpuMatrix'
x %*% y
Arguments
x |
A gpuR object |
y |
A gpuR object |
Value
The inner product of the two gpuVector objects x and y.
The result of matrix-vector multiplication between the gpuMatrix y
and the gpuVector x.
The result of multiplying the vclMatrix 'x' and the vclVector 'y'.
A vclMatrix object, the result of multiplying
the vclMatrix 'x' and the regular R matrix 'y'.
A vclMatrix object, the result of multiplying the regular
R matrix x and the vclMatrix y.
A gpuMatrix object which is the result of multiplying
the two gpuMatrix objects 'x' and 'y'.
Author(s)
Charles Determan Jr.
Outer Product
Description
The outer product of two gpuR vector objects
Usage
## S4 method for signature 'gpuVector,gpuVector'
X %o% Y
## S4 method for signature 'vclVector,vclVector'
X %o% Y
Arguments
X |
A gpuR object |
Y |
A gpuR object |
Value
The outer product of the two gpuVector objects X and Y.
Author(s)
Charles Determan Jr.
Arith methods
Description
Methods for the base Arith methods S4groupGeneric
Usage
## S4 method for signature 'gpuVector,gpuVector'
Arith(e1, e2)
## S4 method for signature 'numeric,gpuVector'
Arith(e1, e2)
## S4 method for signature 'gpuVector,numeric'
Arith(e1, e2)
## S4 method for signature 'gpuVector,missing'
Arith(e1, e2)
## S4 method for signature 'gpuVector,gpuMatrix'
Arith(e1, e2)
## S4 method for signature 'vclMatrix,vclMatrix'
Arith(e1, e2)
## S4 method for signature 'vclMatrix,matrix'
Arith(e1, e2)
## S4 method for signature 'matrix,vclMatrix'
Arith(e1, e2)
## S4 method for signature 'vclMatrix,numeric'
Arith(e1, e2)
## S4 method for signature 'numeric,vclMatrix'
Arith(e1, e2)
## S4 method for signature 'vclMatrix,missing'
Arith(e1, e2)
## S4 method for signature 'vclMatrix,vclVector'
Arith(e1, e2)
## S4 method for signature 'vclVector,vclVector'
Arith(e1, e2)
## S4 method for signature 'numeric,vclVector'
Arith(e1, e2)
## S4 method for signature 'vclVector,numeric'
Arith(e1, e2)
## S4 method for signature 'vclVector,missing'
Arith(e1, e2)
## S4 method for signature 'vclVector,vclMatrix'
Arith(e1, e2)
## S4 method for signature 'gpuMatrix,gpuMatrix'
Arith(e1, e2)
## S4 method for signature 'gpuMatrix,matrix'
Arith(e1, e2)
## S4 method for signature 'matrix,gpuMatrix'
Arith(e1, e2)
## S4 method for signature 'gpuMatrix,numeric'
Arith(e1, e2)
## S4 method for signature 'numeric,gpuMatrix'
Arith(e1, e2)
## S4 method for signature 'gpuMatrix,missing'
Arith(e1, e2)
## S4 method for signature 'gpuMatrix,gpuVector'
Arith(e1, e2)
Arguments
e1 |
A gpuR object |
e2 |
A gpuR object |
Value
A gpuR object
Author(s)
Charles Determan Jr.
Compare vector and gpuVector elements
Description
Methods for comparison operators
Usage
## S4 method for signature 'vector,gpuVector'
Compare(e1, e2)
## S4 method for signature 'gpuVector,vector'
Compare(e1, e2)
Arguments
e1 |
A vector/gpuVector object |
e2 |
A vector/gpuVector object |
Value
A logical vector
Author(s)
Charles Determan Jr.
gpuR Math methods
Description
Methods for the base Math methods S4groupGeneric
Usage
## S4 method for signature 'gpuVector'
Math(x)
## S4 method for signature 'vclMatrix'
Math(x)
## S4 method for signature 'vclVector'
Math(x)
## S4 method for signature 'gpuMatrix'
Math(x)
Arguments
x |
A gpuR object |
Details
Currently implemented methods include:
"sin", "cos", "tan", "asin", "acos", "atan", "sinh", "cosh", "tanh", "log10", "exp", "abs", "sqrt", "sign"
Value
A gpuVector object resulting from the
element-wise mathematical operation performed on
the input gpuVector object x.
A vclMatrix object.
A vclVector object.
A gpuMatrix object.
Author(s)
Charles Determan Jr.
gpuR Summary methods
Description
Methods for the base Summary methods S4groupGeneric
Usage
## S4 method for signature 'gpuVector'
Summary(x, ..., na.rm = FALSE)
## S4 method for signature 'vclMatrix'
Summary(x, ..., na.rm = FALSE)
## S4 method for signature 'vclVector'
Summary(x, ..., na.rm = FALSE)
## S4 method for signature 'gpuMatrix'
Summary(x, ..., na.rm = FALSE)
Arguments
x |
A gpuR object |
... |
Additional arguments passed to method (not currently used) |
na.rm |
a logical indicating whether missing values should be removed ( not currently used) |
Value
For min or max, a length-one vector
Extract gpuR object elements
Description
Operators to extract or replace elements
Usage
## S4 method for signature 'gpuMatrix,missing,missing,missing'
x[i, j, drop]
## S4 method for signature 'gpuMatrix,missing,numeric,missing'
x[i, j, drop]
## S4 method for signature 'gpuMatrix,numeric,missing,missing'
x[i, j, ..., drop = TRUE]
## S4 method for signature 'gpuMatrix,numeric,numeric,missing'
x[i, j, drop]
## S4 replacement method for signature 'gpuMatrix,numeric,missing,numeric'
x[i, j, ...] <- value
## S4 replacement method for signature 'igpuMatrix,numeric,missing,integer'
x[i, j] <- value
## S4 replacement method for signature 'gpuMatrix,missing,numeric,numeric'
x[i, j] <- value
## S4 replacement method for signature 'igpuMatrix,missing,numeric,integer'
x[i, j] <- value
## S4 replacement method for signature 'gpuMatrix,numeric,numeric,numeric'
x[i, j] <- value
## S4 replacement method for signature 'igpuMatrix,numeric,numeric,integer'
x[i, j] <- value
## S4 method for signature 'gpuVector,missing,missing,missing'
x[i, j, drop]
## S4 method for signature 'gpuVector,numeric,missing,missing'
x[i, j, drop]
## S4 replacement method for signature 'gpuVector,numeric,missing,numeric'
x[i, j] <- value
## S4 replacement method for signature 'gpuVector,numeric,missing,integer'
x[i, j] <- value
## S4 method for signature 'vclMatrix,missing,missing,missing'
x[i, j, drop]
## S4 method for signature 'vclMatrix,missing,numeric,missing'
x[i, j, drop]
## S4 method for signature 'vclMatrix,numeric,missing,missing'
x[i, j, ..., drop = TRUE]
## S4 method for signature 'vclMatrix,numeric,numeric,missing'
x[i, j, drop]
## S4 replacement method for signature 'vclMatrix,missing,numeric,numeric'
x[i, j] <- value
## S4 replacement method for signature 'ivclMatrix,missing,numeric,integer'
x[i, j] <- value
## S4 replacement method for signature 'vclMatrix,numeric,missing,numeric'
x[i, j, ...] <- value
## S4 replacement method for signature 'ivclMatrix,numeric,missing,integer'
x[i, j] <- value
## S4 replacement method for signature 'vclMatrix,numeric,numeric,numeric'
x[i, j] <- value
## S4 replacement method for signature 'ivclMatrix,numeric,numeric,integer'
x[i, j] <- value
## S4 replacement method for signature 'vclMatrix,missing,missing,matrix'
x[i, j] <- value
## S4 replacement method for signature 'vclMatrix,missing,missing,vclMatrix'
x[i, j] <- value
## S4 replacement method for signature 'vclMatrix,missing,numeric,vclMatrix'
x[i, j] <- value
## S4 replacement method for signature 'vclMatrix,missing,missing,numeric'
x[i, j] <- value
## S4 replacement method for signature 'vclMatrix,missing,missing,vclVector'
x[i, j] <- value
## S4 replacement method for signature 'vclMatrix,missing,numeric,vclVector'
x[i, j] <- value
## S4 method for signature 'vclVector,missing,missing,missing'
x[i, j, drop]
## S4 method for signature 'vclVector,numeric,missing,missing'
x[i, j, drop]
## S4 replacement method for signature 'vclVector,numeric,missing,numeric'
x[i, j] <- value
## S4 replacement method for signature 'ivclVector,numeric,missing,integer'
x[i, j] <- value
## S4 replacement method for signature 'vclVector,logical,missing,numeric'
x[i, j] <- value
## S4 replacement method for signature 'vclVector,missing,missing,numeric'
x[i, j] <- value
## S4 replacement method for signature 'vclVector,missing,missing,vclVector'
x[i, j] <- value
## S4 replacement method for signature 'vclVector,numeric,missing,vclVector'
x[i, j] <- value
## S4 replacement method for signature 'vclVector,missing,missing,vclMatrix'
x[i, j] <- value
## S4 replacement method for signature 'vclVector,numeric,missing,vclMatrix'
x[i, j] <- value
Arguments
x |
A gpuR object |
i |
indices specifying rows |
j |
indices specifying columns |
drop |
missing |
... |
Additional arguments |
value |
data of similar type to be added to gpuMatrix object |
Value
A gpuR object of the same type as the input x, containing
the elements specified by the indices i and j.
Author(s)
Charles Determan Jr.
Convert object to a gpuMatrix
Description
Construct a gpuMatrix of a class that inherits
from gpuMatrix
Usage
as.gpuMatrix(object, type)
Arguments
object |
An object that is or can be converted to a
|
type |
A character string specifying the type of gpuMatrix. Default is NULL where type is inherited from the source data type. |
Value
A gpuMatrix object
Author(s)
Charles Determan Jr.
Convert object to a gpuVector
Description
Construct a gpuVector of a class that inherits
from gpuVector
Usage
as.gpuVector(object, type)
## S4 method for signature 'vector'
as.gpuVector(object, type = NULL)
Arguments
object |
An object that is or can be converted to a
|
type |
A character string specifying the type of gpuVector. Default is NULL where type is inherited from the source data type. |
Value
A gpuVector object
Author(s)
Charles Determan Jr.
Convert object to a vclVector
Description
Construct a vclVector of a class that inherits
from vclVector
stuff
Usage
as.vclVector(object, type = NULL, ...)
## S4 method for signature 'vector'
as.vclVector(object, type = NULL)
## S4 method for signature 'vclMatrix'
as.vclVector(object, type = NULL, shared = FALSE)
Arguments
object |
An object that is or can be converted to a
|
type |
A character string specifying the type of vclVector. Default is NULL where type is inherited from the source data type. |
... |
Additional arguments to as.vclVector methods |
shared |
Logical indicating if memory should be shared with x |
Value
A vclVector object
A vclVector object.
Author(s)
Charles Determan Jr.
Does device have 'double' support?
Description
Function to query if device (identified by index) supports double precision
Usage
assert_has_double(
device_idx,
context_idx,
severity = getOption("assertive.severity", "stop")
)
Arguments
device_idx |
An integer indicating which device to query |
context_idx |
An integer indicating which context to query |
severity |
How severe should the consequences of the assertion be? |
Value
Returns nothing but throws an error if device does not support double precision
Author(s)
Charles Determan Jr.
See Also
Matrix Blocks
Description
This doesn't create a copy, it provides a child class that
points to a contiguous submatrix of a gpuMatrix or
vclMatrix. Non-contiguous blocks are currently not supported.
Usage
block(object, rowStart, rowEnd, colStart, colEnd)
## S4 method for signature 'vclMatrix,integer,integer,integer,integer'
block(object, rowStart, rowEnd, colStart, colEnd)
## S4 method for signature 'gpuMatrix,integer,integer,integer,integer'
block(object, rowStart, rowEnd, colStart, colEnd)
Arguments
object |
A |
rowStart |
An integer indicating the first row of block |
rowEnd |
An integer indicating the last row of block |
colStart |
An integer indicating the first column of block |
colEnd |
An integer indicating the last column of block |
Details
This function allows a user to create a gpuR matrix object that references a continuous subset of columns and rows of another gpuR matrix object without a copy.
NOTE - this means that altering values in a matrix block object will alter values in the source matrix.
Value
A block of the input gpuMatrixBlock or vclMatrixBlock object.
Author(s)
Charles Determan Jr.
cgpuMatrix Class
Description
An complex float type matrix in the S4 gpuMatrix
representation.
Value
If the gpuMatrix object is of type 'complex float', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a complex float matrix.
Author(s)
Charles Determan Jr.
See Also
gpuMatrix-class,
igpuMatrix-class,
dgpuMatrix-class
Cholesky Decomposition of a gpuR matrix
Description
Compute the Choleski factorization of a real symmetric positive-definite square matrix.
Usage
## S3 method for class 'vclMatrix'
chol(x, ...)
## S3 method for class 'gpuMatrix'
chol(x, ...)
Arguments
x |
A symmetric, positive-definite gpuR matrix object. |
... |
arguments to be passed to or from methods |
Value
Default - the upper triangular factor of the Choleski decomposition, i.e. the matrix R such that R'R = x.
Note
This an S3 generic of chol. The default continues to point to the default base function.
No pivoting is used.
The argument upper is additionally accepted representing a boolean
which will indicate if the upper or lower (FALSE) triangle
should be solved.
Author(s)
Charles Determan Jr.
See Also
Row and Column Sums and Means of gpuMatrix
Description
Row and column sums and of gpuMatrix objects
Usage
## S4 method for signature 'gpuMatrix'
colSums(x)
## S4 method for signature 'gpuMatrix'
rowSums(x)
## S4 method for signature 'gpuMatrix'
colMeans(x)
## S4 method for signature 'gpuMatrix'
rowMeans(x)
Arguments
x |
A gpuMatrix object |
Value
A gpuvector containing the sum of each column in the gpuMatrix.
A gpuvector containing the sum of each row in the gpuMatrix.
A gpuvector containing the mean of each column in the gpuMatrix.
A gpuvector containing the mean of each row in the gpuMatrix.
Author(s)
Charles Determan Jr.
Row and Column Sums and Means of vclMatrix
Description
Row and column sums and of vclMatrix objects
Usage
## S4 method for signature 'vclMatrix'
colSums(x)
## S4 method for signature 'vclMatrix'
rowSums(x)
## S4 method for signature 'vclMatrix'
colMeans(x)
## S4 method for signature 'vclMatrix'
rowMeans(x)
Arguments
x |
A |
Value
A vclVector object
Author(s)
Charles Determan Jr.
Row and Column Names
Description
Retrieve or set the row or column names of a gpuR matrix object
Usage
colnames(x, do.NULL, prefix)
## Default S3 method:
colnames(x, do.NULL = TRUE, prefix = "col")
## S3 method for class 'gpuMatrix'
colnames(x, ...)
## S4 replacement method for signature 'gpuMatrix'
colnames(x) <- value
## S3 method for class 'vclMatrix'
colnames(x, ...)
## S4 replacement method for signature 'vclMatrix'
colnames(x) <- value
Arguments
x |
A gpuR matrix object |
do.NULL |
logical. If |
prefix |
for create names. (not currently used) |
... |
Additional arguments |
value |
A character vector to assign as row/column names |
Value
A character vector of column names of the gpuMatrix object.
Covariance (gpuR)
Description
Compute covariance values
Usage
## S4 method for signature 'vclMatrix,missing,missing,missing'
cov(x, y = NULL, use = NULL, method = "pearson")
## S4 method for signature 'vclMatrix,vclMatrix,missing,missing'
cov(x, y = NULL, use = NULL, method = "pearson")
## S4 method for signature 'vclMatrix,missing,missing,character'
cov(x, y = NULL, use = NULL, method = "pearson")
## S4 method for signature 'vclMatrix,vclMatrix,missing,character'
cov(x, y = NULL, use = NULL, method = "pearson")
## S4 method for signature 'gpuMatrix,missing,missing,missing'
cov(x, y = NULL, use = NULL, method = "pearson")
## S4 method for signature 'gpuMatrix,gpuMatrix,missing,missing'
cov(x, y = NULL, use = NULL, method = "pearson")
## S4 method for signature 'gpuMatrix,missing,missing,character'
cov(x, y = NULL, use = NULL, method = "pearson")
## S4 method for signature 'gpuMatrix,gpuMatrix,missing,character'
cov(x, y = NULL, use = NULL, method = "pearson")
Arguments
x |
A gpuR object |
y |
A gpuR object |
use |
Not used |
method |
Character string indicating with covariance to be computed. |
Value
A gpuMatrix/vclMatrix containing the symmetric covariance values.
Author(s)
Charles Determan Jr.
gpuMatrix Crossproduct
Description
Return the matrix cross-product of two conformable matrices using a GPU. This is equivalent to t(x) or x device and host is required.
Usage
## S4 method for signature 'gpuMatrix,missing'
crossprod(x, y)
## S4 method for signature 'gpuMatrix,gpuMatrix'
crossprod(x, y)
## S4 method for signature 'gpuMatrix,matrix'
crossprod(x, y)
## S4 method for signature 'matrix,gpuMatrix'
crossprod(x, y)
## S4 method for signature 'gpuMatrix,missing'
tcrossprod(x, y)
## S4 method for signature 'gpuMatrix,gpuMatrix'
tcrossprod(x, y)
## S4 method for signature 'matrix,gpuMatrix'
tcrossprod(x, y)
## S4 method for signature 'gpuMatrix,matrix'
tcrossprod(x, y)
Arguments
x |
A gpuMatrix |
y |
A gpuMatrix |
Value
A gpuMatrix
Author(s)
Charles Determan Jr.
Current Context
Description
Get current context index
Usage
currentContext()
Value
An integer reflecting the context listed in listContexts
Current Device Information
Description
Check current device information
Usage
currentDevice()
Value
list containing
device |
Character string of device name |
device_index |
Integer identifying device |
device_type |
Character string identifying device type (e.g. gpu) |
Return Current Platform
Description
Find out which platform is currently in use
Usage
currentPlatform()
Value
platform |
Name of the current platform |
platform_index |
Index of current platform |
See Also
Custom OpenCL Kernels
Description
Compile a custom function using a user provided OpenCL kernel
Usage
custom_opencl(kernel, cl_args, type)
Arguments
kernel |
path to OpenCL kernel file |
cl_args |
A |
type |
The precision on which the kernel is compiled. Options include
|
Value
This function does not return a value directly, but it compiles
the provided OpenCL kernel and performs necessary operations for using it.
cvclMatrix Class
Description
An complex float type matrix in the S4 vclMatrix
representation.
Value
If the vclMatrix object is of type 'complex float', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a complex float type matrix
Author(s)
Charles Determan Jr.
See Also
vclMatrix-class,
ivclMatrix-class,
fvclMatrix-class
Copy a "gpuR" object
Description
This is needed to make a duplicate of a gpuR object
Usage
deepcopy(object, ...)
## S4 method for signature 'gpuVector'
deepcopy(object)
## S4 method for signature 'vclMatrix'
deepcopy(object, source = FALSE)
## S4 method for signature 'vclVector'
deepcopy(object)
## S4 method for signature 'gpuMatrix'
deepcopy(object)
Arguments
object |
A gpuR object |
... |
Additional arguments |
source |
A boolean indicating if source matrix should be copied (only relevant for 'block' and 'slice' objects). |
Details
This is needed to make a duplicate of a gpuR object
(i.e. gpuMatrix, gpuVector,
vclMatrix, vclVector because
the traditional syntax would only copy the pointer of the object.
Value
A gpuR object
A deep copy of the input gpuVector object.
A deep copy of the input vclMatrix object.
A deep copy of the input vclVector object.
A deep copy of the input gpuMatrix object.
Author(s)
Charles Determan Jr.
See Also
Calculate Determinant of a Matrix on GPU
Description
det calculates the determinant of a matrix.
Usage
## S4 method for signature 'vclMatrix'
det(x)
## S4 method for signature 'gpuMatrix'
det(x)
Arguments
x |
A gpuR matrix object |
Value
The determinant of x
Note
This function uses an LU decomposition and the det
function is simply a wrapper returning the determinant product
Author(s)
Charles Determan Jr.
Detect Available OpenCL enabled CPUs
Description
Find out how many CPUs available
Usage
detectCPUs(platform_idx = NULL)
Arguments
platform_idx |
An integer value indicating which platform to query. If NULL it will iterate over all platforms and sum results |
Value
An integer representing the number of available CPUs
See Also
Detect Available GPUs
Description
Find out how many GPUs available
Usage
detectGPUs(platform_idx = NULL)
Arguments
platform_idx |
An integer value indicating which platform to query. If NULL it will iterate over all platforms and sum results |
Value
An integer representing the number of available GPUs
See Also
Detect Number of Platforms
Description
Find out how many OpenCL enabled platforms are available.
Usage
detectPlatforms()
Value
An integer value representing the number of platforms available.
See Also
Check GPU double precision support
Description
This function checks the GPU device extensions for the variable cl_khr_fp64 which means the device supports double precision.
Usage
deviceHasDouble(
gpu_idx = currentDevice()$device_index,
context_idx = currentContext()
)
Arguments
gpu_idx |
An integer value indicating which gpu to query. |
context_idx |
An integer value indicating which context to query. |
Value
A boolean designating whether the device supports double precision
See Also
Check device type
Description
Check what type a device is given platform and device indices
Usage
deviceType(device_idx = NULL, context_idx = currentContext())
Arguments
device_idx |
An integer value indicating which device to query. |
context_idx |
An integer value indicating which context to query. |
Value
A character string indicating the device type
dgpuMatrix Class
Description
A double type matrix in the S4 gpuMatrix
representation.
Value
If the gpuMatrix object is of type 'double', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a double type matrix
Author(s)
Charles Determan Jr.
See Also
gpuMatrix-class,
igpuMatrix-class,
fgpuMatrix-class
dgpuVector Class
Description
An double vector in the S4 gpuVector
representation.
Value
If the gpuVector object is of type 'double', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a double typed vector
Author(s)
Charles Determan Jr.
See Also
gpuR Matrix Diagonals
Description
Extract or replace the diagonal of a matrix
Usage
## S4 method for signature 'vclMatrix'
diag(x)
## S4 replacement method for signature 'vclMatrix,vclVector'
diag(x) <- value
## S4 method for signature 'gpuMatrix'
diag(x)
## S4 replacement method for signature 'gpuMatrix,gpuVector'
diag(x) <- value
Arguments
x |
A gpuR matrix object |
value |
A vector object (gpuR) |
Value
A gpuRvector object of the matrix diagonal of x. The
replacement form returns nothing as it replaces the diagonal of x.
Note
If an identity matrix is desired, please see identity_matrix.
Author(s)
Charles Determan Jr.
See Also
gpuMatrix/vclMatrix dim method
Description
Retrieve dimension of object
Usage
## S4 method for signature 'vclMatrix'
dim(x)
## S4 method for signature 'gpuMatrix'
dim(x)
Arguments
x |
A gpuMatrix/vclMatrix object |
Value
A numeric vector of length 2: the number of rows and columns in the gpuR object 'x'.
Author(s)
Charles Determan Jr.
GPU Distance Matrix Computations
Description
This function computes and returns the distance matrix computed by using the specified distance measure to compute the distances between the rows of a data matrix.
Usage
distance(x, y, method = "euclidean")
## S4 method for signature 'vclMatrix'
dist(x, method = "euclidean", diag = FALSE, upper = FALSE, p = 2)
## S4 method for signature 'vclMatrix,vclMatrix'
distance(x, y, method = "euclidean")
## S4 method for signature 'gpuMatrix'
dist(x, method = "euclidean", diag = FALSE, upper = FALSE, p = 2)
## S4 method for signature 'gpuMatrix,gpuMatrix'
distance(x, y, method = "euclidean")
Arguments
x |
A gpuMatrix or vclMatrix object |
y |
A gpuMatrix or vclMatrix object |
method |
the distance measure to be used. This must be one of "euclidean" or "sqEuclidean". |
diag |
logical value indicating whether the diagonal of the distance matrix should be printed |
upper |
logical value indicating whether the upper triangle of the distance matrix |
p |
The power of the Minkowski distance (not currently used) |
Value
a gpuMatrix/vclMatrix containing the pairwise distances
between rows of 'x' and 'y', based on the specified method.
dvclMatrix Class
Description
An integer type matrix in the S4 vclMatrix
representation.
Value
If the vclMatrix object is of type 'double', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a double type matrix
Author(s)
Charles Determan Jr.
See Also
vclMatrix-class,
ivclMatrix-class,
fvclMatrix-class
dvclVector Class
Description
An double vector in the S4 vclVector
representation.
Value
If the vclVector object is of type 'double', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a double typed vector
Author(s)
Charles Determan Jr.
See Also
gpuMatrix Eigen Decomposition
Description
Computes the eigenvalues and eigenvectors for gpuMatrix objects.
Usage
## S4 method for signature 'gpuMatrix'
eigen(x, symmetric, only.values = FALSE, EISPACK = FALSE)
## S4 method for signature 'vclMatrix'
eigen(x, symmetric, only.values = FALSE, EISPACK = FALSE)
Arguments
x |
A gpuMatrix object |
symmetric |
logical indication if matrix is assumed to be symmetric. If not specified or FALSE, the matrix is inspected for symmetry |
only.values |
if TRUE, returns only eigenvalues (internals still currently calculate both regardless) |
EISPACK |
logical. Defunct and ignored |
Details
This function currently implements the qr_method function
from the ViennaCL library. As such, non-symmetric matrices are not
supported given that OpenCL does not have a 'complex' data type.
Neither the eigenvalues nor the eigenvectors are sorted as done in the base R eigen method.
Value
values |
A |
vectors |
A |
Note
The sign's may be different on some of the eigenvector elements. As noted in the base eigen documentation:
Recall that the eigenvectors are only defined up to a constant: even when the length is specified they are still only defined up to a scalar of modulus one (the sign for real matrices).
Therefore, although the signs may be different, the results are functionally equivalent
fgpuMatrix Class
Description
A float type matrix in the S4 gpuMatrix
representation.
Value
If the gpuMatrix object is of type 'float', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a float matrix.
Author(s)
Charles Determan Jr.
See Also
gpuMatrix-class,
igpuMatrix-class,
dgpuMatrix-class
fgpuVector Class
Description
An float vector in the S4 gpuVector
representation.
Value
If the gpuVector object is of type 'float', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a float typed vector
Author(s)
Charles Determan Jr.
See Also
fvclMatrix Class
Description
An integer type matrix in the S4 vclMatrix
representation.
Value
If the vclMatrix object is of type 'float', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a float matrix.
Author(s)
Charles Determan Jr.
See Also
vclMatrix-class,
ivclMatrix-class,
dvclMatrix-class
fvclVector Class
Description
An float vector in the S4 vclVector
representation.
Value
If the vclVector object is of type 'float', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a float typed vector
Author(s)
Charles Determan Jr.
See Also
Device Information
Description
Get basic information about selected device (e.g. GPU)
Usage
gpuInfo(device_idx = NULL, context_idx = currentContext())
cpuInfo(device_idx = NULL, context_idx = currentContext())
Arguments
device_idx |
An integer value indicating which device to query. |
context_idx |
An integer value indicating which context to query. |
Value
deviceName |
Device Name |
deviceVendor |
Device Vendor |
numberOfCores |
Number of Computing Units (which execute the work groups) |
maxWorkGroupSize |
Maximum number of work items per group |
maxWorkItemDim |
Number of dimensions |
maxWorkItemSizes |
Maximum number of works items per dimension |
deviceMemory |
Global amount of memory (bytes) |
clockFreq |
Maximum configured clock frequency of the device in MHz |
localMem |
Maximum amount of local memory for each work group (bytes) |
maxAllocatableMem |
Maximum amount of memory in a single piece (bytes) |
available |
Whether the device is available |
deviceExtensions |
OpenCL device extensions available |
double_support |
Logical value if double type supported |
Author(s)
Charles Determan Jr.
See Also
detectPlatforms detectGPUs detectCPUs cpuInfo
Construct a gpuMatrix
Description
Construct a gpuMatrix of a class that inherits
from gpuMatrix
Usage
gpuMatrix(data = NA, nrow = NA, ncol = NA, type = NULL, ...)
## S4 method for signature 'matrix'
gpuMatrix(data, type = NULL, ctx_id = NULL)
## S4 method for signature 'missing'
gpuMatrix(data, nrow = NA, ncol = NA, type = NULL, ctx_id = NULL)
## S4 method for signature 'numeric'
gpuMatrix(data, nrow, ncol, type = NULL, ctx_id = NULL)
## S4 method for signature 'integer'
gpuMatrix(data, nrow, ncol, type = NULL, ctx_id = NULL)
Arguments
data |
An object that is or can be converted to a
|
nrow |
An integer specifying the number of rows |
ncol |
An integer specifying the number of columns |
type |
A character string specifying the type of gpuMatrix. Default is NULL where type is inherited from the source data type. |
... |
Additional method to pass to gpuMatrix methods |
ctx_id |
An integer specifying the object's context |
Value
A gpuMatrix object
Author(s)
Charles Determan Jr.
gpuMatrix Class
Description
This is the 'mother' class for all gpuMatrix objects. It is essentially a wrapper for a basic R matrix (possibly to be improved). All other gpuMatrix classes inherit from this class but there are no current circumstances where this class is used directly.
There are multiple child classes that correspond
to the particular data type contained. These include
igpuMatrix, fgpuMatrix, and
dgpuMatrix corresponding to integer, float, and
double data types respectively.
Value
An object of class 'gpuMatrix' with the specified slots.
Slots
Common to all gpuMatrix objects in the package
address:Pointer to data matrix
.context_index:Integer index of OpenCL contexts
.platform_index:Integer index of OpenCL platforms
.platform:Name of OpenCL platform
.device_index:Integer index of active device
.device:Name of active device
Note
R does not contain a native float type. As such,
the matrix data within a fgpuMatrix-class
will be represented as double but downcast when any
gpuMatrix methods are used.
May also remove the type slot
Author(s)
Charles Determan Jr.
See Also
igpuMatrix-class,
fgpuMatrix-class,
dgpuMatrix-class
Construct a gpuVector
Description
Construct a gpuVector of a class that inherits
from gpuVector
Usage
gpuVector(data, length, type = NULL, ...)
## S4 method for signature 'vector,missing'
gpuVector(data, type = NULL, ctx_id = NULL)
## S4 method for signature 'missingOrNULL,ANY'
gpuVector(data, length, type = NULL, ctx_id = NULL)
## S4 method for signature 'numeric,numericOrInt'
gpuVector(data, length, type = NULL, ctx_id = NULL)
Arguments
data |
An object that is or can be converted to a
|
length |
A non-negative integer specifying the desired length. |
type |
A character string specifying the type of gpuVector. Default is NULL where type is inherited from the source data type. |
... |
Additional method to pass to gpuVector methods |
ctx_id |
An integer specifying the object's context |
Value
A gpuVector object
Author(s)
Charles Determan Jr.
gpuVector Class
Description
This is the 'mother' class for all gpuVector objects. All other gpuVector classes inherit from this class but there are no current circumstances where this class is used directly.
There are multiple child classes that correspond
to the particular data type contained. These include
igpuVector.
Value
An object of class 'gpuVector' with the specified slots.
Slots
Common to all vclMatrix objects in the package
address:Pointer to data matrix
.context_index:Integer index of OpenCL contexts
.platform_index:Integer index of OpenCL platforms
.platform:Name of OpenCL platform
.device_index:Integer index of active device
.device:Name of active device
Author(s)
Charles Determan Jr.
See Also
Skip test for CPUs
Description
Function to skip testthat tests if no valid CPU's are detected
Usage
has_cpu_skip()
Value
Returns nothing but gives a message if no CPU is available.
Skip test for GPU double precision
Description
Function to skip testthat tests if the detected GPU doesn't support double precision
Usage
has_double_skip()
Value
Returns nothing but gives a message if GPU doesn't support double precision.
Skip test for GPUs
Description
Function to skip testthat tests if no valid GPU's are detected
Usage
has_gpu_skip()
Value
Returns nothing but gives a message if no GPUs available.
Skip test for multiple GPUs with double precision
Description
Function to skip testthat tests if their aren't multiple detected GPU with double precision
Usage
has_multiple_double_skip()
Value
Returns nothing but gives a message if there are less than 2 GPUs with double precision.
Skip test in less than 2 GPUs
Description
Function to skip testthat tests if less than 2 valid GPU's are detected
Usage
has_multiple_gpu_skip()
Value
Returns nothing but gives a message if only one GPU is available.
Identity Matrix on Device
Description
Creates an identity matrix directly on the current device (e.g. GPU)
Usage
identity_matrix(x, type = NULL)
Arguments
x |
A numeric value indicating the order of the identity matrix |
type |
A character string specifying the type of gpuMatrix. Default
is derived from |
Value
A vclMatrix object
Note
This function was only created for vclMatrix objects as
the copy from CPU to gpuMatrix is trivial using the base
diag function.
Author(s)
Charles Determan Jr.
igpuMatrix Class
Description
An integer type matrix in the S4 gpuMatrix
representation.
Value
If the gpuMatrix object is of type 'integer', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a integer typed matrix
Author(s)
Charles Determan Jr.
See Also
gpuMatrix-class,
igpuMatrix-class,
dgpuMatrix-class
igpuVector Class
Description
An integer vector in the S4 gpuVector
representation.
Value
If the gpuVector object is of type 'integer', returns TRUE, if not, returns an error message.
Slots
address:An integer vector object
Author(s)
Charles Determan Jr.
See Also
Inplace Function Wrapper
Description
Applies the provided function in-place on the first object passed
Usage
inplace(f, x, y)
## S4 method for signature ''function',vclMatrix,vclMatrix'
inplace(f, x, y)
## S4 method for signature ''function',vclMatrix,missing'
inplace(f, x, y)
## S4 method for signature ''function',numeric,vclMatrix'
inplace(f, x, y)
## S4 method for signature ''function',vclMatrix,numeric'
inplace(f, x, y)
## S4 method for signature ''function',gpuMatrix,gpuMatrix'
inplace(f, x, y)
## S4 method for signature ''function',gpuMatrix,missing'
inplace(f, x, y)
## S4 method for signature ''function',numeric,gpuMatrix'
inplace(f, x, y)
## S4 method for signature ''function',gpuMatrix,numeric'
inplace(f, x, y)
## S4 method for signature ''function',vclVector,vclVector'
inplace(f, x, y)
## S4 method for signature ''function',vclVector,missing'
inplace(f, x, y)
## S4 method for signature ''function',vclVector,numeric'
inplace(f, x, y)
## S4 method for signature ''function',numeric,vclVector'
inplace(f, x, y)
## S4 method for signature ''function',gpuVector,gpuVector'
inplace(f, x, y)
## S4 method for signature ''function',gpuVector,missing'
inplace(f, x, y)
## S4 method for signature ''function',gpuVector,numeric'
inplace(f, x, y)
## S4 method for signature ''function',numeric,gpuVector'
inplace(f, x, y)
Arguments
f |
A function |
x |
A gpuR object |
y |
A gpuR object |
Value
No return, result applied in-place
Author(s)
Charles Determan Jr.
ivclMatrix Class
Description
An integer type matrix in the S4 vclMatrix
representation.
Value
If the vclMatrix object is of type 'integer', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a integer typed matrix
Author(s)
Charles Determan Jr.
See Also
vclMatrix-class,
ivclMatrix-class,
dvclMatrix-class
ivclVector Class
Description
An integer vector in the S4 vclVector
representation.
Value
If the vclVector object is of type 'integer', returns TRUE, if not, returns an error message.
Slots
address:An integer vector object
Author(s)
Charles Determan Jr.
See Also
gpuMatrix/vclMatrix length method
Description
Retrieve number of elements in object
Usage
## S4 method for signature 'gpuVector'
length(x)
## S4 method for signature 'vclMatrix'
length(x)
## S4 method for signature 'vclVector'
length(x)
## S4 method for signature 'gpuMatrix'
length(x)
Arguments
x |
A gpuMatrix/vclMatrix object |
Value
Length of the gpuVector object x.
The total number of elements in the vclMatrix object 'x'.
The length of the vclVector based on its data type.
A numeric value
Author(s)
Charles Determan Jr.
Available OpenCL Contexts
Description
Provide a data.frame of available OpenCL contexts and associated information.
Usage
listContexts()
Value
data.frame containing the following fields
context |
Integer identifying context |
platform |
Character string listing OpenCL platform |
platform_index |
Integer identifying platform |
device |
Character string listing device name |
device_index |
Integer identifying device |
device_type |
Character string labeling device (e.g. gpu) |
gpuR Logarithms and Exponentials
Description
log computes logarithms, by default natural logarithms
and log10 computes common (i.e. base 10) logarithms. The general form
log(x, base) computes logarithms with base base.
exp computes the exponential function.
Usage
## S4 method for signature 'gpuVector'
log(x, base = NULL)
## S4 method for signature 'vclMatrix'
log(x, base = NULL)
## S4 method for signature 'vclVector'
log(x, base = NULL)
## S4 method for signature 'gpuMatrix'
log(x, base = NULL)
Arguments
x |
A gpuR object |
base |
A positive number (complex not currently supported by OpenCL): the base with respect to which logarithms are computed. Defaults to the natural log. |
Value
A gpuVector object with the element-wise natural
logarithm of the elements of the input gpuVector object x.
A vclMatrix object.
A vclVector object.
A gpuR object of the same class as x
Compute the Norm of a Matrix
Description
Computes a matrix norm of x. The norm can be the one \("O"\) norm, the infinity \("I"\) norm, the Frobenius \("F"\) norm, the maximum modulus \("M"\) among elements of a matrix, or the “spectral” or "2"-norm, as determined by the value of type.
Usage
## S4 method for signature 'vclMatrix,character'
norm(x, type)
## S4 method for signature 'vclMatrix,missing'
norm(x, type)
## S4 method for signature 'gpuMatrix,character'
norm(x, type)
## S4 method for signature 'gpuMatrix,missing'
norm(x, type)
## S4 method for signature 'ANY,missing'
norm(x, type)
## S4 method for signature 'ANY,character'
norm(x, type)
Arguments
x |
A gpuR matrix object |
type |
character string, specifying the type of matrix norm to be computed. |
Value
The matrix norm, a non-negative number
Author(s)
Charles Determan Jr.
See Also
The Number of Rows/Columns of a gpuR matrix
Description
nrow and ncol return the number of rows or columns
present in x respectively.
Usage
## S4 method for signature 'vclMatrix'
nrow(x)
## S4 method for signature 'vclMatrix'
ncol(x)
## S4 method for signature 'gpuMatrix'
nrow(x)
## S4 method for signature 'gpuMatrix'
ncol(x)
Arguments
x |
A gpuMatrix/vclMatrix object |
Value
An integer.
An integer of length 1
Author(s)
Charles Determan Jr.
Permuting functions for gpuR objects
Description
Generate a permutation of row or column indices
Usage
permute(x, MARGIN, order)
Arguments
x |
A |
MARGIN |
dimension over which the ordering should be applied, 1 indicates rows, 2 indicates columns |
order |
An integer vector indicating order of rows to assign |
Value
A gpuR object
Author(s)
Charles Determan Jr.
OpenCL Platform Information
Description
Get basic information about the OpenCL platform
Usage
platformInfo(platform_idx = 1L)
Arguments
platform_idx |
An integer value to specify which platform to check |
Value
platformName |
Platform Name |
platformVendor |
Platform Vendor |
platformVersion |
Platform OpenCL Version |
platformExtensions |
Available platform extensions |
Author(s)
Charles Determan Jr.
Parallel Maxima and Minima
Description
pmax and pmin take one or more
vectors as arguments and return a single vector giving the 'parallel'
maxima (or minima) of the argument vectors
Usage
pmax(...)
pmin(...)
## S3 method for class 'vclVector'
pmin(..., na.rm = FALSE)
Arguments
... |
a vclVector object |
na.rm |
a logical indicating whether missing values should be removed. |
Value
A vclMatrix object.
See Also
POCL Version Check
Description
Versions of POCL up to 0.15-pre have a bug which results in values being returned when NA values should be (e.g. fractional powers of negative values)
Usage
pocl_check()
Value
Returns nothing but gives a message if the POCL version is too old.
S3 print for gpuMatrix objects
Description
prints a gpuMatrix object that is truncated to fit the screen
Usage
## S3 method for class 'gpuMatrix'
print(x, ..., n = NULL, width = NULL)
Arguments
x |
A gpuMatrix object |
... |
Additional arguments to print |
n |
Number of rows to display |
width |
Number of columns to display |
Value
This function prints a truncated summary of a gpuMatrix object 'x' to fit the screen
Reconstruct the Q or R Matrices from a gpuQR Object
Description
Returns the components of the QR decomposition.
Usage
## S4 method for signature 'gpuQR'
qr.R(qr, complete = FALSE)
## S4 method for signature 'gpuQR'
qr.Q(qr, complete = FALSE)
Arguments
qr |
|
complete |
not currently used |
Value
qr.Q returns all of Q,
qr.R returns all of R
Author(s)
Charles Determan Jr.
See Also
The QR Decomposition of a gpuR matrix
Description
qr computes the QR decomposition of a gpuR matrix
Usage
## S3 method for class 'gpuMatrix'
qr(x, ..., inplace = FALSE)
## S3 method for class 'vclMatrix'
qr(x, ..., inplace = FALSE)
Arguments
x |
A gpuR matrix |
... |
further arguments passed to or from other methods |
inplace |
Logical indicating if operations performed inplace |
Value
A list containing the QR decomposition of the matrix of class
gpuQR. The returned value is a list with the following components:
qr a matrix with the same dimensions as
x.betas vector of numeric values containing additional information of
qrfor extractingQandRmatrices.
Note
This an S3 generic of qr. The default continues to point to the default base function.
Furthermore, the list returned does not contain the exact same elements
as qr. The matrix storage format applied herein doesn't match
the base compact form. The method also doesn't return qraux,
rank, or pivot but instead returns betas
Author(s)
Charles Determan Jr.
See Also
Set Context
Description
Change the current context used by default
Usage
setContext(id = 1L)
Arguments
id |
Integer identifying which context to set |
Value
It does not return anything. This function is designed to set the current context to the specified context ID.
See Also
Set Context for Specific Device Type
Description
This function find the first context that contains a device of the specified type.
Usage
set_device_context(type)
Arguments
type |
A character vector specifying device type |
Value
An integer indicating previous context index
Setup OpenCL Arguments
Description
Generates a data.frame of argument definitions
for use in custom_opencl
Usage
setup_opencl(objects, intents, queues, kernel_maps = NULL)
Arguments
objects |
character vector of gpuR objects to be passed |
intents |
character vector specifying 'intent' of gpuR objects.
options include |
queues |
list of character vectors reflecting equal length to |
kernel_maps |
The corresponding arguments names in the provided OpenCL kernel corresponds to the gpuR objects passed and contains a character vector of which kernels the object will be enqueued. |
Value
A data.frame with columns:
object:The name of the gpuR object.
intents:The intent of the object, specified as 'IN', 'OUT', or 'INOUT'.
queues:A character vector reflecting equal length to
objects, where each element reflects a kernel function defined in an OpenCL kernel file.map:The corresponding arguments names in the provided OpenCL kernel corresponding to the gpuR objects passed.
Vector Slices
Description
This doesn't create a copy, it provides a child class that
points to a contiguous subvector of a gpuVector or
vclVector. Non-contiguous slices are currently not supported.
Usage
slice(object, start, end)
## S4 method for signature 'gpuVector,integer,integer'
slice(object, start, end)
## S4 method for signature 'vclVector,integer,integer'
slice(object, start, end)
Arguments
object |
A |
start |
An integer indicating the start of slice |
end |
An integer indicating the end of slice |
Details
This function allows a user to create a gpuR vector object that references a continuous subset of columns and rows of another gpuR vector object without a copy.
NOTE - this means that altering values in a vector slice object will alter values in the source vector.
Value
A gpuVectorSlice or vclVectorSlice object.
Author(s)
Charles Determan Jr.
Solve a System of Equations for gpuR objects
Description
This function solves the equation a %*% x = b for
x, where b can be either a vector or a matrix.
Usage
## S4 method for signature 'vclMatrix,vclMatrix'
solve(a, b, ...)
## S4 method for signature 'vclMatrix,missing'
solve(a, b, ...)
## S4 method for signature 'gpuMatrix,gpuMatrix'
solve(a, b, ...)
## S4 method for signature 'gpuMatrix,missing'
solve(a, b, ...)
Arguments
a |
A gpuR object |
b |
A gpuR object |
... |
further arguments passed to or from other methods |
Value
A gpuR object
Author(s)
Charles Determan Jr.
Singular Value Decomposition of a gpuR matrix
Description
Compute the singular-value decomposition of a gpuR matrix
Usage
svd(x, nu, nv, LINPACK)
## S3 method for class 'vclMatrix'
svd(x, nu, nv, LINPACK)
## S3 method for class 'gpuMatrix'
svd(x, nu, nv, LINPACK)
Arguments
x |
A gpuR matrix |
nu |
ignored |
nv |
ignored |
LINPACK |
ignored |
Value
The SVD decomposition of the matrix. The returned value is a list with the following components:
d a vector containing the singular values of
xu a matrix whose columns contain the left singular vectors of
x.v a matrix whose columns contain the right singular vectors of
x.
Note
This an S3 generic of svd. The default continues to point to the default base function.
Author(s)
Charles Determan Jr.
See Also
Synchronize Device Execution
Description
This pauses execution until the processing is complete on the device (CPU/GPU/etc.). This is important especially for benchmarking applications.
Usage
synchronize()
Value
This does not return anything.
Author(s)
Charles Determan Jr.
Examples
## Not run:
mat <- vclMatrix(rnorm(500^2), ncol = 500, nrow = 500)
system.time({mat %*% mat})
system.time({mat %*% mat; synchronize()})
## End(Not run)
gpuR matrix transpose
Description
Given a gpuR matrix x, t returns the
transpose of x
Usage
## S4 method for signature 'vclMatrix'
t(x)
## S4 method for signature 'gpuMatrix'
t(x)
Arguments
x |
A |
Value
The transpose of the input vclMatrix object.
A gpuR matrix
Author(s)
Charles Determan Jr.
vclMatrix Crossproduct
Description
Return the matrix cross-product of two conformable matrices using a GPU. This is equivalent to t(x) or x device and host is required.
Usage
## S4 method for signature 'gpuVector,gpuVector'
tcrossprod(x, y)
## S4 method for signature 'gpuVector,missing'
tcrossprod(x, y)
## S4 method for signature 'vclMatrix,missing'
crossprod(x, y)
## S4 method for signature 'vclMatrix,vclMatrix'
crossprod(x, y)
## S4 method for signature 'vclMatrix,matrix'
crossprod(x, y)
## S4 method for signature 'matrix,vclMatrix'
crossprod(x, y)
## S4 method for signature 'vclMatrix,vclVector'
crossprod(x, y)
## S4 method for signature 'vclVector,vclMatrix'
crossprod(x, y)
## S4 method for signature 'vclMatrix,missing'
tcrossprod(x, y)
## S4 method for signature 'vclMatrix,vclMatrix'
tcrossprod(x, y)
## S4 method for signature 'matrix,vclMatrix'
tcrossprod(x, y)
## S4 method for signature 'vclMatrix,matrix'
tcrossprod(x, y)
## S4 method for signature 'vclMatrix,vclVector'
tcrossprod(x, y)
## S4 method for signature 'vclVector,vclMatrix'
tcrossprod(x, y)
## S4 method for signature 'vclVector,vclVector'
tcrossprod(x, y)
## S4 method for signature 'vclVector,missing'
tcrossprod(x, y)
Arguments
x |
A |
y |
A |
Value
A vclMatrix object of the transpose of the outer
product of the two objects x and y.
Author(s)
Charles Determan Jr.
Get gpuR object type
Description
typeof determines the type (i.e. storage mode) of a
gpuR object
Usage
## S4 method for signature 'gpuMatrix'
typeof(x)
## S4 method for signature 'gpuVector'
typeof(x)
## S4 method for signature 'vclMatrix'
typeof(x)
## S4 method for signature 'vclVector'
typeof(x)
Arguments
x |
A gpuR object |
Value
A character string indicating the type of the gpuR object.
Author(s)
Charles Determan Jr.
Construct a vclMatrix
Description
Construct a vclMatrix of a class that inherits
from vclMatrix. This class points to memory directly on
the GPU to avoid the cost of data transfer between host and device.
Usage
vclMatrix(data = NA, nrow = NA, ncol = NA, type = NULL, ...)
## S4 method for signature 'matrix'
vclMatrix(data, type = NULL, ctx_id = NULL)
## S4 method for signature 'missing'
vclMatrix(data, nrow = NA, ncol = NA, type = NULL, ctx_id = NULL)
## S4 method for signature 'numeric'
vclMatrix(data, nrow, ncol, type = NULL, ctx_id = NULL)
## S4 method for signature 'integer'
vclMatrix(data, nrow, ncol, type = NULL, ctx_id = NULL)
Arguments
data |
An object that is or can be converted to a
|
nrow |
An integer specifying the number of rows |
ncol |
An integer specifying the number of columns |
type |
A character string specifying the type of vclMatrix. Default is NULL where type is inherited from the source data type. |
... |
Additional method to pass to vclMatrix methods |
ctx_id |
An integer specifying the object's context |
Value
A vclMatrix object
Author(s)
Charles Determan Jr.
vclMatrix Class
Description
This is the 'mother' class for all vclMatrix objects. These objects are pointers to viennacl matrices directly on the GPU. This will avoid the overhead of passing data back and forth between the host and device.
As such, any changes made to normal R 'copies' (e.g. A <- B) will be propogated to the parent object.
There are multiple child classes that correspond
to the particular data type contained. These include
ivclMatrix, fvclMatrix, and
dvclMatrix corresponding to integer, float, and
double data types respectively.
Value
An object of class 'vclMatrix' with the specified slots.
Slots
Common to all vclMatrix objects in the package
address:Pointer to data matrix
.context_index:Integer index of OpenCL contexts
.platform_index:Integer index of OpenCL platforms
.platform:Name of OpenCL platform
.device_index:Integer index of active device
.device:Name of active device
Note
R does not contain a native float type. As such,
the matrix data within a fvclMatrix-class
will be represented as double but downcast when any
vclMatrix methods are used.
May also remove the type slot
Author(s)
Charles Determan Jr.
See Also
ivclMatrix-class,
fvclMatrix-class,
dvclMatrix-class
Construct a vclVector
Description
Construct a vclVector of a class that inherits
from vclVector. This class points to memory directly on
the GPU to avoid the cost of data transfer between host and device.
Usage
vclVector(data, length, type = NULL, ...)
## S4 method for signature 'vector,missing'
vclVector(data, length, type = NULL, ctx_id = NULL)
## S4 method for signature 'missing,ANY'
vclVector(data, length, type = NULL, ctx_id = NULL)
## S4 method for signature 'numeric,numericOrInt'
vclVector(data, length, type = NULL, ctx_id = NULL)
## S4 method for signature 'vclMatrix,missing'
vclVector(
data,
length = NULL,
type = NULL,
ctx_id = NULL,
col = NULL,
row = NULL
)
Arguments
data |
An object that is or can be converted to a
|
length |
A non-negative integer specifying the desired length. |
type |
A character string specifying the type of vclVector. Default is NULL where type is inherited from the source data type. |
... |
Additional method to pass to vclVector methods |
ctx_id |
An integer specifying the object's context |
col |
index of column to extract from |
row |
index of row to extract from |
Value
A vclVector object
Author(s)
Charles Determan Jr.
vclVector Class
Description
This is the 'mother' class for all vclVector objects. All other vclVector classes inherit from this class but there are no current circumstances where this class is used directly.
There are multiple child classes that correspond
to the particular data type contained. These include
ivclVector.
Value
An object of class 'vclVector' with the specified slots.
Slots
Common to all vclMatrix objects in the package
address:Pointer to data matrix
.context_index:Integer index of OpenCL contexts
.platform_index:Integer index of OpenCL platforms
.platform:Name of OpenCL platform
.device_index:Integer index of active device
.device:Name of active device
Author(s)
Charles Determan Jr.
See Also
zgpuMatrix Class
Description
An complex double type matrix in the S4 gpuMatrix
representation.
Value
If the gpuMatrix object is of type 'complex double', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a complex double matrix.
Author(s)
Charles Determan Jr.
See Also
gpuMatrix-class,
igpuMatrix-class,
dgpuMatrix-class
zvclMatrix Class
Description
An complex double type matrix in the S4 vclMatrix
representation.
Value
If the vclMatrix object is of type 'complex double', returns TRUE, if not, returns an error message.
Slots
address:Pointer to a complex double type matrix
Author(s)
Charles Determan Jr.