-
Notifications
You must be signed in to change notification settings - Fork 2
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
120 changed files
with
5,897 additions
and
564 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,2 +1,42 @@ | ||
| # cucheb - CUDA accelerated large sparse eigensolvers # | ||
| This is a package written in C++ for accelerating large sparse eigenvalue computations using NVIDIA GPUs. | ||
| Jared L. Aurentz and Vasileios Kalantzis, October 2015 | ||
|
|
||
| ## Introduction ## | ||
| __cucheb__ is a collection of C++ subroutines for accurately and | ||
| efficiently solving large sparse matrix eigenvalue problems using | ||
| NVIDIA brand GPUs. These methods are well suited for computing | ||
| eigenvalues of 2D/3D discretization matrices that arise in | ||
| elliptic and parabolic PDEs. | ||
|
|
||
| ### Current features ### | ||
| __cucheb-v0.1.0__ has the following features: | ||
| - double precision eigensolvers for real symmetric matrices | ||
|
|
||
| ## Installation ## | ||
| __cucheb__ is built on top of the [NVIDIA CUDA Toolkit](https://developer.nvidia.com/cuda-toolkit) | ||
| and a small number of C++ standard libraries. You must have the toolkit | ||
| installed before the library can be built. | ||
|
|
||
| ### Linux ### | ||
| To install on a Linux machine simply move into the __cucheb__ root directory, | ||
| edit the file __make.inc__ to suit your system and type: | ||
| ``` | ||
| make install | ||
| ``` | ||
| This creates a shared object library __libcucheb.so._version___ and copies | ||
| it into the user specified installation directory. The installation does not | ||
| create any symbolic links or export any library paths. | ||
|
|
||
| ## Removing cucheb ## | ||
| If the source directory has not been removed simply move into the __cucheb__ | ||
| root directory and type: | ||
| ``` | ||
| make uninstall | ||
| ``` | ||
| If the source directory has been removed the install directory will have to | ||
| be removed explicitly by the user. | ||
|
|
||
| ## Questions and issues ## | ||
| If you have any questions or encounter any issues while using __cucheb__ | ||
| please file an issue on the [__cucheb__ issues](https://github.com/jaurentz | ||
| /cucheb/issues) page of Github. |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,35 +1,186 @@ | ||
| /** \mainpage My Personal Index Page | ||
| * | ||
| * \section intro_sec Introduction | ||
| * | ||
| * This is the introduction. | ||
| * | ||
| * \section install_sec Installation | ||
| * | ||
| * \subsection step1 Step 1: Opening the box | ||
| * | ||
| * etc... | ||
| */ | ||
|
|
||
| #ifndef __cucheb_h__ /* __cucheb_h__ */ | ||
| #define __cucheb_h__ | ||
| #include <cuchebdependencies.h> | ||
|
|
||
| #include <iostream> | ||
| #include <string> | ||
| #include <sstream> | ||
| #include <omp.h> | ||
| //#include <cula_lapack.h> | ||
| //#include <cula_lapack_device.h> | ||
| #include <lapacke.h> | ||
| using namespace std; | ||
| #include <cucheberror.h> | ||
| #include <cuchebstats.h> | ||
| #include <cuchebpoly.h> | ||
| #include <cuchebop.h> | ||
| #include <cuchebsolve.h> | ||
| #include <cuchebeigs.h> | ||
| #include <cuchebmatrix.h> | ||
| #include <cucheblanczos.h> | ||
|
|
||
| /* header file for cucheb data type */ | ||
| #ifndef __cucheb_h__ | ||
| #define __cucheb_h__ | ||
|
|
||
|
|
||
|
|
||
| /* cuchebutils subroutines */ | ||
| /* rotation generator */ | ||
| int cuchebutils_rotation(const double a, const double b, double* c, double* s, | ||
| double* nrm); | ||
|
|
||
| /* function to perform banded symmetric bulge chase */ | ||
| int cuchebutils_chasebulge(int n, int bwidth, double* bands, int ldbands, | ||
| double* bulge, double* vecs, int ldvecs); | ||
|
|
||
| /* eigenvalues and eigenvectors of 2x2 symmetric matrix */ | ||
| int cuchebutils_2x2symeig(double a1, double a2, double b, double* e1, double* e2, | ||
| double* c, double* s); | ||
|
|
||
| /* reduce banded symmetric matrix to tridiagonal */ | ||
| int cuchebutils_bandsymred(int n, int bwidth, double* bands, int ldbands, | ||
| double* vecs, int ldvecs); | ||
|
|
||
| /* eigenvalues and eigenvectors of banded symmetric matrix via QR */ | ||
| int cuchebutils_bandsymqr(int n, int bwidth, double* bands, int ldbands, | ||
| double* evals, double* vecs, int ldvecs); | ||
|
|
||
|
|
||
|
|
||
| /* cuchebstats subroutines */ | ||
| /* standard print cuchebstats object */ | ||
| int cuchebstats_print(cuchebstats* ccs); | ||
|
|
||
| /* print cuchebstats objects to file */ | ||
| int cuchebstats_fileprint(const string& fname, int nummats, string* matnames, | ||
| cuchebstats* ccstats); | ||
|
|
||
|
|
||
|
|
||
| /* cuchebpoly subroutines */ | ||
| /* instantiate cuchebpoly object */ | ||
| int cuchebpoly_init(cuchebpoly* ccp); | ||
|
|
||
| /* destroy cuchebpoly object */ | ||
| int cuchebpoly_destroy(cuchebpoly* ccp); | ||
|
|
||
| /* standard print cuchebpoly object */ | ||
| int cuchebpoly_print(cuchebpoly* ccp); | ||
|
|
||
| /* long print cuchebpoly object */ | ||
| int cuchebpoly_printlong(cuchebpoly* ccp); | ||
|
|
||
| /* second kind Chebyshev points */ | ||
| int cuchebpoly_points(double a, double b, cuchebpoly* ccp); | ||
|
|
||
| /* convert values to coefficients */ | ||
| int cuchebpoly_coeffs(cuchebpoly* ccp); | ||
|
|
||
| /* evaluate poly at scalar */ | ||
| double cuchebpoly_clenshaw(cuchebpoly* ccp, double x); | ||
|
|
||
| /* threshold coefficients */ | ||
| int cuchebpoly_chop(cuchebpoly* ccp); | ||
|
|
||
| /* routine for creating point filter */ | ||
| int cuchebpoly_pointfilter(double a, double b, double rho, int order, cuchebpoly* ccp); | ||
|
|
||
| /* routine for creating step filter */ | ||
| int cuchebpoly_stepfilter(double a, double b, double c, double d, int order, cuchebpoly* ccp); | ||
|
|
||
| /* routine for creating smart filter */ | ||
| int cuchebpoly_smartfilter(double a, double b, double c, double d, cuchebpoly* ccp); | ||
|
|
||
| /* routine for creating gaussian filter */ | ||
| int cuchebpoly_gaussianfilter(double a, double b, double rho, double tau, cuchebpoly* ccp); | ||
|
|
||
|
|
||
|
|
||
| /* cuchebmatrix subroutines */ | ||
| /* instantiate cuchebmatrix object */ | ||
| int cuchebmatrix_init(const string& mtxfile, cuchebmatrix* ccm); | ||
|
|
||
| /* destroy cuchebmatrix object */ | ||
| int cuchebmatrix_destroy(cuchebmatrix* ccm); | ||
|
|
||
| /* print cuchebmatrix object */ | ||
| int cuchebmatrix_print(cuchebmatrix* ccm); | ||
|
|
||
| /* longprint cuchebmatrix object */ | ||
| int cuchebmatrix_printlong(cuchebmatrix* ccm); | ||
|
|
||
| /* gpuprint cuchebmatrix object */ | ||
| int cuchebmatrix_gpuprint(cuchebmatrix* ccm); | ||
|
|
||
| /* routine for sorting entries */ | ||
| int cuchebmatrix_sort(cuchebmatrix* ccm); | ||
|
|
||
| /* routine for converting to csr format */ | ||
| int cuchebmatrix_csr(cuchebmatrix* ccm); | ||
|
|
||
| /* routine for mv multiply on GPU */ | ||
| int cuchebmatrix_mv(cuchebmatrix* ccm, double* alpha, double* x, double* beta, | ||
| double* y); | ||
|
|
||
| /* routine for poly mv multiply on GPU */ | ||
| int cuchebmatrix_polymv(cuchebmatrix* ccm, cuchebpoly* ccp, double* x, double* y); | ||
|
|
||
| /* routine for estimating spectral interval */ | ||
| int cuchebmatrix_specint(cuchebmatrix* ccm); | ||
|
|
||
| /* routine for estimating spectral interval */ | ||
| int cuchebmatrix_specint(cuchebmatrix* ccm, cucheblanczos *ccl); | ||
|
|
||
| /* lanczos routine for intervals */ | ||
| int cuchebmatrix_lanczos(double lbnd, double ubnd, | ||
| int bsize, int numvecs, int stepsize, | ||
| cuchebmatrix* ccm, cucheblanczos* ccl); | ||
|
|
||
| /* lanczos routine for intervals with statistics variable */ | ||
| int cuchebmatrix_lanczos(double lbnd, double ubnd, | ||
| int bsize, int numvecs, int stepsize, | ||
| cuchebmatrix* ccm, cucheblanczos* ccl, | ||
| cuchebstats* ccstats); | ||
|
|
||
| /* filtered lanczos routine for interval */ | ||
| int cuchebmatrix_filteredlanczos(double lbnd, double ubnd, int bsize, | ||
| cuchebmatrix* ccm, cucheblanczos* ccl); | ||
|
|
||
| /* same routine as above but with statistics variable */ | ||
| int cuchebmatrix_filteredlanczos(double lbnd, double ubnd, int bsize, | ||
| cuchebmatrix* ccm, cucheblanczos* ccl, | ||
| cuchebstats* ccs); | ||
|
|
||
| /* expert lanczos routine for intervals */ | ||
| int cuchebmatrix_expertlanczos(double lbnd, double ubnd, int degree, | ||
| int bsize, int numvecs, int stepsize, | ||
| cuchebmatrix* ccm, cucheblanczos* ccl, | ||
| cuchebstats* ccstats); | ||
|
|
||
|
|
||
| /* cucheblanczos subroutines */ | ||
| /* instantiate cucheblanczos object */ | ||
| int cucheblanczos_init(int bsize, int numvecs, cuchebmatrix* ccm, cucheblanczos* ccl); | ||
|
|
||
| /* destroy cucheblanczos object */ | ||
| int cucheblanczos_destroy(cucheblanczos* ccl); | ||
|
|
||
| /* print cucheblanczos object */ | ||
| int cucheblanczos_print(cucheblanczos* ccl); | ||
|
|
||
| /* set cucheblanczos starting vectors */ | ||
| int cucheblanczos_startvecs(cucheblanczos* ccl); | ||
|
|
||
| /* arnoldi run using cuchebmatrix */ | ||
| int cucheblanczos_arnoldi(int nsteps, cuchebmatrix* ccm, cucheblanczos* ccl, | ||
| cuchebstats* ccstats); | ||
|
|
||
| /* filtered arnoldi run using cuchebmatrix */ | ||
| int cucheblanczos_filteredarnoldi(int nsteps, cuchebmatrix* ccm, cuchebpoly* ccp, | ||
| cucheblanczos* ccl, cuchebstats* ccstats); | ||
|
|
||
| /* compute ritz values */ | ||
| int cucheblanczos_ritz(cuchebmatrix* ccm, cucheblanczos* ccl); | ||
|
|
||
| /* compute rayleigh quotients */ | ||
| int cucheblanczos_rayleigh(cuchebmatrix* ccm, cucheblanczos* ccl); | ||
|
|
||
| /* sort evals in interval */ | ||
| int cucheblanczos_sort(double lb, double ub, cucheblanczos* ccl); | ||
|
|
||
| /* sort evals by largest modulus */ | ||
| int cucheblanczos_sort(cucheblanczos* ccl); | ||
|
|
||
| /* check convergence */ | ||
| int cucheblanczos_checkconvergence(cucheblanczos* ccl); | ||
|
|
||
| cuchebStatus_t cuchebSetGridBlocks(int n, dim3 *blockSize, dim3 *gridSize); | ||
|
|
||
| cuchebStatus_t cuchebDinit(int n,double *x,int incx,double val); | ||
|
|
||
| #endif /* __cucheb_h__ */ |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,18 @@ | ||
| #include <stdio.h> | ||
| #include <stdlib.h> | ||
| #include <math.h> | ||
| #include <string.h> | ||
| #include <iostream> | ||
| #include <fstream> | ||
| #include <sstream> | ||
| #include <iomanip> | ||
| #include <algorithm> | ||
| #include <vector> | ||
| using namespace std; | ||
|
|
||
| #include <mmio.h> | ||
| #include <cuda.h> | ||
| #include <cublas_v2.h> | ||
| #include <cusparse.h> | ||
| #include <cufft.h> | ||
|
|
Oops, something went wrong.