cuda_dxt.h File Reference

Interface of DXT compression CUDA implementation. More...

#include "cuda_wrapper.h"

Include dependency graph for cuda_dxt.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Functions
int	cuda_rgb_to_dxt1 (const void *src, void *out, int size_x, int size_y, cuda_wrapper_stream_t stream)
int	cuda_rgb_to_dxt6 (const void *src, void *out, int size_x, int size_y, cuda_wrapper_stream_t stream)
int	cuda_yuv422_to_yuv444 (const void *src, void *out, int pix_count, cuda_wrapper_stream_t str)
int	cuda_yuv_to_dxt1 (const void *src, void *out, int size_x, int size_y, cuda_wrapper_stream_t stream)
int	cuda_yuv_to_dxt6 (const void *src, void *out, int size_x, int size_y, cuda_wrapper_stream_t stream)

Detailed Description

Interface of DXT compression CUDA implementation.

Author

Martin Jirman jirma.nosp@m.n@ce.nosp@m.snet..nosp@m.cz Based on GLSL code by Martin Srom

Definition in file cuda_dxt.h.

Function Documentation

cuda_rgb_to_dxt1()

int cuda_rgb_to_dxt1	(	const void *	src,
		void *	out,
		int	size_x,
		int	size_y,
		cuda_wrapper_stream_t	stream )

CUDA DXT1 compression (only RGB without alpha).

Parameters

src	Pointer to top-left source pixel in device-memory buffer. 8bit RGB samples are expected (no alpha and no padding). (Pointer must be aligned to multiples of 16 bytes.)
out	Pointer to output buffer in device memory. (Must be aligned to multiples of 8 bytes.)
size_x	Width of the input image (must be divisible by 4).
size_y	Height of the input image (must be divisible by 4). (Input is read bottom up if negative)
stream	CUDA stream to run in, or 0 for default stream.

Returns

0 if OK, nonzero if failed.

CUDA DXT1 compression (only RGB without alpha).

Parameters

src	Pointer to top-left source pixel in device-memory buffer. 8bit RGB samples are expected (no alpha and no padding). (Pointer must be aligned to multiples of 16 bytes.)
out	Pointer to output buffer in device memory. (Must be aligned to multiples of 8 bytes.)
size_x	Width of the input image (must be divisible by 4).
size_y	Height of the input image (must be divisible by 4).
stream	CUDA stream to run in, or 0 for default stream.

Returns

0 if OK, nonzero if failed.

Definition at line 782 of file cuda_dxt.cu.

                                                                      {
    return dxt_launch<false, 1>(src, out, size_x, size_y, (cudaStream_t) stream);
}

References dxt_launch().

Referenced by main(), and worker_thread().

Here is the call graph for this function:

cuda_rgb_to_dxt6()

int cuda_rgb_to_dxt6	(	const void *	src,
		void *	out,
		int	size_x,
		int	size_y,
		cuda_wrapper_stream_t	stream )

CUDA DXT6 (DXT5-YcOcG) compression (only RGB without alpha).

Parameters

src	Pointer to top-left source pixel in device-memory buffer. 8bit RGB samples are expected (no alpha and no padding). (Pointer must be aligned to multiples of 16 bytes.)
out	Pointer to output buffer in device memory. (Must be aligned to multiples of 8 bytes.)
size_x	Width of the input image (must be divisible by 4).
size_y	Height of the input image (must be divisible by 4). (Input is read bottom up if negative)
stream	CUDA stream to run in, or 0 for default stream.

Returns

0 if OK, nonzero if failed.

Definition at line 816 of file cuda_dxt.cu.

                                                                      {
    return dxt_launch<false, 6>(src, out, size_x, size_y, (cudaStream_t) stream);
}

References dxt_launch().

Referenced by main(), and worker_thread().

Here is the call graph for this function:

cuda_yuv422_to_yuv444()

int cuda_yuv422_to_yuv444	(	const void *	src,
		void *	out,
		int	pix_count,
		cuda_wrapper_stream_t	str )

Definition at line 762 of file cuda_dxt.cu.

                                                          {
    // grid and threadblock sizes
    const dim3 tsiz(64, 1);
    int thread_count = pix_count / 4; // we process block of 4 pixels
    const dim3 gsiz((thread_count + tsiz.x - 1) / tsiz.x, 1);
    yuv422_to_yuv444_kernel<<<gsiz, tsiz, 0, (cudaStream_t) str>>>(src, out, pix_count);
    return cudaSuccess != cudaStreamSynchronize((cudaStream_t) str) ? -3 : 0;
}

References cudaSuccess, and yuv422_to_yuv444_kernel().

Here is the call graph for this function:

cuda_yuv_to_dxt1()

int cuda_yuv_to_dxt1	(	const void *	src,
		void *	out,
		int	size_x,
		int	size_y,
		cuda_wrapper_stream_t	stream )

CUDA DXT1 compression (only RGB without alpha). Converts input from YUV to RGB color space.

Parameters

src	Pointer to top-left source pixel in device-memory buffer. 8bit RGB samples are expected (no alpha and no padding). (Pointer must be aligned to multiples of 16 bytes.)
out	Pointer to output buffer in device memory. (Must be aligned to multiples of 8 bytes.)
size_x	Width of the input image (must be divisible by 4).
size_y	Height of the input image (must be divisible by 4). (Input is read bottom up if negative)
stream	CUDA stream to run in, or 0 for default stream.

Returns

0 if OK, nonzero if failed.

CUDA DXT1 compression (only RGB without alpha). Converts input from YUV to RGB color space.

Parameters

src	Pointer to top-left source pixel in device-memory buffer. 8bit RGB samples are expected (no alpha and no padding). (Pointer must be aligned to multiples of 16 bytes.)
out	Pointer to output buffer in device memory. (Must be aligned to multiples of 8 bytes.)
size_x	Width of the input image (must be divisible by 4).
size_y	Height of the input image (must be divisible by 4).
stream	CUDA stream to run in, or 0 for default stream.

Returns

0 if OK, nonzero if failed.

Definition at line 799 of file cuda_dxt.cu.

                                                                      {
    return dxt_launch<true, 1>(src, out, size_x, size_y, (cudaStream_t) stream);
}

References dxt_launch().

Here is the call graph for this function:

cuda_yuv_to_dxt6()

int cuda_yuv_to_dxt6	(	const void *	src,
		void *	out,
		int	size_x,
		int	size_y,
		cuda_wrapper_stream_t	stream )

Definition at line 821 of file cuda_dxt.cu.

                                                                      {
    return dxt_launch<true, 6>(src, out, size_x, size_y, (cudaStream_t) stream);
}

References dxt_launch().

Here is the call graph for this function: