* @brief Compute new cluster centers based on their center of gravity. * * @param blk The image block color data to compress.

See header for documentation. */

Load a reference result set from a CSV file on disk. Args: file_path: The input file path.

* @brief Load a KTX compressed image using the local custom loader. * * @param filename The name of the file to load. * @param[out] i

* @brief Load a non-astc image file from memory. * * @param filename The file to load, or a pattern for array loads. * @param di

* @brief Unpack an LDR LA color that uses delta encoding. * * @param input The packed endpoints (L0, L1, A0, A1). * @param[out] output0

* @brief Unpack an LDR LA color that uses direct encoding. * * @param input The packed endpoints (L0, L1, A0, A1). * @param[out] output0

* @brief Unpack an LDR L color that uses delta encoding. * * Output alpha is 255. * * @param input The packed endpoints (L0, L1). * @par

* @brief Unpack an LDR L color that uses direct encoding. * * Output alpha is 255. * * @param input The packed endpoints. * @param[out]

The main function. Return: int: The process return code.

The main function. Return: int: The process return code.

The main function. Return: The process return code.

The main function. Return: int: The process return code.

The main function. Return: The process return code.

The main function. Return: The process return code.

Modify an existing image on disk to make it landscape aspect. Args: file_path: The path of the image on disk.

Make a single mosaic montage consisting of all of the Kodak images. Args: src_dir_path: The directory path of the Kodak images on di

* @brief Return the max vector of two vectors. */

* @brief Merge two planes of endpoints into a single vector. * * @param ep_plane1 The endpoints for plane 1. * @param ep_plane2

* @brief Return the min vector of two vectors. */

* @brief Return the min vector of two vectors. */

* @brief For one partition compute the best format and quantization for a given bit count. * * @param best_combined_error The best error for

Parse the command line. Return: The parsed command line.

Parse the command line. Return: Namespace: The parsed command line container.

Parse the command line. Return: The parsed command line.

* @brief Parse the command line. * * @param argc Command line argument count. * @param[in] argv Command line argument vector.

Parse the log output for PSNR and performance metrics. Args: image: The test image that was compressed. outp

* @brief Compute bit-mismatch for partitioning in 2-partition mode. * * @param a The texel assignment bitvector for the block. * @param b The t

* @brief Compute bit-mismatch for partitioning in 3-partition mode. * * @param a The texel assignment bitvector for the block. * @param b The t

* @brief Compute bit-mismatch for partitioning in 4-partition mode. * * @param a The texel assignment bitvector for the block. * @param b The t

See header for documentation. */

* @brief Print the config settings in a human readable form. * * @param[in] cli_config Command line config. * @param[in] config Codec confi

* @brief Print a set of diagnostic images showing aspects of the encoding. * * @param context The context to use. * @param image The

* @brief Proxy Windows @c CreateThread underneath a pthreads-like wrapper. */

* @brief Proxy Windows @c WaitForSingleObject underneath a pthreads-like wrapper. */

* @brief Quantize a HDR A color using either delta or direct RGBA encoding. * * @param alpha0 The input unquantized color0 endpoint. *

* @brief Quantize a HDR L color using the large range encoding. * * @param color0 The input unquantized color0 endpoint. * @param

* @brief Quantize a HDR RGBA color using either delta or direct RGBA encoding. * * @param color0 The input unquantized color0 endpoint.

* @brief Quantize a HDR RGB + LDR A color using direct RGBA encoding. * * @param color0 The input unquantized color0 endpoint. * @param

* @brief Quantize a HDR RGB color using RGB + offset. * * @param color The input unquantized color endpoint and offset. * @param[out]

* @brief Quantize a LDR L color. * * @param color0 The input unquantized color0 endpoint. * @param color1 The input unquant

* @brief Quantize a LDR LA color. * * @param color0 The input unquantized color0 endpoint. * @param color1 The input unquan

* @brief Quantize an LDR RGBA color. * * Since this is a fall-back encoding, we cannot actually fail but must produce a sensible result. * For this

* @brief Quantize an LDR RGBA color using scale encoding. * * @param color0 The input unquantized color0 alpha endpoint. * @param c

Generate a summary of hot functions in a callgrind_annotate log. Args: lines: The output of callgrind_annotate. Return:

See header for documentation. */

See header for documentation. */

Download the public domain Kodak image set. To make test set mosaics easier to build we rotate images to make everything landscape.

* @brief Unpack an LDR RGB color that uses delta encoding. * * Output alpha set to 255. * * @param input0 The packed endpoint 0 color. *

* @brief Unpack an LDR RGBA color that uses scaled encoding. * * Note only the RGB channels use the scaled encoding, alpha uses direct. * * @param

* @brief Unpack an LDR RGB color that uses scaled encoding. * * Output alpha is 255. * * @param input0 The packed endpoint 0 color. * @pa

* @brief Unpack an LDR RGB color that uses direct encoding. * * Output alpha set to 255. * * @param input0 The packed endpoint 0 color. *

Run Valgrind on a single binary. Args: image: The path of the image to compress. encoder: The encoder variant to use.

Run the test N times. Args: image: The test image to compress. block_size: The block size to use.

Execute one test image in the test set. Args: encoder: The encoder to test. reference: The reference test results to compare

Execute all tests in the test set. Args: encoder: The encoder to test. reference: The reference test results to compare agai

Save this result set to a CSV file. Args: file_path: The output file path.

* @brief Select texel assignment for a single coordinate. * * @param seed The seed - the partition index from the block. * @param x

* @brief Manually set CPU group and thread affinity. * * This is needed on Windows 10 or older to allow benefit from large core count * systems wit

* @brief Set a 1 plane weight quant as active. * * @param weight_quant The quant method to set. */

* @brief Set a 2 plane weight quant as active. * * @param weight_quant The quant method to set. */

Set relative results compared to an existing reference score. Args: reference: The reference result to compare against.

Set the result status. Args: result: The test result status to set.

convert from FP16 to FP32. */

Conversion routine that converts from FP32 to FP16. It supports denormals and all rounding modes. If a NaN is given as input, it is quietened. */

build a table that decodes both magnitude and value of small ACs in one go.

decode image to YCbCr format

@OPTIMIZE: store non-zigzagged during the decode passes, and only de-zigzag when dequantizing