nvtt
|
Structure containing all the input images from device memory. The image data is reordered by tiles. More...
#include <nvtt_lowlevel.h>
Public Member Functions | |
NVTT_API | GPUInputBuffer (const RefImage *images, ValueType value_type, int numImages=1, int tile_w=4, int tile_h=4, float WeightR=1.0f, float WeightG=1.0f, float WeightB=1.0f, float WeightA=1.0f, nvtt::TimingContext *tc=nullptr, unsigned *num_tiles=nullptr) |
Construct a GPUInputBuffer from 1 or more RefImage structs. | |
NVTT_API | GPUInputBuffer (const CPUInputBuffer &cpu_buf, int begin=0, int end=-1, nvtt::TimingContext *tc=nullptr) |
Construct a GPUInputBuffer from a CPUInputBuffer. | |
NVTT_API | ~GPUInputBuffer () |
Destructor. | |
NVTT_API int | NumTiles () const |
The total number of tiles of the input buffer. | |
NVTT_API void | TileSize (int &tile_w, int &tile_h) const |
Tile Size. | |
NVTT_API ValueType | Type () const |
Value type. The same as the input images of the CPUInputBuffer used for creating this object. | |
Public Attributes | |
Private * | m |
Structure containing all the input images from device memory. The image data is reordered by tiles.
NVTT_API nvtt::GPUInputBuffer::GPUInputBuffer | ( | const RefImage * | images, |
ValueType | value_type, | ||
int | numImages = 1, | ||
int | tile_w = 4, | ||
int | tile_h = 4, | ||
float | WeightR = 1.0f, | ||
float | WeightG = 1.0f, | ||
float | WeightB = 1.0f, | ||
float | WeightA = 1.0f, | ||
nvtt::TimingContext * | tc = nullptr, | ||
unsigned * | num_tiles = nullptr ) |
Construct a GPUInputBuffer from 1 or more RefImage structs.
The input images should use the same value type. images[i].data
should point to GPU global memory here (more specifically, a CUDA device pointer). num_tiles
can be used to return the number of tiles of each input image after reordering.
NVTT_API nvtt::GPUInputBuffer::GPUInputBuffer | ( | const CPUInputBuffer & | cpu_buf, |
int | begin = 0, | ||
int | end = -1, | ||
nvtt::TimingContext * | tc = nullptr ) |
Construct a GPUInputBuffer from a CPUInputBuffer.
begin
/end
can be used to sepcify a range of tiles to copy from the CPUInputBuffer. end = -1
means the end of the input buffer.