Module gpu_sort 

GPU-style parallel sorting algorithms (CPU simulation).

This module provides GPU-pattern algorithms for sorting and related operations on f32/u32 data:

• Bitonic sort (power-of-2 padded)
• LSD radix sort for u32 and f32
• Exclusive prefix sum (Blelloch scan style)
• Parallel histogram
• Counting sort
• Morton Z-order sort for 3-D point clouds
• GpuSortBuffer — key+value buffer with sort_pairs
• Parallel merge

Structs

GpuSortBuffer

A buffer abstraction that holds parallel key and value arrays.

Functions

bitonic_sort

Sort a Vecf32` in ascending order using the bitonic sort algorithm.

bitonic_sort_by_key

Sort a VecTby theu32key returned bykey_fn`.

counting_sort

Counting sort for u32 values bounded by max_val (inclusive).

histogram

Compute a histogram of data into n_bins equal-width bins over \[min_val, max_val).

morton_sort_3d

Sort 3-D f32 points by their Morton Z-order (space-filling curve) code.

parallel_merge

Merge two sorted f32 slices into a new sorted Vecf32`.

prefix_sum

Exclusive prefix sum (Blelloch scan) for u32 values.

radix_sort_f32

Sort f32 values by reinterpreting bits and flipping the sign bit for negatives so that the full IEEE 754 ordering maps to unsigned integer order.

radix_sort_u32

LSD radix sort for u32 using 4 passes of 8-bit digits.