Function convertToRGBA16Float

Source/Falcor/Utils/Image/Bitmap.cpp:249–285

* Converts 96/128bpp to 64bpp RGBA floating-point image. * Note that FreeImage doesn't support 16-bit float formats. */

 * Note that FreeImage doesn't support 16-bit float formats.
 */
static FIBITMAP* convertToRGBA16Float(FIBITMAP* pDib)
{
    const auto type = FreeImage_GetImageType(pDib);
    const uint32_t bpp = FreeImage_GetBPP(pDib);
    FALCOR_CHECK(type == FIT_RGBF || type == FIT_RGBAF, "Image type must be RGB/RGBA with 32-bit float per channel.");
    FALCOR_CHECK(bpp == 96 || bpp == 128, "Image must be 96 or 128bpp.");

    const uint32_t width = FreeImage_GetWidth(pDib);
    const uint32_t height = FreeImage_GetHeight(pDib);

    auto pNew = FreeImage_AllocateT(FIT_RGBA16, width, height);
    FreeImage_CloneMetadata(pNew, pDib);

    const uint32_t src_pitch = FreeImage_GetPitch(pDib);
    const uint32_t dst_pitch = FreeImage_GetPitch(pNew);

    const BYTE* src_bits = (BYTE*)FreeImage_GetBits(pDib);
    BYTE* dst_bits = (BYTE*)FreeImage_GetBits(pNew);

    for (uint32_t y = 0; y < height; y++)
    {
        const FIRGBAF* src_pixel = (FIRGBAF*)src_bits;
        FIRGBA16* dst_pixel = (tagFIRGBA16*)dst_bits;

        for (uint32_t x = 0; x < width; x++)
        {
            // Convert pixels to float16_t directly, while adding a "dummy" alpha of 1.0 if source format doesn't have alpha.
            dst_pixel[x].red = float16_t(src_pixel[x].red).toBits();
            dst_pixel[x].green = float16_t(src_pixel[x].green).toBits();
            dst_pixel[x].blue = float16_t(src_pixel[x].blue).toBits();
            dst_pixel[x].alpha = float16_t(type == FIT_RGBAF ? src_pixel[x].alpha : 1.0f).toBits();
        }
        src_bits += src_pitch;
        dst_bits += dst_pitch;
    }
    return pNew;
}
Bitmap::UniqueConstPtr Bitmap::create(uint32_t width, uint32_t height, ResourceFormat format, const uint8_t* pData)
{
    return Bitmap::UniqueConstPtr(new Bitmap(width, height, format, pData));