See header for documentation. */

* @brief Return @c a with lanes negated if the @c b lane is negative. */

* @brief Return the dot product for the bottom 3 lanes, returning scalar. */

* @brief Return a integer value for a float vector, using truncation. */

See header for documentation. */

* @brief Compute the error of an LDR RGB or RGBA encoding. * * @param uquant0 The original endpoint 0 color. * @param uquant1 The original en

* @brief Get the weight quantization used by this block mode. * * @return The quantization level. */

* @brief Determine the quantized value given a quantization level. * * @param quant_level The quantization level to use. * @param value T

take a -128..127 value and stbi__clamp it and convert to 0..255

returns 1 if "a*b*c + add" has no negative terms/factors and doesn't overflow

* @brief Return fast, but approximate, vector atan(x). * * Max error of this implementation is 0.004883. */

* @brief Return fast, but approximate, vector atan2(x, y). */

* @brief Compute the infilled weight for N texel indices in a decimated grid. * * @param di The weight grid decimation to use. * @param weig

* @brief Compute the infilled weight for N texel indices in a decimated grid. * * This is specialized version which computes only two weights per te

* @brief Clamp a float value between 0.0f and 1.0f. * * NaNs are turned into 0.0f. * * @param v The value to clamp. * * @return The clamped va

initialize a callback-based context

initialize a memory-decode context

Get current position

See header for documentation. */

* @brief Generate a prefix-sum array using the Brent-Kung algorithm. * * This will take an input array of the form: * v0, v1, v2, ... * ... an

* @brief Test if a string ends with a given suffix. */

* @brief SP float bit-interpreted as an integer. * * @param v The value to bitcast. * * @return The converted value. */

* @brief Return a float16 value for a float vector, using round-to-nearest. */

Get an expected reference color for a tile image corner. Args: mode: The color mode (LDR, or HDR) corners: T

See header for documentation. */

Test if a given file format is supported by the library. Args: file_format: The target file extension, excluding the '.'

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

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

* @brief Quantize and unquantize a value ensuring top four bits are the same. * * @param quant_level The quantization level to use. * @par

* @brief Return a float rounded to the nearest integer value. */

* @brief Fast approximation of 1.0 / sqrt(val). * * @param v The input value. * * @return The approximated result. */

returns 1 if the product is valid, 0 on overflow. negative factors are considered invalid.

* @brief Store a single vector lane to an unaligned address. */

* @brief Store a single vector lane to an unaligned address. */

* @brief Store a single vector lane to an unaligned address. */

* @brief Store lowest N (vector width) bytes into an unaligned address. */

Helper to find custom attribute by name

See header for documentation. */

* @brief Clamp a float value between 0.0f and 255.0f. * * NaNs are turned into 0.0f. * * @param v The value to clamp. * * @return The clamped v

* @brief Compute the ideal endpoints and weights for 3 color components. * * @param blk The image block color data to compress.

* @brief Return the dot product for the bottom 3 lanes, returning vector. */

* @brief Return an FP32 NaN value for use in error colors. * * This NaN encoding will turn into 0xFFFF when converted to an FP16 NaN. * * @return

Get the path of a specific reference image on disk. Args: image: The image to load. Return: The pat

* @brief Determine the output file name to use for a sliced image write. * * @param img The source data for the image. * @param filename T

* @brief The number of levels use by an ASTC quantization method. * * @param method The quantization method * * @return The number of levels u

* @brief Get a vector mask indicating lanes decompressing into a UNORM8 value. * * @param decode_mode The color profile for LDR_SRGB settings. *

* @brief Integer bit-interpreted as an SP float. * * @param v The value to bitcast. * * @return The converted value. */

Get total length

* @brief Pack low 8 bits of N (vector width) lanes into bottom of vector. */

* @brief Quantize and unquantize a value ensuring top two bits are the same. * * @param quant_level The quantization level to use. * @para

* @brief Reverse the bytes in a uint32_t value. */

clean up the temporary component buffers

decode a jpeg huffman value from the bitstream

after a restart interval, stbi__jpeg_reset the entropy decoder and the dc prediction

mallocs with size overflow checking

initialize a callback-based context

See header for documentation. */

Returns whether this was successfully opened.

* @brief Factory that returns a 2 component vfloat4. */

Helper to add a string attribute

* @brief Compute the ideal endpoints and weights for 1 color component. * * @param blk The image block color data to compress. * @para

* @brief Detect platform CPU ISA support and update global trackers. */

* @brief Insert a partitioning into an order list of results, sorted by error. * * @param max_values The max number of entries in the best

* @brief Test if a single color channel is constant across the block. * * Constant color channels are easier to compress as interpolating between

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

* @brief Return the max vector of a vector and a scalar. * * If either lane value is NaN, @c b will be returned for that lane. */

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

* @brief mPSNR difference between two values. * * Differences are given as "val1 - val2". * * @param val1 The first color value * @param va