Remove scratch folders.

.gitignore

@@ -2,10 +2,15 @@
 *.pyc
 
 # Setuptools distribution folder
+/acc/
 /dist/
 /datasets_big/
 /benchmark/
 /try/
+/develop/
+/develop_HtH/
+/develop_os/
+/develop_solver/
 
 # HTML documentation
 _build/

acc/__init__.py

@@ -1,8 +0,0 @@
-# -*- coding: utf-8 -*-
-"""
-Created on Sat Feb 21 20:46:31 2015
-
-@author: akusok
-"""
-
-#from gpu_solver import GPUSolver, gpu_solve

acc/cuda_code (copy).cpp

@@ -1,152 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <cuda_runtime.h>
-#include "cublas_v2.h"
-
-// H_(N,L), T_(N,C), HH(L,L), HT(L,C)
-
-class CudaSolver {
-        int L, C;
-        double* devHH;
-        double* devHT;
-        double* one;
-        cublasHandle_t handle;
-    public:
-        CudaSolver( int, int, double*, double* );
-        void get_corr( double* HH, double* HT );
-        void add_data( int N, double* H, double* T );
-        void finalize();
-};
-
-
-// init a zero matrix on GPU to store X'*X, add normalization
-CudaSolver::CudaSolver ( int nL, int nC, double* HH, double* HT ) {
-    cudaError_t cudaStat;
-    cublasStatus_t stat;
-    double* ones;
-
-    L = nL;
-    C = nC;
-
-
-    cudaStat = cudaMalloc((void**)&devHH, L*L*sizeof(*HH));
-    if (cudaStat != cudaSuccess) { printf ("devHH device memory allocation failed"); }
-    fprintf(stdout, "Allocating devHH success\n");
-
-    cudaStat = cudaMalloc((void**)&devHT, L*C*sizeof(*HT));
-    if (cudaStat != cudaSuccess) { printf ("devHT device memory allocation failed"); }
-    fprintf(stdout, "Allocating devHT success\n");
-
-    stat = cublasCreate(&handle);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("CUBLAS initialization failed\n"); }
-    fprintf(stdout, "CUBLAS initialization success\n");
-
-    stat = cublasSetPointerMode(handle, CUBLAS_POINTER_MODE_DEVICE);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("Setting pointer mode failed"); }
-    fprintf(stdout, "Setting pointer mode success\n");
-
-    stat = cublasSetMatrix (L, L, sizeof(*HH), HH, L, devHH, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devHH upload failed"); }
-    fprintf(stdout, "devHH upload success\n");
-
-    stat = cublasSetMatrix (C, L, sizeof(*HT), HT, C, devHT, C);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devHT upload failed"); }
-    fprintf(stdout, "devHT upload success\n");
-
-
-    ones = (double*) malloc (1*sizeof(double));
-    ones[0] = 1.0;
-    cudaMalloc((void**)&one, 1*sizeof(double));
-    cublasSetVector(1, sizeof(*ones), ones, 1, one, 1);
-    fprintf(stdout, "here\n");
-
-};
-
-
-// update covariance matrices with new data
-void CudaSolver::add_data ( int N, double* H, double* T ) {
-    cudaError_t cudaStat;
-    cublasStatus_t stat;
-    double* devH;
-    double* devT;
-
-    cudaStat = cudaMalloc((void**)&devH, N*L*sizeof(*H));
-    if (cudaStat != cudaSuccess) { printf ("devH device memory allocation failed"); goto cleanup; }
-    fprintf(stdout, "Allocating devH success\n");
-
-    cudaStat = cudaMalloc((void**)&devT, N*C*sizeof(*T));
-    if (cudaStat != cudaSuccess) { printf ("devT device memory allocation failed"); goto cleanup; }
-    fprintf(stdout, "Allocating devT success\n");
-
-    stat = cublasSetMatrix (N, L, sizeof(*H), H, N, devH, N);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devH upload failed"); }
-    fprintf(stdout, "devH upload success\n");
-
-    stat = cublasSetMatrix (N, C, sizeof(*T), T, N, devT, N);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devT upload failed"); }
-    fprintf(stdout, "devT upload success\n");
-
-
-    stat = cublasDsyrk(handle, CUBLAS_FILL_MODE_UPPER, CUBLAS_OP_N,
-                       L, N, one, devH, L, one, devHH, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("H'H update failed"); }
-    fprintf(stdout, "H'H update success\n");
-
-//
-//    stat = cublasDgemm(handle, CUBLAS_OP_N, CUBLAS_OP_T,
-//                       L, C, N,
-//                       one, devH,  L,//L
-//                            devT,  C,//C
-//                       one, devHT, L);
-
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("H'T update failed\n"); }
-    fprintf(stdout, "H'T update success\n");
-
-
-   cleanup:
-    cudaFree (devH);
-    fprintf(stdout, "devH freed success\n");
-    cudaFree (devT);
-    fprintf(stdout, "devT freed success\n");
-
-
-};
-
-
-// return current covariance matrices
-void CudaSolver::get_corr ( double* HH, double* HT ) {
-    cublasStatus_t stat;
-
-    stat = cublasGetMatrix (L, L, sizeof(*HH), devHH, L, HH, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devHH download failed"); }   
-
-    stat = cublasGetMatrix (L, C, sizeof(*HT), devHT, L, HT, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devHT download failed"); }   
-};
-
-
-// free memory
-void CudaSolver::finalize( ) {
-    fprintf(stdout, "Solver finalized\n");
-    cudaFree (devHH);
-    cudaFree (devHT);
-    cudaFree (one);
-    cublasDestroy(handle);
-}
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-

acc/cuda_code.cpp

@@ -1,138 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <cuda_runtime.h>
-#include "cublas_v2.h"
-
-// H_(N,L), T_(N,C), HH(L,L), HT(L,C)
-
-class CudaSolver {
-        int L, C;
-        double* devHH;
-        double* devHT;
-        double* one;
-        double* ones;
-        cublasHandle_t handle;
-    public:
-        CudaSolver( int, int, double*, double* );
-        void get_corr( double* HH, double* HT );
-        void add_data( int N, double* H, double* T );
-        void finalize();
-};
-
-
-// init a zero matrix on GPU to store X'*X, add normalization
-CudaSolver::CudaSolver ( int nL, int nC, double* HH, double* HT ) {
-    cudaError_t cudaStat;
-    cublasStatus_t stat;
-
-    L = nL;
-    C = nC;
-
-
-    cudaStat = cudaMalloc((void**)&devHH, L*L*sizeof(*HH));
-    if (cudaStat != cudaSuccess) { printf ("devHH device memory allocation failed"); }
-    fprintf(stdout, "Allocating devHH success\n");
-
-    cudaStat = cudaMalloc((void**)&devHT, L*C*sizeof(*HT));
-    if (cudaStat != cudaSuccess) { printf ("devHT device memory allocation failed"); }
-    fprintf(stdout, "Allocating devHT success\n");
-
-    stat = cublasCreate(&handle);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("CUBLAS initialization failed\n"); }
-    fprintf(stdout, "CUBLAS initialization success\n");
-
-    stat = cublasSetMatrix (L, L, sizeof(*HH), HH, L, devHH, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devHH upload failed"); }
-    fprintf(stdout, "devHH upload success\n");
-
-    stat = cublasSetMatrix (C, L, sizeof(*HT), HT, C, devHT, C);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devHT upload failed"); }
-    fprintf(stdout, "devHT upload success\n");
-};
-
-
-// update covariance matrices with new data
-void CudaSolver::add_data ( int N, double* H, double* T ) {
-    cudaError_t cudaStat;
-    cublasStatus_t stat;
-    double* devH;
-    double* devT;
-    double one = 1.0;
-
-    cudaStat = cudaMalloc((void**)&devH, N*L*sizeof(*H));
-    if (cudaStat != cudaSuccess) { printf ("devH device memory allocation failed"); goto cleanup1; }
-//    fprintf(stdout, "Allocating devH success\n");
-
-    cudaStat = cudaMalloc((void**)&devT, N*C*sizeof(*T));
-    if (cudaStat != cudaSuccess) { printf ("devT device memory allocation failed"); goto cleanup2; }
-//    fprintf(stdout, "Allocating devT success\n");
-
-    stat = cublasSetMatrix (N, L, sizeof(*H), H, N, devH, N);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devH upload failed"); }
-//    fprintf(stdout, "devH upload success\n");
-
-    stat = cublasSetMatrix (N, C, sizeof(*T), T, N, devT, N);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devT upload failed"); }
-//    fprintf(stdout, "devT upload success\n");
-
-
-    stat = cublasDsyrk(handle, CUBLAS_FILL_MODE_UPPER, CUBLAS_OP_T,
-                       L, N, &one, devH, N, &one, devHH, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("H'H update failed\n"); }
-//    fprintf(stdout, "H'H update success\n");
-
-    stat = cublasDgemm(handle, CUBLAS_OP_T, CUBLAS_OP_N,
-                       L, C, N,
-                       &one, devH,  N,
-                            devT,  N,
-                       &one, devHT, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("H'T update failed\n"); }
-//    fprintf(stdout, "H'T update success\n");
-
-
-   cleanup2:
-    cudaFree (devT);
-//    fprintf(stdout, "devT freed success\n");
-   cleanup1:
-    cudaFree (devH);
-//    fprintf(stdout, "devH freed success\n");
-
-};
-
-
-// return current covariance matrices
-void CudaSolver::get_corr ( double* HH, double* HT ) {
-    cublasStatus_t stat;
-
-    stat = cublasGetMatrix (L, L, sizeof(*HH), devHH, L, HH, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devHH download failed"); }   
-
-    stat = cublasGetMatrix (L, C, sizeof(*HT), devHT, L, HT, L);
-    if (stat != CUBLAS_STATUS_SUCCESS) { printf ("devHT download failed"); }   
-};
-
-
-// free memory
-void CudaSolver::finalize( ) {
-    fprintf(stdout, "Solver finalized\n");
-    cudaFree (devHH);
-    cudaFree (devHT);
-    cublasDestroy(handle);
-}
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-

acc/cuda_code.h

@@ -1,12 +0,0 @@
-#include <stdio.h>
-#include <stdlib.h>
-#include <cuda_runtime.h>
-#include "cublas_v2.h"
-
-class CudaSolver {
-    public:
-        CudaSolver( int, int, double* HH, double* HT );
-        void get_corr( double* HH, double* HT );
-        void add_data( int N, double* H, double* T );
-        void finalize();
-};

acc/cuda_solver.c

@@ -1 +0,0 @@
-#error Do not use this file, it is the result of a failed Cython compilation.