// notes:
// Slow CPU based decompression : Should look into also using HW based decompression with this interface
//
CodecError CCodec_ASTC::Decompress(CCodecBuffer& bufferIn, CCodecBuffer& bufferOut, Codec_Feedback_Proc pFeedbackProc, CMP_DWORD_PTR pUser1, CMP_DWORD_PTR pUser2)
{
m_xdim = bufferIn.GetBlockWidth();
m_ydim = bufferIn.GetBlockHeight();
m_zdim = 1;
CodecError err = InitializeASTCLibrary();
if (err != CE_OK) return err;
// Our Compressed data Blocks are always 128 bit long (4x4 blocks)
const CMP_DWORD imageWidth = bufferIn.GetWidth();
const CMP_DWORD imageHeight = bufferIn.GetHeight();
const CMP_DWORD imageDepth = 1;
const BYTE bitness = 8;
const CMP_DWORD CompBlockX = bufferIn.GetBlockWidth();
const CMP_DWORD CompBlockY = bufferIn.GetBlockHeight();
CMP_BYTE Block_Width = bufferIn.GetBlockWidth();
CMP_BYTE Block_Height = bufferIn.GetBlockHeight();
const CMP_DWORD dwBlocksX = ((bufferIn.GetWidth() + (CompBlockX - 1)) / CompBlockX);
const CMP_DWORD dwBlocksY = ((bufferIn.GetHeight()+ (CompBlockY - 1)) / CompBlockY);
const CMP_DWORD dwBlocksZ = 1;
const CMP_DWORD dwBufferInDepth = 1;
// Override the current input buffer Pitch size (Since it will be set according to the Compressed Block Sizes
// and not to the Compressed Codec data which is for ASTC 16 Bytes per block x Number of blocks per row
bufferIn.SetPitch(16 * dwBlocksX);
// Output data size Pitch
CMP_DWORD dwPitch = bufferOut.GetPitch();
// Output Buffer
BYTE *pDataOut = bufferOut.GetData();
const CMP_DWORD dwBlocksXY = dwBlocksX*dwBlocksY;
for(CMP_DWORD cmpRowY = 0; cmpRowY < dwBlocksY; cmpRowY++) // Compressed images row = height
{
for(CMP_DWORD cmpColX = 0; cmpColX < dwBlocksX; cmpColX++) // Compressed images Col = width
{
union FBLOCKS
{
float decodedBlock[144][4]; // max 12x12 block size
float destBlock[576]; // max 12x12x4
} DecData;
union BBLOCKS
{
CMP_DWORD compressedBlock[4];
BYTE out[16];
BYTE in[16];
} CompData;
bufferIn.ReadBlock(cmpColX*4, cmpRowY*4, CompData.compressedBlock, 4);
// Encode to the appropriate location in the compressed image
m_decoder->DecompressBlock(Block_Width, Block_Height, bitness, DecData.decodedBlock,CompData.in);
// Now that we have a decoded block lets copy that data over to the target image buffer
CMP_DWORD outCol = cmpColX*Block_Width;
CMP_DWORD outRow = cmpRowY*Block_Height;
CMP_DWORD outImgRow = outRow;
CMP_DWORD outImgCol = outCol;
for (int row = 0; row < Block_Height; row++)
{
CMP_DWORD nextRowCol = (outRow+row)*dwPitch + (outCol * 4);
CMP_BYTE* pData = (CMP_BYTE*)(pDataOut + nextRowCol);
if ((outImgRow + row) < imageHeight)
{
outImgCol = outCol;
for (int col = 0; col < Block_Width; col++)
{
CMP_DWORD w = outImgCol + col;
if (w < imageWidth)
{
int index = row*Block_Width + col;
*pData++ = (CMP_BYTE)DecData.decodedBlock[index][BC_COMP_RED];
*pData++ = (CMP_BYTE)DecData.decodedBlock[index][BC_COMP_GREEN];
*pData++ = (CMP_BYTE)DecData.decodedBlock[index][BC_COMP_BLUE];
*pData++ = (CMP_BYTE)DecData.decodedBlock[index][BC_COMP_ALPHA];
}
else break;
}
}
}
}
if (pFeedbackProc)
{
float fProgress = 100.f * (cmpRowY * dwBlocksX) / dwBlocksXY;
if (pFeedbackProc(fProgress, pUser1, pUser2))
{
return CE_Aborted;
}
}
}
return CE_OK;
}
CodecError CCodec_ASTC::Compress(CCodecBuffer& bufferIn, CCodecBuffer& bufferOut, Codec_Feedback_Proc pFeedbackProc, CMP_DWORD_PTR pUser1, CMP_DWORD_PTR pUser2)
{
m_AbortRequested = false;
int xsize = bufferIn.GetWidth();
int ysize = bufferIn.GetHeight();
int zsize = 1; //todo: add depth to support 3d textures
m_xdim = bufferOut.GetBlockWidth();
m_ydim = bufferOut.GetBlockHeight();
m_zdim = 1;
CodecError err = InitializeASTCLibrary();
if (err != CE_OK) return err;
#ifdef ASTC_COMPDEBUGGER
CompViewerClient g_CompClient;
if (g_CompClient.connect())
{
#ifdef USE_DBGTRACE
DbgTrace(("-------> Remote Server Connected"));
#endif
}
#endif
#ifdef USE_DBGTRACE
DbgTrace(("IN : BufferType %d ChannelCount %d ChannelDepth %d", bufferIn.GetBufferType(), bufferIn.GetChannelCount(), bufferIn.GetChannelDepth()));
DbgTrace((" : Height %d Width %d Pitch %d isFloat %d", bufferIn.GetHeight(), bufferIn.GetWidth(), bufferIn.GetWidth(), bufferIn.IsFloat()));
DbgTrace(("OUT: BufferType %d ChannelCount %d ChannelDepth %d", bufferOut.GetBufferType(), bufferOut.GetChannelCount(), bufferOut.GetChannelDepth()));
DbgTrace((" : Height %d Width %d Pitch %d isFloat %d", bufferOut.GetHeight(), bufferOut.GetWidth(), bufferOut.GetWidth(), bufferOut.IsFloat()));
#endif;
int bitness = 0; //todo: replace astc_codec_image with bufferIn and rewrite fetch_imageblock()
switch (bufferIn.GetBufferType())
{
case CBT_RGBA8888:
case CBT_BGRA8888:
case CBT_ARGB8888:
case CBT_RGB888:
case CBT_RG8:
case CBT_R8:
bitness = 8;
break;
case CBT_RGBA2101010:
break;
case CBT_RGBA16:
case CBT_RG16:
case CBT_R16:
break;
case CBT_RGBA32:
case CBT_RG32:
case CBT_R32:
break;
case CBT_RGBA16F:
case CBT_RG16F:
case CBT_R16F:
break;
case CBT_RGBA32F:
case CBT_RG32F:
case CBT_R32F:
break;
default:
break;
}
if (bitness != 8)
assert("Unsupported type of input buffer");
astc_codec_image_cpu *input_image = allocate_image_cpu(bitness, xsize, ysize, zsize, 0);
if (!input_image)
assert("Unable to allocate image buffer");
// Loop through the original input image and setup compression threads for each
// block to encode we will load the buffer to pass to ASTC code as 8 bit 4x4 blocks
// the fill in source image. ASTC code will then use the adaptive sizes for process on the input
BYTE *pData = bufferIn.GetData();
int ii = 0;
for (int y = 0; y < ysize; y++) {
for (int x = 0; x < xsize; x++) {
input_image->imagedata8[0][y][4*x ] = pData[ii]; // Red
ii++;
input_image->imagedata8[0][y][4 * x + 1] = pData[ii]; // Green
ii++;
input_image->imagedata8[0][y][4 * x + 2] = pData[ii]; // Blue
ii++;
input_image->imagedata8[0][y][4 * x + 3] = pData[ii]; // Alpha
ii++;
}
}
m_NumEncodingThreads = min(m_NumThreads, (decltype(m_NumThreads))MAX_ASTC_THREADS);
if (m_NumEncodingThreads == 0) m_NumEncodingThreads = 1;
// Common ARM and AMD Code
CodecError result = CE_OK;
int xdim = m_xdim;
int ydim = m_ydim;
int zdim = m_zdim;
uint8_t *bufferOutput = bufferOut.GetData();
// Common ARM and Compressonator Code
int x, y, z, i;
int xblocks = (xsize + xdim - 1) / xdim;
int yblocks = (ysize + ydim - 1) / ydim;
int zblocks = (zsize + zdim - 1) / zdim;
float TotalBlocks = (float) (yblocks * xblocks);
int processingBlock = 0;
for (z = 0; z < zblocks; z++)
{
for (y = 0; y < yblocks; y++)
{
for (x = 0; x < xblocks; x++)
{
int offset = ((z * yblocks + y) * xblocks + x) * 16;
uint8_t *bp = bufferOutput + offset;
EncodeASTCBlock((astc_codec_image *)input_image, bp, xdim, ydim, zdim, x * xdim, y * ydim, z * zdim);
processingBlock++;
}
if (pFeedbackProc)
{
float fProgress = 100.f * ((float)(processingBlock) / TotalBlocks);
if (pFeedbackProc(fProgress, pUser1, pUser2))
{
result = CE_Aborted;
break;
}
}
}
}
CodecError EncodeResult = FinishASTCEncoding();
if (result != CE_Aborted)
result = EncodeResult;
destroy_image_cpu(input_image);
#ifdef ASTC_COMPDEBUGGER
g_CompClient.disconnect();
#endif
return result;
}