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

Add nvassemble tool to create cubemaps, volumes, and texture arrays.

Browse Source
This commit is contained in:
castano
2007-11-16 11:05:17 +00:00
parent 98b2377a11
commit 854da30b8e
11 changed files with 542 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/nvtt/cuda/CompressKernel.cu
View File

@ -294,6 +294,7 @@ __device__ float evalPermutation4(const float3 * colors, float3 color_sum, uint
uint akku = 0;
// Compute alpha & beta for this permutation.
#pragma unroll
for (int i = 0; i < 16; i++)
{
const uint bits = permutation >> (2*i);
@ -329,6 +330,7 @@ __device__ float evalPermutation3(const float3 * colors, float3 color_sum, uint
uint akku = 0;
// Compute alpha & beta for this permutation.
#pragma unroll
for (int i = 0; i < 16; i++)
{
const uint bits = permutation >> (2*i);

6
src/nvtt/cuda/CudaCompressDXT.cpp
View File

@ -109,12 +109,14 @@ void nv::cudaCompressDXT1(const Image * image, const OutputOptions & outputOptio
uint imageSize = w * h * 16 * sizeof(Color32);
uint * blockLinearImage = (uint *) malloc(imageSize);
convertToBlockLinear(image, blockLinearImage);
convertToBlockLinear(image, blockLinearImage); // @@ Do this on the GPU!
const uint blockNum = w * h;
const uint compressedSize = blockNum * 8;
const uint blockMax = 32768; // 49152, 65535
clock_t start = clock();
// Allocate image in device memory.
uint * d_data = NULL;
cudaMalloc((void**) &d_data, min(imageSize, blockMax * 64U));
@ -125,8 +127,6 @@ void nv::cudaCompressDXT1(const Image * image, const OutputOptions & outputOptio
setupCompressKernel(compressionOptions.colorWeight.ptr());
clock_t start = clock();
// TODO: Add support for multiple GPUs.
uint bn = 0;
while(bn != blockNum)