@ -52,16 +52,20 @@ using namespace nvtt;
extern " C " void setupCompressKernel ( const float weights [ 3 ] ) ; extern " C " void setupCompressKernel ( const float weights [ 3 ] ) ;
extern " C " void compressKernelDXT1 ( uint blockNum , uint * d_data , uint * d_result , uint * d_bitmaps ) ; extern " C " void bindTextureToArray ( cudaArray * d_data ) ;
extern " C " void compressKernelDXT1_Tex ( uint bn , uint blockNum , uint w , cudaArray * d_data , uint * d_result , uint * d_bitmaps ) ;
extern " C " void compressKernelDXT1 ( uint firstBlock , uint blockNum , uint w , uint * d_result , uint * d_bitmaps ) ;
extern " C " void compressKernelDXT1_Level4 ( uint blockNum , uint * d_data , uint * d_result , uint * d_bitmaps ) ; extern " C " void compressKernelDXT1_Level4 ( uint blockNum , uint * d_data , uint * d_result , uint * d_bitmaps ) ;
extern " C " void compressWeightedKernelDXT1 ( uint blockNum , uint * d_data , uint * d_result , uint * d_bitmaps ) ; extern " C " void compressWeightedKernelDXT1 ( uint blockNum , uint * d_data , uint * d_result , uint * d_bitmaps ) ;
extern " C " void compressNormalKernelDXT1 ( uint blockNum , uint * d_data , uint * d_result , uint * d_bitmaps ) ; extern " C " void compressKernelDXT3 ( uint firstBlock , uint blockNum , uint w , uint * d_result , uint * d_bitmaps ) ;
extern " C " void compressKernelCTX1 ( uint blockNum , uint * d_data , uint * d_result , uint * d_bitmaps ) ; //extern "C" void compressNormalKernelDXT1(uint blockNum, uint * d_data, uint * d_result, uint * d_bitmaps);
//extern "C" void compressKernelCTX1(uint blockNum, uint * d_data, uint * d_result, uint * d_bitmaps);
# include "Bitmaps.h" // @@ Rename to BitmapTable.h # pragma message(NV_FILE_LINE "TODO: Rename Bitmaps.h to BitmapTable.h")
# include "Bitmaps.h"
/*
// Convert linear image to block linear.
// Convert linear image to block linear.
static void convertToBlockLinear ( const Image * image , uint * blockLinearImage ) static void convertToBlockLinear ( const Image * image , uint * blockLinearImage )
{ {
@ -81,45 +85,49 @@ static void convertToBlockLinear(const Image * image, uint * blockLinearImage)
}
}
}
}
} }
*/
# endif # endif
CudaCompressor : : CudaCompressor ( ) : m_bitmapTable ( NULL ) , m_bitmapTableCTX ( NULL ) , m_data ( NULL ) , m_result ( NULL ) CudaContext : : CudaContext ( ) :
{ bitmapTable ( NULL ) ,
bitmapTableCTX ( NULL ) ,
data ( NULL ) ,
result ( NULL )
{
# if defined HAVE_CUDA # if defined HAVE_CUDA
// Allocate and upload bitmaps.
// Allocate and upload bitmaps.
cudaMalloc ( ( void * * ) & m_bitmapTable , 992 * sizeof ( uint ) ) ;
cudaMalloc ( ( void * * ) & , 992 * sizeof ( uint ) ) ;
if ( m_bitmapTable ! = NULL )
if ( ! = NULL )
{
{
cudaMemcpy ( m_ bitmapTable, s_bitmapTable , 992 * sizeof ( uint ) , cudaMemcpyHostToDevice ) ;
cudaMemcpy ( , s_bitmapTable , 992 * sizeof ( uint ) , cudaMemcpyHostToDevice ) ;
}
}
cudaMalloc ( ( void * * ) & m_bitmapTableCTX , 704 * sizeof ( uint ) ) ;
cudaMalloc ( ( void * * ) & bitmapTableCTX , 704 * sizeof ( uint ) ) ;
if ( bitmapTableCTX ! = NULL )
if ( m_bitmapTableCTX ! = NULL )
{
{
cudaMemcpy ( m_ bitmapTableCTX, s_bitmapTableCTX , 704 * sizeof ( uint ) , cudaMemcpyHostToDevice ) ;
cudaMemcpy ( , s_bitmapTableCTX , 704 * sizeof ( uint ) , cudaMemcpyHostToDevice ) ;
}
}
// Allocate scratch buffers.
// Allocate scratch buffers.
cudaMalloc ( ( void * * ) & m_ data, MAX_BLOCKS * 64U ) ;
cudaMalloc ( ( void * * ) & , MAX_BLOCKS * 64U ) ;
cudaMalloc ( ( void * * ) & m_ result, MAX_BLOCKS * 8U ) ;
cudaMalloc ( ( void * * ) & , MAX_BLOCKS * 8U ) ;
# endif # endif
} }
CudaCo mpressor: : ~ CudaCompressor ( ) CudaCo ntext: : ~ CudaContext ( )
{ {
# if defined HAVE_CUDA # if defined HAVE_CUDA
// Free device mem allocations.
// Free device mem allocations.
cudaFree ( m_data ) ;
cudaFree ( bitmapTableCTX ) ;
cudaFree ( m_result ) ;
cudaFree ( bitmapTable ) ;
cudaFree ( m_bitmapTable ) ;
cudaFree ( data ) ;
cudaFree ( m_bitmapTableCTX ) ;
cudaFree ( result ) ;
# endif # endif
} }
bool CudaCo mpressor : : isValid ( ) const bool CudaCo ntext : : isValid ( ) const
{ {
# if defined HAVE_CUDA # if defined HAVE_CUDA
cudaError_t err = cudaGetLastError ( ) ;
cudaError_t err = cudaGetLastError ( ) ;
@ -129,91 +137,88 @@ bool CudaCompressor::isValid() const
return false ;
return false ;
}
}
# endif # endif
return m_data ! = NULL & & m_result ! = NULL & & m_bitmapTable ! = NULL ;
return bitmapTable ! = NULL & & bitmapTableCTX ! = NULL & & data ! = NULL & & result ! = NULL ;
} }
// @@ This code is very repetitive and needs to be cleaned up.
#if 0
struct CudaCompressionKernel CudaCompressor : : CudaCompressor ( CudaContext & ctx ) : m_ctx ( ctx )
{ {
virtual void setup ( const CompressionOptions : : Private & compressionOptions )
{
setupCompressKernel ( compressionOptions . colorWeight . ptr ( ) ) ;
}
virtual void setBitmapTable ( ) ;
virtual void runDeviceCode ( int count ) ;
virtual void runHostCode ( int count ) ;
} ;
void CudaCompressor : : compressKernel ( CudaCompressionKernel * kernel ) }
void CudaCompressor : : compress ( nvtt : : InputFormat inputFormat , nvtt : : AlphaMode alphaMode , uint w , uint h , void * data , const nvtt : : CompressionOptions : : Private & compressionOptions , const nvtt : : OutputOptions : : Private & outputOptions )
{ {
nvDebugCheck ( cuda : : isHardwarePresent ( ) ) ;
nvDebugCheck ( cuda : : isHardwarePresent ( ) ) ;
# if defined HAVE_CUDA # if defined HAVE_CUDA
// Allocate image as a cuda array.
cudaArray * d_image ;
if ( inputFormat = = nvtt : : InputFormat_BGRA_8UB )
{
cudaChannelFormatDesc channelDesc = cudaCreateChannelDesc ( 8 , 8 , 8 , 8 , cudaChannelFormatKindUnsigned ) ;
cudaMallocArray ( & d_image , & channelDesc , w , h ) ;
const int imageSize = w * h * sizeof ( uint ) ;
cudaMemcpyToArray ( d_image , 0 , 0 , data , imageSize , cudaMemcpyHostToDevice ) ;
}
else
{
# pragma message(NV_FILE_LINE "FIXME: Floating point textures not really supported by CUDA compressors.")
cudaChannelFormatDesc channelDesc = cudaCreateChannelDesc ( 32 , 32 , 32 , 32 , cudaChannelFormatKindFloat ) ;
cudaMallocArray ( & d_image , & channelDesc , w , h ) ;
const int imageSize = w * h * sizeof ( uint ) ;
cudaMemcpyToArray ( d_image , 0 , 0 , data , imageSize , cudaMemcpyHostToDevice ) ;
}
// Image size in blocks.
// Image size in blocks.
const uint w = ( image - > width ( ) + 3 ) / 4 ;
const uint bw = ( w + 3 ) / 4 ;
const uint h = ( image - > height ( ) + 3 ) / 4 ;
const uint bh = ( h + 3 ) / 4 ;
const uint bs = blockSize ( ) ;
const uint blockNum = bw * bh ;
const uint compressedSize = blockNum * bs ;
uint imageSize = w * h * 16 * sizeof ( Color32 ) ;
void * h_result = malloc ( min ( blockNum , MAX_BLOCKS ) * bs ) ;
uint * blockLinearImage = ( uint * ) malloc ( imageSize ) ;
convertToBlockLinear ( image , blockLinearImage ) ; // @@ Do this in parallel with the GPU, or in the GPU!
const uint blockNum = w * h ;
setup ( d_image , compressionOptions ) ;
const uint compressedSize = blockNum * 8 ;
clock_t start = clock ( ) ;
// Timer timer;
// timer.start();
kernel - > setup ( compressionOptions ) ;
kernel - > setBitmapTable ( m_bitmapTable ) ;
// TODO: Add support for multiple GPUs.
uint bn = 0 ;
uint bn = 0 ;
while ( bn ! = blockNum )
while ( bn ! = blockNum )
{
{
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
cudaMemcpy ( m_data , blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
compressBlocks ( bn , count , w , h , alphaMode , compressionOptions , h_result ) ;
kernel - > runDeviceCode ( count , m_data , m_result ) ;
kernel - > runHostCode ( count ) ;
// Check for errors.
// Check for errors.
cudaError_t err = cudaGetLastError ( ) ;
cudaError_t err = cudaGetLastError ( ) ;
if ( err ! = cudaSuccess )
if ( err ! = cudaSuccess )
{
{
nvDebug ( " CUDA Error: %s \n " , cudaGetErrorString ( err ) ) ;
//nvDebug("CUDA Error: %s\n", cudaGetErrorString(err));
if ( outputOptions . errorHandler ! = NULL )
if ( outputOptions . errorHandler ! = NULL )
{
{
outputOptions . errorHandler - > error ( Error_CudaError ) ;
outputOptions . errorHandler - > error ( Error_CudaError ) ;
}
}
}
}
// Copy result to host, overwrite swizzled image.
cudaMemcpy ( blockLinearImage , m_result , count * 8 , cudaMemcpyDeviceToHost ) ;
// Output result.
// Output result.
kernel - > outputResult ( outputOptions . outputHandler ) ;
if ( outputOptions . outputHandler ! = NULL )
if ( outputOptions . outputHandler ! = NULL )
{
{
outputOptions . outputHandler - > writeData ( blockLinearImage, count * 8 ) ;
outputOptions . outputHandler - > writeData ( h_result, count * bs ) ;
}
}
bn + = count ;
bn + = count ;
}
}
clock_t end = clock ( ) ;
//timer.stop();
//printf("\rCUDA time taken: %.3f seconds\n", float(end-start ) / CLOCKS_PER_SEC);
//printf("\rCUDA time taken: %.3f seconds\n", timer.elapsed( ) / CLOCKS_PER_SEC);
free ( blockLinearImage ) ;
free ( h_result ) ;
cudaFreeArray ( d_image ) ;
# else # else
if ( outputOptions . errorHandler ! = NULL )
if ( outputOptions . errorHandler ! = NULL )
@ -221,92 +226,88 @@ void CudaCompressor::compressKernel(CudaCompressionKernel * kernel)
outputOptions . errorHandler - > error ( Error_CudaError ) ;
outputOptions . errorHandler - > error ( Error_CudaError ) ;
}
}
# endif # endif
}
# endif // 0
}
void CudaCompressor : : setImage ( const Image * image , nvtt : : AlphaMode alphaMode ) void CudaCompressor DXT1: : setup ( cudaArray * image , const nvtt : : CompressionOptions : : Private & compressionOptions )
{ {
m_image = image ;
setupCompressKernel( compressionOptions . colorWeight . ptr ( ) ) ;
m_alphaMode = alphaMode ;
bindTextureToArray( image ) ;
} }
void CudaCompressorDXT1 : : compressBlocks ( uint first , uint count , uint w , uint h , nvtt : : AlphaMode alphaMode , const nvtt : : CompressionOptions : : Private & compressionOptions , void * output )
/// Compress image using CUDA.
void CudaCompressor : : compressDXT1 ( const CompressionOptions : : Private & compressionOptions , const OutputOptions : : Private & outputOptions )
{ {
nvDebugCheck ( cuda : : isHardwarePresent ( ) ) ;
// Launch kernel.
# if defined HAVE_CUDA compressKernelDXT1 ( first , count , w , m_ctx . result , m_ctx . bitmapTable ) ;
// Allocate image as a cuda array.
// Copy result to host.
cudaChannelFormatDesc channelDesc = cudaCreateChannelDesc ( 8 , 8 , 8 , 8 , cudaChannelFormatKindUnsigned ) ;
cudaMemcpy ( output , m_ctx . result , count * 8 , cudaMemcpyDeviceToHost ) ;
}
cudaArray * d_image ;
const int imageSize = m_image - > width ( ) * m_image - > height ( ) * sizeof ( uint ) ;
cudaMallocArray ( & d_image , & channelDesc , m_image - > width ( ) , m_image - > height ( ) ) ;
cudaMemcpyToArray ( d_image , 0 , 0 , m_image - > pixels ( ) , imageSize , cudaMemcpyHostToDevice ) ;
void CudaCompressorDXT3 : : setup ( cudaArray * image , const nvtt : : CompressionOptions : : Private & compressionOptions )
{
setupCompressKernel ( compressionOptions . colorWeight . ptr ( ) ) ;
bindTextureToArray ( image ) ;
}
// Image size in blocks.
void CudaCompressorDXT3 : : compressBlocks ( uint first , uint count , uint w , uint h , nvtt : : AlphaMode alphaMode , const nvtt : : CompressionOptions : : Private & compressionOptions , void * output )
const uint w = ( m_image - > width ( ) + 3 ) / 4 ;
{
const uint h = ( m_image - > height ( ) + 3 ) / 4 ;
// Launch kernel.
const uint blockNum = w * h ;
compressKernelDXT3 ( first , count , w , m_ctx . result , m_ctx . bitmapTable ) ;
const uint compressedSize = blockNum * 8 ;
void * h_result = malloc ( min ( blockNum , MAX_BLOCKS ) * 8 ) ;
// Copy result to host.
cudaMemcpy ( output , m_ctx . result , count * 16 , cudaMemcpyDeviceToHost ) ;
}
//clock_t start = clock();
void CudaCompressorDXT5 : : setup ( cudaArray * image , const nvtt : : CompressionOptions : : Private & compressionOptions )
{
setupCompressKernel ( compressionOptions . colorWeight . ptr ( ) ) ;
setupCompressKernel ( compressionOptions . colorWeight . ptr ( ) ) ;
bindTextureToArray ( image ) ;
}
uint bn = 0 ;
void CudaCompressorDXT5 : : compressBlocks ( uint first , uint count , uint w , uint h , nvtt : : AlphaMode alphaMode , const nvtt : : CompressionOptions : : Private & compressionOptions , void * output )
while ( bn ! = blockNum )
{
/*// Launch kernel.
compressKernelDXT5 ( first , count , w , m_ctx . result , m_ctx . bitmapTable ) ;
// Copy result to host.
cudaMemcpy ( output , m_ctx . result , count * 16 , cudaMemcpyDeviceToHost ) ; */
// Launch kernel.
if ( alphaMode = = AlphaMode_Transparency )
{
{
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
// compressWeightedKernelDXT1(first, count, w, m_ctx.result, m_ctx.bitmapTable);
}
else
{
// compressKernelDXT1_Level4(first, count, w, m_ctx.result, m_ctx.bitmapTable);
}
// Launch kernel.
// Compress alpha in parallel with the GPU.
compressKernelDXT1_Tex ( bn , count , w , d_image , m_result , m_bitmapTable ) ;
for ( uint i = 0 ; i < count ; i + + )
{
//ColorBlock rgba(blockLinearImage + (first + i) * 16);
//OptimalCompress::compressDXT3A(rgba, alphaBlocks + i);
}
// Check for errors.
// Copy result to host.
cudaError_t err = cudaGetLastError ( ) ;
cudaMemcpy ( output , m_ctx . result , count * 8 , cudaMemcpyDeviceToHost ) ;
if ( err ! = cudaSuccess )
{
nvDebug ( " CUDA Error: %s \n " , cudaGetErrorString ( err ) ) ;
if ( outputOptions . errorHandler ! = NULL )
// @@ Interleave color and alpha blocks.
{
outputOptions . errorHandler - > error ( Error_CudaError ) ;
}
}
// Copy result to host, overwrite swizzled image.
}
cudaMemcpy ( h_result , m_result , count * 8 , cudaMemcpyDeviceToHost ) ;
// Output result.
if ( outputOptions . outputHandler ! = NULL )
{
outputOptions . outputHandler - > writeData ( h_result , count * 8 ) ;
}
bn + = count ;
}
//clock_t end = clock();
//printf("\rCUDA time taken: %.3f seconds\n", float(end-start) / CLOCKS_PER_SEC);
free ( h_result ) ;
# else
if ( outputOptions . errorHandler ! = NULL )
{
outputOptions . errorHandler - > error ( Error_CudaError ) ;
}
# endif
}
// @@ This code is very repetitive and needs to be cleaned up.
#if 0
/// Compress image using CUDA.
/// Compress image using CUDA.
void CudaCompressor : : compressDXT3 ( const CompressionOptions : : Private & compressionOptions , const OutputOptions : : Private & outputOptions ) void CudaCompressor : : compressDXT3 ( const CompressionOptions : : Private & compressionOptions , const OutputOptions : : Private & outputOptions )
@ -337,16 +338,16 @@ void CudaCompressor::compressDXT3(const CompressionOptions::Private & compressio
{
{
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
cudaMemcpy ( m_ data, blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
cudaMemcpy ( m_ ctx. data, blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
// Launch kernel.
// Launch kernel.
if ( m_alphaMode = = AlphaMode_Transparency )
if ( m_alphaMode = = AlphaMode_Transparency )
{
{
compressWeightedKernelDXT1 ( count , m_ data, m_ result, m_ bitmapTable) ;
compressWeightedKernelDXT1 ( count , m_ ctx. data, m_ ctx. result, m_ ctx. bitmapTable) ;
}
}
else
else
{
{
compressKernelDXT1_Level4 ( count , m_ data, m_ result, m_ bitmapTable) ;
compressKernelDXT1_Level4 ( count , m_ ctx. data, m_ ctx. result, m_ ctx. bitmapTable) ;
}
}
// Compress alpha in parallel with the GPU.
// Compress alpha in parallel with the GPU.
@ -369,7 +370,7 @@ void CudaCompressor::compressDXT3(const CompressionOptions::Private & compressio
}
}
// Copy result to host, overwrite swizzled image.
// Copy result to host, overwrite swizzled image.
cudaMemcpy ( blockLinearImage , m_ result, count * 8 , cudaMemcpyDeviceToHost ) ;
cudaMemcpy ( blockLinearImage , m_ ctx. result, count * 8 , cudaMemcpyDeviceToHost ) ;
// Output result.
// Output result.
if ( outputOptions . outputHandler ! = NULL )
if ( outputOptions . outputHandler ! = NULL )
@ -428,16 +429,16 @@ void CudaCompressor::compressDXT5(const CompressionOptions::Private & compressio
{
{
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
cudaMemcpy ( m_ data, blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
cudaMemcpy ( m_ ctx. data, blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
// Launch kernel.
// Launch kernel.
if ( m_alphaMode = = AlphaMode_Transparency )
if ( m_alphaMode = = AlphaMode_Transparency )
{
{
compressWeightedKernelDXT1 ( count , m_ data, m_ result, m_ bitmapTable) ;
compressWeightedKernelDXT1 ( count , m_ ctx. data, m_ ctx. result, m_ ctx. bitmapTable) ;
}
}
else
else
{
{
compressKernelDXT1_Level4 ( count , m_ data, m_ result, m_ bitmapTable) ;
compressKernelDXT1_Level4 ( count , m_ ctx. data, m_ ctx. result, m_ ctx. bitmapTable) ;
}
}
// Compress alpha in parallel with the GPU.
// Compress alpha in parallel with the GPU.
@ -460,7 +461,7 @@ void CudaCompressor::compressDXT5(const CompressionOptions::Private & compressio
}
}
// Copy result to host, overwrite swizzled image.
// Copy result to host, overwrite swizzled image.
cudaMemcpy ( blockLinearImage , m_ result, count * 8 , cudaMemcpyDeviceToHost ) ;
cudaMemcpy ( blockLinearImage , m_ ctx. result, count * 8 , cudaMemcpyDeviceToHost ) ;
// Output result.
// Output result.
if ( outputOptions . outputHandler ! = NULL )
if ( outputOptions . outputHandler ! = NULL )
@ -516,10 +517,10 @@ void CudaCompressor::compressDXT1n(const nvtt::CompressionOptions::Private & com
{
{
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
cudaMemcpy ( m_ data, blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
cudaMemcpy ( m_ ctx. data, blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
// Launch kernel.
// Launch kernel.
compressNormalKernelDXT1 ( count , m_ data, m_ result, m_ bitmapTable) ;
compressNormalKernelDXT1 ( count , m_ ctx. data, m_ ctx. result, m_ ctx. bitmapTable) ;
// Check for errors.
// Check for errors.
cudaError_t err = cudaGetLastError ( ) ;
cudaError_t err = cudaGetLastError ( ) ;
@ -534,7 +535,7 @@ void CudaCompressor::compressDXT1n(const nvtt::CompressionOptions::Private & com
}
}
// Copy result to host, overwrite swizzled image.
// Copy result to host, overwrite swizzled image.
cudaMemcpy ( blockLinearImage , m_ result, count * 8 , cudaMemcpyDeviceToHost ) ;
cudaMemcpy ( blockLinearImage , m_ ctx. result, count * 8 , cudaMemcpyDeviceToHost ) ;
// Output result.
// Output result.
if ( outputOptions . outputHandler ! = NULL )
if ( outputOptions . outputHandler ! = NULL )
@ -585,10 +586,10 @@ void CudaCompressor::compressCTX1(const nvtt::CompressionOptions::Private & comp
{
{
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
uint count = min ( blockNum - bn , MAX_BLOCKS ) ;
cudaMemcpy ( m_ data, blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
cudaMemcpy ( m_ ctx. data, blockLinearImage + bn * 16 , count * 64 , cudaMemcpyHostToDevice ) ;
// Launch kernel.
// Launch kernel.
compressKernelCTX1 ( count , m_ data, m_ result, m_ bitmapTableCTX) ;
compressKernelCTX1 ( count , m_ ctx. data, m_ ctx. result, m_ ctx. bitmapTableCTX) ;
// Check for errors.
// Check for errors.
cudaError_t err = cudaGetLastError ( ) ;
cudaError_t err = cudaGetLastError ( ) ;
@ -603,7 +604,7 @@ void CudaCompressor::compressCTX1(const nvtt::CompressionOptions::Private & comp
}
}
// Copy result to host, overwrite swizzled image.
// Copy result to host, overwrite swizzled image.
cudaMemcpy ( blockLinearImage , m_ result, count * 8 , cudaMemcpyDeviceToHost ) ;
cudaMemcpy ( blockLinearImage , m_ ctx. result, count * 8 , cudaMemcpyDeviceToHost ) ;
// Output result.
// Output result.
if ( outputOptions . outputHandler ! = NULL )
if ( outputOptions . outputHandler ! = NULL )
@ -643,4 +644,4 @@ void CudaCompressor::compressDXT5n(const nvtt::CompressionOptions::Private & com
# endif # endif
} }
# endif // 0
castano
15 years ago