drewcassidy/nvidia-texture-tools
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Add missing changes.

Browse Source
This commit is contained in:
castano 2008-02-28 22:34:18 +00:00
parent ebe47a0f78
commit 3dae7f9b70
2 changed files with 20 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
src/nvtt/cuda/CudaCompressDXT.cpp
View File

@ -48,7 +48,8 @@ using namespace nvtt;
#if defined HAVE_CUDA #if defined HAVE_CUDA
#define MAX_BLOCKS 32768 // 49152, 65535 #define MAX_BLOCKS 8192U // 32768, 65535
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 compressKernelDXT1(uint blockNum, uint * d_data, uint * d_result, uint * d_bitmaps);
@ -79,12 +80,15 @@ static void convertToBlockLinear(const Image * image, uint * blockLinearImage)
#endif #endif
CudaCompressor::CudaCompressor() CudaCompressor::CudaCompressor() : m_bitmapTable(NULL), m_data(NULL), m_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**) &m_bitmapTable, 992 * sizeof(uint));
if (m_bitmapTable != NULL)
{
cudaMemcpy(m_bitmapTable, s_bitmapTable, 992 * sizeof(uint), cudaMemcpyHostToDevice); cudaMemcpy(m_bitmapTable, s_bitmapTable, 992 * sizeof(uint), cudaMemcpyHostToDevice);
}
// Allocate scratch buffers. // Allocate scratch buffers.
cudaMalloc((void**) &m_data, MAX_BLOCKS * 64U); cudaMalloc((void**) &m_data, MAX_BLOCKS * 64U);
@ -102,7 +106,16 @@ CudaCompressor::~CudaCompressor()
#endif #endif
} }
bool CudaCompressor::isValid() const
{
#if defined HAVE_CUDA
if (cudaGetLastError() != cudaSuccess)
{
return false;
}
#endif
return m_data != NULL && m_result != NULL && m_bitmapTable != NULL;
}
// @@ This code is very repetitive and needs to be cleaned up. // @@ This code is very repetitive and needs to be cleaned up.
@ -119,7 +132,7 @@ void CudaCompressor::compressDXT1(const Image * image, const OutputOptions::Priv
uint imageSize = w * h * 16 * sizeof(Color32); uint imageSize = w * h * 16 * sizeof(Color32);
uint * blockLinearImage = (uint *) malloc(imageSize); uint * blockLinearImage = (uint *) malloc(imageSize);
convertToBlockLinear(image, blockLinearImage); // @@ Do this on the GPU! convertToBlockLinear(image, blockLinearImage); // @@ Do this in parallel with the GPU, or in the GPU!
const uint blockNum = w * h; const uint blockNum = w * h;
const uint compressedSize = blockNum * 8; const uint compressedSize = blockNum * 8;

2
src/nvtt/cuda/CudaCompressDXT.h
View File

@ -37,6 +37,8 @@ namespace nv
CudaCompressor(); CudaCompressor();
~CudaCompressor(); ~CudaCompressor();
bool isValid() const;
void compressDXT1(const Image * image, const nvtt::OutputOptions::Private & outputOptions, const nvtt::CompressionOptions::Private & compressionOptions); void compressDXT1(const Image * image, const nvtt::OutputOptions::Private & outputOptions, const nvtt::CompressionOptions::Private & compressionOptions);
void compressDXT3(const Image * image, const nvtt::OutputOptions::Private & outputOptions, const nvtt::CompressionOptions::Private & compressionOptions); void compressDXT3(const Image * image, const nvtt::OutputOptions::Private & outputOptions, const nvtt::CompressionOptions::Private & compressionOptions);
void compressDXT5(const Image * image, const nvtt::OutputOptions::Private & outputOptions, const nvtt::CompressionOptions::Private & compressionOptions); void compressDXT5(const Image * image, const nvtt::OutputOptions::Private & outputOptions, const nvtt::CompressionOptions::Private & compressionOptions);