--- /dev/null
+/*
+ Copyright (C) 2012 fmaj7b5.info
+
+ This program is free software: you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation, either version 2 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program. If not, see <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef FM7b5_DEVICE_MEMORY_CUH
+#define FM7b5_DEVICE_MEMORY_CUH
+
+#include <stdexcept>
+#include <type_traits>
+#include <cuda.h>
+
+#include "MDView.cuh"
+
+namespace FM7b5
+{
+ namespace memory
+ {
+ //! Linear 1-D memory
+ template <class T>
+ class Linear
+ {
+ public:
+ typedef T value_type;
+ typedef typename std::add_pointer<T>::type pointer_type;
+ typedef typename std::add_const<pointer_type>::type const_pointer_type;
+ typedef typename std::add_reference<T>::type reference_type;
+
+ Linear(const size_t count = 0)
+ : m_count(count)
+ {
+ m_extent = sizeof(value_type);
+ md_data = static_cast<pointer_type>(allocate(m_extent * count));
+ }
+
+ ~Linear()
+ {
+ deallocate(md_data);
+ }
+
+ void* allocate(const size_t byte)
+ {
+ void* p(nullptr);
+ cudaError_t status;
+
+ status = cudaMalloc(&p, byte);
+ if (status != cudaSuccess) {
+ throw std::bad_alloc(cudaGetErrorString(status));
+ }
+
+ return p;
+ }
+
+ void deallocate(void* p)
+ {
+ cudaFree(p);
+ }
+
+ void copy_from(const void* const src, const size_t byte)
+ {
+ cudaMemcpy(md_data, src, byte, cudaMemcpyHostToDevice);
+ }
+
+ void copy_to(void* const dst, const size_t byte)
+ {
+ cudaMemcpy(dst, md_data, byte, cudaMemcpyDeviceToHost);
+ }
+
+ size_t size() const { return m_extent * m_count; }
+ size_t count() const { return m_count; }
+ pointer_type data() const { return md_data; }
+
+ MDView<value_type, 1> ref() const
+ {
+ return MDView<value_type, 1>(md_data, &m_extent);
+ }
+
+ protected:
+ size_t m_extent;
+ size_t m_count;
+ pointer_type md_data;
+ };
+
+ //! Linear 2-D memory
+ template <class T>
+ class LinearPitch
+ {
+ public:
+ typedef T value_type;
+ typedef typename std::add_pointer<T>::type pointer_type;
+ typedef typename std::add_const<pointer_type>::type const_pointer_type;
+ typedef typename std::add_reference<T>::type reference_type;
+
+ LinearPitch(const size_t width, const size_t height)
+ : m_width(width), m_height(height)
+ {
+ md_data = static_cast<pointer_type>(allocate(width, height, &m_extent[0]));
+ m_extent[1] = sizeof(value_type);
+ }
+
+ ~LinearPitch()
+ {
+ deallocate(md_data);
+ }
+
+ void* allocate(const size_t width, const size_t height, size_t* pitch)
+ {
+ void* p(nullptr);
+ cudaError_t status;
+
+ status = cudaMallocPitch(&p, pitch, sizeof(value_type) * width, height);
+ if (status != cudaSuccess) {
+ throw std::bad_alloc(cudaGetErrorString(status));
+ }
+
+ return p;
+ }
+
+ void deallocate(void* p)
+ {
+ cudaFree(p);
+ }
+
+ void copy_from(const void* const src, const size_t width, const size_t height, size_t spitch = 0)
+ {
+ if (spitch < 1) {
+ spitch = width;
+ }
+ cudaMemcpy2D(md_data, pitch(), src, spitch, width, height, cudaMemcpyHostToDevice);
+ }
+
+ void copy_to(void* const dst, const size_t width, const size_t height, size_t dpitch = 0)
+ {
+ if (dpitch < 1) {
+ dpitch = width;
+ }
+ cudaMemcpy2D(dst, dpitch, md_data, pitch(), width, height, cudaMemcpyDeviceToHost);
+ }
+
+ size_t size() const { return m_extent[1] * m_height; }
+ size_t width() const { return m_width; }
+ size_t height() const { return m_height; }
+ size_t pitch() const { return m_extent[0]; }
+ pointer_type data() const { return md_data; }
+
+ MDView<value_type, 2> ref() const
+ {
+ return MDView<value_type, 2>(md_data, m_extent);
+ }
+
+ protected:
+ size_t m_width;
+ size_t m_height;
+ size_t m_extent[2];
+ pointer_type md_data;
+ };
+ }
+}
+
+#endif /* FM7b5_DEVICE_MEMORY_CUH */
projects / cuda.git / blobdiff